Research Methods in Politics

Research Methods in Politics
A practical guide
Roger Pierce
Los Angeles • London • New Delhi • Singapore

To
David, Kate and Tom Pierce
© Roger Pierce 2008
First Published 2008
Apart from any fair dealing for the purposes of research or private study, or criticism or
review, as permitted under the Copyright, Designs and Patents Act, 1988, this publication
may be reproduced, stored or transmitted in any form, or by any means, only with the
prior permission in writing of the publishers, or in the case of reprographic reproduction,
in accordance with the terms of licences issued by the Copyright Licensing Agency.
Enquiries concerning reproduction outside those terms should be sent to the publishers.
SAGE Publications Ltd

1 Oliver’s Yard
55 City Road
London EC1Y 1SP
SAGE Publications Inc.

2455 Teller Road
Thousand Oaks, California 91320
SAGE Publications India Pvt Ltd

B 1/I 1 Mohan Cooperative Industrial Area
Mathura Road
New Delhi 110 044
SAGE Publications Asia-Pacific Pte Ltd

33 Pekin Street #02-01
Far East Square
Singapore 048763
Library of Congress Control Number: 2007934984
British Library Cataloguing in Publication data
A catalogue record for this book is available from the British Library
ISBN 978-1-4129-3550-0
ISBN 978-1-4129-3551-7 (pbk)
Typeset by CEPHA Imaging Pvt. Ltd., Bangalore, India

Printed and bound in Great Britain by TJ International Ltd
Printed on paper from sustainable resources
Contents
About This Book v
Preface x
PART I INTRODUCTION 1
1 Introduction 3
2 Power in Research, Ethics, Data Protection and Bias 9
3 The Philosophy and Principles of Research 22
PART II METHODOLOGIES 39
4 Qualitative Versus Quantitative Methods: A Relevant

Argument? 41
5 Collecting Data Sets: Case Studies, Experimental,

Comparative, Longitudinal and Action Research Methods 51
PART III COLLECTING INFORMATION 67
6 Critically Evaluating Published Research 69
7 Evaluating Information: Validity, Reliability, Accuracy,

Triangulation 79
8 Completing a Literature Review: Accessing Published (β)

Information 100
9 Asking Questions: Effective Elite Interviews, Other Interviews,

Vignettes, Projective Questions, and Focus Groups 117
10 Questionnaire Surveys 140

iv Contents
11 Observation 161
PART IV DATA ANALYSIS 173
12 Analysing Research Data: The Process 175
Part IVA Quantitative Analysis 181
13 Calculating and Interpreting Descriptive Statistics 183
14 Using and Understanding Inferential Statistics 197
15 Testing for Association 206
16 Applying Factor Analysis and Other Advanced Techniques 220
PART IVB Qualitative Analysis 239
17 Analysing Qualitative Information: Classifying, Coding

and Interpreting Information 241
18 Using Content Analysis 263
19 Understanding and Adopting Discourse and Narrative

Analysis 279
PART V COMMUNICATING RESEARCH 307
20 Writing-up 309
Glossary of Terms 319
Key Formulae and Symbols 329
Bibliography 330
Index 335
About This Book
This textbook has been written for students and researchers in Departments of
Politics at UK and other ‘western’ universities where English is the language of
instruction.3 It is designed to provide an introductory text for undergraduates, an
intermediate text for graduates following ‘taught’ Masters’ programmes and first-
year postgraduate researchers. It is also designed to provide an additional teaching
resource for busy teachers – especially those for whom teaching research methods is
a duty rather than matter of choice. It seeks to meet the wide variety of readers’ prior
training and training needs by adopting a multi-level approach. In this way, it seeks
to bridge the present gap in the literature between the good introductory texts and
the more advanced texts requiring substantial mathematical training, knowledge or
sociolinguistics.4,5 It also seeks to be inclusive by including and giving equal coverage
to quantitative (numeric) and qualitative ‘talk and text’ methods. It recognises the
criticism of research methods texts of ‘eurocentricity’ by acknowledging that, while
the underlying principles of research may be (almost) universal, the underlying
philosophic roots and techniques must vary between different contexts.
The book assumes a mixed readership of home English-speaking students and
overseas students for whom English is a second language. Similarly, it assumes that
some students – especially of ‘taught courses’ – may come from other disciplines
and be unused to terms more specific to Politics. To meet the special needs of some
students without patronising others, a system of comprehensive endnotes has been
used to explain terms and to cite references fully.6 (Regrettably, today’s printing
software precludes the use of footnotes.) Technical terms are written in italics. Where
key technical terms are introduced for the first time, they are displayed in bold text,
defined and discussed. Briefer definitions of terms are given in the glossary. Whilst
the Harvard system of referencing is used, this has been augmented by providing
the date of first publication so that the chronological significance is retained, e.g.:
Hobbes, T. (1651/1996) Leviathan. Oxford, Oxford University Press.
Teaching and learning strategy
Teaching and learning are often conceived as two sides of the same coin. Some
students still prefer to learn passively from teaching through formal lectures. Others
prefer to gain (arguably deeper) insights through a combination of individual study
vi About This Book
(including reading and writing), exploration and heuristic learning (including direct
practical experience). Most benefit from collective discussions at seminars or informal
discussions with colleagues.
The pedagogic 7 principle underlying this textbook is that we all learn best – in terms
of acquiring knowledge and skills – through a mix of formal teaching, independent
study, experience, and practice, in which the proportions will vary from student to
student. So a blended (or distributed) teaching and learning strategy has been adopted.
This offers readers a mix of:
1. text (supplemented by diagrams and other illustrations) communicating the underlying

principles and to introduce various techniques
2. worked examples demonstrating how approaches and techniques can be applied
3. sources for independent study
4. topics for seminar and informal discussion
5. case studies providing practical assignments for individual or group work.
To assist readers, supplementary case studies and packages of PowerPoint slides for
each of the chapters are available via the website www.sagepub.co.uk/pierce.
How to use the book
Readers are not expected or encouraged to read this book from cover-to-cover.
You are unlikely to really assimilate knowledge for its own sake. You are more
likely to assimilate best what and when you ‘need to know’. So a strategy of timely,
selective reading is recommended. And while this book seeks to be inclusive and
authoritative, it makes no claims to be a definitive text. Users are therefore expected
to follow the principles of best scholarship by considering and comparing other
narratives before firming up their views. The relevant additional literature includes:
• Burnham, P., Grillard, K., Grant, W. and Layton-Henry, Z. (2004) Research Methods in
Politics. Basingstoke: Palgrave Macmillan. This 308-page textbook has been written by
members of the Department of Politics at the University of Warwick. It draws on their
experience as teachers of research methods and their critique of the limitations of the
application to politics of the many research guides written by sociologists.
• Harrison, L. (2001) Political Research: An Introduction. London: Routledge.This 180-page
textbook has been written by Lisa Harrison, senior lecturer in Politics at the University of
the West of England. As its title states, it has been written primarily as an introductory
text for students.
• Marsh, D., and Stoker, G. (eds.) (2002) Theory and Methods in Political Science.
Basingstoke: Palgrave. This 368-page textbook contains separate chapters by leading
About This Book vii
academics in Politics at universities in UK, US and Australia. Its strength lies in the
coverage which it offers of the main approaches and issues and their implications for
research methods.
Where best to begin?
A three-stage approach is recommended using the following proforma. You can

copy this from www.sage.co.uk/pierce. It lists the chapter titles, topics and degrees
of knowledge:
1. Assess your starting research skills.

2. In conjunction with your research supervisor, identify your research training needs. You
can skip or skim those chapters and topics with which you are already familiar.
3. Concentrate on those other chapters and topics where your research training needs
are greatest. This will enable you to choose and develop the research skills which best
serve your research interests.
Training needs self-evaluation proforma
Chapter Chapter Title Existing Knowledge and Understanding

No. Topics None Substantial
0 1 2 3 4 5
PART I INTRODUCTION
1 Introduction
Political science/ politics/political 0 1 2 3 4 5
studies
2 Power in Research, Ethics, Data Protection and Bias
Ethics in research 0 1 2 3 4 5
Data protection 0 1 2 3 4 5
The Research Effect (Hawthorne) 0 1 2 3 4 5
Bias 0 1 2 3 4 5
3 The Philosophy and Principles of Research
Positivism 0 1 2 3 4 5
Empiricism 0 1 2 3 4 5
Behaviouralism 0 1 2 3 4 5
Naturalism 0 1 2 3 4 5
Feminism 0 1 2 3 4 5
Marxism 0 1 2 3 4 5
Inductive research 0 1 2 3 4 5
(Continued)
viii About This Book

0 1 2 3 4 5
Deductive research 0 1 2 3 4 5
Grounded research 0 1 2 3 4 5
PART II METHODOLOGIES
4 Qualitative versus Quantitative Methods: A Relevant Argument?
Qualitative methods: claims and 0 1 2 3 4 5
criticisms
Quantitative methods: claims and 0 1 2 3 4 5
criticisms
Mixed methods 0 1 2 3 4 5
Research design(s) 0 1 2 3 4 5
5 Collecting Data Sets: Case Studies, Experimental, Comparative, Longitudinal
and Action Research Methods
Case studies 0 1 2 3 4 5
Comparative research 0 1 2 3 4 5
Longitudinal research 0 1 2 3 4 5
Action research 0 1 2 3 4 5
PART III COLLECTING INFORMATION
6 Critically Evaluating Published 0 1 2 3 4 5
Research
7 Evaluating Information: Validity, Reliability, Accuracy, Triangulation
Validity, reliability and accuracy 0 1 2 3 4 5
Primary and secondary sources 0 1 2 3 4 5
Triangulation 0 1 2 3 4 5
Sampling 0 1 2 3 4 5
8 Completing a Literature Review: Accessing Published (β) Information
Information search 0 1 2 3 4 5
Completing a critical literature 0 1 2 3 4 5
review
9 Asking Questions: Effective Elite Interviews, Other Interviews, Vignettes,
Projective Questions, and Focus Groups
Elite interviews 0 1 2 3 4 5
Group meetings 0 1 2 3 4 5
Vignettes 0 1 2 3 4 5
Focus groups 0 1 2 3 4 5
10 Questionnaire Surveys
Survey sampling 0 1 2 3 4 5
Designing and coding 0 1 2 3 4 5
questionnaires
Projective questions 0 1 2 3 4 5
Designing vignettes 0 1 2 3 4 5
11 Observation 0 1 2 3 4 5
About This Book ix

0 1 2 3 4 5
PART IV DATA ANALYSIS
12 Analysing Research Data: The 0 1 2 3 4 5
Process
IVA Quantitative Analysis
13 Calculating and Interpreting Descriptive Statistics
MS Excel 0 1 2 3 4 5
SPSS 0 1 2 3 4 5
Descriptive statistics
Mean, median, mode, outliers, 0 1 2 3 4 5
range, deviance
Grouped frequency distribution 0 1 2 3 4 5
Standard deviation 0 1 2 3 4 5
14 Using and Understanding Inferential Statistics
Standard error of the mean 0 1 2 3 4 5
Confidence limits 0 1 2 3 4 5
15 Testing for Association
Correlation 0 1 2 3 4 5
Significance 0 1 2 3 4 5
Linear regression analysis 0 1 2 3 4 5
16 Applying Factor Analysis and Other Advanced Techniques
Factor analysis 0 1 2 3 4 5
Bernouli distribution 0 1 2 3 4 5
Time series analysis 0 1 2 3 4 5
IVB Qualitative Analysis
17 Analysing Qualitative Information: Classifying, Coding and Interpreting
Information
Coding text 0 1 2 3 4 5
18 Using Content Analysis 0 1 2 3 4 5
19 Understanding and Adopting Discourse and Narrative Analysis
0 1 2 3 4 5
PART V COMMUNICATING RESEARCH
20 Writing-up 0 1 2 3 4 5
Preface
This book encourages researchers to begin their reports with a personal statement in
the form of a preface. This serves several purposes. First, it tells the readers why
the researcher has (really) chosen the research topic: what it means to the author
and their identity. Second, it discloses (the inevitable) starting biases, prejudices and
hunches – the intellectual and emotional baggage that the researcher carries. Third,
the preface acknowledges those other people who have influenced the work.
Writing the preface is a difficult task. It requires reflexivity (intellectual self-
awareness) and honesty. Overall, the personal statement should clarify matters at
the beginning for both the researcher and the reader. In particular, it underlines the
personal – the identity of the researcher that, in the discipline of Politics and other
social sciences, exerts a major influence on the choice of research, the findings and,
crucially, their interpretation. The researcher is therefore, in the language of research,
an independent variable (a causal factor or driver which affects the outcome).
In this case, the textbook has been written to ‘close the chapter’ on my ten-
year career as a mature graduate, doctoral student and, latterly, teacher of research
methods in the Department of Politics at the University of York. It seeks to bring
together in a single volume materials developed from many sources and academic
specialities over many years. In my previous work in urban regeneration, I helped
shape new environments and policies. They provide enduring evidence of effort and
success – or otherwise. Hopefully, this book serves a similar purpose. Si monumentum
requiris, etc.8
My baggage is essentially that of a practitioner as distinct from an academic.9 So
my approach seeks to be eminently practical: to show readers how to carry out
different types of research and techniques to high standards of scholarship. But it is
not a-theoretical: the various methodologies are grounded in theoretical principles
that must be properly understood if the tools are to be applied correctly. After all,
There is nothing as useful as a good theory.10
My particular research interests are: power, deference and complicity and the explanations
of Gramsci, Lukes, Bourdieu and Foucault.
This textbook has been shaped by many other people. Special thanks are due to:
my former mentor, Professor Mark Evans; to Dr Adrian Leftwich who provided the
original brief; to Dr Roger MacGinty and Dr Simon Parker who encouraged me to
develop this text from my module guides. To Liz Harrison who proof-read the drafts;
Preface xi
and, to Patrick Brindle of Sage who encouraged me to submit a proposal. However,

the greatest influence has been exerted by the consumers: the students who provided
formal course evaluation and informal comments; the graduate teaching assistants11 ;
and the research subjects (people) with whom (rather than on whom) the research
methods were developed.
However, any errors or omissions are entirely my own.
Dr Roger Pierce
York
Notes
1 Panegyric, public speech.

2 Pasquinade, lampoon, libel, satire.
3 Politics is used throughout this book to refer to the subject otherwise called ‘political science’,
‘political studies’, ‘politics and government’ etc. A capital P is used throughout to distinguish the
discipline and study of Politics from its more general use to describe everyday activity by
politicians and electors.
4 For example, Harrison, L. (2001) Political Research: An Introduction. London: Routledge.
5 For example: Pennings, P., Keman, H., and Kleinnijhuis, J. (2006) Doing Research in Political
Science: An Introduction to Comparative Methods and Statistics. London: Sage.
6 ‘footnotes [and endnotes] are the humanists’ rough equivalent of the scientist’s report on data:
they offer empirical support for the stories told and arguments presented . . .’ Grafton, A. (1997)
The Footnote. London: Faber. p.vii.
7 From pedagogy, ‘the science of teaching’.
8 Si monumentum requiris circumspice, ‘If you seek his monument, see around you’ inscription in
St Paul’s Cathedral London to its architect, Sir Christopher Wren.
9 Dictionary definitions of ‘academic’ include: ‘of no practical use’.
10 Kurt Lewin (1890–1947), US-naturalised German psychologist, founder of ‘action research’ and
mentor to Carl Festinger (‘cognitive dissonance’).
11 Especially Fiona Aspinall, Christine Hamieh and Sharleene Bibbings.
The Researcher
‘He will have views and prospects to himself perpetually soliciting his eye, which he can
no more help standing still to look at than he can fly; he will moreover have various …
Accounts to reconcile;
Anecdotes to pick up;
Inscriptions to make out;
Stories to weave in; Traditions to sift
Personages to call upon;
Panegyrics1 to paste up at his door; Pasquinades2 at that…’
Rev. Laurence Sterne (1996/1760) Tristram Shandy. Ware: Herts.,
Wordsworth., p. 27.
Part I
Introduction
Chapter 1
Introduction
This chapter begins, like the others that follow, with a clear statement of the teaching
and learning objectives (or purpose). They are shown below:
Teaching and learning objectives:
1. To answer the question: What is research?

2. To identify key features of research specific to Politics.
3. To set out the structure and contents of the textbook.
What is research?
Research is essentially a process of systematic inquiry. Its core activities are:
• goal orientated and purposeful

• inquisitive – searching for answers to specific questions – especially ‘why?’ and ‘how?’
• careful, systematic and methodical
• original.
Additionally, we can claim that academic research can be distinguished from other
research by
• its central concern for theory involving either testing or extending existing theory
(deductive research), or developing new theory (inductive research).
Academic research essentially involves a systematic process. It begins with a

research question. This is followed by a literature review, the collection or discovery
of information, analysis, interpretation and conclusions. It can be both inventive
and creative in terms of designing the research process and framing new theory.
However, research may also involve serendipity – the happy knack of making
discoveries by accident. For example, penicillin was ‘discovered’ in 1928 when
Fleming noted that a petri dish had been contaminated by mould. However, rather
4 Research Methods in Politics
than throwing out the dish, he first examined the contents and discovered that the
mould had prevented the formation of Staphylococcus bacilli. The discovery thereby
evidenced Pasteur’s earlier remark that ‘chance only favours the prepared mind’
(Greenfield et al, 2001: 302).1 Today, specialist ‘serendipity software’ is available
to search and compare data sets to identify potential associations for subsequent
investigation.
Similarly, research on research has shown that intellectual curiosity is the main
driver of effective research which, in turn, stimulates and sustains the essential
concentration and motivation (Mace, 1962: 29).2 In a post-modern world, it
may also be argued that scepticism (the doctrine of the Philosophic School of
Sceptics that real knowledge of the world is unattainable) is also an essential
component. There is therefore a new readiness to challenge accepted theories, ‘truth’
and ‘facts’.
At a time when so much is already ‘known’, the question has to be asked: why
undertake research? In particular, what real contributions can a single-semester,
undergraduate research project make to the knowledge and understanding of our
world? In reply, it can be argued that the real benefit of undergraduate, graduate
and doctoral research lies in the contribution the research activity makes to your
intellectual and personal development in terms of:
• testing the applicability and relevance of theory to new contexts

• promoting a better understanding of theoretical concepts
• developing analytical and interpretative skills
• learning how best to design investigative processes and manage projects
• uniquely, the opportunity given via undergraduate group research projects to gain
experience of effective team-working – the mantra of modern management.
In other words, research provides an opportunity and arena for education for both
academic or non-academic careers.
But, what is Politics? Does the discipline require or pre-suppose a unique approach to
research training that favours specialist, in-house, faculty-wide, generic research training?
Politics is a relatively new academic discipline in the family of social sci-
ences. It has roots in philosophy, history, law, geography, economics, sociology,
psychology and, in the sub-field of voting behaviour, quantitative (statistical)
analysis and mathematical modelling. It has therefore been described as: ‘the
junction subject of the social sciences’ (Burnham et al, 2004: 8).3 Alternatively,
it can be seen as a ‘mongrel subject’ or ‘crossover discipline’ that draws on
others promiscuously. It can be both prescriptive (normative) and descriptive. Two
broad and six constituent approaches have been identified in the ‘diverse and …
cosmopolitan’ discipline: the formal operation of politics in government (and
other arenas) embracing: behaviouralism; rational (public) choice theory and
(new) institutional analysis; and, politics as a social process pervading all levels of
Introduction 5
society (associated with feminism); interpretative theory (anti-foundationalism); and,

Marxism (Marsh & Stoker, 2002: 3).4 While there is no official definition of Politics,
an official view was negotiated by a committee of academics to provide a ‘benchmark
statement’:
[Politics is concerned with] … developing a knowledge and understanding of government

and society. The interaction of people, ideas and institutions provides the focus to
understand how values are allocated and resources distributed at many levels, from
the local to the sectoral, national, regional and global. The analyses of ‘who gets what,
when, how, why and where’ are central, and pertain to related questions of power, justice,
order, conflict, legitimacy, accountability, obligation, sovereignty and decision-making.
Politics encompasses philosophical, theoretical, institutional and issue-based concerns
relating to governance.
(QAA, 2000: 2)5
This overarching view of Politics might, at first sight, be seen to imply that all
available research approaches and techniques can be adopted in Politics and that the
researcher’s tool bag is vast. But this is not the case. In particular, power (which
many commentators regard as central to the discipline) is not readily quantifiable or
measurable.6 It as another ‘essentially contested concept’ (Lukes, 1974: 9).7 Indeed,
power may not be observable (Bachrach & Baratz, 1962).8 Similarly, conflict need
Illustration 1 Ask yourself, Father, what do we really mean by ‘power’?

not necessarily be overt. And historic events cannot be re-run. So, those laboratory-
based, scientific research approaches and techniques which rely on the repeated
observation of phenomena under controlled conditions are largely inappropriate to
Politics.
This textbook concentrates on those research methods which seek to discern and
interpret the underlying meanings, causes and consequences of conflict and power
at the level of supra-state, state, government, party, class and other identity groups,
and people in either the past or present-day. So a special characteristic of research
in Politics is that it collects information from both archives (historical records) and
fieldwork (field research).
Textbook structure and contents
The textbook is structured in five parts and their associated chapters which, generally
speaking, follow the research process and degrees of complexity.
Part I (Introduction) begins with this scoping chapter. It is followed, in Chapter 2,
by a review of power in research, ethics, data protection and the research
effect. It concludes, in Chapter 3, with a review of the underlying philosophy
and principles of research, including the concept of causality, provided by the
main schools of positivism, empiricism, behaviouralism, naturalism and feminism
newly mentioned and distinguishes between inductive, deductive and grounded
research.
Part II (Methodologies) identifies the underlying principles and theories from
which various methods have been developed. It begins, in Chapter 4, with a review
of the dualism of qualitative and quantitative research, the claims and counter-claims
of their disciples and the role of mixed methods. This concludes by identifying
the main approaches to research design. This is followed, in Chapter 5, by a
critical review of case studies, experimental, comparative, longitudinal and action
research.
Part III (Collecting information) begins, in Chapter 6, with advice on how to
critically evaluate published research. This is followed by guidance on how to test
the quality of information in terms of the gold standards of validity, reliability
and accuracy, the distinction between primary and secondary sources, and the use
of triangulation and sampling in Chapter 7. The next chapter (8) concentrates
on accessing secondary sources and completing an effective literature review.
This is followed, in Chapter 9, by guidance on how to organise and complete
interviews with political elites and other individuals, group meetings and focus
groups, projective questions and vignettes. Chapter 10 provides practical advice on
designing and conducting questionnaire surveys. Part III concludes with a critical
review of other observational methods including the role of the new popular science
of body language in Chapter 11.
Introduction 7
Part IV introduces methods for analysing and interpreting the information

collected. It begins, in Chapter 12, by reviewing the generic process of analysis,
introducing a hierarchy of analysis and distinguishing between predominantly
quantitative and qualitative methods of analysis.
Quantitative methods start, in Chapter 13, by reviewing the main descriptive
statistics (mean, standard deviation, etc.) and their use. It is followed, in Chapter 14
by a review of the main inferential statistics (standard error of the mean, etc.) in
which conclusions about populations are inferred from samples. The next chapter
introduces the concepts of association, correlation, significance and regression
analysis. The quantitative methods conclude, in Chapter 16, by introducing cluster
and factor analysis (which enable underlying factors to be identified and labelled),
time series analysis and the binomial distribution. The formulae built into MS
Excel are used in the early chapters to calculate descriptive and inferential statistics.
SPSS statistical software is used for the more sophisticated calculations in the later
chapters.
Qualitative methods begins by offering guidance on how to transcribe and code
‘talk and text’ in Chapter 17. Chapter 18 introduces new approaches to content
analysis. The section ends by introducing discourse analysis and narrative analysis
and demonstrating their use in Chapter 19.
Finally, Part V provides advice on communicating research. In particular, it
considers, in Chapter 20 the writing-up process from the initial research proposal to
the final research report. Despite its (traditional) location at the end of the book, it
argues that research begins by writing the research proposal and continues in parallel
with the research process. So writing-up is both the ‘alpha’ (beginning) and ‘omega’
(end) of research. The text is supported by three appendices: a glossary of terms,
statistical formulae and bibliography.
Questions for discussion or assignments
1. What is Politics? Is ‘political science’ more appropriate, a misnomer or an

oxymoron?
2. What is research?
3. What is the current research agenda of Politics? What topics would you wish to
see added? Why?
4. Is power central to Politics?
5. Which research methods are generic to social science? Which others, if any, are
specific to Politics?
6. What are the implications of post-modernism for academic research? Is there a
distinction between ‘truth’ and ‘facts’?
FURTHER READING
Burnham, P., Grillard, K., Grant, W. and Layton-Henry, Z. (2004) Research

Methods in Politics. Basingstoke: Palgrave Macmillan. pp. 1–29. This very
readable introduction identifies and discusses the origins of the discipline of
Politics, the dominant paradigms and their main methodological implications.
Held, D. and Leftwich, A. (1984) Chapter 8: A Discipline of Politics? In Leftwich, A.
(ed.) What is Politics? Its Activity and Study. Oxford: Blackwell. pp. 139–159.
This book provides an excellent collection of essays providing different
perspectives on Politics. In the final chapter, Leftwich assesses these
interpretations with particular reference to the centrality of conflict.
Leftwich, A. (2004) Thinking Politically: On the politics of Politics. In Leftwich, A.
(ed.) (2004) What is Politics? Cambridge: Polity Press. pp. 1–22. This textbook
offers new and additional essays on Politics. Leftwich introduces the debate
with a scoping review which concludes that: ‘what unites political analysts is a
concern for the provenance, forms, distribution, use, control, consequences
and analysis of political power. What separates them is the difference of focus
and the levels and frameworks of analysis …’ p. 20.
Marsh, D. and Stoker, G. (eds.) (2002) Theory and Methods in Political Science.
Basingstoke: Palgrave. pp. 1–16. This short introduction by the editors seeks ‘to
get readers into the foothills of understanding the political science range’, p. 16.
Table 1 provides a very good summary of the characteristics of the main
approaches, pp. 6–7.
Notes
1 Greenfield, S., Singh, S., Tallack, P. et al (2001) The Science Book. London: Cassell & Co. p. 302.
2 Mace, C.A. (1962) The Psychology of Study. Harmondsworth: Penguin. p. 29.
3 Burnham, P., Grillard, K., Grant, W. and Layton-Henry, Z. (2004) Research Methods in Politics.
Basingstoke: Palgrave Macmillan. p. 8.
4 Marsh, D. and Stoker, G. (eds.) (2002) Theory and Methods in Political Science. Basingstoke:
Palgrave. pp. 3–11.
5 QAA (Quality Assurance Agency for UK Higher Education) (2000) Politics and International
Relations’ Benchmark Statement. QAA, Gloucester cited in Leftwich, A. (ed.) (2004) What is
Politics? Cambridge: Polity Press. p. 20. Note how this constipated ‘official view’ draws centrally on
Lasswell, H. (1958) Who Gets What, When, How. New York: Meridian.
6 For example, Robertson, D. (1993) Politics. London: Penguin.
7 Lukes, S. (1974) Power: A Radical View. London: Macmillan. p. 9.
8 See Bachrach, P., and Baratz, M. (1962) Two Faces of Power. In American Political Science
Review, 56: 947–52, and Lukes, S. (1974) Power: A Radical View. London: Macmillan.
Chapter 2
Power in Research, Ethics, Data
Protection and Bias
‘The researchers take, hit, and run. They intrude into their subjects’ privacy, disrupt
their perceptions, utilise false pretences, manipulate the relationships, and give little in
return.’
(Reinharrz, 1984: 95)1
1. To develop an understanding of the asymmetric power relations between

researcher and those researched and the scope for harm.
2. To recognise the types of potential harm.
3. To understand the context and scope of the Nuremberg Code.
4. To identify the implications for researchers of data protection legislation.
5. To gain knowledge of the different types of ‘research effects’.
6. To understand how the researcher’s own values can be a source of bias
requiring ‘reflexivity’.
Asymmetric power relations in research
The researcher (you) choose the topic, research question, design and means of data
collection and its interpretation. You will choose sources selectively for accessibility,
validity, reliability and accuracy. You will seek to exploit the resources and, for that
matter, the prestige and goodwill of the university. Some feminists likened traditional
research to a ‘rape model’. As you will see, the major resources in political research
are archives and people. The people may be agents, actors or bystanders. You seek
their collaboration on the basis of the benefit to society of the research and the
benefits of the interviewees (the subjects). But within the survey or interview
situation, power relations between the researcher and those researched are unequal.
Any process of question-and-answer favours the questioner. In particular, you are
likely to be more experienced and can leave the field at the end of the interview or
research project. So the potential for causing real harm, however unintentional or
as agent provocateur, is very great.
Types of potential harm
Harm may be physical, financial, social and psychological

Physical harm can be caused where you reveal the subject’s identity to repressive
groups. They may be agents of the state, criminals, terrorists or relatives. You
may name subjects or enable others to identify them through details of your
journey, fieldwork notes, photographs, or apparently casual, chance conversation
with other people. Alternatively, you may promote a new recognition by subjects
of circumstances or institutions which they previously accepted as ‘natural’. This
new consciousness may lead to conflict between and within groups. Financial harm
can be caused where the subject’s identification leads to loss of their job, home,
support or reputation. Social harm can be caused where the subject’s position or role
within the family and wider social networks is affected adversely. Psychological harm
can be caused where you unlock painful memories or feelings of guilt, or create a
‘false memory’. Raising false expectations, altering identities, or imposing western,
middle class values can cause great harm.
But, conversely, you can also be harmed or exploited (particularly by elites). You
may also face dilemmas when, for example, you uncover instances of child abuse,
sexual exploitation or conspiracies to carry out acts of criminality or terrorism.
What should you do – especially where you have given the subjects undertakings
of confidentiality? One dubious practice is to use the ‘don’t ask, don’t tell’ caveat
where you warn them that you do not wish to hear of, or see, any illegal activity
which you are obliged to report to the authorities. But, in this way, you may allow
their abuse of others to go undetected.
Research can therefore pose important ethical challenges especially where the
researcher and the subjects are members of different cultural groups. Some universal
framework and local institutions are desirable.
The Nuremberg Code (1947)
The trial of the Nazi war leaders at Nuremberg in 1947 was followed by the trial of
scientists, doctors and administrators who had carried out experiments on prisoners
in the concentration camps. They were charged with ‘crimes against humanity’.
Their defence was threefold: first, that they had not undertaken acts which were
contrary to German law at the time; second, that the practices were consistent with
the prevailing national ideology that Jews, Gypsies and Slavs were anti-state groups;
and, third, that the experiments were justified by their scientific value to mankind.
Power in Research, Ethics, Data Protection and Bias 11
One example cited in their defence was the immersion of prisoners in cold water to
determine the effect of lack of heat, body mass, age and gender on survival times.
The defences were dismissed. The argument of scientific value was thrown out on
the grounds that the prisoners were not representative of the population: they had
been systematically starved and beaten to the point that their very will to survive
had been extinguished. The defendants were executed or imprisoned.
During the trial, a group of US philosophers, doctors and lawyers was tasked to
provide a universal ethical code for medical experiments. The Nuremberg Code was
adopted by the Great Powers in 1947. It is reproduced in Box 2.1. The code has since
BOX 2.1 The Nuremberg Code
The Nuremberg Code (1947)

Permissible Medical Experiments
The great weight of the evidence before us to effect that certain types of medical experiments
on human beings, when kept within reasonably well-defined bounds, conform to the
ethics of the medical profession generally. The protagonists of the practice of human
experimentation justify their views on the basis that such experiments yield results for
the good of society that are unprocurable by other methods or means of study. All agree,
however, that certain basic principles must be observed in order to satisfy moral, ethical and
legal concepts:
1. The voluntary consent of the human subject is absolutely essential . This means
that the person involved should have the legal capacity to give consent; should be
so situated as to be able to exercise free power of choice, without intervention of any
element of force, fraud, deceit, duress, overreaching, or other ulterior form of constraint
or coercion; and should have sufficient knowledge and comprehension of the elements
of the subject matter involved as to enable him to make an understanding and
enlightened decision. This latter element requires that before the acceptance of an
affirmative decision by the experimental subject there should be made known to him
the nature, duration, and purpose of the experiment; the method and means by which
it is to be conducted; all inconveniences and hazards reasonably to be expected; and
the effects upon his health or person which may possibly come from participation in
the experiment.
The duty and responsibility for obtaining the quality of consent rests upon each
individual who initiates, directs, or engages in the experiment. It is a personal duty and
responsibility which may not be delegated to others with impunity.
(Continued )
2. The experiment should be such as to yield fruitful results for the good of society ,
unprocurable by other methods or means of study, and not random and unnecessary
in nature.
3. The experiment should be so designed and based on the results of animal
experimentation and a knowledge of the natural history of disease or other problem
under study that the anticipated results justify the performance of the experiment.
4. The experiment should be so conducted as to avoid all unnecessary physical
and mental suffering and injury .
5. No experiment may be conducted where there is an a priori reason to believe that
death or disabling injury will occur; except perhaps in those experiments where the
experimental physicians also serve as subjects.
6. The degree of risk should never exceed that determined by the humanitarian
importance of the problem to be solved by the experiment
7. Proper preparations should be made and adequate facilities provided to protect the
experimental subject against even the remote possibility of injury, disability or death.
8. The experiment should be conducted only by scientifically qualified persons. The
highest degree of skill and care should be required through all stages of the experiment
of those who conduct or engage in the experiment.
9. During the course of the experiment the human subject should be at liberty
to bring the experiment to an end if he has reached the physical and
mental state where continuation of the experiment seems to him to be
impossible.
10. During the course of the experiment the scientist must be prepared to terminate
the experiment at any stage, if he has probable cause to believe, in the exercise of
good faith, superior skill and careful judgement required of him, that a continuation
of the experiment is likely to result in injury, disability, or death to the experimental
subject.
been strengthened by other codes, e.g., the Helsinki Code, to provide additional
guidance to scientists. It established four principles for research for whatever purpose:
voluntary informed consent; benefit to society; protection from unnecessary harm of subjects;
and, the right of subject to end experiments.
Today, UK’s ESRC2 and other research funders ask for research proposals to
include a statement identifying the specific ethical problems likely to be encountered
and how they will be resolved. Ideally, the applicant’s statement should cite the
university and department’s family of ethical research codes, and the value of frequent
supervision.
Data protection legislation and academic research
The experiences of World War II and the post-war fears of totalitarianism (captured
in George Orwell’s Animal Farm and 1984) led to the United Nations Declaration of
Human Rights and the European Convention of Human Rights & Fundamental Freedoms
to protect the individual from government.3 However, fears were heightened again
by the increasing use of computers by government and business from the early 1960s
and their capacity to hold vast amounts of data on individuals. There was widespread
concern for the accuracy and security of that information (especially HIV and other
medical records) and its use for purposes for which it had never been intended.
Growing concerns for political freedom and privacy led to the convergence of human
rights and data protection protocols at supra-national level which were brought into
effect as national legislation.
The use of personal data for research was the subject of Article 108 of the Treaty
of Rome, 1957 and EEC Directive 95/46. Article 6 of the Directive provided
derogations for processing and long-term storage of data for ‘historical, statistical or
scientific purposes’ subject to ‘compatibility with the original purpose of collection’.
Specific issues that member states were asked to resolve were, on one hand, the right
of access by individuals to information held on them, and, on the other, restrictions
on their use of the data, considerations of national security, and the use of data
generated by medical research by private organisations. In Britain, the directive was
given effect by the Data Protection Act, 1998. This required ‘all persons who process
or use personal data’ in electronic, paper, tape or video formats to follow eight data
protection principles. The principles were that all data must be:
• obtained and processed lawfully

• obtained for a lawful purpose
• adequate, relevant and not excessive
• accurate and kept-up-to-date
• not kept longer than is necessary
• processed in accordance with the subject’s rights
• kept safe from unauthorised access, loss or destruction
• not transferred outside the European Economic Area (i.e. EU and EFTA) unless
safeguards are as adequate.4
S.33 of the Act gives exemptions for personal data kept for ‘research purposes’.
S.33(5) allows ‘permitted disclosures’ to others for ‘research purposes’, to the subject
or their agents, or at the subject’s request. But data need not be disclosed where it
would require ‘disproportionate effort’ or ‘where the risk to the rights and freedoms
of subjects is low’. The implications for academic research were considered by the
Association of Colleges in 1999 who advised members that:
Data collected for the purpose of one piece of research can be used for other
research without breaching the regulations on incompatible purposes and can be kept
indefinitely. This means that staff and students involved in research can keep records
of questionnaires and contacts so that the research can be revisited at a later date, or
so that a research project looking into an associated area, could reanalyse the data.
However, in order to avoid breaking subject access rules, researchers must ensure that
the final results of the research do not identify the individual.
British universities must register with the regulating authority: the Data Protection
Registrar. A senior university official must be designated as the Data Protection
Co-ordinator and submit annual returns to the Registrar. Generally speaking, the
head of department is designated as Data Protection Officer. They may delegate
administrative responsibility to a Data Protection Manager. Researchers should
notify the Data Protection Co-ordinator of:
• the title of the research project

• name and contact address of the researcher
• data to be collected
• purpose of the research
• with whom the data will be shared
• content
• format
• data security (including anonymity).
The implications for the UK-based researcher are that:
• data must be held securely. This means that data on subjects and interview transcripts
must be stored separately in lockable cabinets or, if held on PCs or laptops etc. password-
protected
• personal data on subjects should only be held where essential
• individuals should be anonymous or not otherwise identifiable; strategies include:
– ‘anonymisation’, e.g., Mr. A said …
– positional descriptions, e.g., minister A said …
– deception, e.g. Ms A said … (where sex is not relevant)
– replacing place names by generic pseudonyms, e.g. Northtown
– changing descriptions slightly
• personal data should not be disclosed, transferred or copied (including by email),
especially outside the EEA, without prior agreement by the recipient to confidentiality
and authorisation
• potentially reusable data should be archived centrally, e.g. at ESRC’s archive at The
University of Essex.
The penalties for negligence may include fines and the withdrawal of funding support
for the research. The greatest penalty will be the loss of the researcher’s, department’s
and university’s reputation and loss of future research opportunities as subjects refuse
to co-operate.
The ‘research effects’
You will inevitably change your behaviour in the presence of power. You are
unlikely to maintain the same tone of voice and general manner in all situations,
for example, teachers do not speak to their head teacher in the same way that they
address schoolchildren. Your actual language and tone will vary according to whom,
when and where you are speaking. In the same way, researchers and their subjects
will also change their behaviour, especially in face-to-face situations. This change
of behaviour is termed the‘research effect’. The main types of research effect are
the Hawthorne effect, the Placebo effect, the John Henry effect, the Halo effect, Experimental
effects, the Pygmalion effect and the Peacock effect.
Hawthorne effect
The Hawthorne effect takes its name from the Hawthorne Plant of the Western
Electric Company in Cicero, Illinois. There, in the relay assembly room, engineers
experimented in 1927 with the effects on productivity on the introduction
of improved lighting, mechanical ventilation and humidity control. Separate
workspaces and groups were designated as experimental or control areas,where
the lighting conditions etc. were varied in the experimental areas and unchanged
in the control areas. The engineers were delighted when productivity in the
experimental areas increased as lighting and ventilation were improved. However,
when sceptical colleagues challenged the experimenters to reduce lighting and
ventilation, they found that productivity remained high. And it had also increased
in the control area. The help of the Harvard Business School was sought. The senior
investigator, Elton Mayo wrote that:
A highly competent group of Western Electric engineers refused to accept defeat when
experiments to demonstrate the effect of illumination on work seemed to lead nowhere.
The conditions of scientific experiment had apparently been fulfilled – experimental
room; control room; changes introduced one at a time; all other conditions held
steady. And the results were complexing: Roethlisberger5 gave two instances – lighting
improved in the experimental room, production went up; but it also rose in the control
room.
The opposite of this: lighting diminished from 10 to 3 foot-candles in the experimental

room and production went up again; simultaneously in the control room, with illumination
constant, production also rose.6
(Mayo, 1949: 60)
The explanation given by Mayo was that productivity had increased in both the
experimental and control workspaces because the workers there were the focus of
the researchers’ attention. The research was subsequently extended until 1932. Mayo
and his colleagues also argued that the research showed that aptitude is only weakly
related to performance, and that the workplace is a social system in which informal
organisation of work-groups and their attitudes are critical to production. In effect,
the work-group decided what was a ‘fair-day’s work’.
Mayo termed this behaviour change as the ‘research effect’. It has become better
known as the Hawthorne effect. However, his explanation has been challenged. Some
industrial psychologists argue that Mayo’s findings were wishful thinking: Mayo and
his colleagues were biased in their interpretation by the researchers’ promotion of
the (softer) human relations school of industrial relations (which favoured humane
treatment of worker and workplace democracy) in the face of the (harsher) scientific
management prevailing in the US.7 While the studies were extensive, the samples
were relatively small and not representative. The generalisability of their findings
could therefore be challenged. By re-analysing the data, critics argued that changes
in performance should better be attributed to the Depression, the substitution of
two particular workers and the threat of disciplinary action.8 Whether or not this
challenge to the Hawthorne folklore can be sustained, the episode does illustrate
the potential impact of the researcher’s own values on the research proposal, the
collection of data and its analysis (which is considered later in this chapter).
Placebo effect
The Placebo effect pre-dates Hawthorne. The term is derived from the Latin word
‘placebo’ meaning ‘I shall be pleasing’ which was used by physicians to describe inert
remedies to which some patients responded by meeting their psychological needs for
medical attention. The term has been extended to describe the practice in medical
drug trials, and other experiments, of giving placebos to the control group so that
the real efficacy of the trial drug can be calculated. In these cases, random, double-blind
trials are conducted in which the patients and medical observers are unaware who
is receiving the trial drug and who is receiving the placebo.
John Henry effect

The John Henry effect is the converse of the Hawthorne effect. It happens when the
control group compete with the experimental group and achieve greater changes in
behaviour or outcomes. It can be seen as a super-placebo effect.
Halo effect
The Halo effect is the response of subjects to novelty where the press and other media
(including advertising) have created an expectation.
Experimental effect
The Experimental effect is similar. Here expectations are raised by the researcher.
Pygmalion effect
The Pygmalion effect is a sub-type of expectancy change. It can be a form of ‘self-
fulfilling prophecy’. It is most often cited in education where the teacher’s expressed
expectations of pupils, whether high or low, can greatly affect their performance.
It may also occur where, in an interview, the questioner implies that specific
responses are to be expected because they reflect prevailing norms, i.e. male, white,
middle-class values.
Peacock effect
The Peacock effect is the name given to the behaviour of some male birds among
females when they display their plumage or attributes. Some young men may behave
similarly in the presence of young women.
How can the impact of Hawthorne and other research effects on research be
reduced? Four strategies can be adopted:
• anticipating and discounting it

• reducing expectations
• adopting a longer timescale (so the subjects become inured to observation)
• adopting covert observation (however, covert observation is contrary to the princi-
ples of voluntary, informed consent and control by the subject). This raises ethical
dilemmas.
Research effects will inevitably bias the research. Bias is essentially a predisposition
or prejudice for or against a theory, person, group or institution which may distort or
skew cognisance and interpretation of phenomena. However, your greatest source
of bias may be your very own researcher’s bias.
Researcher’s bias
You are unlikely to begin your with an open mind. Your choice of research
topic, question and starting hypothesis will reflect deep-seated values and prejudices.
The greatest danger is of ‘wishful thinking’ or ‘self-fulfilling prophecies’ in which

you ‘see what you want to see’ and infer conclusions. But, on the other hand, your
biases can also motivate and sustain the research. Remember Beatrice Webb’s advice
that:
Finally, [the researcher] must realise that he is biased, and somehow or other he must
manage to discover this bias …
‘Know thyself’ is the maxim uniquely imperative on the investigator of social institutions.
For the greatest obstacle to the advancement of knowledge – is an obstacle in the mind
of the student, the presence of biases.
(Webb, B., and Webb, S., 1975: 31 and 44)9
Minimising bias requires reflexivity – intellectual self-awareness through self-

examination. You should record the outcome of this examination at the beginning
of the research report so that the reader can learn from the outset ‘where you’re
coming from’ and discount the bias. But then your readers will also have their own
biases and pre-judge the text accordingly.
1. Identify the extent of your own prejudices using the Table 2.1 below. You can
copy the full table from the web-site www.sagepub.co.uk/pierce. Add your own
categories. Distinguish between your immediate (private) feelings and your (public)
thoughts. Try to identify their source. Are they derived from others (e.g. parents
or friends), intellectual inquiry or experience? How have they changed over the
years? Why?
2. You have been asked to undertake a survey of electors to find out whether non-
voters in general elections differ in their (other) characteristics from voters and
whether they are serial abstainers. What ethical difficulties do you anticipate? How
would you tackle them? What are your own starting prejudices?
3. In the course of your fieldwork, you discover a continuing case of child sexual
abuse. You have given an undertaking of confidentiality. What should you do?
4. Obtain a copy of your own department’s code of research ethics. What criticisms
would you make? How should it be improved? OR, if a code is ‘pending’, write a
draft.
Table 2.1 Prejudices: a self-completion proforma
To what extent are you Strongly Against Neutral In Favour Strongly in

prejudiced in favour or Against Favour
against the groups listed
below? −2 −1 0 +1 +2
University students ! ! ! ! !
University teachers ! ! ! ! !
University administrators ! ! ! ! !
Social scientists ! ! ! ! !
Scientists ! ! ! ! !
Doctors ! ! ! ! !
Police officers ! ! ! ! !
Members of the armed ! ! ! ! !
forces
Unemployed people ! ! ! ! !
Disabled people ! ! ! ! !
Conservative party ! ! ! ! !
Liberal-Democrat party ! ! ! ! !
New Labour ! ! ! ! !
Old Labour ! ! ! ! !
Green party ! ! ! ! !
British National Party ! ! ! ! !
Communist party ! ! ! ! !
Republican party ! ! ! ! !
Democrat party ! ! ! ! !
President George W Bush ! ! ! ! !
Prime Minister Tony Blair ! ! ! ! !
Prime Minister Gordon ! ! ! ! !
Brown
Bill Clinton ! ! ! ! !
Hillary Clinton ! ! ! ! !
Iraq war/invasion ! ! ! ! !
UN ! ! ! ! !
USA ! ! ! ! !
EU ! ! ! ! !
UK ! ! ! ! !
Scots people ! ! ! ! !
Irish people ! ! ! ! !
English people ! ! ! ! !
The British Empire ! ! ! ! !
Imperialism ! ! ! ! !
Freedom ! ! ! ! !
(Continued )
Table 2.1 Cont’d
To what extent are you Strongly Against Neutral In Favour Strongly in

prejudiced in favour or Against Favour
against the groups listed
below? −2 −1 0 +1 +2
Democracy ! ! ! ! !
Liberty ! ! ! ! !
Terrorism/armed struggle ! ! ! ! !
Atheism ! ! ! ! !
Agnosticism ! ! ! ! !
Deism ! ! ! ! !
Roman Catholicism ! ! ! ! !
Protestantism ! ! ! ! !
Islam ! ! ! ! !
Buddhism ! ! ! ! !
Hinduism ! ! ! ! !
Secularism ! ! ! ! !
Nationalism ! ! ! ! !
Patriotism ! ! ! ! !
Conflict ! ! ! ! !
Competition ! ! ! ! !
Markets ! ! ! ! !
Regulation ! ! ! ! !
Trade unions ! ! ! ! !
Children ! ! ! ! !
Middle-aged people ! ! ! ! !
Older people ! ! ! ! !
Young adults ! ! ! ! !
Marriage ! ! ! ! !
Divorce ! ! ! ! !
Other people ! ! ! ! !
Yourself ! ! ! ! !
FURTHER READING
Berg, B. (2001) Chapter 3: Ethical Issues. In Qualitative Research for the Social
Sciences. London: Allyn & Bacon. pp. 39–65. Berg reviews the historical
development of ethical concerns in research from a US perspective with
particular emphasis on the role of Institutional Review Boards (IRBs). He
identifies how and why disadvantaged groups are more accessible to
researchers than privileged elites.
Burnham, P., Grillard, K., Grant, W. and Layton-Henry, Z. (2004) Chapter 11:
Ethics and Political Research. In Research Methods in Politics.
Basingstoke: Palgrave Macmillan. pp. 250–69. The authors provide a synoptic

review of the specific ethical issues in Politics research including the role of
gatekeepers, fraud and professional codes of conduct. They compare the codes
of the Social Research Association, American Political Science Association and
the UK Political Science Association (PSA). They also raise the ethical issue of
whether an institution should impose its own code on researchers.
De Vaus, D.A. (2001) Chapter 19: Ethics in survey research. In Surveys in Social
Research. London: Routledge. pp. 330–50. The author reviews the principles of
ethical research with particular emphasis on the decisions required by
‘informed consent’ and duties and responsibilities to sponsors, funders and
research colleagues.
Notes
1 Reinharrz (1984) p. 95 cited by Oakley, A. (2000) Experiments in Knowing: Gender and Methods in
the Social Sciences. Cambridge: Polity Press. pp. 37–8.
2 The Economic and Social Research Council (ESRC) funds research and training in social and
economic issues. It claims an international reputation both for providing high-quality research on
issues of importance to business, the public sector and government and for commitment to training
excellence, which produces world-class social scientists. ESRC is an independent organisation,
established by Royal Charter, but receives most funding through the Government’s Office of
Science and Innovation. Its budget of £181 million (2007/2008) funds over 2,500 researchers in
academic institutions and policy research institutes throughout the UK.
3 Significantly, whilst the British (Labour) government supported the application of The Convention to
recently-liberated, continental states, it resisted its application to the UK on the grounds that history
had shown that it was unnecessary and that it might impede the government’s programme of
post-war regeneration including large-scale nationalisation and economic planning. The
Convention was finally only adopted by the Human Rights Act, 1998. Jacobs, F.G. (1978) European
Convention on Human Rights. Oxford; Clarendon Press. p. 214.
4 EFTA: European Free Trade Area: formerly a seven-state free trade area now reduced to the
non-EU states of Iceland, Liechtenstein, Norway and Switzerland.
5 Roethlisberger, F.J. and Dickson, W.J. were Mayo’s principal associates.
6 Mayo, E. (1949) Hawthorne and the Western Electric Company. London: Routledge.:p. 60
7 Scientific management was originated in the 1890s by Taylor, W.F. who argued that productivity of
workers could be increased by the marriage of science, engineering and classical economics to
achieve greater organisation, specialisation, de-skilling of tasks and the use of bonuses. It is
associated with ‘time-and-motion’ regimes.
8 Franke, R.H. and Kaul, J.D. (1978) The Hawthorne experiments: first statistical interpretation,
American Sociological Review, 43: 623–43.
9 Webb, S. and B. (1932/1975) Methods of Social Study. LSE: Cambridge University Press. The
Webbs were founder members of the Fabian Society and the London School of Economics.
Chapter 3
The Philosophy and Principles
of Research
1. To develop an understanding of the main philosophic and theoretical prin-

ciples underlying research approaches including positivism, empiricism,
behaviouralism, naturalism, feminism and postmodernism.
2. To introduce the concept of ‘causality’.
3. To introduce and distinguish between inductive, deductive and grounded
research.
Introduction
Qualitative and quantitative methods are regarded by many practitioners as simply

alternative sets of designs and techniques for carrying out research. However, both
qualitative and quantitative methods stem from fundamentally divergent ontological
and epistemological paradigms having their roots in philosophy. Ontology is, simply
expressed, the branch of philosophy (‘thinking about thinking’) devoted to the nature
of being. It considers such questions as: do minds exist? In contrast, epistemology
is the branch concerned with theories of knowledge. Theory is, simply stated,
a statement of general principles of the underlying relationships in phenomena or
events. Theory may be expressed as laws, propositions, arguments or hypotheses
(tentative explanations). Historically, theory has generally been descriptive in
terms of describing and, therefore, explaining relationships. Some descriptive
theories have been elevated to the status of laws (for example, Newton’s law of
universal gravitation) when they are accepted as having been universally verified
by observation. Newton’s law was subsequently superseded by Einstein’s theory of
relativity. The title ‘law’ is therefore rarely given to theories. Alternatively, theory
may be normative where it proposes what ought to be the relationship. For example,
the theory of egalitarianism is that all people should be treated equally regardless of
origin or circumstances. A proposition is an unproven, generalised explanation.
The Philosophy and Principles of Research 23
An hypothesis is a specific proposition which is presented for testing by research.

It can be derived from other theory or a researcher’s ‘hunch’ (informed guess based
on theoretical insights or observation).
Arguably, the essential operational distinction between quantitative and qualita-
tive research is whether the researcher is a neutral, objective observer studying a
person as an object, or as an independent variable engaging subjectively the person
as a subject or client. Positivism and naturalism are prototypical of the competing
ontologies.
Positivism
Until the Age of Enlightenment (Eighteenth century), significant events were

attributed to God, his miracles, or magic. So Hobbes and other philosophers were
forced to invoke God and king to legitimate their writings. While David Hume
challenged this practice, the French philosopher, Auguste Comte (1798–1859) while
developing the ideas of Saint-Simon is credited with ‘secularising philosophy’ and
developing a ‘religion of humanity’.1 He sought to place the French Revolution –
whose Terror undermined British support for the Enlightenment – into an historical
context as an intermediate phase of the development of knowledge. He argued that
all societies were fated to move from a theological stage of ‘fictitious knowledge’
(in which all otherwise inexplicable phenomena were attributed to spiritual forces)
via an intermediate, metaphysical stage to a positive stage. The underlying beliefs
of the theological stage were medieval faith and custom in which the family was
the social base.2 Those of the metaphysical stage were Enlightenment philosophy,
the ‘scientific revolution’ and the nation state. The state of knowledge of the
positivist stage was scientific when rational logic and humanity prevailed. When
these purer forms of explanations were reached, there would be no miracles or
magic: phenomena would be explained by physical causes. Explanation would
be confined to verifiable and measurable correlations between phenomena which
revealed universal laws.
The essence of positivism is the application of natural science methods (i.e. those
used in astronomy, maths, physics, chemistry and quantitative biology) to social
sciences.3 Comte advocated a value-free form of enquiry which would eliminate
any bias. However, his work has been criticised for accepting the dominance of
white society and providing opportunities for social engineering.4 Positivism would
therefore enable Man to control all other forms of nature. In its time, positivism
was ‘progressive’. It became the dominant paradigm in the social sciences until the
1960s when it became increasingly a pejorative term (i.e. a term of abuse).5
Comte’s work was developed by John Stuart Mill, Emile Durkheim, Ernst Mach
(whose name is commemorated in aerodynamics) and the interwar ‘Vienna Circle’
of ‘logical positivists’ including Wittgenstein and Popper (author of The Poverty
of Historicism) and other, mainly Jewish, mathematician-philosophers. These positivists
sought to apply scientific measurements and objectivity to generate causal laws.

They relied on specific assumptions about social reality. Social reality was ordered
rather than random: cosmos rather than chaos. This ordered system was based on
the rationality of self-interested individuals who, ceteris paribus (‘other things being
equal’), behaved in the same way. The reasoning capacity of individuals (i.e. their
rationality) enabled them to distinguish fact from fiction. So positivist explanations
could only be entirely logical and derived from the facts observed. Observable facts
were entirely different and a better foundation for understanding behaviour than
abstract theories, ideas and feelings. Positivist researchers are entirely objective and
without political or religious bias.
It was the application of positivism which enabled the study of social life to claim
the title of social science and the study of politics to legitimate nineteenth century
claims to the title ‘Political Science’.6 Popper’s major contribution to positivism was
to replace the test of verifiability (an early, central doctrine of the Vienna Circle) by
falsifiability, falsification or fallibilism. No theoretical generalisation could be finally
verified as truthful or certain: a single unexpected outcome could challenge an
hypothesis. Mere prima facie (‘at first sight’) proof was insufficient: absolute certainty
was essential. This could never be attained. A researcher could never claim or show
that they could not be wrong. Furthermore, an hypothesis could not be accepted
until the ‘alternative hypothesis’ were dismissed. After initial claims to inductivism,
positivism became essentially deductive (moving from the general to the particular),
hypothesis-testing research which tended to rely on numeric data to make objective
claims. This change reflected the recognition that researchers could never collect
the universe of information potentially available. Furthermore, selecting a research
question or topic implied a theoretical starting position; the researcher’s mind was
rarely a ‘tabula rasa’ (Locke’s empty slate). Subsequently, positivism has been criticised
for adopting a ‘mechanical model’ of man. More recently, it has also become
criticised for being heavily gendered, i.e. perpetuating the historic dominance of
men and the subordination of women.
Types of positivism include logical empiricism, behaviourism and behaviouralism.
Empiricism
Empiricism is, simply stated (and at its most extreme), the doctrine that the only
source of real knowledge (‘the facts’) is experience gained by the senses through
observation and experiment. Only the observable can be observed and measured.
Emile Durkheim (1858–1917) is widely regarded as a foremost empiricist in his
advocacy of the scientific study of social facts (objects of analysis particular to a
social science discipline). The empirical research for which he is best known is his
comparative study of suicide to determine its causes (Suicide, 1897). He regarded
suicide as the ultimate outcome, an event unique to the individual. Using official
records and by analysing these for ‘empiric regularities’ (patterns and correlations),
he sought to replace the traditional explanations of moral weakness or sin, by societal

explanations. He identified three common factors as potentially causal: religious
denomination; membership of rural or urban society; and existence of dependents.
These explained four main types of suicide: altruistic; egotistic; anomic; and fatalistic.
He later developed a functionalist view of religion whereby ministry reproduced
institutions, e.g. marriage.
Behaviouralism must be distinguished from Behaviourism.
Behaviourism
Behaviourism is a school of psychology associated with Harvard professor B.F.
Skinner (1904–90). He was the architect of the controllable ‘Skinner box’ and
‘programmed learning’. Its key tenet is that only observable behaviour may be
scientifically studied. So, while behaviour may include speech, intentions and
mental processes must be disregarded as ‘unobservable’. Behaviourism is associated
with animal experiments and the application of causal relations, e.g. Pavlov’s
dogs, to operant conditioning and behaviour modification. This emphasis on
experimentation and statistical analysis led some empiricists to subscribe to the ideas
of eugenics (the science of heredity associated with proposals for scientific breeding
to reproduce ‘higher’ intellectual characteristics and to exclude genetic weaknesses)
promoted by Darwin’s cousin, Galton (‘whenever you can, count’). Indeed, the
UCL professor, Professor Spearman, who devised the widely used Spearman’s
coefficient of correlation, r, was a leading eugenicist (and early socialist). Today,
Skinner’s contribution has become superseded by greater support for ‘cognitive
learning’.
Behaviouralism
Behaviouralism (note special spelling) was the term adopted by social and political
scientists for the application of positivism and empiricism to test and extend
explanatory theory at the level of state, party or other group, or individual.
In particular, the new approach offered a progressive approach to those academics
critical of the discipline’s obsession with the ‘old institutionalism’ of the study of the
structures of government. It became the dominant approach of politics – especially
in the US – in the post-war period. The approach is characterised by its greater
use of: theoretically derived hypotheses; sampling; early use of computers for data
collection and analysis; statistics; and belief in replication. Its principal exponent
remains Robert Dahl whose 1950s study of New Haven, Connecticut (described in
Who Governs?) remains the classic example of behaviouralist research. It demonstrates
the strengths and weaknesses of the approach.7
Dahl sought to explore two, rival hypotheses: the elite theory view (‘of European
origin’) that a unified oligarchy governed US cities (as argued by Hunter from his
study of Atlanta and C. Wright Mills’ Power Elite); and the (alternative) pluralist
theory that urban political systems were polyarchic. In his introduction, Dahl
noted the extreme variations of income and educational attainment, property
ownership and material environments in the US and asked: ‘Given the existence
of inequalities like these, who actually governs in a democracy?’. The research
problem which Dahl faced was how to observe and measure power and influence.
He chose what he called an ‘eclectic approach’ which spread the risk of unreli-
ability of data over six separate sources of information and methods of analysis,
namely:
1. To study changes in the socio-economic characteristics of incumbents in city

offices in order to determine whether any rather large historical changes may
have occurred in the sources of leadership;
2. To isolate a particular socio-economic category and then determine the nature
and extent of participation in local affairs by persons in this category;
3. To examine a set of ‘decisions’ in different ‘issue-areas’ in order to determine what
kinds of persons were the most influential according to one operational measure
of relative influence, and to determine patterns of influence;
4. To survey random samples of participants in different sub-areas in order to
determine their characteristics;
5. To survey random samples of registered voters in order to determine the
characteristics of those who participate in varying degrees and in varying ways
in local affairs;
6. To study changes in patterns of voting among different strata in the community.
(Dahl, 1969: 331)8
The cornerstones of his research were observation of ‘what actually happened’

in terms of the decisions made at meetings, and interviews with business and
political leaders. Dahl concluded that different, separate groups dominated different
policy areas. Furthermore, power, in terms of decision making, was openly
exercised by elected officials who were accountable and thereby sensitive to
public opinion. Therefore political power was not organised into a single pyramid
with a ruling elite at its apex. Instead, political power in New Haven was
marked by pluralism, openness and competition. But Dahl’s conclusions were
confounded by the riots which destroyed inner areas of New Haven in 1965.
This raises wider questions of research: are answers the results of the methodology
used?
Dahl’s approach was famously criticised by Bachrach and Baratz (after Schattschnei-
der) who argued that there was a second, unobservable dimension of power: the
power to control the agenda (Bachrach and Baratz, 1962).9 So, Dahl’s observation
that the officer’s recommendations to the City Planning Commission were generally
approved overlooked the possibility that other reports were excluded from the
agenda. Critics argued that potentially-controversial reports were withdrawn by

non-elected officials at their private, agenda-setting meeting with the senior officials.
Alternatively, other recommendations were never aired because they would be likely
to be rejected and harm relations between commissioners and advisers.10 Stephen
Lukes argued (drawing on Gramsci’s writings on hegemony) that there was a ‘third
dimension of power’:
the most supreme and insidious use of power which prevents grievances by shaping
people’s perceptions, cognitions and preferences in such a way that they accept their
role in the order of things.
(Lukes, 1974: 24).11
Furthermore, the ‘counterfactual’ (what you would have expected to happen but
didn’t) was - like Sherlock Holmes’ ‘dog that didn’t bark’ in Conan Doyle’s Silver
Blaze – evidence of the exercise unseen of third-dimensional power.12 But the
de-bunking of Dahl’s empiricist, behavioural model was not accomplished alone
by new, theoretical argument. As a noted pluralist and supporter of the ‘American
dream’, he has been criticised for wishful thinking and ‘seeing what he wanted to
see’. More recently, a systematic re-analysis of Dahl’s data has suggested serious
methodological and interpretative weaknesses. These were that: the survey samples
were not representative; Yale University was a much greater power-player than
(its professor) reported; he too readily accepted the ‘truth’ of interview statements
by the mayor and others which other documentary evidence contradicts; and he
was not sufficiently sceptical (Domhoff, 2005).13
Behaviouralism has become widely criticised in recent years. Three main criticisms
have been made. First, the emphasis on deduction meant that research became
‘boxed-in’ from developing new theory. Second, it tended to concentrate on
more readily measurable phenomena rather than other, more relevant topics
(Sanders, 2002: 52).14 And, third, important theoretical concerns and nuances were
unexplored or overlooked.
Naturalism
Naturalism has two, almost uniquely, contradictory conceptualisations. The first,

the minority view held by Giddens, is that naturalism refers to the adoption in
the social sciences of models of inquiry from the natural sciences. So naturalism is
effectively a new, non-pejorative name for positivism. But the dominant view is that
naturalism rejects positivism in favour of a free-standing method of inquiry which
is essentially humanistic and hermeneutic (after Hermes, interpretative) and concerned
with social meanings, actor’s beliefs, motives, purposes and reasons, which lead
to ‘social actions’, rather than frequency. While positivism emphasises scientific,

controlled, replicable experimentation, naturalism seeks to study everyday life in
naturally-occurring situations. Naturalism is experiential: researchers seek to see the
world through the eyes of the ‘subject’ and to understand what people feel, interpret
and do. Moods, ideas, identity and beliefs are the stuff of naturalistic research. People
have a capacity for language and use it to construct and reproduce their social worlds.
Society, unlike inanimate material, is always changing. But individuals are neither
bound by universal rules nor independent of their contexts; they are structured by
institutions. Because there can be no universal organising principles, then there can
be no research hypotheses. Naturalistic research is primarily inductive. The five main
principles of naturalistic research are therefore:
• meeting subjects in their own habitats of home or work, etc

• asking them their views and meanings
• asking them in such a way that they can reply in their own language, dialect and words
• achieving depth of inquiry
• addressing their social context and situation.
Feminism
Feminism offers another, additional critique of positivism: ‘that the concern with
measurement and control underlying quantitative, experimental ways of knowing
developed is part of the same social process which enables men to exercise power
over women’ (Oakley, 2000: 16).15 Feminists point to the historical association
between quantification and masculinity. Social science defined society in terms of
male values which reflected historic power relations in society where women were
simplistically categorised. Feminists argued that the relation between researchers
and researched reproduced the power relations between men over women. They
argued that relations between researchers and subjects should be consensual and that
greater weight should be given to ethical considerations. The subjects should be
empowered by the research process. Research should therefore be an emancipatory
project: ‘research for women not about them’. Feminist perspectives of research were
strengthened by increasing public opposition to experimental research, especially
involving animals, to which women were the main opponents. But the eminence
of qualitative methodology has been marked paradoxically by the divergence
between those traditional and new feminists. The traditionalists continue to promote
exclusively qualitative method. The new feminists (and others) argue that the real
conditions of women and other exploited, powerless groups can be more effectively
exposed by greater use of quantitative data, especially by harder, objective data on
the distribution of low pay, poverty and unwaged work (especially parenthood).
Prototypical examples of new feminist research in Politics are the published works
of Norris (1995, 1999) and Lovenduski (1995).16
Post-modernism
Post-modernism initially developed in the 1970s from a critique of modernism

and positivism by intellectuals in the arts and architecture disenchanted by the
failure of modernity. This critique was given a stronger intellectual core by social
scientists, drawing on existentialism, nihilism and anarchism, who argued that
the founding concepts of the Modern Movement, industrialisation, urbanisation
and the individual as producer, were fast becoming obsolete. They saw instead a
new, post-industrial age in which the old divisions between economic classes had
become eroded by new lifestyles in the ‘information societies’ when the individual
had become a consumer able to exercise choice in the market. Post-modernism
challenged universal truths and the meta-narratives of Marxism, liberalism and other
utopian ideas. There is no certainty: any knowledge is local, contingent and partial.
The scientific researcher is unable to be objective. The individual is not rational.
Language is not neutral: meaning is modulated by dominant interests. There is
no division between the arts and social sciences. The complexity of the social
world cannot be captured by research or causality deduced. Intuition, emotion and
imagination should not be discouraged. Instead, research should engage, stimulate
and evoke (Neuman, 2003: 89).17 On the other hand, post-modernism in the
discipline of Politics has been criticised as relativist, pessimistic and anti-progressive
(Heywood, 2001: 101–03).18
The significance of postmodernism for research in Politics has been the impetus
given to the closer study of ‘talk and text’ and ‘interpretative social research’ with
its emergent disciplines of hermeneutics, ethnography and qualitative sociology.
It therefore favours participative fieldwork and real involvement of researchers
with their subjects. It revives the tradition of ‘trying to see the world through the
subjects’ eyes’ by verstehen (empathetic understanding) advocated by Max Weber
(1864–1920).
Causality
Post-modernism does not entirely dismiss the concept of causality (‘cause and
effect’ or causation) as a basis for explanatory theory: it argues that the world
is chaotic, rather than a closed system, in which causes and consequences cannot
be identified. Many post-modernists therefore adopt functional, structural, historical
or interpretative explanations where, for example, a functional explanation of poverty
might be because it ‘works’. But causality is central to all the main approaches used
in Politics research where causes are sought to explain effects. In emergent, everyday
English, causes are often referred to a ‘drivers’. But, in academic research, causes
and effects are generally called variables where:
• causes are termed independent variables or causal events

• effects are termed dependent variables, consequent events, or outcomes.
Cause and effect appears straightforward. For example, from the report or observa-
tion: A knocks over B, then you can claim that A is the cause of B’s falling over. But
how do you know that A is not responding to B’s behaviour or seeking to protect
C? The social world and political world are incredibly complex. Unlike, say, nuclear
physics, the subjects of Politics research are sentient human beings exercising some
agency (personal autonomy) within a structure (rules of the game). The degree of
agency will vary with the individual context. Assume, for example, that a university
teacher, A, appears increasingly exasperated by student B. A raises a clenched fist
directly to B’s nose. What will happen? Well, you don’t know what will happen.
You can only speculate. A may or may not hit B. B may adopt a passive response
and accept ‘whatever happens next’. Or B may take pre-emptive action by either
moving or striking first. They may strike back if hit. But the choices of A and
B are both contingently structured by the teacher-student power relation, gender
relations, academic regulations and criminal law.19
There is therefore a substantial risk that the researcher may identify, incorrectly,
causes merely because they appear to immediately precede apparent effects. This
fallacy is termed post hoc, ergo propter hoc (‘after this, therefore because of this’). This
problem was considered by the Scottish Enlightenment philosopher, David Hume
(1711–76). He dismissed the claim that causes necessitate the events that are their
effects. He also refuted the argument that cause could be determined by reason alone.
The belief that flames are hot can only be based on sensory experience, custom and
practice. He developed eight ‘general rules’ which must be met before ‘probable
causation’ could be claimed.
1. The cause and effect must be contiguous in space and time.

2. The cause must be prior to the effect.
3. There must be a constant union betwixt cause and effect. ’Tis chiefly this quality,
that constitutes the relation.
4. The same cause always produces the same effect, and the same effect never
arises but from the same cause …
5. Where several different objects produce the same effect, it must be by means of
some quality which we discover to be common amongst them.
6. The difference in the effects of two resembling objects must proceed from that
particular, in which they differ …
7. Beware not to draw such a conclusion from a few experiments …
8. An object which exists for any time in its full perfection without an effect, is not the
full cause of that effect …20
After Hume, five precise conditions have been specified for causation: covariation;
constancy of association; cause must take place before the effect; independent and dependent
variables must be discrete (rather than continuous); and non-spuriousness.21
Covariation describes the phenomenon where two variables appear to change
at the same time, for example, unemployment and poverty appear to increase or
decrease at the same time. The covariation can be plotted graphically and tested for
association (the degree of consistent relationship) by the statistical test of correlation.
However, ‘association does not necessarily equal causation’.22 And association does
not identify the independent and dependent variable. There are several ways that
variables may change at the same time. In social science, the cause is normally
indicated as x (from the Greek xenon meaning unknown) and the effect y. The causal
effect of independent variable x, on the dependent variable y, can be represented as
shown by the diagram:
x→y
However, another, independent variable, x2 may also be affecting y by the process
termed addivity:
x1 → y
↑
x2
Alternatively, x2 may be affecting x1 directly by the process of intervention:
x1 → y
↑
x2
Or x2 may act as a type of catalyst on x1 and y. This is termed interaction:
x1 → y
↑
x2
Finally, the association between x1 and y may be a spurious relation where they
are both being affected by x2 :
x1 y
x2
In practice, few causal relations in Politics research involve only one cause. Most
effects have many causes; and these causes will affect many effects. There is also the
argument that, even so, causes are only the tangible manifestation and agent of the
underlying distribution of power within society:
The flaw in the pluralist heaven is that the heavenly chorus sings with a strong upper-
class accent.
(Schattschneider, 1961: 35).23
Hume explained cause and effect in terms of billiard balls in which moving ball A
hits static ball B which is then propelled along the ‘green baize’. However, today
we would claim that, in Politics, the baize is not level and smooth but furrowed and
pitted, structured by institutions that affect the trajectory of the ball and frustrate the
trajectories of A and B.
Inductive, deductive and grounded research
These methodologies generally follow the explanatory theory adopted.

Inductive research has been defined as: ‘the inference from the particular to the
general’ (Honderich, 1995: 403).24 It is used to ‘build’ theory and is traditionally
the methodology used in the natural sciences where observation of phenomena
is followed by a search for new explanatory theory, e.g. the behaviour of super-
conductors at low temperatures. Inductive research can also be adopted to find
answers and then explanations for questions of ‘what if ?’ For many years, ‘the
problem of induction’ attracted the attention of philosophers like Hume who argued
that all inferences from observation relied on the premise of the future always
resembling the past. But the roots of modern-day induction lie in the philosophy of
idealism which sees the mind as the only true source of ideas. The methodology is
widely used by sociology researchers where they are trying to understand the social
meanings (beliefs, attitudes, and reasons) of their subjects exhibited through their
behaviour. These researchers adopt Weber’s verstehen and interpretative methods.
The idealised inductive research process may therefore be illustrated as:
1. select the topic

2. choose the research question
3. collect data (through repeated observation of phenomenon)
4. interpret the data
5. develop a theoretical explanation of data
6. collect modified data to test initial explanation
7. reiterate (repeat) until conclusions can be made which best meet the test of falsifiability.
Conversely, deductive research can be explained as seeking inferences from the

general to the particular. It is the methodology adopted by positivist, empiricist and
behaviouralist researchers. While this methodology is associated with quantitative
enquiry, it is widely used in Politics research (especially by undergraduates, and
‘taught’ graduates) to apply and test the application of theoretical models to new
(often overseas) contexts. Examples include the application of policy transfer theory
to the transfer between developing states, and Stone’s urban regime theory to UK
cities. The idealised model is therefore:
1. select the theory, e.g. urban regime theory

2. frame the research question: e.g. ‘can the regeneration of UK cities be attributed to
urban regimes?’
3. state the hypothesis: (This might be:‘No: in practice many regimes have been concocted
of the “usual suspects” to meet government criteria for aid; other factors are more
significant’.)
4. collect selected data relevant to the research question: e.g. urban regimes in successful
and unsuccessful urban regeneration
5. analyse and interpret the data
6. confirm or infirm (disprove) the hypothesis.
However, in practice, the distinctions between deductive and inductive research are
blurred. In particular, the claim by inductive social researchers that their research
is a-theoretically grounded may be questioned. Can social scientists ever really ‘rid
their minds’ of the grand theories that they acquired in their earlier years? If they
really had no preconceptions or hunches, then how or why did they come to select
the phenomena for study? Can a young, white, male, middle-class, western social
researcher ever really ‘see the world through the eyes’ of an old, white, working-
class western woman let alone a Third World subject speaking an entirely different
language?
Grounded research was proposed by Barney Glaser and Anselm Strauss of
Columbia University in the 1960s as a new ‘third way’ derived from their critique
of the dominant tradition of deductive research in US sociology (Glaser and Strauss,
1995).21 In their judgement, US social researchers were pre-occupied with testing
hypotheses derived from the ‘grand masters’ of sociology: Marx, Weber, Tonnies
and Durkheim. Consequently, new researchers were not developing new theory to
explain the new social dynamics of a fast-changing, urban society. Most research
added only detail or embellishment. Social research had become insular and was
effectively overlooking new concerns. They were also critical of some of the claims
of a-theorocity of inductive researchers. At the same time, they were critical of
the cost and resource demands of large-sample quantitative data. They wanted to
develop a new approach which harnessed the skills and other resources of social
researchers to generate new, exciting, relevant and imaginative research.
The solution they devised was grounded research where the findings would
be entirely grounded in the data collected. The essence of grounded theory lay in
the selection by researchers of almost semi-randomised variables (for example, by
examining the contents of the ‘returned book trolley’ in the university library).
Rather than hypothesise specific causal relations, the researchers should first seek
out evidence of any relation. The initial sample and data would be limited. Their
findings would be grounded in that data, tested and re-formulated in further samples
until theoretical saturation were reached, when the additional data failed to reveal
any further significant findings. Importantly, ‘less was more’. It would no longer be
necessary to collect, at vast financial and opportunity costs, all the data before analysis
were attempted (with the temptation to overstate the findings to justify the costs).
One noted variant of grounded research is its application to deductive research by
adopting small, iterative samples to test hypotheses. However, despite the appealing
propositions put forward by Glaser and Strauss for ground-breaking ‘quick and dirty’
research, the criticism remains that, in practice, the selection of variables for scrutiny
invariably reflects the values and theoretical biases of the researchers. Additionally,
how can the researcher claim that the process is a-theoretical without reviewing
existing theory and the associated literature? How can they know that their research
is new and original?
The idealised grounded research model is:
1. select the variables

2. collect data from a small sample
3. advance an explanation
4. test against larger samples
5. revise the initial explanations
6. reiterate until theoretical saturation is reached.
Grounded research has great appeal to most British Politics researchers because it
avoids number-crunching and can be cited to justify a ‘thin’ theoretical review and
limited immersion in the field. Most researchers develop variants appropriate to their
research question.
1. What is positivism? What are its strengths and weaknesses?

2. Critically compare positivism and naturalism.
3. Explain Popper’s concept of ‘falsifiability’. In what circumstances might it be

employed in Politics research?
4. ‘Grounded research’: another useless American import? Defend.
5. You have been asked to explore the relationship between social class, party
identification and turnout in the UK. Which research methods would you adopt?
Why?
6. What is verstehen? Is it ever possible? In what circumstances is it most likely and
least likely to be achieved?
FURTHER READING
Dahl, R. (1969) Who Governs? Democracy and Power in an American City. New
Haven: Yale University Press. pp. A new edition was published in 2005, ISBN
03001 03921. This is a well-written and detailed account of Dahl’s research
method and his interpretation of its findings. It remains a seminal work and
model for present-day researchers.
De Vaus, D.A. (2001) Chapter 2: Theory and Social Research. In Surveys in
Social Research. London: Routledge. pp. 11–26. This short account
summarises the interaction of theory and research, the process of theory
building and testing, and the sources of theory.
Glaser, B.G. and Strauss, A.L. (1967/1995) The Discovery of Grounded Theory:
Strategies for Qualitative Research. New York: Aldine de Gruyter. This very pacy
account provides the critique of contemporary research in the early 1960s from
which grounded theory was developed. As the American Sociological Review
claims: ‘The authors successfully transmit the sense of adventure, air of
excitement and of positive apprehension over what is discovered as one tracks
down clues and sorts [them] among attractive alternatives’.
Gubrium, J.F. and Holstein, J.A. (1997) The New Language of Qualitative Method.
Oxford: Oxford University Press. pp. 19–94. This 244-page textbook provides
incisive chapters on naturalism, ethnomethodology, emotionalism and
post-modernism including extracts from many case studies.
Marsh, D. and Furlong, D. (2002) Chapter 1: A Skin is Not a Sweater: Ontology
and Epistemology in Political Science. In Marsh, D. and Stoker, G. (eds.) Theory
and Methods in Political Science. Basingstoke: Palgrave Macmillan. pp. 17–41.
This chapter reviews the centrality of ontology and epistemology and the
divergent perspectives of positivism, interpretivism, and constructivism, and
demonstrates their application to case studies of globalisation and multi-level
governance.
Neuman, W.L. (2003) Chapter 4: The Meanings of Methodology. In Social

Research Methods: Qualitative and Quantitative Approaches. London: Allyn
and Bacon. pp. 68–94. This chapter describes the historical development of
methodologies in social science. The author identifies the differences between
the main approaches by asking eight questions. They include: what is the basic
nature of human beings? What is the relationship between science and
common sense? The answers are tabulated on Table 4.1 on p. 91. This will be
of particular values to researchers new to Politics and seeking to understand
the complex approaches.
Smith, M.J. (2003) Chapter 3: The Emergence of the Social Science. In Social
Science in Question. London: Sage. pp. 75–116. This is a core text of the Open
University’s social science modules. It describes and evaluates the origins of
positivism and its contribution to emergent social sciences and the
development of empiricism. A section is devoted to explaining Popper’s
‘falsificationist solution’ to the ‘problem of induction’. The chapter ends with a
rich collection of readings from Comte, Bentham and Eysenck.
Notes
1 Lacey, A. (1995) In Honderich,T. The Oxford Companion to Philosophy. Oxford: Oxford University
Press. pp. 145, 705–6.
2 Smith, M.J. (2003) Social Science in Question. London: Sage. p. 79.
3 The ‘natural sciences’ are generally considered to be astronomy, physics, maths, chemistry and
quantitative biology.
4 Smith, M.J. (2003) Social Science in Question. London: Sage. p. 81–2.
5 Williams, R. (1983) Keywords: A Vocabulary of Culture and Society. London: Fontana. pp. 238–9:
‘[positivism] becomes a swear word, by which nobody is swearing’.
6 The first ‘school of political science’ was established at Columbia in 1880: Burnham, P.,
Gilland, K., Grant, W. and Layton-Henry, Z. (2004) Research Methods in Politics. Basingstoke:
Palgrave Macmillan. p. 12.
7 Dahl, R. (1969) Who Governs? Democracy and Power in an American City. New Haven: Yale
University Press. A new edition was published in 2005, ISBN 03001 03921.
8 Dahl, R. (1969) Who Governs? Democracy and Power in an American City. New Haven: Yale
University Press. p. 331.
9 Bachrach, P. and Baratz, M. (1962) The two faces of power. In American Political Science Review,
56. ‘Dahl thereby excludes the possibility that in the community in question there is a group
capable of preventing contests from arising on issues of importance to it … [Dahl is] unable
adequately to differentiate between a ‘key’ and ‘routine’ political decision’.
10 This author’s own experience supports the claim that, in UK local government, the real decisions
on planning policy (but not planning applications) are made at the ‘pre-agenda’ meeting between
the (elected) executive members and the ‘officers’ (unelected officials).
11 Lukes, S. (1974) Power: A Radical View. London: Macmillan.
12 I experienced a counterfactual in my research on hegemony and social exclusion in rural England
where I note that the estate villages appeared to lack the war memorials that were characteristic
of the greens or crossroads of villages elsewhere. I found that the memorials had actually been
located inside the village churches – despite most of the casualties having been members of
Methodist chapels.
13 Domhoff, D.W. (2005) Power in America: Who really ruled in Dahl’s New Haven.
http:/sociology.ucsc.edu/whorulesamerica/power/new_haven.html (20.06.2006).
14 Sanders, D. (2002) Chapter 2: Behaviouralism. In Marsh, D. and Stoker, G. (eds.) Theory and
Methods in Political Science. Basingstoke: Palgrave Macmillan. p. 52.
15 Oakley, A. (2000) Experiments in Knowing: Gender and Methods in the Social Sciences.
Cambridge: Polity Press. p. 16.
16 Evans, G. and Norris, P. (1999) Critical Elections. London: Sage, and Norris, P. and, Lovenduski, J.
(1995) Political Recruitment: Gender, Race and Class in the British Parliament. Cambridge:
Cambridge University Press.
17 Neuman, W.L. (2003) Social Research Methods. Boston: Pearson Education. p. 89. Neumann
develops an alternative classification of social science into three paradigms: positivist, critical
social science and interpretative social science. He explains the latter as: ‘the systematic analysis
of socially meaningful action through the direct detailed observation of people in natural settings
in order to arrive at understandings and interpretations of how people create and maintain their
social worlds’ (p. 71).
18 Heywood, A. (2000) Key Concepts in Politics. Basingstoke: Palgrave Macmillan. pp. 101–3.
19 In UK criminal law, assault in ‘assault and battery’ refers to the perceived threat of battery by the
‘victim’.
20 Hume, D. (1739) A Treatise on Human Nature. 1: 467–8.
21 Glaser, B.G. and Strauss, A.L. (1967/1995) The Discovery of Grounded Theory: Strategies for
Qualitative Research. New York: Aldine de Gruyter.
22 Voight, P. (1999) Dictionary of Statistics and Methodology. London: Sage. p. 58
23 Schattschneider, E.E. (1961) The Semisovereign People: A Realists’s View of Democracy in
America. New York: Holt, Rinehart and Winston. p. 35.
24 Honderich, T. (ed.) (1995) The Oxford Companion to Philosophy. Oxford: Clarendon Press.
pp. 405–6.
Part II
Methodologies
Chapter 4
Qualitative Versus Quantitative
Methods: A Relevant Argument?
‘Quantitative research is hard and reliable … qualitative research is deep and rich’
(Bryman, 1996:94)1
‘Qualitative and quantitative methods are more than just differences between research
strategies and data collection procedures. These approaches represent fundamentally
different epistemological frameworks for conceptualising the nature of knowing, social
reality, and procedures for comprehending those phenomena’.
(Filstead, 1979:45)2
1. To understand the origins, merits, strengths and weaknesses, claims and

counter-claims of qualitative and quantitative research.
2. To consider whether ‘mixed methods’ are incompatible with best research
scholarship.
3. To enable each student to reach and defend their own preferences.
Introduction
Qualitative and quantitative methods have historically been portrayed as mutually

exclusive research approaches promoted by competing specialists appealing to the
hearts and minds of new students.
The arguments and antagonisms have become very tired over the years as each
new cohort of students is asked to critically compare the competing claims. However,
the traditional divergence of opinion has been challenged by the increasing
interest expressed by Politics researchers in ‘mixed methods’ and the availability
of sophisticated CAQDAS (Computer-Assisted Qualitative Data Analysis System)
software. This uses quantitative methods to analyse qualitative data.3 These include
programs which systematically analyse ‘talk and text’. However, they depend
on systems of coding designed and applied by the researcher. So the analysis is

not therefore automatic. They also rely on assumptions, especially content analysis
software, about the significance of the frequency and regularity of words which are
essentially positivist.
The full, definitive discussion of the claims and counter-claims of qualitative and
quantitative methods is provided by Alan Bryman’s Quantity and Quality in Social
Research, London, Routledge, 1996/88. Read it.
This chapter seeks to summarise briefly the origins, characteristics, merits,
strengths and weaknesses, claims and counter-claims of qualitative and quantitative
methods. You will already be aware of the particular pitfalls and incoherence
of comparing two or more concepts. The conventional method of comparative,
descriptive narrative can become easily weighed down with repeated use of ‘on one
hand’ and ‘on the other hand’. A weakness of text is its linear, one-dimensional
presentation of information. This makes ready comparison difficult and time-
consuming. An alternative approach, which this book advocates for writing research
reports, is to tabulate (set out in a table) for ready comparison the core characteristics
of the concepts. Tabulation enables specific aspects to be extracted for detailed
discussion. This approach is used in Table 4.1 to highlight the essential differences
between the two methods.
The strengths and weaknesses of the two research methods are not mirror-images
of each other and are best considered separately.
Quantitative methods: strengths
The greatest strength of quantitative method lies in its general acceptance by others
as being rational, logical, planned and systematic. The findings are regarded as
credible. This method is therefore particularly favoured by public, research-funding
bodies keen to justify their investments to a sceptical public. It is also preferred
by news media whose audiences, they claim, regard percentages as ‘hard news’.
Therefore quantitative method, however complex the statistical methods employed,
is regarded as being straightforward and providing the facts. The researcher is seen
as dispassionate, objective and, therefore, trustworthy. It employs very large samples
designed to reflect and be representative of the population being studied. The use of
questionnaires ensures that every member of the sample is asked the same question in
the same manner. Supporters argue that attitudes can be measured by using scales, e.g.
the Likert scale of strongly agree, agree, etc. (see Chapter 9). Face-to-face contact
is not necessary and may contaminate the data. Research objects are best kept at
arm’s length. Geographically remote or immobile people can be surveyed by postal
survey, telephone or internet. Quantitative method makes best use of computers
and other new technology. So the time-scale for data analysis and the publication of
findings can be relatively short. Furthermore, the data can be re-examined, audited
Qualitative Versus Quantitative Methods: A Relevant Argument? 43
Table 4.1 Characteristics of qualitative and quantitative methods
Characteristics Quantitative Research Qualitative Research

Origins Natural science Social science
Philosophic roots Positivism Naturalism, feminism
Research design Deductive Inductive
Systematic Flexible
Contribution to theory Theory testing Theory building
Seeks explanation and Seeks explanation and solutions
universal laws
Researcher Observer, outsider Participant, insider
Researcher’s attitude Objective Subjective, empathetic
People are regarded as: Objects Subjects
Location Research centre or Field based
laboratory
Samples or cases Random Non-random
Large Small
Focus on universes Focus on minorities
Focus The individual The group (family, clan, peer group,
work group)
Records Frequency Social meanings
Data Numeric concept indicators Non-numeric concepts
Data collection Surveys and structured Personal interviews and
questionnaires unstructured interviews
Represented by Tables and charts Transcripts
Analysis Statistical Non-statistical
Generalisability High None claimed
Costs High Low
Findings Nomothetic (general laws) Ideographic (specific to historical
and cultural context)
and re-analysed or used for other purposes. The process of analysis is overt in which
the scope for bias by the researcher is deliberately minimised. The method enables
research projects to be carried out by teams in which specialist talents can be properly
exploited and work sub-contracted to agencies. Quantitative research effectively
conforms to the modern-day business model.
In terms of academic research, quantitative method is particularly suited to
the development of grand, meta and micro theory by testing logical hypotheses.
Applying Popper’s tests of falsifiability ensures that the research hypothesis can only
be accepted after all alternative hypotheses have been disproved. It offers the best
means of identifying and comparing the distribution between people, places and
times of phenomena such as party membership, voting, income, poverty, housing
conditions, and changing attitudes. It can also be used to identify clusters of
relatively small-scale phenomena and to analyse these statistically to identify whether
particular concentrations may be attributed to chance or potentially localised causes.

Quantitative method can mine and analyse existing data banks, for example,
census data and voting records. It is particularly useful in electoral research which
focuses increasingly on the problem in western states of declining turnout in what
commentators have termed the ‘period of post-democracy’ (Crouch, 2004).4 One
notable study used the data available from the British Election Study (BES) and
local government records to explain potential causes of the drop in UK turnout
in the 1997 general election (when, after 18 years in opposition, Labour replaced
the Conservative party in government). The drop, from 77.9% (in 1992) to 71.6%,
could be attributed mainly to the combination of the closeness of party ideology
and the extent to which the outcome was in no doubt (Heath and Taylor, 1999).5
The apparent certainty of the result was reported by news media using the evidence
of opinion polls – another quantitative device. In this way, ‘hard news’ became
self-fulfilling.
Weaknesses and criticisms of quantitative methods

These are mainly fivefold. First, the underlying doctrine of positivism is, as already
stated in Chapter 3, contestable in its application to the social world. Second, it is too
detached, remote and clinical to really understand and explore the complex social and
political world. Quantitative research is amoral. Third, its use by the social sciences
does not meet the high standards of the natural sciences in which its reputation
and claims lie. It is ‘bad science’. The samples, despite claims of randomness, are not
statistically reliable. There are too many variables for causality ever to be determined,
especially when the measures are superficial. Measures of strength of attitude are
fallible: is ‘strongly object’ twice the value of ‘object’? Fourth, quantitative research
in Politics research relies on the ability to express concepts as measurable indicators.
But, as already discussed, how can power be measured? Furthermore, official data
is often ‘doctored’ to serve government interests. To what extent are records of,
say, civil disturbances and arrests really representative of underlying dissent? Can we
really weigh an iceberg by measuring its tip? The necessity for measurable concept-
indicators means that quantitative research is confined to researching measurable
variables rather than more important issues. Lastly, the reliance on observation limits
the range and depth of research to what is both observable and measurable.
The stereotypical quantitative researcher is a clean-shaven, male, white-coated,
ascetic, Times-reading Dr Strangelove working in an ‘established university’.
Defence and counter-claims

Advocates of quantitative research rebut the criticisms by defending the applicability
of positivism and empiricism to most aspects of the social world. It is less imperfect
than qualitative research. In purist terms, it is the evil of two lessers. But samples
can be devised and statistical analysis applied to produce findings within specified
confidence limits. Questionnaires can be designed with sufficient checks and controls
to identify bias by respondents and questioners. Most concepts are capable of being
represented by measurable indicators. Subconscious feelings and motivations can be
identified by using projective questioning (see Chapter 9). Quantitative method accepts
its own fallibility which its users consciously seek to minimise. Most variables are
capable of observation by trained researchers using modern technology. At the end
of the day, quantitative research is reliable and concrete and has ‘put the science’
into Politics to create political science.
Qualitative methods: strengths
This is the dominant approach adopted in UK Politics research. This choice reflects
in part the lack of confidence of many UK researchers in their quantitative skills.
In the main, qualitative method is preferred because it is considered best suited to
the study, understanding and explanation of the complexities of social and political
life. The strength of this method lies in its unique capacity, through in-depth
interviewing and observation, to learn and understand the underlying values of
individuals and groups. It better enables theory to be created by induction. By
learning the social meanings that the subjects apply to their world, researchers
are better able to ‘see the world through the subject’s eyes’. So researchers can
identify and understand the interpretative lens that subjects adopt, and, therefore,
the dominant powers and institutions that frame the view and tint the lenses.
Essentially, politics takes more than one person: it is an intra- and inter-group
activity. Qualitative method enables the focus to be shifted from the individual
to the group(s) and to learn (following the theories of Goffman and others) how
meanings are negotiated between members and the group dynamics involved.6
It allows comparisons and distinctions to be drawn between what the individual
says in the privacy of a personal interview and what they say, or don’t say, in
a group. Furthermore, the varied opportunities for participant observation, along a
continuum between thresholder and full member, allow degrees of access to the group.
The method also enables minorities to be researched who would otherwise be
missed by sample surveys because they are either small in number or might be
unwilling to identify themselves. Covert groups include homosexuals, drug-users,
gang-members, and terrorists and their supporters. The method avoids distance and
objectivity by seeking verstehen (empathetic understanding). Many of the researchers
are also (critical) activists. They seek to expose exploitation and to improve the lives
of their subjects through policy change. They want to assist their subjects to see
their circumstances through other perspectives and lenses, and to develop their own
interpretations. They seek to empower their subjects through the research process.
There is therefore a normative dimension to qualitative method that quantitative
method lacks and, for that matter, avoids. The method also offers a greater focus on
verbal and other communications and the application of linguistics to the analysis.
Sophisticated (digital) recording devices enable the data to be shared with other
researchers. New, miniaturised technology allows fieldwork to be carried out in the
homes and workplaces of their subjects in a more surreptitious way. So the subjects
are not inhibited by the equipment and contingent variables are less affected.
Criticisms and weaknesses of qualitative methods

The archetypal criticism of qualitative methods is that the data collected is largely
anecdotal or exaggerated. The method is also ‘soft’: it lacks the intellectual and
operational rigour of quantitative method. The researcher is likely to ‘go native’
(over-identify with the subjects) or, alternatively suffer, from observer drift (where
obvious deviances become accepted norms that are under-recorded). So the data and
findings are highly unreliable. The claims to induction are spurious: the research
question and choice of issues and concepts must be influenced by prior learning.
The researcher is weighed down by bias. By participating with their subjects, the
researcher will inevitably contaminate the social field and determine outcomes, or at
least influence them in some way, rather than observe them. Furthermore, despite
their distaste for concept-indicators, qualitative researchers inevitably adopt concepts
which involve some form of proxy to make them intelligible to the subjects. Verstehen
is elusive. How can a young, white, middle-class male researcher ‘see the world
through the eyes’ of a female with the same background let alone a person from a
wholly different culture? But, even if the data is sound, there is the insurmountable
problem of interpretation. The researcher must interpret in turn the interpretation
by the subject of their world. The scope for misinterpretation is huge. How can the
researcher know or show that their interpretation is inaccurate? There is also the
ethical danger that, by participating closely in oppressed minorities, the researcher
may act as an agent provocateur who leads the group into harm’s way and then retires
to the sanctuary of their university.
The stereotypical qualitative researcher is seen by critics as long-haired, earring-
wearing, leather and denim-clad, Guardian-reading, smoking, Dr Howard Kirk
(The History Man) in a ‘new university’. Or, worse still, he is a woman.
Defence and counter-claims on behalf of qualitative

research
Advocates reject the criticisms made as wholly exaggerated. The social sciences are
not a soft option. Qualitative research is intellectually demanding and rigorous.
Researchers are fully aware of the weaknesses and dangers of close identification
between subject and researcher. Protocols and close supervision are required. Data is
not anecdotal. It is no longer accepted on trust: the researcher is required to make
Illustration 2 The great divide: Quantitative versus Qualitative Research. The

stereotypes
available full transcripts and other records of interviews and group meetings for
corroboration. There is a full paper trail available for audit purposes and external
examination. The research method provides the only means by which overlooked or
concealed minorities can be studied. In this way, their needs for special recognition
can be addressed by policy-makers before continued rejection leads to outright
dissent. The emphasis on studying subjects in their own habitat means that contingent
factors can be identified and the research effect (see Chapter 2) minimised. It recognises
that language is never neutral: ‘every word is a bias’ (Nietzsche). The normative
opportunities should be welcomed rather than criticised: research should improve
the circumstances of its subjects. The method acknowledges the central role of
the researcher in the research rather than pretending that this can be eliminated.
Quantitative method does not readily enable group behaviour to be studied. Overall,
qualitative research provides high quality data and findings, and deep, meaningful
insights into underlying values, fears and motivations of agents and actors in the
political world.
Mixed methods
Given the separate strengths and weaknesses of quantitative and qualitative methods,
then using them together – mixed methods – would appear to offer the best of
both worlds. But, given their opposing ontology and epistemology (see Chapter 3),
the criticism could be made of the lack of scholarship and the inevitable theoretical
weaknesses of seeking to mix oil and water. Alternatively, the criticism could be
rebutted as being too esoteric and academic and of overlooking the advantages,
especially of time and other costs, of bringing the two methods together when
the opportunities arise. The justifications cited for a mixed approach could include
synergy, corroboration of sources and data triangulation (see Chapter 7).
The received view among research practitioners tends to be that the differences
between the methods have been exaggerated. However, there remains strong
objection to a ‘mix and match’ approach. Instead, the use of combined methods
can be better accepted as appropriate where one method dominates and the other
is used in a secondary, supportive way.
Two examples of appropriate combined methods can be given. Both concern
research into voting behaviour.
EXAMPLE 1: Election turnout
UK election turnout has fallen since 1951.The research question is: why has turnout fallen? It
adopts a positivist, deductive approach in which Crouch’s theory of post-democracy is used
and tested. The research seeks to test Crouch’s argument that turnout has fallen because
globalisation and the decline of industrial production have loosened the previous sense
of antagonistic social identity and class. So the appeal of unions and political parties has
weakened and reduced support for active participation. Instead participation has once again
become an elite activity by ‘self-referential political class more concerned with forging links
with wealthy business interests’ (Crouch, 2004).7 In this case, the proxy-indicators of active
participation selected are: membership of parties and unions; attendance at political etc.
meetings; holding office; and, voting. The data will be collected from national published
records of voting and party, etc., membership since 1951. This population data will be
augmented by a national, postal questionnaire sample of a stratified sample of say 2,000
electors (following a pilot survey). It may offer hard, reliable evidence of changes in voting
behaviour, class consciousness and party identity and participation. Analysis is likely to
show a strong correlation between the decline in participation and loosening of political
identity. But it will lack compelling evidence of causal relations. In this case, the data could
be supported by interviews with a small sample of activists and non-activists to learn of their
personal experiences and to tease out their changing motivations. This qualitative data could
either be sought before the questionnaire survey to help frame the questions and alternative
answers, or afterwards to explore particular findings. In either event, ‘quotable quotes’ could
be incorporated into the research report to illustrate the findings and to create interest. The
report is likely to conclude with a discussion of the implications of the findings for ‘democracy’
and electoral reform.
EXAMPLE 2: Increasing abstention
The second example also concerns the fall in UK turnout since 1951. The research question
remains the same although expressed in a slightly different way: why are more people
abstaining from voting? But this time, the research is inductive: there is no hypothesis to be
tested. The researcher is not seeking one explanation but a number of reasons applicable
only to the persons concerned. In this case, the researcher will interview a sample of people
in depth to learn what participation and voting means to them. The sample will not seek to
be representative of the population. Whilst the researcher will hope to build explanations
there is no expectation that a law of electoral behaviour will emerge. The researcher will also
interview and observe meetings of groups of people to learn how and if group norms are
negotiated and peer pressure exerted. In this case, these interviews could be supported by
evidence of the extent of abstention and classifications drawn from official records and the
British Election Survey. This quantitative data will quantify the incidence and extent of the
behaviour rather than to give explanations.This research is unlikely to stop at the explanations
gleaned. It will consider the implications for the individual and groups and to suggest how
the processes of participation and articulation should be improved to benefit abstainers.
1. What do you understand by quantitative and qualitative methods?

2. Summarise the main claims and counter-claims of quantitative and qualitative
methods. Which do you regard as more appropriate to research in Politics? Why?
3. You are asked to carry out research to test Galbraith’s ‘culture of contentment’
thesis.8 Which method would you use? Why?
4. Mixed methods: can they be logically consistent? Are they ever justified? Choose
and defend two examples which support your position.
FURTHER READING
Bryman, A. (1988/96) Quantity and Quality in Social Research. London:

Routledge. This text provides the most authoritative review of qualitative and
quantitative methods, their claims and counter-claims and learned discussion
on mixed methods.
Harrison, L. (2001) Political Research: An Introduction. London: Routledge.
pp. 74–7. This book includes a very concise analysis of the research methods in
terms of whether they produce different results and answers to research

questions.
Reed, M. and Marsh, D. (2002) Chapter 11: Combining Quantitative and
Qualitative Method. In Marsh, D. and Stoker, G. (eds.) Theory and Methods in
Political Science. Basingstoke: Macmillan, pp. 231–48. This book follows
separate critical accounts of qualitative methods (pp. 197–215) and quantitative
methods (pp. 216–30) with an authoritative discussion of mixed methods
(‘a false dichotomy?’) including two case studies.
Seale, C. (1999) The Quality of Qualitative Research. London: Sage.
Notes
1 Brymam, A. (1998/96) Quantity and Quality in Social Research. London: Routledge.

2 Filstead, W.J. (1979) Qualitative Methods: a needed perspective in evaluation research. In Cooke,
T.D. and Reichardt, C.S. (eds.) Qualitative and Quantitative Methods in Evaluation Research.
Beverly Hills: Sage. p. 45.
3 Popular software includes Nud*st and ATLAS.ti.
6 Erving Goffman (1922–82) Canadian-US sociologist who specialised in studying face-to-face
interaction within the social interactionist tradition in which he identified the significance of ritual in
everyday activity
7 Crouch, C. (2004) Post-Democracy. Cambridge: Polity.
8 Galbraith, J.K. (1992) The Culture of Contentment. London: Penguin. Galbraith argued that US
voters were predominantly socially and economically advantaged electors who voted in favour of
groups which offered to reduce taxation and therefore welfare provision for others in need of
intervention.
Chapter 5
Collecting Data Sets: Case Studies,
Experimental, Comparative,
Longitudinal and Action Research
Methods
1. To consider the distinctive characteristics of case studies, experimental,

comparative, longitudinal and action research methods.
2. To identify their strengths and weaknesses.
3. To learn where each is best used.
Introduction
Case studies, experimental, comparative, longitudinal and comparative research are

all different methods that you can use to gather data. This data may either be
existing, documented information, or new information which is required to test
your (deductive) hypothesis, or develop (inductively) new theoretical explanations.
Simply described, a case study method is a ‘sample of one’ event, instance, state
or sub-unit at one point in time. Experimental method involves a ‘sample of two’:
a control sample and an experimental sample drawn (ideally randomly) from
the population under study. The experimental sample is made subject to a change
in some independent variable whilst conditions are held constant in the control
sample. For example, the independent variable may be new, additional forms of news
information, financial incentives or cost reductions by way of, say, postal voting.
Comparative method involves obtaining data from a population or sample of
equivalent states or sub-units, sectors or groups, at the same points in time. It is the
principal method used in comparative politics and international relations. Longitu-
dinal method seeks data from the observation of a cohort, a group of people, states
or organisations, etc. sharing one or more common characteristics, over an extensive
period of time. The common characteristics may include age, education, place or
specific condition. Action research method essentially involves the researcher

obtaining data through direct participation in a group of actors experiencing
particular conditions. The method therefore allows you to share and experience the
institutional and other barriers the group faces in the struggle towards transformation.
Choice of method
The choice of research method will generally be determined by five factors:
1. academic or other institutional requirements

2. the time and other resources available to the researcher
3. the sub-discipline of Politics
4. accessibility of the data sets
5. whether (mainly) quantitative or qualitative data is sought.
Generally speaking, university regulations for final-year, single-term research

projects by undergraduates and ‘taught’ graduates will be satisfied by a single
case study. This will provide sufficient opportunity for students to complete original
research which demonstrates their theoretical understanding and their ability to
obtain and analyse high-quality, in-depth data and offer profound insights. However,
doctoral research generally requires data to be sought from a number of sources
which combine both depth and extent. In this way, the set of multiple sources should
demonstrate whether the particular findings are unique or, hopefully, universal and
the impact of contextual variables.
A single-term or single-semester timescale will normally limit the method
employed to a single, in-depth case study, or to a comparative study of either a
limited number of variables from a large sample, or a larger number of data from a
smaller sample. At the other end of the scale, longitudinal research will require the
necessary time to track and observe members of the sample over the longest possible
periods.
Different data collection methods are better suited to particular sub-disciplines of
Politics. Comparative politics, international relations and development studies tend
to use comparative method. Gender studies are more likely to use action research.
Students of conflict studies are more likely to adopt an in-depth case study of a
particular geographical area.
Accessibility will also influence the choice of method. It means both access to
the group or set under study and to the relevant information about them. Groups
may be inaccessible because they are physically remote, concealed, nomadic, or lack
sufficient identity amongst themselves. Even when the groups are readily contactable
it may be difficult to access information about them. For example, information on
income, health, participation, exclusion and illegal activities is difficult to access.
Collecting Data Sets 53
Additionally, the specific phenomenon or indicator may not be readily observable

or measurable.
Pre-disposition to quantitative or qualitative method will also influence, if not
entirely determine, the choice of method for collecting data. The limited availability
of numeric information from a single case and the requirements for generalisability
(the wider applicability of its conclusions) will generally require a large sample of
comparable cases to be used to achieve the law of large numbers.1 Conversely, the
requirements for ‘deep and rich’ data in qualitative method are more likely to be
met by a single case study or, if resources permit, a small group of cases selected for
their apparent similarity or difference.
The single case study
The single case study is the essential ‘building block’ of empiric research. It must
provide the basic minimum of information to enable the research question to be
answered and the research hypothesis to be tested. It is a sample of one. Historical
research may define the case study and its scope. For example, the research question
‘why did the TUC halt the British general strike of 1926’, will concentrate on the
specific events leading to the general strike and its conduct. The difficulty here
will be to decide when the ‘events leading to’ actually began and the main actors
involved over this timescale. The law of diminishing returns may be evidenced when
the marginal cost of further interviews becomes greater than their benefit to the
research, and where opportunity costs are high. On the other hand, research into, for
example the causes of strikes in the UK car industry post 1997 may focus, in the
first instance, on a single company or a plant within that company.
Self-containment and typicality are important requirements for an effective case
study. Self-containment means that the variables being researched should be clearly
distinguishable and be relatively unaffected by complex external factors. The case
study should also be typical of the range of other potential case studies. So conscious
choice – rather than blind or random selection – is required. It is therefore important
that the initial preferred case study should itself be carefully ‘researched’ before the
choice is confirmed. This initial examination should show that the case is not affected
by other, exceptional factors. It should also show whether the necessary access to key
individuals and other sources of data is likely to be achievable. You should also find
out whether the group or location has been researched before: frequent prior research
may have ‘poisoned the well’ of information and left people uncertain of their ‘social
reality’. Insertion and immersion (i.e. prolonged exposure) in the proposed case study
early in the research process is therefore essential if disappointment, or even failure,
is to be avoided.
However good the fieldwork, the quality of data and its analysis, the great
weakness of a single case study remains that its generalisability – and therefore its
significance – remains unproven. The explanation may be unique to that case.

A solution may be to adopt a thick/thin approach or case study and a quarter. Here,
the findings of an in-depth case study are tested against another typical group or
situation. For example, suppose that a case study of ‘large-scale voluntary transfer’
of council houses in a northern, English, industrial town had concluded that the
high support by tenants there could be attributed to the involvement of a national
housing charity as a mediating body.2 In reaching this conclusion, the researchers
had established that other factors such as rents, repairs backlogs, key personalities and
support by tenants for the council, were much less significant. The findings of this
study could then be tested by a ‘thin’ study of another, similar case where a mediator
had been appointed to find out whether their involvement was regarded as critical
to the outcome. Alternatively, the findings could be tested against a similar ‘thin’
case where a mediator had not been appointed to see whether support was lower.
The use of this solution may be particularly effective for ‘taught masters’ students.
Experimental research
Experimental research is rarely used in politics research. Here, the necessary

observation of an experimental group and control group under laboratory-type con-
ditions – let alone the gold-standard of random double-blindness – is difficult to
achieve. To be successful, large-scale institutional support may be required.
One particular type of experimental research is the pilot project where a change
of policy is tested on a particular group or area before the initiative is ‘rolled-out’
(extended). A good example of pilot testing is the research carried out by the UK
Electoral Commission in 2002–4 into the effect of introducing all-postal voting
to replace previous forms of ballots. This initiative was introduced to counter the
decline of electoral turnout since 1951, especially in local government elections
where this had fallen to less than a third of electors. Low turnout was considered to
undermine the legitimacy of councillors in government/local government relations.
The traditional form of voting required electors to vote in person at the specified,
local polling station on a single date unless the elector could justify the necessity for
a proxy vote (on grounds of disability) or postal vote (on the grounds of absence).
All-postal voting meant that the ballot paper was delivered direct to the elector well
in advance of the election and could be returned by post before the polling day or
delivered to the polling station.
Following a selection process, 32 local authorities were chosen to test the
practicality of all-postal voting and its effect on turnout. Each was the subject
of a statutory report by the Commission. The reports show that, overall, the
level of turnout in the pilot areas increased to 50%. For example, turnout in the
City of Newcastle upon Tyne increased from 32% (in 2003) to 49.9%.3,4 Similar
increases were reported in the other pilot areas. On this evidence, the (unstated)
hypothesis – that the change from traditional to all-postal balloting would increase
turnout – appeared confirmed. The outcome supported the rational choice
interpretation of voting behaviour (in which reducing barriers and costs of voting
would increase participation). However, other factors may have contributed to the
increase in turnout. In particular, the local authorities undertook substantial publicity
campaigns in the local press, radio, television, billboards and additional mail-outs to
electors. So significant Hawthorne and Pygmalion effects cannot be discounted. I also
suspect that pilot tests generally tend to realise higher improvements beyond those
likely to be attained when the initiatives are extended nationally. This pilot effect
may be attributed to the original selection of the pilots for their greater suitability
and the additional resources made available including the greater involvement of
leading officials. Subsequently, complaints were made to the police that applications
for postal voting in marginal wards had risen by 500%. The police investigations
led to trials and convictions in a small number of cases. Subsequently, in August
2004, the Commission recommended that the system of all-postal voting should be
discontinued because, in the absence of the reforms it had recommended for voter
registration, postal voting was vulnerable to abuse.5 Yet its earlier statutory reports
had stated that ‘it was unaware of any allegations of fraud and malpractice …’.
Comparative research
Comparative research, or comparative method, is the approach most widely adopted

by comparative politics, international relations, public policy and developmental
politics. Indeed, in North America, comparative politics is regarded as essentially the
study of ‘foreign states’. The underlying principle of comparative research is that, by
comparing two or more cases, researchers can identify causal variables which could
not be deduced from a single case. It is essentially positivistic and relies on the system
of analysis afforded by J S Mill’s methods of difference and similarity (Mill, 1843).6 But
whilst comparative research is positivistic, it need not necessarily be quantitative.
A good example of early comparative research is Engel’s The Condition of the
Working Class in England (1845).7 After reviewing the contemporary development
of the main industrial sectors and creation of the proletariat, he examined in turn
the ‘great towns’. He described conditions in the ‘working men’s housing areas’
in, for example, London, Dublin, Edinburgh, Glasgow, Birmingham, Salford and
Manchester, Nottingham, Leeds and the smaller industrial towns of Yorkshire and
Lancashire. He concluded that:
To sum up the facts thus cited. The great towns are chiefly inhabited by working people,
since in the best case there is one bourgeois for two workers, often for three …; these
workers have no property of their own, and live wholly for wages, which usually go from
hand to mouth … every working man is therefore constantly exposed to loss of work
and food … The dwellings of the workers are everywhere badly planned, badly built, and
kept in the worse condition, badly ventilated, damp, and unwholesome … The average
is nearer the worst case than the best.
(Engels, 1993/1845: 85–6)
His comparison of the ‘great towns’ enabled him to conclude that the conditions
were universal within British industrial towns and cities. The causes were therefore
not local, specific to different sectors of industry or otherwise contextual. He
concluded that the cause of these common conditions were state-wide: competition
between workers and Irish immigration. He then investigated these causes in turn,
extending his research to the ‘mining proletariat’ and the ‘agricultural proletariat’
and concluded that ‘The war of the poor against the rich … will become direct and
universal [i.e. not confined to Britain]’ (Engels, 1993: 302).
Engels’ research demonstrates many of the purposes (reasons) for undertaking
comparative politics research:
1. to provide rich, contextual descriptions which identify clearly the observed similarities
and differences between cases and places
2. to identify and develop systems of classification and typologies which generate data
sets appropriate to the research question
3. to distinguish independent variables (causes) from other variables
4. to test hypotheses and, thus, the validity of explanatory theories
5. to develop predictive capacity where, for example, there is evidence of stages of
development.
The selection of cases for comparative research depends on the number of variables
to be examined, the availability and access to related, comparable data, data format
(numeric or non-numeric) and the means of analysis to be adopted. Elementary
algebra has taught you that the value of two variables cannot be found from a single
equation: two, simultaneous equations are required. At least one case is required
per variable. Similarly, algebra and logic has taught you that you cannot compare
apples and pears: the data to be compared must be measures or descriptions of the
same variable. They must also be equivalent to the variables being examined. The
cases to be selected must offer the same data sets over the same periods. But data
need not be numeric or quantifiable. It can equally well be binary (e.g. yes/no,
democracy/non-democracy, etc.) or nominal (e.g. the name of the state).
The method of comparison to be adopted will determine the cases to be selected,
and the size of the sample relative to the size of the population. The three most
widely-used methods of comparison are Mills’ method of agreement, method of difference
and method of concomitant variations.
Where the method of agreement is to be used, different cases which have

the same outcome (result, effect) are selected. The potential independent variables
(causes, contributory factors) are sought and listed. Where one (or more) variable is
common to each case, then the researcher will conclude that this is the independent
variable. This can be demonstrated below:
Case Variables Outcome
1 a b c y
2 a c d y
3 a b d y
In this group of cases, only a is common to all cases where the outcome is y. So
a appears to be the cause/independent variable.
The method of difference is the converse of the method of agreement. It is
generally regarded as more practicable (easier). Here, similar cases are chosen for
their different outcomes, e.g. single-party or coalition governments. The potential
variables are sought and listed. Where the absence or presence of one variable is
associated with specific outcomes, then that variable may be concluded to be causal:

1 a b c d y
2 a b c d y
3 a b c d y
4 p b c d z
In this example, variables b, c and d can be seen not to effect the outcome whereas
variable a appears to be the cause of outcome y and variable p appears to cause
outcome z.
The method of concomitant variations is used where the magnitude of the
potential cause appears associated with the magnitude of the outcome. For example:

1 a b c d y
2 2a b c d 2y
3 3a b c d 3y
In this group of cases, the magnitude of a appears to influence the magnitude of y.

In all these examples, the findings are qualified by saying ‘appears to be’. Firm
conclusions can only be drawn when other explanations – e.g. spurious correlations
or chance – have been eliminated.
The individual cases must also be comparable. So samples of states may be drawn
from the EU (or EEA if comparison between EU and other European states is
sought), OECD, G8, states sharing a geographical area (e.g. South America), or states
having a similar history (e.g. former British colonies). Or alternatively, a comparison
can be made between offices (e.g. foreign ministers), events (e.g. devaluations, civil
wars), policies (e.g. privatisation), organisations (e.g. NGOs) and ideologies.
The size of sample will also have a bearing on the degree of certainty to be attached
to the findings. Ideally, the entire population of, say, OECD states should be chosen.
Population-wide data is now more readily available especially via the internet. Where
a sample of the population is used, then there may be a suspicion that cases which
do not support the hypothesis have been deliberately excluded. For example, the
comparative research carried out by Przeworski et al (to answer the research question
‘what makes democracies endure?’) used a large sample of 135 states over the
period between 1950 and 1990 (Przeworski et al, 1996).8 But a comparatively small
number of UN member states were omitted. They included the oil-producing Gulf
states, where, with western support, feudalism continued to prevail. The necessary,
compelling defence against criticisms of selection bias was not made.
The potential samples can be categorised as small-n or large-n, i.e. a small or
large sample of the population, N. As previously noted, the size of sample must be
no less than the number of potential independent variables to be investigated. A
small-n sample is more appropriate for qualitative research where a more intensive
level of investigation is proposed, generating rich, thick descriptions, and where the
focus of analysis is differences or similarities of outcomes. The small sample enables
Mills’ method of similarity or difference to be used. These methods are termed
most similar systems design (MSSD) or most different systems design
(MDSD). Comparative research using small samples is called case-orientated
analysis. Generally speaking, the optimum size for small-n samples is 15.
In contrast, a large-n sample is used to enable a limited number of potential
independent variables to be identified by statistical methods (including regression
analysis and discriminant analysis) from a very large number of cases. It is termed
variable-orientated analysis.
The decision whether to adopt small or large samples for comparative politics
research will, as ever, depend upon time and other resources and access to reliable
data. Each has its strengths and weaknesses. Large samples enable computer-driven
statistical techniques to be adopted and levels of confidence to be determined.
A higher level of abstraction is permitted. Unusual cases are more likely to be readily
identified for further examination. However, the level of abstraction may be too
high and superficial and the available data may not properly represent the variables
under consideration. Data collection standards may also vary between cases. A more
profound understanding of advanced statistical analysis may be required. At the end of
the day, the acceptance of the research findings will depend on the reader’s readiness
to accept statistical methods. By comparison, small samples offer thick description, a
lower level of abstraction, intensity and analysis of differences and similarities which
the reader is more likely to follow and accept. But the suspicion of selection bias
may be difficult to refute. One solution may be to combine both methods in which,
say, a large sample is first used to identify apparently key variables, which are then
Table 5.1 Selected average annual rates of growth of GNP per capita,
1965–90
Democratic State Change GNP Non-democratic Regime GNP Change

per Head (%) (%)
Jamaica −1.3 Libya −3.0
Venezuela −1.0 Zaire −2.2
Senegal −0.6 Zambia −1.9
Trinidad 0.0 Nigeria 0.1
Costa Rica 1.4 Algeria 2.4
India 1.9 Brazil 3.3
Sri Lanka 2.9 Thailand 4.4
Mauritius 3.2 Indonesia 4.5
Malaysia 4.0 China 5.8
Singapore 6.5 Taiwan 7.0
Botswana 8.4 South Korea 7.1
Source: Council for Economic planning and Development (1992); World Bank (1992) cited by
Leftwich, 2000: 132i9
tested against a smaller sample from which the influence of the variables may be
more properly understood.
One topic of intensive comparative politics research is the relationship between
democracy and development. This is illustrated in Table 5.1.9
The democracy-development link is the holy grail of western, comparative
politics and development research. The findings thus far vary. For example, Moore’s
comparative qualitative study (1966) of the paths to modernisation of Britain, USA,
France, India, Japan and China concluded that, where the new bourgeoisie and
progressive landowners combined, then democracy was more likely (Britain and
USA).10 But where the traditional elites were unable to meet the peasants’ demands,
then communism would prevail (China). Finally, where traditional elites were
able to frustrate democratic institutions, then fascism was possible (Japan). In the
post-war period, the prevailing view, especially among colonial powers, was that
democracy would hinder development. Democracy (decolonisation) would best
be the child of development only when higher levels of literacy, civil society and
class formations had been achieved. Indeed as Table 5.1 shows, states achieving
higher levels of economic growth 1965–90 included China, Taiwan and South
Korea. A critical, comprehensive review of the published comparative research is
provided by Landman (2003).11 He distinguishes between large-n and small-n samples
and qualitative and quantitative inquiry. He shows that the large-n studies carried
out by such luminaries as Lipset, Dahl and Przeworski support both ‘weak’ and
‘strong’ conclusions where, respectively, democracy is associated with development
and development causes democracy. Wealthy democracies are more likely to endure.
Small-n, quantitative research either supports the ‘weak version’ (i.e. association) or
no relationship. Small-n, qualitative research generally concludes that democracy is

the product of discrete historical events. Single-country studies conclude that: ‘case-
specific factors, particularly political culture, condition the relationship’ (Landman,
2003: 91, Table 4.5). So, the variance among the conclusions reached would
suggest that the type of comparative research undertaken can influence the findings.
However, Landman rejects this view. He argues that the variations can be attributed
to the absence of a contemporaneous trajectory of development, selection bias and
selection. The different comparative methods should be regarded as complementary:
inferences from large samples should be investigated by an analysis of a smaller
sample of states. Deviant cases should then be examined separately by single-case
studies (Landman, 2003: 90–2). But Landman’s prescription is a counsel of perfection
(unattainably high standards) which is unlikely to be achieved by a single, doctoral
researcher working alone.
Longitudinal research
Longitudinal research – also called panel survey research or tracking

research – enables researchers to monitor changes in behaviour or opinion over
time. Statistical time series analysis can be used on quantitative longitudinal data,
e.g. unemployment data, to distinguish between long-term trends, seasonal variations
and irregular events (see Chapter 16). The magnitude of irregular events enables the
impact of policy change (or other intervention) and the time-lag involved (between
announcement, introduction and effect) to be identified, e.g. the impact of New
Deal-type job creation initiatives on unemployment.
Longitudinal research essentially involves obtaining data over time. It may take
the form of personal life histories or narratives when an individual is asked to
describe life events from earliest memory. Whilst this can be a rich source of data,
access can be difficult and necessary confidence and empathy (rapport) difficult
to develop. It can be very time-consuming especially when recordings have to be
transcribed. The data may suffer from faulty memory, selective memory, constructed
memory and the tendency for subjects to select unrepresentative, exceptional or
dramatic events to maintain the researcher’s interest. They may also exaggerate their
own role. Constructed memory occurs when you describe events which you
cannot remember but of which you have been frequently reminded by parents, e.g.
your first steps. Alternatively, personal life histories can be augmented or rely on
personal diaries, letters or other records.
Longitudinal research may be retrospective or prospective. Retrospective research
looks back over previous periods of time whereas prospective research studies
changes from the present-day onwards. The ideal model of longitudinal research
involves the cross-sectional interviewing or surveying of a cohort or panel of
subjects. A cohort is a group of subjects sharing some common characteristic, e.g.
date of birth, or membership of a particular school class. In this type of survey, each
of the subjects is interviewed separately for every year or period of years. Those
who leave (by death or migration) are not replaced. You will find it very difficult
to complete successfully in urban areas where the residents frequently move homes
and jobs, or change family names on marriage or divorce.
The most widely-used longitudinal research data is collected by panel surveys.
A panel is a set of subjects chosen to be representative of the population under
study. For example, the government-sponsored British Household Panel Survey
(BHPS) – begun in 1991 – involves an annual survey of all the adults within 5,000
randomly selected British households.12 However, the term panel can be misleading:
unlike common usage, the members of the panel do not meet together. Each
is interviewed separately using a highly-structured questionnaire. When children
in the selected households reach 16 they are added to the panel to replace those
who are lost through death, illness or disappearance (which may include either
withdrawing consent or moving contact address without telling the researchers).
BHPS is carried out by the Institute for Social and Economic Research incorporating
the ESRC Research Centre on Micro-Social Change at Essex University. Each
cross-sectional survey is termed a wave. The BHP survey for 2006 was called
‘Wave 14’. Ideally, a panel survey should seek to ask the panel members the same
core questions at every wave to enable comparison to be made of changes of
attitude or behaviour over time. However, new questions can be added. The panel
may also be increased to ensure better representation of minorities. For example,
in Wave 9, the membership was increased in Wales to enable attitudes towards
the Welsh language and national identity to be monitored. The data gathered
by BHPS is available for inspection and use at the ESRC Data Archive at Essex
University.13
Another authoritative source of longitudinal research data is the British Election
Survey (BES) funded by ESRC and the Electoral Commission, and carried out at
each British general election. A new panel is recruited by random sampling at every
general election. Members of the panel are interviewed before and after the election.
BES 2005 interviewed 4,700 British electors. It gave special attention to the issue of
(falling) turnout and (increasing) abstention.14 Panel survey data is also collected – on
a near-monthly basis - by the EU commission and published in the Eurobarometer.
New panels are recruited by stratified, simple, random sampling (see Chapter 7).
Eurobarometer 65.2, published in Spring 2006, included data from a British panel of
1,312 electors.15 A particular concern for the EU is tracking the movement of pro-
and anti-EU sentiment among nationalities in the 25 member states. Elsewhere, the
Panel Study of Income Dynamics (PSID) provides a comparable example of US
longitudinal research.
Longitudinal Politics research can be time-consuming and costly. Sponsorship
by some government research agency is therefore likely to prove essential in order
to provide a panel large enough to produce good, representative data. The great
advantage is that this data is available for re-analysis and use by the Politics researcher.
Why collect new information when good data is already available?
Action research
Action research is derived from critical social theory whose central aim is
the emancipation of society (rather than mere description or explanation). The
researcher focuses on the myths and contradictions, e.g. ‘women’s work’. A common
maxim is: ‘reports don’t change policies … it is people who will change them’.
Action researchers therefore tend to seek out underprivileged groups and work
with and among them to enable them to recognise and overcome the institutional
barriers which prevent them from self-actualising (achieving self-fulfilment). The
groups may be Travellers, migrant families, illegal immigrants, gays and lesbians,
former prisoners, and other, pariah groups. The researcher is a passionate, committed
activist. Researching the group means sharing their lives, experiences and indignities.
It can be a high-risk undertaking for you and for your co-participants. The risk is
particularly great for the participants when they challenge the institutions which
have served dominant groups. The danger is that you may lead the group into
harm’s way or raise expectations which cannot be achieved. The involvement of an
experienced supervisor is therefore essential if some form of emancipation, rather
than repression, is really to be achieved.
The task facing the action researcher is difficult. You will face problems obtaining
access to the subject group and establishing mutual trust. Funding is problematic.
The UK Community Development Programmes lost funding when they criticised
the government’s funding ministry and its local allies. Your task is to enable the
subject group to ‘think (and discuss) the unthinkable’, to identify potential actions
and to implement change.The arena of action research is the group meeting. Your
role here is to facilitate (enable) group identity to form and a new consciousness
of their social reality to emerge. In particular, you should gently prod the group
to re-examine institutions which have become regarded as ‘natural’ or ‘for the
best’. Examples may include the disproportionate burden of childcare borne by
women, restricted opportunities to education, or the classification and exclusion of
groups from mainstream programmes. Another role for you will be to carry out
those tasks which the group is unable, as yet, to accomplish. They may include
researching similar groups and activities, arranging meetings, taking notes, printing
and distributing leaflets, lobbying elites and representing the group at meetings with
dominant others. You should also train the group to carry on these activities for
themselves as part of the essential capacity-building exercise. This should form
part of your exit strategy which should also include a celebration of the group’s
activities. This event, like other social ritual, marks and announces to all the transfer
of responsibility to the group.
Because action research focuses on outcomes rather than process, there are
relatively few research reports. Publication is an anathema to many social action
researchers who would regard their own preferment as further exploitation of those
they seek to assist. However, writing-up and publishing action research projects is
justifiable for the insights they provide into the why’s and how’s of domination and
the success (or otherwise) of action to dismantle barriers. The increasing emphasis
in western polities on participation has led to a widening of research interest in
and funding for quasi-action research. A recent publication, Effective participation in
anti-poverty and regeneration work and research (Beresford and Hoban, 2005), assesses
the lessons drawn from UK initiatives to involve people with direct experience of
poverty in anti-poverty programmes.16 The study concluded that:
‘Powerlessness is central to people’s experiences of poverty and disadvantage …

capacity building to develop people’s confidence, self-esteem and understanding
supports their empowerment and participation. It is not the same as skill development …’
But there remains a strong suspicion that increasing western government interest in
public involvement seeks to smooth the process of policy change by co-opting and
thus neutralising potential losers.
1. You are researching the causes of local government corruption in England. Your
case study has identified new causes which are not included in the current
literature. What steps should you take to establish whether these causes were
unique to your case study or more widespread?
2. Can experimental research be wholly discounted in Politics? How might the impact
of various electoral innovations on turnout be evaluated?
3. You wish to carry out comparative research on the relationship between democratic
reform, economic policy change and changes in GNP per capita in the former
Warsaw Pact states in Eastern Europe after 1991. What sample and method of
analysis would you use? Why?
4. Gypsies in your locality are widely criticised for their exploitation of welfare services.
You regard the basis for this criticism and related hostility as a myth. A national
charity offers funding for social action research to improve the conditions of
Gypsies and other Travellers. Draft a research proposal setting out how you
propose to access this group, what roles you expect to pay, indicators of success
and your exit strategy.
FURTHER READING
Comparative research
Hopkin, J. (2002) Chapter 12: Comparative Methods. In Marsh, D. and Stoker, J.
(eds.) Theory and Methods in Political Science. Basingstoke: Palgrave
Macmillan. pp. 249–67. This chapter provides a good introduction to the
development of comparative politics, the comparative advantages of small-n
and large-n samples and the competing claims of quantitative and qualitative
comparative research.
Landman, T. (2003) Issues and Methods in Comparative Politics: An Introduction.
London: Routledge. This 292-page textbook is devoted to comparative research
in Politics. It begins by identifying and examining the reasons for using
comparisons, theory and method, units of analysis, and levels of analysis. This
is followed by comparing comparisons where the comparative research on a
range of significant topics is compared and evaluated. The topics are: economic
development and democracy, violent political dissent and social revolution,
non-violent political dissent and social movements, transitions from democracy,
institutional design and democratic performance, and human rights. It ends by
identifying the new challenges facing the sub-discipline.
Leftwich, A. (2000) States of Development: On the Primacy of Politics in
Development. Cambridge: Polity. pp. 127–51. This excellent, rich chapter uses
comparative data and published comparative research to abstract and examine
five main conditions for democratic consolidation: legitimacy; adherence to
rules-of-the-game; policy restraint by winning parties; poverty as an obstacle;
and ethnic, cultural or religious cleavages as constraints. He concludes
controversially by arguing that: ‘democratic politics is seldom the politics of
radical economic change’ (p. 150).
Mayer, L.C. (1989) Redefining Comparative Politics: Promise versus Performance.
London: Sage. pp. 1–46. The author identifies the epistemology of social
science and its implications for comparative research. The role of comparative
research is to suggest explanation of differences and similarities at state level.
He argues that the unique role of comparative research lies in its capacity to
‘identify two distinct objects for analysis: the attributes of individuals and the
attributes of whole systems’ (p. 42).
Pennings, P., Keman, H. and Kleinnijenhuis, J. (2006) Doing Research in Political
pp. 1–51. Pennings et al is a challenging text. However, it uniquely provides
hard, technical explanation of the concepts behind comparative methods within
the discipline of Politics.
Przeworski, A., Alvarez, M., Cheibub, J.A. and Limongi, F. (1996) What makes
democracies endure? In Journal of Democracy, 7(1): 35–55. This is a very good
example of the application of comparative methods to international data sets.
However, as previously noted, the exclusion of Gulf-states from the large-n
sample compromises the generalisability of the conclusions.
Longitudinal research
Gilbert, N. (2003) Researching Social Life. London: Sage. pp. 275–9. This short
extract provides an authoritative explanation of panel, longitudinal and cohort
studies and an introduction to the main data sets available from US, UK and EU
sources.
Pennings, P., Keman, H. and Kleinnijhuis, J. (2006) Doing Research in Political
pp. 166–79. This extract provides a higher-level introduction to the mathematics
of time series analysis and ways of diagnosing and reducing the potential
problems of autocorrelation and heteroscedasticity.
Robson, C. (1993) Real World Research. Oxford: Blackwell. pp. 50–1. This very
brief extract outlines the design of effective longitudinal research.
Action research
Harvey, L. (1990) Critical Social Research. London: Unwin Hyman. pp. 19–32.
Lee, R.L. (1995) Dangerous Fieldwork. London: Sage. pp. 1–13. In this
introduction, Lee scopes the types of danger that researchers face when
working with ‘outlaw bikers’, youth gangs, people infected with HIV, and
informants in inherently dangerous occupations. He argues that proper risk
analysis can effectively minimise the dangers facing researchers and prevent
them from being inhibited – or even paralysed – by their fears in the research
setting.
Stringer, E. (1996) Action Research: A Handbook for Practitioners. London: Sage.
pp. 142–60. Stringer has become a foremost authority on participatory or
community-based action research especially in the discipline of education. In
this extract, he sets out strategies for effective action research to practitioners.
Notes
1 The ‘Law of Large Numbers’ states that the probability that a sample shares the characteristics of
the population increases as the size of sample (i.e. number of cases) increases.
2 The Thatcher government introduced ‘large-scale voluntary transfer of housing stock’ in 1989 as
an extension of its ‘right-to-buy’ policy of assisting tenants of council houses (municipal social
housing) to buy their own homes (at a discount). This policy was retained by the New Labour
government.
3 www.electoralcommission.org.uk/about-us/statutoryreports.cfm 25/07/06
4 Elections staff at the city council reported that, at the all-seats all-postal voting local council and
EU elections in May 2004, turnout was 46.8%. The results ended the Labour party’s hegemony
over the city council when the Liberal-Democrats took control.
5 ‘Delivery democracy? Postal voting. Executive summary.
www.electoralcommission.org.uk/files/dms/Deliveringdemocracy 4473-10934 E S W.pdf
6 Mill, J.S. (1843) A System of Logic. London: Longman.
7 Engels, F. (1845/1993) The Condition of the Working Class in England. Oxford: Oxford University
Press.
8 Przeworski, A., Alvarez, M., Cheibub, J.A. and Limongi, F. (1996) What makes democracies
endure? In Journal of Democracy, 7 (1) pp. 35–55. The research concluded that democratic
government is more likely to endure where per capita income is high, income inequalities
moderate or falling, economies maintained or increased levels of growth, and there is no recent
history of democratic government being overthrown. Furthermore, parliamentary democratic
governments are more likely to endure than presidential systems (where deadlock between
president and parliament can lead to paralysis and a coup d’etat by the defence forces).
9 Leftwich, A. (2000) States of Development; On the Primacy of Politics in Development.
Cambridge: Polity.
10 Moore, B. (1966) Social Origins of Dictatorship and Democracy. Boston: Beacon Press.
11 Landman, T. (2003) Issues and Methods in Comparative Politics: An Introduction. London:
Routledge.
12 For more information on the British Household Panel Survey, see www.esds.ac.uk
13 www.data-archive.ac.uk
14 The survey questions and data can be inspected at www.essex.ac.uk
15 Eurobarometer data can be accessed at http://ec.eurpoa.eu
16 Beresford, P. and Hoban, M. (2005) Effective participation in anti-poverty and regeneration work
and research. York: Joseph Rowntree Foundation. The full report can be accessed at
www.jrf.org.uk/knowledge/findings/social policy/0395.asp
Part III
Collecting Information
Chapter 6
Critically Evaluating Published
Research
1. To learn how to critically evaluate published research using research com-

ponent analysis.
2. To apply Rose’s ABCDE model (1982) for deciphering research.1
Introduction
As a child and student, you are likely to have developed and improved your own
writing skills – consciously or otherwise – by reading books and other publications
recommended to you. Similarly, reading others’ research reports will help you
to develop your own research report-writing skills. However, going further, by
consciously, critically evaluating others’ research reports, you can draw valuable
lessons into effective research design, data collection, analysis and interpretation,
and how best to communicate your findings. Critical capacity is an essential part of
academic scholarship (which owes its origin to its original function of training priests
to identify heresies). In this way, you can therefore appreciate better the successes or
otherwise of other researchers before you begin your own research. In any event,
the literature review of texts and research reports is an essential first step in research
(which is considered later in Chapter 8).
Some overseas students are inhibited from criticising published works – especially
by great authors – because they have been taught to respect authority and the
expert opinion contained within university libraries. However, in the UK and
other western universities, you must offer justified criticism of the literature relevant
to your topic. But, as an overseas student, you may also be concerned that your
criticisms of your own government’s policy or accepted textbooks may lead to you
or your family being punished. In this special case, you can ask for those criticisms
to be excluded from public copies of your dissertation.
This chapter introduces two ways of evaluating published research: research
component analysis (RCA); and, Rose’s ABCDE validity analysis.
Research component analysis
Research component analysis dissects the published research into 15 or more

components which are assessed separately. The absence or weakness of any one is
likely to be significant but not necessarily fatal. The analytical components are:
1. when
2. where
3. who
4. abstract
5. research question and rationale
6. theory and literature review
7. hypothesis
8. methodology
9. data collection: what, when, how
10. data presentation
11. data analysis, interpretation and discussion
12. conclusions
13. implications
14. style
15. referencing and bibliography.
When concerns the data of publication. The assumption is that recently-published

research is more likely to incorporate the latest data, thinking and state of the art
than earlier research, say, published ten years ago. Research with a publication date
pre-2000 is now ‘last century’. Ceteris paribus (all other things being equal), most
recent research can be considered more valuable than older publications.
But where can be more significant than when. Research published in a refereed,
academic journal is likely to be of greater quality and authority than research, say,
published as a departmental working paper. The process of refereeing means that,
before the research is published, it will have been sent to two or more academic
referees for their assessment. They will have been chosen for their reputation as
commentators or researchers in the field of research. Their anonymous evaluations
will have been forwarded by the editor to the researcher who will be told whether
the research is accepted for publication, refused or whether changes are sought in
the light of the referees’ observations. Refereed UK journals of political science
include: British Journal of Politics & International Relations, British Journal of Political
Science, Comparative Politics, European Journal of Political Research, Journal of Public Policy,
Party Politics, Political Quarterly, Political Research Quarterly, Political Science Quarterly,
Political Studies, Political Studies Review, Governance, Politics, Politics and Policy, Public
Administration, Journal of Theoretical Politics, Public Administration and Development. PhD
Critically Evaluating Published Research 71
dissertations will have been subject to a similar process of examination by leading

academic specialists and modification; although unpublished, they will be a rich,
authoritative source. You can inspect these at the university libraries where they are
held, request copies or microfilm, or, in some cases, read on-line.
Who refers to the provenance and attribution of the research, i.e. claimed authors.
Many journal articles are attributed to more than one author. They can be a
leading academic researcher and a research associate (normally named last). In those
circumstances, then it is likely that the leader won the funding whilst the associate
undertook and wrote-up the research. The source of funding may also be significant.
Funding by government research agencies – ESRC, AHRB, MRC2 – signifies
that the research topic is considered of national significance. Similarly, EU or UN
funding is indicative of higher status research. Funding by these bodies means that
the researchers are likely to have been given access to decision-makers and other
privileged sources of information that may not have been available to others. The
academic department may also be significant. Research undertaken within a UK
department given a 5* rating in the most recent RAE is likely to be of a higher
standard than that of another with a lower rating.3
The abstract provides a synopsis of the research undertaken and its conclusions.
It should ideally attract your attention and stimulate your interest. Are the topic and
conclusions relevant to your proposed research?
Research question and rationale. The centrality of the research question
distinguishes and elevates research from essays or other scholastic activities. The
research question should be clearly identifiable and unambiguously stated. The
rationale – the reason(s) for undertaking the study should also be given. They should
show that the research question relates to a topic of acknowledged importance which
should ideally be topical and a significant problem, i.e. a barrier to policy objectives.
The rationale should also set out the context.
The theory and literature review should demonstrate that the researcher has
taken the necessary steps to cite and critically evaluate the relevant explanatory
theories and the main literature. The review should pick out contradictions, overlaps
and gaps. The review should justify the selection of any theory which is to provide
the conceptual lens of inquiry and identify any uncertainties which the research
should address.
The hypothesis – the researcher’s initial answer to the research question –
must be given where a deductive methodology is to be adopted. It should cite
the causal, independent variables inferred from the theoretical and literature review
and their relationship with the outcome. The hypothesis may be single or double-
tailed. A single-tailed hypothesis will specify the directional relationship of the
independent variables x, and the outcome y, e.g. as x increases, y will also increase.
A ‘double-tailed’ hypothesis will merely assert that, as x increases, y will change.
The methodology should state and show consistency between the underlying
philosophy (positivistic or otherwise), inductive or deductive approach, quantitative
or qualitative method, and method of data collection (case study, comparative

study, etc.).
Data collection should identify what data was collected, why, by whom and
when (date and duration). The selection of data is critical. Where the variables
and concepts are not directly measurable, then the selection of the indicators
used must be justified. A number of indicators may be adopted to represent the
concept under scrutiny. For example, overseas aid may be expressed either as the
donor’s budget allocation, locally purchased goods and services, technical assistance,
etc. This may show that, whilst state A allocated most funds to overseas aid, state
B invested a higher volume of aid locally. Indicators will also have been chosen
to reflect local changes, e.g. changes in infant mortality. No single measure is
likely to be sufficient. The description of the data collection process should also
describe and justify the choice of samples made, questionnaires designed and elites
interviewed.
Data presentation concerns how key data is presented in the research report
and how it is illustrated and supported. Sufficient data must be given to enable the
reader to share and have confidence in the researcher’s analysis and conclusions.
Graphs and diagrams are particularly helpful. ‘Quotable quotes’ from interviewees
are especially useful.
Data analysis, discussion and interpretation should be considered separate
from its presentation. The method of analysis should be described and its selection
justified. Where appropriate, the confidence limits should be made clear. The discussion
should review the potential relations and explanations between variables leading to
their interpretation by the researcher.
The conclusions should summarise the research and state whether the initial
hypothesis was confirmed, qualified or infirmed. Other important findings should
be given. Research that entirely confirms the starting hypothesis must be treated
cautiously.
The implications should say ‘what the research means’, the lessons learned and
advances claimed, overall significance and the scope for follow-on research projects.
Style refers to the general ‘readability’ of the research. Does the report do justice
to the research undertaken? In plain English, is it boring? How does the author
maintain (or lose) pace and interest? Does the researcher exaggerate the significance
of the research or their role?
Referencing and bibliography should provide proper acknowledgement of
others’ earlier works and a ‘paper trail’ of texts which others can follow. The type
of referencing system adopted – Harvard or footnotes etc. – is less important than its
consistency of application.
Finally, re-read the abstract. Does it fairly reflect the contents and importance of
the research report?
Research component analysis can be criticised for its ‘checklist’ approach and
its greater concern for the individual components than their linkages and overall
coherence. By comparison, Rose’s ABCDE deciphering method offers a more holistic

approach.
Rose’s deciphering model (1982)
A Theory: an explanatory statement about social phenomena. ←
↓ ↑
B Theoretical Proposition: specific propositions to be ←

investigated in the study. ↑
↓ ↑
C Operationalisation: decisions made on how to carry out ←
empirical work; techniques of data collection; ↑
↓ sampling; concepts and indicators; variables; units. ↑
D Fieldwork: collecting data; practical problems of ↑
implementing Stage C decisions. ↑
↓ ↑
E Results: data analysis leads to findings; interpretation leads back ↑
to C, B, and A
Figure 6.1 Rose’s ‘model for the research process’

Source: (Rose, 1982: 14)4
Despite its longevity (1982), Rose’s deciphering model has not been bettered.
It is premised on a Weber-type ideal model of empiric research. This model has five
stages and elements named – somewhat prosaically in an era before memorable
acronyms became widely adopted – A, B, C, D, E. The model for theory-testing
research is:
A is the theory which is being tested. The propositions in B are the hypotheses
in the form of testable statements about the causal relationships between the
independent variables and the outcome. Rose classifies stage C as the ‘pivotal stage in
the research process’. Here, the theoretical concepts are transformed into measurable
indicators (concept indicators) on which data is collected in Stage D and analysed in

Stage E
The concept-indicator link is highly problematic. For example, Downs applies
rational choice theory in his Economic Theory of Democracy (1957) to develop three
hypotheses.5 His second hypothesis is that:
every citizen rationally attempts to maximise his utility income, including that portion of
it derived from government activity.
(Downs, 1957: 297).
From this hypothesis, he derives 15 ‘testable propositions’. They include the

relatively straightforward proposition (19) that:
the percentage of low-income citizens who abstain in elections is higher than the
percentage of high-income citizens who abstain (ceteris paribus).
(1957: 299).
This proposition can be tested at micro- and meta-levels. A large sample of voters
could be randomly selected and asked details of their income and voting. But,
income data can be a very sensitive area. The prospect of a high level of response
is poor. A more acceptable means of asking people’s income might be to adopt
thresholds of, say, less than £12,000 for low income and more than £35,000 for
high incomes. In those cases, the thresholds have become indicators of income. The
meta-level of analysis would compare the (readily available) data on ward turnout
with socio-economic data from the census. In this case, turnout (or its absence)
could be used as an indicator of abstention whilst census data on house tenure, social
class, education and occupation, etc. would be used as indicators of income. Using
multiple regression analysis, the relationship between turnout and socio-economic
variables, and the power of the equation can be calculated.
Rose uses his model of the research process to provide a three-stage framework
for critically evaluating research:
Central to this method of evaluation is the concept of validity. Derived from
the Latin, validus meaning ‘strong’, validity in the context of research means:
the extent to which a measure, indicator or method of data collection possesses the
quality of being sound or true as far as can be judged. … in the social sciences generally,
the relationship between indicators and measures and the underlying concepts they are
taken to measure is often contested.
(Jary & Jary, 1995: 714)6
Relationship to other theory and research
External validity
A Theory: ←
↓ ↑ Internal theoretical validity
B Theoretical Proposition ←
↓ ↑
C Operationalisation ←
↓ ↑ Internal empirical validity
D Fieldwork ↑
↓ ↑
E Results ↑
Figure 6.2 Research Evaluation

Source: Rose, 1982: 32, Figure 2.2
Rose suggests that the first stage involves evaluating the internal empirical
validity by assessing the extent to which the conclusions are fairly drawn from
the data collected. The second stage involves assessing the internal theoretical
validity. This examines whether the theoretical propositions have been fairly drawn
from the chosen theory and whether the stated concepts are fairly represented
by the indicators chosen. The final stage examines the external validity – the
strength of the relationship between the research, the wider literature and the ‘real
world’.
A criticism of this method of evaluation is that, by working backwards through
the research process, it ‘puts the cart before the horse’, i.e. if the research does not
relate well to the literature (or the indicators to the concepts, etc.) then the internal
empirical validity doesn’t really matter.
Rose develops separate ‘strategies’ for evaluating research reports based on
quantitative and qualitative data. The strategy for evaluating theory-testing,
quantitative research has six steps:
1. summarising the research into ABCDE stages

2. assessing the operationalisation in C: concept-indicators, samples, units and variables
3. adding further descriptive details of: the researcher’s theoretical exposition and theory-
evidence linkages; and tables of data presented, reliability and accuracy, and interpre-
tation
4. assessing the internal empirical validity by examining the ‘fit’ between the data and its
interpretation
5. assessing the internal theoretical validity, i.e. the relationship between A (theory ) and
B (hypothesis)
6. concluding with a general assessment which consolidates the separate assessments
made and makes an assessment of the overall contribution of the research to the
discipline. (Rose, 1982: 104–5)
The strategy for evaluating theory-building research utilising qualitative

data is very different. It can be reduced to three steps:
1. summarising the research report into its ABCDE stages stating: its natural history
(origins and development); the data and methods of collection; sampling; method of
analysis; and presentation of results
2. evaluating in turn the validity of: the concept-indicator links; theory; and sampling and
generalisation
3. concluding by reviewing the consistency (or otherwise) of the various validities and other
factors and assessing of the overall contribution of the new theory to the discipline.
(Rose, 1982: 130–2)
Rose can be criticised for the assumptions made about the neat dichotomy of theory-
testing/quantitative and theory-building/qualitative research and for unnecessarily
complicating the evaluation strategy.
Both research component analysis and Rose’s ABCDE provide good, highly-
structured methods for analysing research which are particularly useful for first-time
or inexperienced researchers. With practice, you will learn how to home directly
on the strengths and weaknesses of published research and to answer the question:
does it do what it says on the tin?
Questions for discussion and assignments
1. Consider the concepts of: conflict; freedom; liberty; hegemony; and consciousness
of identity. What concept-indicators would you suggest for these? Why? What data
would you collect?
BOX 6.1
UWUK
Department of Politics
University of Watersea
Dr Adrian Helvetica
Editor
New Politics Digest
[date]
Dear Arial
[title of draft research report]
Thank you for inviting me to referee this paper.
Overall, my view is that, whilst this excellent paper breaks new ground, there is a number
of areas where minor revisions and additions would be helpful.
Let me begin by summarising what I believe are the research question, theoretical
framework, the hypothesis and the key concepts involved. The research question etc., is ….
[200–300 words].
The strengths of the paper are … [300–400 words].
The weaknesses are … [300–400 words].
I would also like to comment on the style of the paper … [200–300 words].
The paper is, of course, suitable for publication in your journal in its present form. However,
I would suggest that a number of small changes are made. They are … [200–300 words].
I look forward to meeting you at the Washington conference.
Best wishes,
Prof [your exam number]

2. [For undergraduates] You are a distinguished academic. You have been asked by
the editor of a refereed journal to act as a referee for a draft research report. Select
one of the two published reports:
Bevir, M. and Rhodes, R.A.W. (2006) Prime Ministers, Presidentialism and
Westminster Smokescreens, Political Studies, 54 (4): 671–90
Chaney, P. (2006) Critical Mass, Deliberation and the Substantive Representation
of Women: Evidence from the UK’s Devolution Programme, Political Studies, 54 (4):
691–714.
Write your assessment using the letter (above):
3. [For graduates] Select a research report on a subject that interests you from
a refereed journal in your university library. Evaluate it critically using research
component analysis or Rose’s strategies. Attach a copy of the research paper
to your assignment showing what you concluded were the various analytical
components. How could the research and the report be improved?
FURTHER READING
Rose, G. (1982) Deciphering Sociological Research. Basingstoke: Macmillan.

pp. 155.
Notes
1 Rose, G. (1982) Deciphering Sociological Research. London: Macmillan.

2 ESRC, AHRB and MRC are the acronyms of the government-funded Economics & Social
Research Council, Arts & History Research Council and Medical Research Council.
3 RAE: Research Assessment Exercise: a process of academic review which concentrates on
research performance and which provides ratings which determine allocations of government
funding. 5* is the highest rating.
4 Rose, G. (1982) Deciphering Sociological Research. London: Macmillan. Figure 2.1, p.14.
5 Downs, A. (1957) An Economic Theory of Democracy, New York: Harper & Brothers.
6 Jary, D. and Jary, J. (1995) Sociology, Glasgow: HarperCollins.
Chapter 7
Evaluating Information: Validity,
Reliability, Accuracy,
Triangulation
‘Do not feel absolutely certain of anything.’

Bertrand Russell, 19511
1. To consider why information should be assessed

2. To understand the distinction between ‘primary’ and ‘secondary sources’ of
information
3. To learn what is meant by the validity, reliability, and accuracy of information
4. To consider some warnings about ‘official data’
5. To consider further the distinction between ‘facts’ and ‘truth’
6. To understand the origin of triangulation and its application to research
7. To consider methods of sampling which can be used to collect data.
Introduction
In Chapter 6, you read how published research reports can be assessed. The
research component analysis and Rose’s ABCDE model examined the completeness
and coherence of the research process adopted. They also considered the validity
or otherwise of the relationships between theory and hypothesis, concepts and
indicators, empiric data and analysis, and conclusions. Research essentially involves
the gathering or collection of data that addresses the research question and enables
theory to be tested or developed. So the data from which answers to the research
question are to be drawn must be appropriate in terms of its relevance and efficacy –
‘fitness for purpose’. Much of this information will be drawn from published sources
that will be supplemented as necessary by new information specially collected for the
research project. So this chapter therefore suggests how best you can assess existing
data and seek additional material.
Many textbooks use information and data interchangeably. Some complicate
matters by treating ‘data’ as a plural noun and therefore writing ‘the data are …’.
While this is grammatically correct (for data is indeed the plural of the datum), it can
sound odd to students untrained in Latin conjugation. The author, Kingsley Amis
notably described such Latin correctness as the practice of ‘wankers’ as opposed to
‘berks’ who used slipshod English (1977).2 So this book follows everyday practice of
treating data as singular. Politicians also tend to use the word ‘evidence’ to describe
what they would wish us to regard as ‘conclusive, compelling information’ which
either proves or, in its absence, disproves allegation.
But is there a real difference between data, information and evidence? Certainly, the
dictionary meanings are similar. But some distinction is useful. Researchers tend to
speak of data as the mass of disordered, raw material from which information
(knowledge) is abstracted to provide evidence to support argument and conclusions.
(Information technologists adopt a similar distinction by defining information as
processed data sets attaining meaning.) Information informs. Evidence supports
conclusions.
So it is helpful to conceive of research as involving three stages. First, the raw
data is gathered. Second, the data is organised and distilled into information. Third,
evidence is abstracted from the information through processes of analysis and testing.
But neither information nor evidence is self-evident: the material seldom ‘speaks
for itself’. Some interpretation is required. However, when interpretation is re-
interpreted, some distortion of the original is inevitable. So some distinctions, criteria
and tests are useful to weed out distortions and ‘untruths’. The distinctions adopted
are between primary and secondary sources of information. The criteria used are validity,
reliability and accuracy. The main test adopted is triangulation.
Primary and secondary sources
The value of this distinction depends on which of the different definitions of

primary and secondary is adopted. Some authorities adopt the definition that primary
information is data generated specifically for the research project whilst secondary
information is data collected for other research. But, in this book, the ‘majority view’
prevails: that data is distinguished at the outset by its provenance (source). Primary
data is original, unedited and ‘first-hand’ whilst secondary data is ‘second-hand’,
edited and interpreted material. However, the distinction between the information
that you generate in the course of our research and that which you have abstracted
from other sources is valuable. I will therefore term this (after Huxley) α data
Evaluating Information: Validity, Reliability, Accuracy, Triangulation 81
and β data.3 Wherever possible, Politics researchers prefer to use primary, eye-
witness data recorded at the time by participants or privileged observers. The main
sources of primary data used by Politics researchers are fourfold:
1. contemporary documentary (written) records including minutes, letters, emails and

diaries
2. your interviews with key individuals, ‘agents’ and ‘actors’
3. numerical records, e.g. election results, census data
4. your own observation and records of interviews, etc. and other events.
Other sources are popular songs, poems, paintings and cartoons, photographs,
graffiti, murals (e.g. N. Ireland), T-shirts and videos. But beware, all records,
however ‘primary’ incorporate some degree of bias, perception, interpretation, and
editing, whether contextual, cultural, curatorial or deliberate.
Written primary records include accounts of meetings, minutes, diaries, letters,
reports, telephone transcripts, telegrams, emails, and newspaper reports, etc. But
how reliable, accurate and truthful are they? Who actually prepared them and why?
Arguably, all accounts are partial because they are functional, i.e. designed to fulfil
a purpose. But whose purpose? Most public records reflect the interpretation of
those holding power. Foucault argued that the victors write history. Alternatively,
how reliable are the diaries of (former British Labour Cabinet ministers) Richard
Crossman, Barbara Castle or Tony Benn? Did their cabinet colleagues know that
they were keeping diaries and change their behaviour accordingly?
Crossman offers telling insights into official records:
Thursday, 28 July 1966

One of the disconcerting features of the recent crisis has been the Cabinet Secretariat’s
habit of suppressing whole sections of the minutes on the grounds that they are too
secret to circulate. But this morning they didn’t do that.The section on prices and incomes
was reported at enormous length and most of what we said has been very adequately
summarised. Of course, this means that the Cabinet Secretariat regards the whole
subject as fraught with danger and was careful to record the arguments of the opponents.
Cabinet minutes are highly political and the way they are written has enormous effect.
By eliminating whole sections from the discussion and reporting other sections in full,
the Secretariat can greatly affect the way a decision is interpreted in Whitehall.
(Crossman, 1976: 590)4
But the BBC’s former Political Editor, John Cole wrote of Crossman that:
… [Crossman] had a brilliant mind, was a great polemicist, and a subtle – though
sometimes self-defeating – operator. But if you were Constable Plod seeking a reliable
witness, he would not be your first choice. I sometimes wondered if he knew how
to distinguish what he said to the Prime Minister from what the Prime Minister said
to him.
(Cole, 1996: 64)5
But even PC Plod can be an unreliable witness: Churchill’s bodyguard, Detective

Inspector Walter Thompson was criticised by Churchill’s biographer, Roy Jenkins
for exaggerating his importance in events (Jenkins, 2001: 552).6
Inspector Thompson in his two volumes of reminiscences is good at capturing the heart
of the matter but less reliable on exact dates, times and places than might have been
hoped for a meticulous detective.
(Jenkins, 2001: 562)
As a general principle, all primary information in the form of records – other than
those that you make through your own observations – should be treated with
caution. A ‘health warning’ is necessary. You should always ask yourself:
1. who prepared the record?

2. why?
3. for whom was it prepared?
4. for whom was it intended?
5. for what purpose was it made?
6. who would have ‘corrected’ or otherwise altered the record before it was finalised?
A common misconception is to believe that numeric information is more trust-

worthy than other formats because it is less vulnerable to ‘spin’. But, because
numeric records are generally regarded as trustworthy, they attract manipulation.
For example, the TUC and ILO accused the Thatcher government of changing
the definition of unemployment 23 times (between 1979 and 1991) to reduce
the headline figure and therefore conceal the true extent of unemployment.
The government responded to the criticism by saying that each new definition
distinguished further between genuinely unemployed people and others claiming to
be unemployed to obtain benefits. A similar charge was levied later against the New
Labour government that the lower levels of unemployment recorded and reported
had been achieved by accepting more readily claims (on mainly health grounds) for
the (higher) incapacity benefit. You should therefore check numerical records for
any changes of definition and any selective use of periods to enable worst records
to be omitted.
By implication, data that is not primary must be secondary – after the event, second-
hand. But it should not be discarded. Secondary information will include records
gathered from a number of separate, primary sources and may contain authoritative
commentary and analysis. The source’s interpretations and bias are important –
especially of evidence of how events were interpreted at the time and later, and the
moral relativism of value-judgements.
Validity, reliability and accuracy
As you learned in Chapter 6, social science research confers a special meaning to

validity:
the extent to which a measure, indicator or method of data collection possesses the
quality of being sound or true as far as can be judged. … in the social sciences generally,
the relationship between indicators and measures and the underlying concepts they are
taken to measure is often contested.
(Jary & Jary, 1995: 714)7
In effect, the validity of information is its relevance and appropriateness to your

research question and the directness and strength of its association with the concepts
under scrutiny. Often you will have to use best available information whose
validity may be weak. For example, to what extent, if any, does the decline
in ‘sectarian violence’ in N. Ireland post-2001 reflect a lessening of antagonisms
between conflicting groups? Does the election of an opposition party reflect
popular support for its manifesto or criticism of the outgoing government? Do
declining rates of party membership reflect a lessening of interest in health and
education? One measure that intrigues Politics researchers is the counterfactual –
events that don’t happen – as evidence of hegemonic domination.8 But how can
researchers be confident that the absence of an event can be attributed to the
omnipresence of another? One solution to this particular problem of problematic
validity is for you to adopt a wider range of measures to reduce dependence on
any one.
Reliability is, literally, the extent to which you can rely on the source of the data
and, therefore, the data itself. Reliable data is dependable, trustworthy, unfailing,
sure, authentic, genuine, reputable. Consistency is the main measure of reliability. So,
in literary accounts, the reputation of the source is critical. In John Cole’s view,
Richard Crossman was not a reliable diarist. Indicators of reliability will include
proximity to events, (whether the writer was a participant or observer), likely
impartiality, and whether, as the police say, the record was really contemporaneous
or an eventide reflection on the day’s events. Very few politicians admit to
real failings: all too often, their own agenda appears to justify their actions or
to criticise others. Tony Benn’s diaries seek to portray the inner workings of
cabinet government. But Dennis Healey claimed (playfully) that Tony Benn ‘always
seemed to be on the toilet every time a difficult decision had to be made’ (BBC2
interview). Accounts may have been ‘sexed up’ to promote sales. Biographies
may be hagiographic. For example, Michael Foot’s biography of Aneurin Bevan
uncritically portrays the Welshman as a wholly heroic figure, whereas my father –
a fellow native of Blaenau-Gwent – told me how, after 1948, some local trade
unionists called the Ebbw Vale MP ‘Urinal Bevan’. This epithet followed Bevan’s
assertion that:
it is for the [Party] Conference to lay down the policies of the Parliamentary Party, and
for the Parliamentary Party to interpret those policies in the light of the parliamentary
system.
(Foot, 1973: 236)9
In contrast, Grigg’s biography of another, Welsh hero, Lloyd George, provided a

‘warts and all’ portrait (Grigg, 1978).10
Numeric data need not necessarily be reliable. The source – even official statistics –
may not be wholly impartial. Populations may be undercounted (e.g. 2001 census).
The samples used may be insufficient or not randomly selected. Confidence limits
(margin of error) may be omitted. The rate of non-responses to questionnaires may
be disguised. Respondents may not have been wholly truthful in their replies. For
example, on the basis of replies to their questions, most opinion polls (wrongly)
predicted a Labour victory in the 1993 general election. Inappropriate statistical
techniques may have been used. But reliable witnesses may also be inaccurate on
occasions.
Andrew Marr, John Cole’s successor as the BBC’s Political Editor and a former
editor of The Independent is very sceptical of the reliability of modern-day news
reporting by the newspapers and TV news services (Marr, 2004). He blames this
on the competition to drive down costs, consequent reductions in the number of
journalists, and their being confined to their desks where they must too readily
accept the stories ‘fed’ them by professional press officers. He recommends readers
(and researchers) to:
Know [which newspaper] you’re buying. Reporting is so contaminated by bias and

campaigning, and general mischief, that no reader can hope to get a picture of what is
happening without first knowing who owns the paper, and who it is being published for.
The Mirror defines its politics as the opposite of the Sun’s, which in turn is defined by
the geo-politics of Rupert Murdoch – hostile to European federation and the euro … It
is ferociously against Tony Blair, this is because Number Ten has been passing good
stories to the Sun.
(Marr, 2004: 251)
He also warns against news of research from:
hundreds of dodgy academic departments put out … to impress busy newspaper people
and to win themselves cheap publicity which can in turn be used in their next funding
applications.
(2004: 254)
Similarly, Marr explains that TV news editors are:
biased towards exciting or unusual pictures; news that is refreshing or odd; and news
that bears some relation to viewers’ lives.
(2004: 291)
So anything that looks dull, ‘stories about northern European countries, about
buses, about old people, about infrastructure, banking, manufacturing, Whitehall
and regeneration,’ is unlikely to be televised. Marr argues that a task of TV news is
to increase viewing figures – which means also retaining the viewers of the preceding
programme – usually popular light entertainment of the ‘soap’ or ‘chat show’ genre.
Accuracy is sensitivity to change – especially of detail, e.g. dates, numbers,

persons present, etc. Remember that some biographers deliberately add false detailed
information to trap and sue plagiarisers.
Facts and truth
Once again, you will find that adopting a critical distinction between facts and
truth is useful. Facts are the available data. They present incomplete snapshots of
events. Truth is the reality behind the facts. Sometimes the facts may obscure
the truth – perhaps deliberately so. A good example was provided to me by a
leading academic. He privately described how he had critically reviewed a best-
selling account of British rural life where the author had misrepresented the facts by
combining material from a number of interviews to represent a composite figure.
The author had replied to the effect that his critic was unable to distinguish between
the facts and truth.
Interviews
Interviews with political elites provide a major source of information in Politics

research. They may be undertaken by the researcher or, where personal access is not
possible, by watching video recordings of interviews in TV news and documentaries.

But you must never assume that what you are told or hear is reliable and accurate.
For example, a former Prime Minister told me that he strongly supported a specific
White Paper. However, the Minister of State who claimed to have instigated the
new policies told me that the Prime Minister had opposed the White Paper. You
must always question (implicitly) the answers to your questions and look for signs
of deception or self-deception by informants, e.g. the coping strategy of long-term
prisoners who are guilty but believe that they are innocent, i.e. in denial. TV
interviews (i.e. secondary sources) are highly edited – especially field interviews
where a single camera is used or where the interviewee has been granted some
editorial control.
Triangulation is the means adopted by researchers to secure effective corroboration.

However, before this method is described, consider the case study below:
Case Study
Harold Nicholson provides a detailed narrative of the fall of the second Labour government in
1931 and its replacement by a National Government which was to last effectively until 1945
(Nicholson, 1953: 453–469).11 The Labour Prime Minister, Ramsay MacDonald, became the
leader of the National Government in what became named by Labour Party members as the
‘great betrayal’.
Nicholson describes the relevant background as the rapidly deteriorating public finances
caused during the worst years of the Depression when the demand for public expenditure
on unemployment benefit etc. grew whilst income from taxation fell. In response to demands
by the Conservative and Liberal parties (amplified by the Tory press), the Government formed
an independent committee under Lord May. On 31 July 1931, May recommended substantial
cuts of up to 20% in public sector salaries, 20% cut in unemployment benefit and reduction
in the pay of the armed services to 1925 levels. But two of the six May members issued a
minority report dissenting from May’s recommendations on the basis that the costs of the
cuts would fall mainly on the working classes. Nicholson wrote that ‘The rank and file of the
Labour party agreed whole heartedly with [the Minority Report]; MacDonald and Snowden
[Chancellor of the Exchequer] did not’ (1953: 455). Nicholson reports how, later that day,
MacDonald formed a five-man, special, Cabinet Economy Committee to consider how May
could be implemented. The ‘Big Five’ consisted of MacDonald, Snowden (Chancellor), J. H.
Thomas, Arthur Henderson and William Graham.The likely continuing withdrawal of deposits
held in London meant that the government would be unable to fund the public sector deficit
without support from bankers in Paris and New York. The bankers were unwilling to lend the
money unless and until firm proposals were made to balance the UK budget. On 19 August,
the Cabinet’s Economy Committee proposed cuts similar to the May report. Half the deficit
would be met by reductions in unemployment benefit and public sector pay and the other by
increasing taxation etc. According to Nicholson, the proposals were reluctantly approved by a
majority of the Cabinet with the exception of transitional unemployment pay (1953: 457).They
suggested an additional revenue tariff which MacDonald told them would not be accepted by
the Liberals.The TUC – which had created and funded the Labour party – met MacDonald on
20 August. The Council refused to accept cuts in unemployment benefit or public sector pay.
On 22 August, MacDonald proposed a modified version of the programme – including a 10%
reduction in unemployment pay – to the Cabinet. He obtained Cabinet support to ‘enquire’
of the Opposition leaders whether the revised proposals were acceptable. The Opposition
leaders responded that the overseas bankers’ support was critical to wider support for the
package.
The King (George V) had been kept informed by MacDonald of the increasing crisis. He
returned from Balmoral to London on 23 August when he was told by MacDonald that leading
members of the Cabinet would not support the latest proposals (being considered by the
bankers). The King decided that the ‘correct constitutional course’ would be to meet the
leaders of the three main parties: MacDonald (Labour), Baldwin (Conservative) and Samuel
(Liberal, as Lloyd George was in hospital). Nicholson describes how the King stated his
preference for MacDonald and the Labour government to stay in office and to implement
the cuts. If that were not practicable, then the best alternative would involve the formation
of a National Government – headed by MacDonald – with a Cabinet drawn from three
parties and commanding a sufficient majority in the House of Commons to approve the
necessary legislation. The King met the three leaders separately who agreed to join a
National Government if necessary. Once the crisis had been resolved, then new elections
should be held.
Later that day, the Labour Cabinet met. After a long adjournment, they were advised by
telegram from the Government’s agents (the bankers, J. P. Morgan) that the necessary US
public support for a public loan would be problematic until Parliament had approved the
proposals. They suggested a short-term treasury loan. Finally, they enquired whether they
were correct in assuming that the package proposed by the Cabinet had the support of
The Treasury and City. But the Cabinet had not and would not approve the programme.
MacDonald told them that he would report the divisions (eleven in favour: eight opposed) to
the King whom he would ask to convene a meeting of the three party leaders. He would tell
the King that the Cabinet had placed their resignations in his hands. He immediately reported
to the King. Acting unilaterally, MacDonald began that night to plan the new Government
Continued
with what Snowden described ‘an enthusiasm which showed that the adventure was highly
agreeable to him’ (1953: 465). On 24 August, the King met the three party leaders who agreed
to form a National Government. MacDonald tendered his resignation. He was then invited
to lead the National Government. He asked them to prepare a communiqué saying that
the formation of the new Government was being considered. The party leaders agreed that
the new Government would not be a coalition but a ‘co-operation of individuals’. MacDonald
described the proposals to his Cabinet and invited them to join a new ‘Cabinet of Individuals’.
With the exception of J. H. Thomas, Lord Sankey and Philip Snowden, they declined.
Following the resignation of the Labour Cabinet, the new Cabinet was formed on 26 August
with Baldwin as Vice-Premier.
There are a number of other accounts of this episode. One central dispute
among them is whether the King or MacDonald first raised the proposal for a
National Government, i.e. whether MacDonald accepted the King’s proposal to lead
a national government out of patriotism and loyalty, or, proposed the arrangement as
a means of continuing in office and increasing his power whilst appearing to follow
the constitutional requirements of the King. For example, the celebrated Labour
historian, G. D. H. Cole wrote that:
The exact method of the split is vehemently disputed. The Labour Cabinet was still
discussing the outrush of gold and the ‘threat to the pound’ under the influence of the
deliberately exaggerated menaces of Philip Snowden, when it came. They had agreed
to enormous concessions but jibbed (it is stated) at penalising the unemployed. Then it
was put to them that arrangements had already been made, with the King’s consent but
clearly on MacDonald’s initiative for the formation of a ‘National Government’ of Labour,
Conservatives and Liberals.
(Cole, 1938/66: 593)12
So how much importance should the researcher apply to Nicholson’s account

where it differs from others in critical respects? Obviously, the researcher will
seek corroboration from other sources and assiduously compare the various clues.
But just how reliable is Nicholson’s account. Who is Nicholson? What was his
book about? Why did he write it? What documents and witnesses did he have
access to?
The answers are that Sir Harold Nicholson (1886–1968) was a distinguished
diplomat, historian and biographer. He was educated at Wellington and Balliol
College. He attended the Paris Peace Conference (1919) as a diplomat. He retired

from the service to become a writer. In 1931, he stood as an MP for Harold
Mosley’s New Party but left when Mosley formed the British Union of Fascists.
He became a National Labour MP in 1935 but was defeated in 1945. He became
a governor of the BBC. He was a ‘man of independent means’ who married
Vita Sackville-West. They lived at Sissinghurst Castle. Both were bi-sexual and
practised an ‘open marriage’. However, their life-style was not unusual amongst
the upper classes. We can therefore regard Nicholson as an Establishment figure
albeit on its arts and literary wing. His book was entitled King George V and
published in 1953, i.e. 27 years after the King’s death in 1936. He had been
appointed by the Royal Family to write an ‘authorised biography’. So Nicholson
had access to the King’s diaries and the official papers kept by his secretaries. He
also interviewed the King’s secretary, Lord Samuel, Herbert Morrison and other
participants. However, we are unaware of the editorial control exercised by the
Royal Family.
Clearly, the overall purpose of the biography was to make public King George’s
hand in the making of history. But Nicholson could not be accused of presenting
a wholly flattering picture of the King. For example, he observed that George
was neither very clever nor witty: he was a relatively dull man who was
therefore entirely representative of the British people. Overall, we can probably
conclude that Nicholson’s account is probably very accurate in terms of the
detail. However, we cannot be entirely sure about its reliability: the biography
is more likely to portray King George V as a central figure, imposing wise,
constitutional solutions, rather than acquiescing to what other commentators
(like Cole) saw as MacDonald’s devious scheme. No single account can ever be
regarded as wholly reliable or accurate. Other sources must be sought and used.
However, the number of competing sources may be huge. The most widely
used method of selecting sources and materials from the range available is termed
triangulation.
Triangulation
Triangulation is a method developed over the centuries for navigation and surveying.
It provides the basis for satellite navigation. Its origins lie in geometry. A point
can be precisely defined in space by the angle it subtends to a line joining two
other points by the application of the Law of Sines. Any geographical area could be
mapped by first selecting two ‘trig points’ (triangulation points) that are a measured
distance apart and then recording the position of any other point in terms of
the angle subtended. In this way, a third trig point could be established and the
exercise extended. In navigation, triangulation is used to establish a ship’s precise
position by taking bearings on three or more known landmarks. The position is
most accurately determined when the three points are equally located around the
ship, namely
Lighthouse A Lighthouse B
Ship ⇑
Lighthouse C
The same method is used in Politics research to obtain an efficient corroboration

of any crucial account. Triangulation involves seeking accounts from three or
more perspectives.
So, for example, a researcher investigating the General Strike of 1926 would
seek to obtain accounts from the TUC, the Government and at least one source
independent of the two adversaries. Politics researchers face a special challenge:
the two main parties and their perspectives are often diametrically opposed to
each so that any, third-party, independent views are gained from only one side.
To overcome this difficulty, they seek as many independent sources as possible.
Furthermore, Politics researchers seek to triangulate at each level of data media. So
you should seek to triangulate between contemporary written records and news
reports, autobiographies, personal interviews with participants, and other research
narratives. You should also try to find new angles. However, given that each party
will adopt different perspectives, then the ‘truth of the matter’ may be unique to
each participant. You may be able to repudiate some accounts but you may find
that you are unable to offer a definitive version of events. Indeed, the participants
may be unsure of their real motivations or involvement. At the end of your
triangulation, you may well know more about and understand better the particular
event than the participants because you will have accessed records unavailable to
them. But remember the adage that, whilst success has many parents, failure is
an orphan. However, by demonstrating the application of triangulation, you will
be able to show the reader the process by which corroboration has been sought.
You should also be able to pinpoint both gaps in and inconsistencies between the
accounts.
Sampling
You will quickly find that, even when you adopt triangulation, the volume and
potential sources of data in terms of people and records can still remain vast. So
a selective approach is essential. You simply cannot interview every member of
a union or examine every council minute. The age-old solution to this particular
problem is to concentrate your activity on a sample of the population. These everyday
words have special meanings in academic research which warrant explanation. The
population is the universe of all the subjects or cases under study. You must define
your population. It may be all the members of a union, the residents of a city or
town, states of Africa, a particular ethnic or age group and so on. The population
is a set of individuals, cases, states, etc., which share a common characteristic. The
sample is a selection of individuals, cases, states, etc., made from the population.
The sample is, therefore, a subset of the population.
You must define your research population in a sample frame. Your sample frame
is a list or schedule of the population from which the sample will be drawn. It may
be a membership list or a directory of engineering firms.
You can use either probability or non-probability samples.
Non-probability samples
Non-probability samples are samples where members of the population do not
have an equal chance of being selected. They are not statistically reliable. They
cannot generate generalisable data. You select the members of the sample. Non-
probability samples include nomination, snowballs, volunteers, case studies and theoretical
samples (used in grounded research). Non-probability samples will not generate
generalisable information. You have already been introduced to case studies and
theoretical samples in Chapter 5. Non-probability samples are very small. They are
used for qualitative research.
Nomination is the most widely-used form of recruiting a non-probability
sample. Essentially, you ask a local social gatekeeper or intermediary to nominate
(name) a group of people who meet your requirements for a research sample.
A social gatekeeper exercises control over who enters a community. They may
be head teachers, village headmen, employers or heads of family (if you want to
interview children). Essentially, you search them out, write to them explaining who
you are and the nature of your research, and ask them to nominate a cross-section of
local people to interview. Often, they will offer to arrange the meetings. But this offer
of assistance is incompatible with the ethical ideal of voluntary consent. The people
nominated may feel as if they have been commanded to attend. This will inhibit
your meeting: they will feel obliged to say what the gatekeeper would like them to
say. It is better if you contact the named people individually (preferably in writing).
Again you should introduce yourself, explain the nature of your research and its
benefits to them. You should say why you want to speak to them individually or at
a (focus) group meeting, and add that X (the local social gatekeeper) has suggested
that they might be able to help. Intermediaries are not social gatekeepers but
trusted outsiders or thresholders who are respected by the community. Typically,
they are ministers of religion, teachers, nurses or local leaders of voluntary agencies
or NGOs (Non-Governmental Organisations) like Oxfam.
Snowballs grow larger as they are rolled across snowfields. The term snowball
sample is therefore used to describe samples which become larger as each contact
suggests more people to contact. They are particularly useful for researching groups
whose identity is concealed. For example, say you wish to research the likely impact
on crime of a greater liberalisation of laws against illicit ‘hard’ drug use (crack, meth,
etc.). You would be unlikely to make contact with drug-using criminals through
a probability sample of, say, 1000 from the UK population. First, the incidence
of drug-using criminals is relatively small. And, second, they would be unlikely to
reveal themselves to you. The two problems you face here are access and trust. These
can be overcome by using a snowball sample. In the example above, you could
begin by contacting a person who is publicly known to have had some contact
with these criminals. They might be a specialist doctor, a prison visitor, shelter
manager or journalist. Your first task would be to meet them to establish your
bona fides (Latin, meaning ‘good faith’) as an academic researcher and, therefore,
trustworthy. Several meetings might be required. You would ask them to refer
you to other people who might be able to assist you. They might refer you on
to recovering addicts, their relatives, or organisers of self-aid groups. In turn, you
would hope that they would find you sufficiently genuine and trustworthy to provide
further contacts. You would hope that your research trail would end in clandestine
meetings with practising addicts who funded their drug purchases through petty
crime or wider involvement in criminal networks. At the end of your research,
you will not be able to claim that your findings applied to addicts beyond those
you had met: however, you can claim that they apply to all the subjects in your
sample.
A volunteer sample is one where members of the research population volunteer
to take part in your research. You are most likely to seek volunteers where your
sample is likely to undergo a period of discomfort, pain or financial cost. You will
have received emails from university departments – especially Psychology – asking
for volunteers for lab-tests. They may offer a small cash payment to volunteers. For
example, you might seek volunteers from colleagues if you wished to pre-test and
compare the possible impact of positive and negative political advertising in the
UK. But appeals for volunteers need not necessarily be made only to people. You
can also ask for volunteers from organisations. For example, to pursue research on
political networks, you could write directly to each unitary authority asking them
to take part in your research. The great advantage of seeking volunteer samples is
that they can be relied upon to co-operate fully. Conversely, because the volunteers
are self-selecting, then they are more likely to be especially interested in the topic
and therefore may be less likely to be representative of the population as a whole.
Probability Samples
Probability samples are samples in which every member of the defined population
has an equal likelihood of being selected for inclusion. So, in a population of 1,000,
each person has a 1/1000 probability (also expressed as 0.1% or p = 0.001) of
being selected. Probability samples are statistically reliable. This means that they
are capable of generating data which is representative of the population. In other
words, if the average age of members of a probability sample is 30, then you can be
confident that the average age of the population as a whole is also 30. This capacity to
generate accurate representation is called generalisability. Probability samples are
large. They are used in all quantitative research where the population is very large.
The reliability of the data obtained from a sample will increase as the sample
increases in size towards that of the whole population. But it is the size of the
sample which determines its accuracy: the size of the population is less relevant. But
doubling the size of the sample will not double the reliability of the information.
Accuracy is proportional to the square root of the sample size. So, to double the accuracy,
the sample size must be increased fourfold – which will greatly increase the cost of
the sample survey.
There is, therefore, a trade-off between cost and reliability – and time. This is
an example of the so-called law of diminishing returns (or diminishing marginal utility).
It explains why most samples are relatively small, e.g. national opinion polls rarely
use more than 1,500 people, whilst even national, life-or-death medical surveys
rarely exceed 60,000, i.e. a 0.1% sample.
There are many types of probability samples: simple random samples (entirely
random); systematic samples (every nth person); stratified samples (e.g.
50:50, men: women, etc.); multi-stage cluster samples, and, probability
proportionate to size (PPS) samples. Essentially, stratified, multi-stage cluster
and PPS involve pre-designing the sample to reflect the known characteristics of the
population under study, e.g. by gender, age group, ethnicity, social group, residence.
They are also used to make the sample more readily contactable and thereby reduce
costs. For example, if you were proposing to carry out a face-to-face survey of
10,000 people in England, then you could choose 1,000 electoral wards randomly
and then 10 addresses within each (or 200 wards and 50 addresses in each). A further
refinement would be to select randomly 125 wards within each of the 8 standard
regions and 10 residents in each ward. But some regions have larger populations than
others. So you could vary the number of wards pro rata so the likelihood of any ward
being chosen was more equal. But wards also vary significantly in size (depending
on whether they elect one, two, or three members). So the list of wards from which
the sample is to be chosen should reproduce multi-member wards twice or three
times, and so on. In this way, you are constructing a sample with a probability
proportionate to size. In this case, you would have constructed your sample frame
in three layers: region, ward, population. These layers are termed strata.
The more strata employed, the smaller the size of each group and the lower
the reliability of the data. One statistical survey rule – rarely followed – is that the
smallest group should not be lower than 1,000. In practice, a ‘rule-of-thumb’ is
adopted in which the group should be at least 50–100 (Hoinville et al, 1977: 61).13
There is a particular problem with very small minorities, e.g. non-whites in UK rural
county towns. A group of, say, 100 educated males 26–35 may well include less than
five members of minority ethnic communities. But five people are unlikely to be
representative of the many minority ethnic communities. In this case, you should
seek a higher number of participants from these communities and scale down the
data accordingly. Alternatively, if ethnicity is a critical variable, then a very much
higher stratified sample should be sought.
Tables have been developed which relate sample size to the degrees of acceptable
sampling error and levels of confidence (CL). Most Politics research adopts levels of
confidence of 95%. This means that you are confident that, in 95 out of every 100
cases, the characteristic (e.g. party preference) shown by the sample will be shared
by the research population. The sampling error is the inaccuracy arising from the
use of a sample. So, as Table 7.1 shows, if you are willing to accept a sampling error
of 5% either way, you can use a sample of 400. But, if you insist on a sample error
as low as 1% either way, then you must use a sample of 10,000.
For example, say you have adopted a random sample of 2,500 people whether
they support or oppose NATO forces, involvement in Afghanistan. Of these, 36%
say they are supportive. What the table above tells you is that, in this case, you can
be confident that, in 95 out of every 100 members of the population, 36% will be
supportive ± 2%. So, at 95 confidence levels, support will lie between 34% and 38%.
Table 7.1 Sampling errors/sample size of random

samples at 95% confidence levels.
Sampling Error % Sample Size

1 10,000
2 2,500
3 1,100
4 625
5 400
6 277
7 204
8 156
9 123
Abstracted from De Vaus, 2001: 71, Table 5.4
Therefore, if you want to reduce the sampling error by half (to ± 1%), then you
will have to increase your sample to 10,000.
The greatest difficulty in using probability samples is designing the sample frame.
The sample frame is the list of the population from which the sample will be
drawn. Say, for example, you want to carry out questionnaire research of residents
of a city. Twenty years ago, you could have used the electoral roll or telephone
directory. However, to protect the privacy of the public and to encourage more
people to register and therefore to vote, the full electoral roll is no longer publicly
available. The telephone directory is now much less representative than previously
now that a third of ‘subscribers’ choose to go ‘ex-directory’. In any event, the
telephone directory lists heads of household and not their partners. Additionally,
young people are more likely to use only mobile telephones which are unlisted.
Large-scale probability sampling is very expensive. One practicable means
available to (fully-funded) researchers to obtain very large samples is to buy into
one of the very large (100,000) sample omnibus questionnaire surveys undertaken
by market researchers acting on behalf of commercial clients – notably superstore
grocers and financial services. They use random samples obtained from a sample
frame of a national gazetteer of postal addresses. The disadvantage is the relatively
low completion rates and the bias arising from the use of prize incentives.
For these and other reasons (especially costs and the lack of life-or-death
consequences for the population), Politics researchers rarely use probability samples.
Instead, like most private firms, you are most likely to use quota samples.
Quota sampling
Despite being the most widely-used, quantitative sampling technique, quota sampling
is non-probabilistic. However, for calculation purposes, custom allows you to analyse
the data using the same statistical techniques as if it had been obtained using simple
random sampling. A quota sample is a sample of the population which is pre-
designed to be representative. So, for example, if you know that 69% of your
population of UK electors voted in the last general election, then you will design your
sample to have a quota of 69% voters. How do you know whether a person voted?
You ask them. Quota samples are usually recruited in town centres by researchers
who select passers-by to complete their quota of interviews. It is non-probabilistic
because the interviewers select individuals to meet their quota. So each passer-by
does not have the same probability of being selected. Furthermore, because the
weekday city centre contains higher proportions of some groups of the population
and less of others, then the probability of each member of the research population
being selected is unequal. This also explains why you may not have been interviewed
by the ever-present pollsters in your city centre. Their quota of people like you had
already been filled. Or, alternatively, your type of person did not form part of the
sample frame of, say, pensioners. Or, importantly, they may have felt uncomfortable
by your appearance. You will find out when you carry out your own quota sampling
that you are likely to select people with whom you believe that you are likely to
develop rapport. You will find more detailed guidance on administering (carrying
out) an on-street survey in Chapter 8.
Designing a quota sample

You will find that the simplest way to design and use a quota sample is to start by
setting up a 10 by 10 matrix of 100 cells.
For example, suppose you require a representative sample of the research
population in which age, gender and socio-economic group are considered
relevant – as potential independent variables – to your research question. You find
from published sources that:
1. the breakdown of population between males and females is 50:50%

2. between 18–34, 35–59, 60+ years old, the relative distribution is 20:50:30%
3. as a proxy indicator of socio-economic group, the percentage ratio of owner-occupations
to tenant is 60:40%.
Then you can sub-divide the 100-cell matrix into columns for sex and socio-
economic group (assuming the owner/tenant distribution is uniform between
genders and ages) and into rows for age bands, namely:
In this matrix, the highest, right-hand cell (marked ‘X’) will be a woman, aged
18–34 and living in rented accommodation whereas ‘P’ will be a man, aged 35–49
who is an owner-occupier.
You can then ‘scale-up’ the matrix to provide the optimum sample size on the basis
of weighing the advantages of accuracy, reliability and representativeness, against
the resource costs. You can, of course, add further sub-divisions, for example, of
ethnic origin. But remember that, each time you subdivide the sample further, the
sub-groups become smaller and potentially less representative.
So armed with your quota matrix and questionnaire, you can begin your quota
sampling. As you recruit each member of the quota sub-group, you ‘tick them off’
from the matrix. Beginning is easy. However, the technique becomes more difficult
as the number of vacant cells in the matrix reduces.
1. What do you understand by the essential difference between validity, reliability and
accuracy, in terms of data? Why are the distinctions useful?
Table 7.2 Quota sample matrix (100 cells)

Men Women
18–34 X
35–59
60+
Tenants Home-owners Tenants
2. Discuss the case study of the formation of the National Government in 1931,
drawing on other accounts which, together with Nicholson’s biography of George
V, enable an appropriate triangulation to be achieved.
3. Consider your university’s prospectus. Identify examples where numeric or other
authoritative data may have been used selectively to create a best-case presen-
tation to potential students.
4. Arthur Scargill attracted both strong supporters and critics for his role as leader
of the NUM (National Union of Mineworkers) during the miners’ strike of 1983–4.
You have obtained documentary sources from the NUM and the autobiographies
of government ministers and advisers of the day. What other sources would you
seek out to achieve triangulation?
5. Design a quota sample of 400 adults to represent your research population of
adult residents of your university city, or town.You wish to test the potential causal
relationship between ‘green practices’ and age, class and sex. You have learned
that 60% of households separate and sort recyclables from their household refuse.
Find the data on age, class and sex from the census data for your city or town.
What are the advantages and disadvantages of this method?
FURTHER READING
Harrison, L. (2001)Political Research: An Introduction. London: Routledge.

pp. 25–9, 106–12. In the first extract, Harrison discusses and distinguishes
between validity and reliability, and provides additional material on types of
validity including construct validity and content validity. In the second extract,
she examines and discusses alternative sources on existing political data
including the ‘mass media’, party resources, biographical, autobiographical and
political memoirs and the internet.
Neuman, W.L. (2003) Social Research Methods. London: Pearson. pp. 178–87,
137–8. The first extract discusses reliability and validity in the separate contexts
of quantitative and qualitative research. Table 7.1 on p. 183 summarises the
measurement reliability and validity types identified. The second extract
provides a very readable discussion of triangulation of which he provides an
example of four types in Box 6.1 on p. 138.
De Vaus, D.A. (2001) Surveys in Social Research. London: Routledge. pp. 54–79.
This extract begins with a discussion of reliability and validity. The author
introduces three means of assessing validity: criterion validity; content validity;
and construct validity. He also discusses the special problem of how people
may interpret indicators in different ways. The second part of the extract
provides good, practical advice on the separate types of probability and
non-probability samples.
Notes
1 Russell, B. (1971) A Liberal Decalogue. In The Autobiography of Bertrand Russell, 3. 1944–67,

London: George Allen & Unwin. pp. 60–01.
2 Amis, K. (1997) The King’s English – A Guide to Modern Usage, cited by Mount, H. (2006) Amo,
Amas, Amat and All That: How to Become a Latin Love, London: Short Books. p. 40.
3 Huxley, A. (1932) Brave New World.
4 Crossman, R.H. (1976) The Diaries of a Cabinet Minister 1. Minister of Housing 1964–66.
London: Hamish Hamilton and Jonathan Cape. p. 590.
5 Cole, J. (1996) As It Seemed to Me: Political Memoirs. London. Weidenfeld & Nicolson. p. 64.
6 Jenkins, R. (2001) Churchill: A Biography. London: Macmillan.
7 Jary, D. and Jary, J. (1995) Sociology. Glasgow: HarperCollins. p. 714.

8 Lukes, S. (1974) Power: A Radical View. London: Macmillan. p. 51.
9 Foot, M. (1973) Aneurin Bevan: A Biography, 2. 1945–60, London: Davis-Poynter.
10 Grigg, J. (1978) Lloyd-George: The People’s Champion. London: Eyre Methuen.
11 Nicolson, N.(1953) King George V. London: Routledge. Chapter XXVII. pp 453–69
12 Cole, G.D.H. (1938/66) The Common People. London: Methuen.
13 Hoinville, G., Jowell, R. et al (1977) Survey Research Practic, London: Heinemann. p.61, cited in
De Vaus, D.A. (2001) Surveys in Social Research. London: Routledge. p. 73
Chapter 8
Completing a Literature Review:
Accessing Published (β)
Information
1. To understand why researchers undertake literature reviews.

2. To learn how to structure and present a literature review.
3. To consider the various search processes to access published (β) information.
Introduction
The literature review is a major part of any research. But many literature reviews
are done badly. Poor reviews often consist of one description of a text after another.
They lack any original criticism or evidence that the student has actually pondered
what they have read. There remains a suspicion that some students haven’t actually
read the texts they cite, but merely summarised the abstracts or dust-covers. They
are also very boring – especially to the external examiners who will have been
chosen because they are authorities on the research topic. Perhaps the fault lies in
the expectation by higher education that students will have been taught how to do
a literature review at secondary school and vice versa. There is also the view that
doing a literature review is ‘common sense’ which requires no preliminary guidance.
I strongly disagree with these views.
Why do a literature review?
The review is not undertaken for its own sake, or as a kind of academic penance
(although it may feel that way at times). The primary purpose of the literature review
is to establish the state of current ‘knowledge’ – or argument – about your research topic.
Knowledge includes views, concepts, theories, understanding, evidence, schisms
Completing a Literature Review: Accessing Published (β) Information 101
and schools, and the main authors. In the first place, it involves you finding out
what has already been written (or said). It also involves finding out what you don’t
know and possible unknowns. As Donald Rumsfeld famously said:
Reports that say that something hasn’t happened are always interesting to me because,
as we know, there are known knowns; there are things that we know we know. We also
know that there are known unknowns; that is to say we know that there are some things
that we do not know. But there are also unknown unknowns – the one’s we don’t know
that we don’t know.
(Donald Rumsfeld, US Secretary of State for Defense, 2003)
Reaction against Rumsfeld’s part in the Iraq war has meant that his statement was
widely ridiculed, whereas, in hindsight, it is worthy of Foucault.
Reviewing the literature prevents you from ‘reinventing the wheel’ and enables
you to contribute to new knowledge or understanding. However, in Politics,
there are few uncontested, ‘knowledge claims’. Instead, there are a number of –
often competing – explanations of phenomena which reflect opposing paradigms,
information and periods of time. But those explanations are not beyond criticism.
Your task as a Politics researcher is, therefore, to identify: who are the main
theorists, commentators and texts; how the arguments have developed over time; the
areas of agreement and disagreement; the overlaps and the ‘gaps in the literature’. The
literature review should also summarise and evaluate published criticism and, most
importantly, offer fresh, original criticism. Your criticism should identify loopholes,
illogical deductions and conclusions, which are not supported by the information
from which they are allegedly derived. A literature review is not merely a chronicle
of who wrote what and when, but a forensic examination of texts to identify one
or more critical elements where current understanding is unclear or contested and
which the new research can address. After all, it is this lack of clarity or information
that is the justification for you undertaking the research and for the reader spending
their time.
The explosion of text, in the form of books, journals and the Internet, makes
literature reviews more difficult than before. Conversely, the Internet has made
searching for publications very much easier. But you will be unable to read every
available text. A selective approach is essential in terms of what texts – and which
parts of those texts – you should read and what records you should make. The ‘law
of diminishing returns’ applies to literature. And, conversely, reading ever more
texts can become a displacement activity to justify putting off the start of fieldwork.
A good supervisor can spot when you are blaming the literature review for general
lack of progress (or effort). As a rule-of-thumb, the literature review is likely to
occupy between a quarter and a third of the time available for the research project.
A quarter of the research report is likely to be set aside for the theory and literature
review. Traditionally, the first year of a three-year PhD research project was devoted
to ‘reading the literature’. Whilst the proportion of time remains the same, current
practice is to integrate the literature review with the search for empiric data in a
type of grounded research with a near-continual process of writing-up-as-you-go that
begins with the original research proposal. The objective of the review is to identify,
criticise and synthesise the most recent, relevant, authoritative texts. Authority is critical.
It generally means those texts that are widely accepted by the academic community
as works of considerable scholarship. Finding and accessing these texts requires a
systematic search of literary sources.
Searching the literature
The literature search necessarily precedes the literature review. So, where should you
begin to search for authoritative texts? A starting point is your course modules, your
core texts and the bibliographies provided by their authors. Then you must (not
‘should’) move on to the bibliographic databases of journals (and their predecessors the
printed bibliographies) provided by the university library. As already noted, journals
are much shorter and more concise than books, and will have generally been vetted
before publication by referees selected from the leading academic specialists in the
topic field. Bibliographic databases allow you to search for articles which include
specific words in the title or abstract or, alternatively, written by a particular author.
You can limit your search to a specific range of years, a particular set of journals, and
so on. The database is likely to have been designed to enable you to follow a link
directly from the database entry to the article itself, where the journal is available in
electronic format. Useful databases for UK-based Politics researchers include (listed
alphabetically):
ASSIA (Applied Social Sciences Index and Abstracts) www.csa.com/factsheets/

assia-set-c.php
BIDS (Bath Information and Data Services) www.bids.ac.uk
British Humanities Index www.csa1.co.uk/factsheets/supplements/bhi.php
IBSS (International Bibliography of the Social Sciences) www.ibss.ac.uk
PAIS International (Public Affairs Information Service) www.pais.org
Social Sciences Citation Index (SSCI) www.isinet.com
Each journal article will also contain its own bibliography. The Social Sciences Cita-
tion Index allows you to move from the details of one article to all the other articles
that have been cited and also to identify which articles are cited most often. You must
log in to use these databases. Many have been acquired by the US information firm,
CSA based in Maryland, or other commercial providers. But researchers based in UK
universities can normally access the databases free-of-charge by registering at their

university with ATHENS which will provide the necessary password etc. Others
can use the databases by taking advantage of the free, short-term trial membership
provided by CSA and others.
Other relevant material on your topic may have been published on the Internet
without having been submitted to an academic or professional journal. When using
a commercial search engine like Google or Yahoo, it can be difficult to identify
which web sites carry authoritative texts or other information. A general health
warning is necessary. To overcome this problem, some UK academic institutions
have established their own searchable directories of web sites that meet acceptable
standards of scholarship. These include RDN (Research Discovery Network,
www.rdn.ac.uk), Intute (www.intute.ac.uk) which includes an excellent ‘virtual
seminar’ on how to make electronic searches, and its social science sub-set SOSIG
(Social Sciences Information Gateway, www.sosig.ac.uk).
Masters and doctoral dissertations should also be revealed in RDN, etc. searches.
But, they are unlikely to be made available on-line. Some universities will – on
request – send your university library a microfilm copy (on short-term loan) which
you will be able to print at your university library using a special photocopying
machine. Where microfilm is not available, then you will have to visit the university
library where the dissertation is held and where you may be allowed to photocopy it.
Access to official records
Government and local government maintain extensive records of periodic surveys –

including the decennial census and electoral rolls – and records of policies and
decisions. Original census returns for 1841–1901 are now available for inspection
and copying at the relevant local reference libraries. The 1901 census returns can
now be examined on-line at www.1901census.nationalarchives.gov.uk/. For 1911
onwards, individual census returns are not yet available. However, aggregate data
is available from 1911 onwards at: county; local authority; electoral ward; parish
and enumeration district (ED). ED’s are the basic building block of the UK census
and consist of around 200 households. Whilst county, district and ward boundaries
change, ED boundaries can endure and therefore provide a basis for historical
comparison. Census data includes: households, household sizes, age of residents,
housing and facilities, socio-economic class, economic activity and, from 2001,
ethnicity and health. The census is supplemented by a 10% household survey that
includes information on type of employment. The 2001 census can be examined
on-line at www.statistics.gov.uk/census.
The Freedom of Information Act, 2000 enables UK residents – regardless of
nationality – to seek specific information from public bodies including: central
government; local government; NHS trusts; police authorities and advisory bodies.
You must make the request in writing (which may include email) to the organisation
which you believe holds the information. The organisation is required to answer in
20 working days (i.e. four weeks). A small charge may be made for photocopying.
However, where the cost of providing the information is more than £600 to
central government or £450 for other bodies, then the request can be refused.
Some records will be protected from scrutiny on grounds of necessary secrecy,
or the Data Protection Act. Where a request is refused, then you can appeal
to the independent Information Commissioner. The onus remains on the body
holding the information to justify not making it available. For further details, see:
www.dca.gov.uk/index.htm.
Local councils, NHS trusts and Police authorities are required to open their
meetings to members of the public. So you can attend and note what happens.
However, cameras and tape-recorders, etc. are not allowed. But you should be
issued with a copy of the agenda papers (if not, ask the committee clerk/democracy
officer). Reports will identify the relevant ‘background papers’ which you can ask
to inspect. But you will be excluded from hearing those ‘exempt items’ concerning
commercially sensitive or personal information.
The UK’s national archives are held at the National Archives for England, Wales
and the United Kingdom (formerly the Public Records Office) at Kew, London,
where they are available for public inspection. The records held cover the British
Isles, former British Empire colonies and Commonwealth states. The National
Archives includes the Family Records Centre (at Islington). A new search engine
at www.nationalarchives.gov.uk/search enables you to search all official and private
archives in the UK. Researchers do not have to register or obtain permits before
they visit Kew or Islington.
All other states maintain comparable systems. The US national archives are held by
the National Archives and Records Administration in Washington (www.archives/gov)
which also maintains regional archives. US states and city authorities have archives.
Alliances and supra-state groups also maintain records. For example, EU archives
can be accessed via ec.europa.eu/historical_archives/
UK county councils (and some unitary authorities) keep extensive local archives.
These are a valuable source of information. The professional archivists know their
records and are generally very helpful. It may be worth making an appointment
before a visit to arrange for the specialist member of staff and relevant records to be
made available. Trade unions, large companies, the churches, etc. all maintain large
archives. The county councils and unitary authorities also have specialist ‘research
and information units’ who may be able to run cross-tabulations of census data etc.
using SASPAC and other software.
The Land Registry is also available for public inspection. This holds records of
the titles, extent and ownership of all properties that have been sold. A small charge
(currently £3.00) is made for each title search. The Land Registry remains the only
current authoritative record of land ownership in the UK. However, information on
land that has never been sold – for example, the core estates of the great aristocratic
landowners – is not available.
Companies House in Cardiff holds information on all current and some past public
companies, including their annual reports and accounts. Records can be obtained
on-line at www.dupont.co.uk/companies-house.htm. A small charge (currently
£1.99) is made for each search.
Reading the literature
Whether your search has revealed a very large number of texts or otherwise, you
will have to be selective in your reading and note taking. Consider the advice given
to historian William Woodruff in his first tutorial in Oxford. He was asked to write
an essay on Inclosure. He asked his tutor which books he should read. Eight volumes
were recommended. When he asked whether it was really necessary to read all of
them, his tutor replied:
Gracious me no, Woodruff. You’d be mad to. You don’t read books; you gut them; it’s the
gist you’re after. If you feel that an author has nothing important to say, drop him. You’ll
get the nub of things pretty quickly. Anyway, you won’t find half the books I’ve given you.
There are other students preparing essays, you know … Even looking for a book you
can’t find will teach you something … what you are after is the gist, remember.
(Woodruff, 2003: 130–1, my emphasis)1
You ‘gut’ the literature by reading the abstract first. If this is relevant, then read the
introduction and conclusion and – only if necessary – individual chapters. Learn to
skim the text to identify relevant passages for slow, careful reading.
You will need to make a rolling bibliography of texts that you have read and those
that you intend to read. You will also need to make a record of those texts to be cited
in your theory and literature review. A traditional method is to record the relevant
information for each publication – author, year of publication, title, publisher’s
address and name and a summary of the main arguments and page references – on a
standard 150 × 100 mm (6’ × 4’) record card. The cards can be kept in an alphabetical
index where they can be abstracted and grouped into themes, topics or approaches.
But there is a limit to the information that can be held on a single record card.
An alternative approach is to keep an electronic record of each text in a separate
file within a bibliography folder and a hard copy. These records are, therefore, not
limited by space to bibliographic citation and key arguments. They can include key
passages identified by their page numbers. In this way, a key text can be reduced to
four pages (or less) of quotations (which, incidentally, can be traded with colleagues
for similar records of their summaries). The passages can then be copied and pasted
directly into the research report or discussion paper.
Writing the literature review
Having reduced the literature to a series of records, abstracts and summaries, you
can now begin the literature review. The trend among inexperienced students is
to write-up the literature as a series of descriptions of what each author wrote
in chronological order. Such approaches are over-long; at best overly descriptive
and, at worst, obvious plagiarism, confusing and boring. In particular, they lose the
interest of the writer and reader, and tend to give least attention to the most recent
texts. Instead, a critical synthesis is required which focuses on the ideas and arguments
rather than the authors. Indeed, I would argue that, with the exception of the ‘grand
theorists’ – Marx, Weber, etc. – the names of the (lesser) authors should generally
be omitted from the text, and given in Harvard references or footnotes. You must
give references to acknowledge the sources of information and to show where readers
can find them. They should be clear, consistent and follow an ordered structure.
A guide is given in Box 8.1.
BOX 8.1 Referencing: a guide
A reference consists of two parts: a notation or insertion in the text; and a full description of
the source.
There are two systems for notations and insertions: the footnote system and author-date
(Harvard ).
The footnote system uses a supertext notation in conjunction with footnote or endnote,
e.g.1 (In MS Word you would use the Insert/Reference/Footnote routine.) You can either
insert the footnote number immediately after the source or quotation to be referenced or
at the end of the sentence (after the full stop). The full description of the source should be
given in the footnote or endnote. When you use the number system of referencing, the full
description of the text in the footnote or endnote should follow the convention of:
Author, Initials (Year of Publication used) Title of Book. Place of Publication: Publisher.
Pages.
The Harvard system is more widely used than the numbering system. It consists of:
(Name of Author, Year of Publication used: page number) e.g. (Smith, 2007: 57). Harvard
is particularly useful where you want to acknowledge several texts and authors, e.g. (Smith,
2007: 57; Waterman, 2007: 132–5).
In all systems of referencing, you must provide a full description of the text or other source in
the Bibliography (list of sources) at the end of your research report.The convention is Author
name, Initials. (Date of Publication) Title and sub-title. Place of Publication: Publisher, e.g.
Burnham, P., Gilland, K., Grant, W. and Layton-Henry, Z. (2004) Research Methods in
Politics. Basingstoke: Palgrave Macmillan.
Where you draw your reference from an edited textbook, then the full description given
should be:
Budge, I. (1997) Chapter 1: Party Policy and Ideology: reversing the 1950s? In Evans, G.,
and Norris, P. (eds.) Critical Elections: British Parties and Voters in Long-Term Perspective.
London: Sage.
Where a journal is cited, then the reference should be:

Author of Article, Initials (Year of Publication) Title of Article, In Title of Journal, Volume,
Number: Pages, e.g.:
Przeworski, A., Alvarez, M., Cheibub, J.A., and Limongi, F. (1996) ‘What makes
democracies endure?’ In Journal of Democracy, 7 (1): 35–55.
Where you cite an Internet source then the reference given should be:
Retrieval Method, e.g. http://, host and domain name, Path and File Name (Date of Visit).
Example
In 2003, I completed a short literature review on power. The first pages are
reproduced at the end of the chapter.
Ideally, the literature should be brought together in distinct groups or typologies.
One very effective way of communicating these is to tabulate the arguments by
schools or sub-topics. In this way, you can clearly distinguish the major differences
between groups of texts. The review can therefore concentrate on the main
arguments between the schools on key topics that you can then make the subject
of detailed scrutiny. In this way, you can achieve real depth – especially in terms of
original criticism – and show your mastery of the material. But remember that you
must justify your criticism: to argue that a text is ‘fundamentally flawed’, you must
demonstrate how. Where the synthesis reveals that the arguments have developed
over the years between and within schools of thought, then this relationship can be
mapped diagrammatically. Table 8.2 in the Questions for discussion or Assignments
shows how I was able to map the development of western philosophies on property.
Remember that tables and diagrams (and photographs) drastically reduce the word
count and add interest.
The literature review can be the most difficult part of the research project and the
most likely to be re-written as you seek to reduce its length. Three drafts are likely.
At the end of the first draft, ask yourself:
• Has the literature search revealed all the main sources? How do you know?
• Has the search identified those texts that are the most relevant, authoritative and
recent? How?
• Have you ‘gutted’ the most important texts and uncovered their ‘gists’?
• In writing the literature review, have you been able to organise the texts into distinct
schools or approaches?
• Have you clearly identified the agreements between different schools, their disagree-
ments, overlaps and, crucially, gaps in the literature?
• Have you identified and evaluated the key criticism already made by commentators?
• Have you provided original, penetrating and pungent criticism?
• In your criticism, have you clearly identified what is ‘known’, contested and ‘unknown’?
• Have you clearly identified the theoretical perspective to be adopted, the contested area
or gap in the literature to be addressed in the fieldwork, and a refined hypothesis to test?
If the answer to any of these questions is no, then you must revise – or wholly
rewrite – your literature review. Where the research is taking place over several
years, then you will have to revisit your literature review to identify any new
texts and amend your review accordingly. Your bibliography will also need
revising.
Plagiarism
Plagiarism is failing to acknowledge the source of material used in your research

report or assignment, and therefore taking credit for other people’s work. It is a form
of cheating and can attract heavy penalties. Wherever you include part of another
textbook or other source, you must use single quotation marks at the beginning and
end of the passage and a full footnote or Harvard reference. Where the passage is
greater than thirty words, then it should be placed in an indented paragraph and use
a smaller font. In both cases, you may italicise the text.
Plagiarism can be detected readily by supervisors primarily because the language
is very different in style and vocabulary, strikingly error-free and is likely to be
of a higher standard than your previous work. There are many other signs. Most
universities now ask that written work is also submitted in electronic format so that
it may be searched by plagiarism-detecting software.
A similar offence is to include references to textbooks and sources which you
haven’t actually read. Once again, this is readily identifiable. Don’t do it.
Table 8.1 Property ownership: the evolving arguments
Common Ownership State God & Church God & king Private Ownership
Plato 375BC (Republic)

Plato 360BC (Laws)
Aristotle (348–322BC)
Cicero (106–43BC)
Seneca (48)
Early Christian Church St Augustine (354–430)
Early Christian Church
Early Christian Church

Rufinus (1158)
Albert Magnus (1206–80)
St Thomas Aquinas (1224–74)
William of Ockham (1285–1347)
Melachton (1521) Fortesque (1470)

Moore (1516)
Erasmus (1516) Luther (1535)
Calvin (1559)
Ponet (1556)
Grotius (1625)
Levellers (1646)
Filmer (1680) Hobbes (1651)
Locke (1690)
Rousseau 1755 Hume (1739)
Paine (1796) Smith (1776)
Jacobins (1797) Burke (1790)

Proudhon (1840) Owen (1840) Hegel (1821)
Marx/Engels (1848) Mill (1848)
Mill 1873 Maine (1873)
George (1880)
Spencer (1884)
Communism Libertarianism
Fascism Capitalism
‘Free Markets’
‘Property-Owning Democracy’
1. You have been asked to undertake a literature review on ‘social exclusion in the
UK’ (or another topic set by your teacher). Describe how you would undertake the
literature search and the priority to be given to particular sources. How would you
‘read’ the relevant texts? What records would you make of the key texts and how?
Describe the structure of your literature review.
2. Select a literature review from a journal article. Review it critically. How would you
improve how it was written and presented?
FURTHER READING
Hart, C. (1998) Doing a Literature Review: Releasing the Social Science

Research Imagination. London: Sage. This 230-page textbook remains the
premier guide for Politics researchers seeking to take a disciplined approach to
their literature reviews. It provides practical guidance on how to: search out
existing literature on a topic; effectively analyse arguments and ideas; and map
out arguments. Hart advocates the use of network diagrams to map out the
development of ideas and philosophy by different schools and authors. He
provides a good example of DNA development on p. 169. I followed his advice
to demonstrate how different and competing concepts of property had
developed in western society since Biblical times. This is shown in Table 8.1.
BOX 8.1 Example of a literature review
Power
Introduction
Power is a universal experience. Sometimes it is recognisable, tangible, noisy. At other

times, it is silent, barely discernible, scarcely felt. So, what is power ?
Lukes famously described power as an ‘essentially contested concept’(Lukes, 1974: 9). But
power might better be called an ‘essentially contestable concept’. It is not ‘under-theorised’:
there are many theories and theorists. They include Hobbes, Hume, Machiavelli, Marx
and Gramsci. More recent theorists include Bourdieu, De Jouvenel, Foucault, Clegg, Mann,
Wrong and the authors of what have been called the three dimensions of power : Dahl;
Schattschneider; Bachrach and Baratz; and Lukes. But whilst there are many theories,
most are either partial or offer views from different perspectives which reflect the normative
values or disciplines of the authors. Given the proliferation of theories and perspectives,
then exceptionally in this dissertation, their authors are identified in the text. Arguably, the
perspectives are also culturally and historically eurocentric (Foucault, 1973). But many of
the conceptualisations are supportive or compatible. For example, most support the view
that power is relational and relative. There is no absolute power. There is usually some
resistance – conscious or unconscious, overt or covert – to the exercise of power. There is
also complicity. Power is contingent. It is contextual. It is also fluid. It is generally a means to
objectives rather than an end in itself (which raises the question whether all means of attaining
objectives are forms of power). Power is a generalising medium, a characteristic akin to –
but different from – a resource and money (Parsons, 1963: 232–62). But the argument that
power, like money, is a ‘circulating medium’ (Pareto, 1935; Parsons, 1963; Habermas, 1987)
is not widely supported.
Power can be held and exercised by and between individuals, groups of individuals,
by organisations (‘embodied institutions’), and states (geopolitics). Power can be either
extensive, comprehensive or intensive (De Jouvenel, 1958: 159–69) where, as extensiveness
increases, comprehensiveness and intensity fall (Arendt, 1951). Power is exercised within
realms or scopes (Wrong, 1997: 10). Society is stratified by power where the strata reflect
the distribution of power.
Power may be distributive (conflictual – involving zero-sum outcomes), or collective
(co-operative, non-zero-sum outcomes), (Parsons, 1963). Distributive and collective powers
are dialectical : organisation is the manifestation of collective power which is inherently
oligarchic, where leaders exercise distributive power over members (Michels, 1949; Lowi,
1971; Mann, 1998). One consequence of collective power is that, paradoxically, individuals
become subordinated by the delegation they extend to officials (Bourdieu, 1999: 107–16).
However, the individual rarely belongs to only one organisation (Truman, 1951: 508).
Similarly, organisations rarely act alone: organisations are ‘functionally promiscuous’:
‘organisational outflanking’ is a common strategy contributing to the turbulence of power
relations (Mann, 1998: 7).
Collective power is the product of size (membership), solidarity and organisation (Wrong,
1998: 237). Solidarity derives from mutual identity, consciousness of kind and rituals of
belonging. But, as membership increases, intensity of common identity may diminish and
‘organisational diseconomies of co-ordination’ may multiply. The organisations created are
‘living machines’, sui generis characterised by their own objectives and internal conflicts
among members and officials (Weber, 1947: 140). Political mobilisation is the ultimate form
of collective power – where the organisation seeks control of the state (Wrong, 1998: 144).
Alternatively, parties or unions are seen as ‘class persuaders’ providing the structural cement
Continued
of ‘class reality’ rather than as an organisational expression, class or other groupings (Sartori,
1969 in Wrong, 1998: 167).
By most accounts, power over is a special (distributive) type of power to (of which
power against is a defensive type of power to). There is also general agreement on
the bases and forms of power. The bases – sources or domains (Dahl, 1961), power
resources (Wrong, 1997) – include ‘collectivity’ (the greatest), wealth, reputation, fear,
respect, success, nobility and excellence (according to Hobbes, 1651/1996: 58–59). Others
have added legitimate position derived from office, personal appeal, skill or specialised
knowledge (Dahl, 1961). Additional power (re)sources include property, class, credentials,
capacity for reasoning, information and knowledge, control of discourses, historiography
and discursive media (Foucault, 1978), ‘systematic luck’ (Dowding, 1996: 170), speech,
bearing, accent, clothes, how the mouth is pursed (Bourdieu, 1999: 17) and legitimacy –
moral, traditional or rational-legal (Weber, 1986). Thus power can be ‘original ’ (intrinsic)
or ‘instrumental ’ (acquired) (Hobbes, 1651/1996: 58). However, postmodernists argue that
knowledge, rather than being a source of power, is its product (Foucault, 1978).The argument
appears compromised by the secrecy of the modern state. Alternatively, the counter-
argument can be made that the state gathers, screens and guards the data from which
knowledge is fabricated to reinforce its dominant discourse; the ‘stem’ of the word statistics is
state.
Authority is problematic. The essential characteristic of authority is the general approval
and acceptance of those over whom it is exercised (Weldon, 1953 in Friedrich, 1958: 31).
But some argue that authority can only be derived from credentials or received wisdom
which confer ‘evidence of a capacity to reason, weigh alternatives and exercise delegation’:
authority is a source of power (Friedrich, 1958: 31; De Jouvenel, 1958: 159–69).
They believe that the quality of communication is the ‘essence of authority’ (Friedrich,
1958: 34–7). Thus authority can be exercised without power, e.g. Roman Senate, or
vice versa, e.g. Nero (Friedrich, 1958). Context may be instrumental (Bourdieu, 1999).
Most regard authority as a special form of power (Wrong, 1997) inseparable from
legitimacy (Weber, 1986). Others regard legitimacy as consent to authority in which
authority is a legitimate use of power and therefore different from power (Giddens,
1998: 339).
Power is a universal experience. Most people exercise power over others at some time –
especially as parents – and are also the subject of others’ power (although they may not
recognise its source or influence). Perhaps this universality explains the broad compatibility
of distinguished definitions: ‘an intuitive idea of power, is … A has power over B to the
extent that he can get B to do something that he would otherwise not do’ (Dahl, 1957:
201–05); ‘the probability that an actor within a social relationship will be in a position to
carry out his own will despite resistance, regardless of the basis on which this probability
rests’ (Weber, in Lukes, 1986: 28–36); ‘the capacity of persons or collectivities to “get
things done”’ (Parsons, 1963: 232); ‘the production of intended effects’ (Russell, in Lukes,
1986: 19–21 ); ‘the capacity to achieve outcomes … to make a difference’ (Giddens, 1984
in Clegg, 1984: 138 ); ‘the capacity to produce intended and foreseen effects in others’
(Wrong, 1997: 2);‘the ability to pursue and attain goals through mastery of one’s environment’
(Mann, 1998).
Power is, therefore, widely held to be primarily a capacity. Power can thus be
dispositional (capacity) or episodic (exercise of power) (Ryle, 1949, In Clegg, 1998: 73).
Intentionality is considered by many commentators to be a requisite of the exercise of
power (Russell, 1986: 19–21; Wrong, 1997, In Clegg, 1998: 73). Others argue that a ‘law
of anticipated reactions’ is evident where subjects anticipate and comply-in-advance which
makes redundant the exercise of power (Friedrich, 1937). This phenomenon challenges
the classical notion of power as necessarily observable in the sequence of stimulus-
delay response – ‘push-and-shove’ (Hobbes, 1691; Dahl, 1957). Intentionality is therefore
contested.
Power can be negative (prohibitory) or positive (especially Foucault) or both. But the
issue of overt and covert power is contested. Behaviouralists argue that only the observable
exercise of power is verifiable and substantive (Dahl, 1957). Others criticise this view as
‘one-dimensional’. They argue that power has ‘two faces’ and can also be exercised unseen
by non-decisions and by preventing issues from reaching agenda (Bachrach and Baratz,
1962: 947–52 after Schattschneider, 1961). Those with power determine both the rules-
of-the-game and which game is played (Schattschneider, 1961). Some argue that power
can be exercised unseen and unknown through hegemony - the cultural domination of civil
society by discourses which lead subjects unquestioningly to carry out activities which are
contrary to their interests (Gramsci, 1935 In Forgacs, 1988: 192; Lukes, 1974; Foucault,
1978). Lukes termed this: ‘three-dimensional power . . . the most insidious use of power . . .
by influencing, shaping and determining the very wants’ of those over which it was exercised
(Lukes, 1974: 23).
But others argue that three-dimensional power is unachievable (‘How is it done? With
mirrors?’ Clegg, 1998: 165). This challenges a major tenet of psychology that the major
determinants of personality, identity, attitudes and behaviour are the family, the state and
the economy. The (arrested) development and resource-dependence of children on the
institutions of parents and school makes them subject to exceptionally long periods of
control. However, ‘brain-washing’ has no enduring efficacy and may be a ‘coping strategy’
(Brown, 1983: 24). Foucault alternatively argues that control of thought and attitudes
is first achieved through control of physical behaviour, e.g. children being seated at
desks, soldiers marching, etc. (So, forms of dress which surround the body – ties and
rings – are, after Saussure, signs of others’ control.) Supporters of three-dimensional
power argue that this can be evidenced – paradoxically – through counter-factuals (Lukes,
1974) or counter-actuals (Dowding, 1996), e.g. Conan Doyle’s ‘dog that didn’t bark’ in
Silver Blaze. In any event, the consequences of the intended exercise of power will
generally have unforeseen or counter-intended consequences (Friedrich, 1937; Foucault,
1978).
Continued
The dissension concerning hegemonic or three-dimensional power arises from opposing

views on the existence or otherwise of objective interests or moral relativism and the
incidence of internalisation of (other) norms (Gramsci, 1971; Lukes, 1974). Critics argue
that there are no real or objective interests but only alternative interests or discourses
(Clegg, 1998; Foucault, 1978). Alternatively, hegemony is regarded as more akin to business
leadership than cultural domination (Whitt, 1979: 81–100). The paradox of emancipation
describes how the oppressed are so propagandised as to be incapable of any emancipation
(Benton, 1981: 161–84). These arguments challenge the classical view that interests are
derived primarily from needs (after Smith, 1776). They are universal and absolute in a
hierarchy extending from survival to self-actualisation (Maslöw, 1954: 80–106). Alternatively,
the argument is made that needs and, therefore, interests reflect the subject’s existential
perception of dominant norms and experience which are themselves the product of dominant
discourses (Bohin, 2002). Hence, self-sacrifice – rather than self-preservation – can counter-
intuitively become the highest form of self-actualisation (e.g. Sartre’s The Roads to Freedom,
1950). Hence also the prisoner’s complicity in his own incarceration because of his learned
helplessness. Prisoners and guards are incarcerated alike by Bentham’s panotpicon and by
themselves (Foucault, 1980: 156).
There is more agreement on the forms of power. These are generally accepted to include
force, persuasion, manipulation and (among those who do not see it as a source of
power) authority (Wrong, 1998). Force may be physical or psychic (ritual degradation) and
violent or non-violent. Manipulation is ‘hidden persuasion’. The power-holder’s intentions
are concealed or disguised; the subjects believe that they are exercising a free choice. In
contrast, persuasion is overt involving appeals and arguments. It is less likely than force to
provoke resistance by the subjects. Whilst persuasion involves the tested acceptance of the
power-holder’s judgement, authority involves the untested acceptance of the power-holder’s
views: source and reputation are significant determinants. The outcome is identification –
which is likely to be ephemeral – rather than enduring internalisation. Authority is power ex
cathedra. In this account, authority rests on the subject’s acceptance of the power-holder’s
claim to legitimacy (Wrong, 1998; Weber, 1986).
The sub-types of authority (as power) are: coercive authority, induced authority, legitimate
authority, competent authority and personal authority. Coercive authority is the most
extensive form of power. It depends on the subject’s conviction of the power-holder’s
capability and readiness to use force. It is a major component of political power and of
state institutions who have (only comparatively recently) monopolised control of the forces
of coercion. Both the Mafia and terrorists challenge the state’s monopoly of violence.
States emphasise the growth of organised crime to justify the growth of (their) state-
wide forces of coercion (Foucault, 1999). But coercion stimulates both opposition – real
or imagined – and development of systems of surveillance. Induced authority involves
offering inducements to subjects. Whilst there is little or no resistance initially, diminishing
marginal utility (‘law of diminishing returns’) progressively undermines the effectiveness
of the inducement. Indeed, the practice may provoke resistance if the value of the
inducement is not raised and especially when inducements are withdrawn. Hence the
combination of coercive and inducive authority used in the form of ‘throffers’ (Dowding,
1996: 56–8).
Legitimate authority requires a bilateral acceptance by power-holder and subject of the
right-to-command and duty-to-obey, whereas competent authority (by this account) arises
from the subject’s belief in the superior knowledge or decision-making of the power-holder,
e.g. Socrates’ patient-doctor relationship. Personal power is unique to the power-holder and
subject; its bases are love and admiration or predispositions to dominate or submit. A special
type of extensive personal power is charismatic authority (Weber, 1986). Arguably, all forms
of authority are overlain by personal authority which power-holders craft to bolster their
power.
Each form of authority is unstable and subject to decay or de-mystification. Power-
holders must develop strategies of metamorphosis and diversification.The conqueror’s initial
coercive authority metamorphoses into legitimate authority as the conquered adjust to the
new power-holder. Arguably, the adage that ‘political power grows out of the barrel of a gun’
(Mao, 1938/72: 61) refers principally to the long-term consequences of coercion. However,
acceptance by subjects of the power-holder’s legitimate use of force is greatest when they
are likely to be unaffected by its use and where its use will be directed against ‘others’. The
professor who gives an incomprehensible lecture increases his academic authority (Foucault,
1999).
Power can be considered as a type of influence or, alternatively, all influences can be
considered as power. Understanding is obscured by the ‘polyvalence’ of words (De Jouvenel,
1958). Power has no transitive verb. Whilst its ultimate causality – agency or structure – may
be contested, its ultimate purpose must be the achievement of the power-seeker’s mastery
of the present and future environment – in its widest sense – social, economic, political,
technical, physical. Arguably, the ultimate power is social power – ‘mastery over other
people’ and, thus, their resources (Mann, 1998: 6). The ultimate means is through control
of the state which territorialises, institutionalises and regulates its population (Mann, 1998)
(although others have seen the state as the instrument of international capitalism (Marx)
and globalisation (Cerny, 1999). Power, unlike money, is hierarchical (Parsons, 1963). The
state stands at the intersections of ideological, economic and political systems and has
sole control of coercive military power (which has internal and external application). Until the
late eighteenth century, ideological (especially religious) and military power were dominant in
Europe (Mann, 1998). Now economic and political systems of power prevail. States have both
despotic power and, increasingly, infrastructural power – especially through state education
(Mann, 1998).
Who exercises the power of the state is contested. Pluralists argue that power is either
widely diffused within the systems or, where elites develop, no single elite dominates
(e.g. Dahl. 1961/89:). Elitists now emphasise the distributive power of a centralised elite
Continued
and the geopolitics between states. Institutional statists argue that states – not elites –
institutionalise social conflict in organisations which develop autonomous interests which
structure future conflicts (e.g. Skocpol, 1979 in Clegg, 1998: 259). The institutions provide
‘regulatory passage points’ (e.g. Parliament) in ‘fields of force’ (Lockwood, 1964: 244–57).
Alternatively, these views are rejected because they fail to recognise the disunity of
state institutions, confederalism, the ‘empty centre’, globalisation and ‘polymorphism’ (the
way states crystallise at different centres in power networks). Instead a cock-up/ foul-up,
‘organisational materialist’ model of the state is argued (Mann, 1998: 41). However, these
views embrace a variety of normative and descriptive viewpoints of a range of divergent
states.
These views also incorporate polarised perspectives of where, ‘in the final analysis’,
power lies and who initiates power. Modernists, after Hume and Hobbes and extending
to Dahl and Lukes, have developed primarily mechanistic concepts of power and causality.
They emphasise individual agency and autocracy and accept the centralising power of the
state through the social contract to achieve protective social order. Power is ‘zero-sum’; ‘the
negation of power of others’ (Clegg, 1998).
The alternative view is that structure prevails. The baize over which Hume’s billiard balls
move is already marked by folds which channel them along pre-determined pathways. The
modern individual is ‘institution-ridden’ and ‘institution-supported ’ (De Jouvenel, 1958: 164).
There was no original social contract. Marxists argue that the interests of the dominant
mode of production will prevail. The lives of people are ‘written’ by those who frame the
dominant discourses and exploit the social sciences to create ‘bio-power’ and systems of
surveillance (Foucault, 1978, 1995). Self-surveillance is the most effective. ‘They’ construct
nodal points which fix meanings essential for political articulation (Laclau and Mouffé, 1985).
Individuals are also caged by the past: ‘Men make their only history … under circumstances
directly encountered, given and transmitted by the past’ (Marx, 1851/1977: 300). They are
also caged by whoever constructs that history: ‘Who controls the past controls the future;
who controls the present controls the past’ (Orwell, 1949/1989: 37).
Alternatively, agency and structure ‘interpenetrate’ to form a ‘duality of structure’ in which
power can be analysed as ‘relations of autonomy and dependence between actors in
which they draw upon and reproduce structural powers of domination’ (Giddens, 1981: 29).
Within organisations, power relations reflect (mutual) resource-dependency, e.g. work-pay
(Mintzberg, 1983: 172). Whilst these and other power relations are asymmetric, subordinates
can extract benefits by emphasising the reciprocal obligations of power relations, e.g. the
‘deferential (farm) worker’ (Newby, 1977: 381–432).
Note
1 Woodruff, W. (2003) Beyond Nab End, London, Abacus.

Chapter 9
Asking Questions: Effective Elite
Interviews, Other Interviews,
Vignettes, Projective Questions,
and Focus Groups
1. To distinguish between the different types of questions and structures of

interviews.
2. To understand who are ‘political elites’, why we interview them and how to
secure an appointment.
3. To consider how the interview should be carried out and recorded.
4. To learn how interviews with other people differ from those with political elites.
5. To learn what is meant by ‘projective questions’ and when they can be used.
6. To consider why, when and where ‘vignettes’ can be used and how they can
be framed.
7. To understand the value of focus groups and other group meetings, and how
to organise and facilitate them.
Questions
Questions are a means to an end: obtaining information in the form of answers.

The answers are rarely straightforward. The subject (interviewee, respondent) may
be evasive. You may also want to question the answers which you have received.
Asking questions is central to both qualitative and quantitative research methods.
They are used to corroborate background and secondary sources, and to collect
new primary information. While face-to-face questioning is, of course, ideal, it is
not necessarily essential. Indeed, physical access and security considerations may
prevent on the spot interviews being held. Many overseas students on MA courses
are unable to access directly elites or others in their own countries to provide data for
their dissertations. So, in order to replace or supplement face-to-face questioning,
increasing use is made of the indirect telecommunications media. Traditionally,

researchers have used surface mail and, to a lesser extent, telephone.1 Now, email
or voice-over-internet protocol (VOIP) is widely used. In practice, a combination
of media is employed. They all begin with the researcher establishing their bona
fides2 as an academic researcher (rather than say, a ‘muck-raking’ journalist posing
as a university researcher) and setting out fully the objectives of the research, its
significance and the contribution that the subject (also termed ‘interviewee’ and
‘respondent’) can make.
Questions can be categorised across a continuum with closed questions and open
questions at each pole. Closed questions essentially seek – or receive – closed
answers. Closed answers are generally short and confined to yes, no or don’t know
or specific answers, e.g. date of birth. Open questions – seek open, lengthy
answers. They are concerned with why and how, beliefs, opinions, forecasts and narratives
(i.e. stories, biographies, etc.). Another similar categorisation of questions concerns
the structure of the interview. Sets of questions constitute interviews. They can
be classified on a continuum of highly-structured to unstructured encounters. In
highly-structured interviews, you must ‘follow’ a tight pre-designed schedule of
questions. Where comparisons are necessary, then you must ask the same questions to
different subjects in an identical way in terms of wording, inflexion and other aspects
of delivery. At the other extreme, the entirely unstructured interview will follow
the course of general conversation in which the next question follows from the
preceding answer, and where the roles of questioner and respondent interchange.
In Politics research, the most widely-used type of interview – especially in elite
interviews – is the semi-structured interview. Here you use a schedule of a limited
number of topic-related questions and, pre-determined, alternative supplementary
questions (which question further aspects of the answer received). For example, you
may ask: ‘and how did you feel about this?’ The format of the semi-structured interview
is essentially one of question-and-discussion. It is the type of interview adopted in
news reviews where both interviewer and subject have their own agendas.
In practice, most productive interviews follow the pattern of good conversations
in which rapport (mutual empathy and confidence) develops. You will find that,
where rapport is achieved, your subject will share their intimate thoughts and
feelings. They therefore make admissions which they might otherwise withhold.
Such interviews will generally range across the continua described.
Generally speaking, qualitative researchers tend to ask open questions in semi-
structured interviews of a small number of people. They may ask their questions
of individuals or groups. In contrast, quantitative researchers are more likely to ask
closed questions in highly-structured interviews of large numbers of respondents
one-at-a-time. The most common type of quantitative interview uses questionnaires
which may be administered (carried out) directly (face-to-face), indirectly (voice-to-
voice) or remotely (by self-completion questionnaire via post or email). Advice on
questionnaire design is provided in Chapter 10.
Asking Questions 119
Elite interviews
Interviews are generally defined as: ‘conversations with a purpose’ (anon). Interviewing
members of the political elite – often shortened to ‘political elites’ – as a routine
activity is unique to Politics research. But it can prove difficult, especially for
first-time or young researchers to gain access to the most appropriate elite or
useful information from the encounter. Whilst gaining access to the most senior
elites may better serve your research interests, you may well be overawed and
inhibited by the combination of their personal presence, the presence of their
‘minders’ and the surroundings (which are often designed to create a sense of
power and prestige). Effective interviewing requires sound preparation, planning
and confidence. So class-based simulation and individual practice can be very valuable
to prepare you for the interview. Remember that the best way to learn is often from
other people’s mistakes in class exercises.
The whole process of arranging and conducting the interview is bound by
institutions (traditions, customs and rules) that vary significantly between states and
cultures. For example, officials in the People’s Republic of China will rarely agree
to be interviewed unless their superior has given formal approval. Even so, they
are obliged to re-state official policy. So, before you seek to interview elites from
a different state or culture to your own, find out – from a fellow student from that
country or from your diplomatic service – what the institutions are, how they have
developed and what purposes they perform. The advice given below is written for
students seeking to interview UK political elites. However, much of the advice will
apply to students seeking to interview elites in other western states.
Who are the members of UK political elites?

Arguably, political elites are people who exercise disproportionately high influence
on the outcome of events or policies in your research area. They may be ministers,
MPs, senior civil servants, business leaders, union leaders, members of think tanks
or financial institutions, learned commentators, journalists, local councillors, chief
executives, ‘gatekeepers’ etc. They may influence outcomes without becoming
directly involved through what Friedrich (1937) termed the law of anticipated
reactions.3
Why do we interview them?

• to confirm (our understanding of) documentary material
• to fill gaps or clarify grey areas
• to check ‘Have I got the story right?’
• to try to understand their perceptions, beliefs, mindsets, i.e. underlying psychology
• to obtain ‘quotable quotes’
• to help identify other actors involved

• to identify networks
• to facilitate: to open the way to interviews with others
• to triangulate, i.e. to corroborate (check, test) accounts from other interviews
Triangulation is often best achieved by interviewing the subject’s ‘opposite number’,

‘shadow’ or ‘parliamentary pair’4 (who is likely to know the subject well and
may provide a penetrating analysis of personalities and precipitating events), a
commentator (e.g. journalist) and a subordinate. Retired individuals are likely to
be more accessible and open but not necessarily more honest: self-justification –
often sincere – is ever-present. Remember what Enoch Powell said: ‘all political
careers end in failure’. You will find that few political elites appear to admit their
failures.
Why do they agree to be interviewed?

• because the potential benefits are attractive and outweigh the costs (indeed, you can ask
them why they agreed to be interviewed)
• because you have persuaded them of the importance of your research and their potential
contribution
• because it-goes-with-the-territory and underlines their status
• because their family and colleagues, etc. have ‘heard it all before’
• because they ‘want to set the record straight’
• because they welcome the opportunity to be reflective
• the rules-of-the-game have been agreed in advance.
The rules of the game for interviews cover whether the interview will be on a one-
to-one basis, whether the replies will be treated in confidence, attribution, circulation
of drafts, publication and recording. Subjects may ask for advance (written) notice
of questions to be asked. Paradoxically, there is some confusion as to what the
various terms ‘in confidence’ etc. actually mean. You are recommended to adopt a
conservative interpretation unless the subject agrees otherwise. Here, in confidence
means what it says: that whatever is said remains a secret between the interviewer
and subject. Off-the-record similarly means that no recording or note is kept
of the interview and that the information is provided to enable the researcher to
understand the context and background better. On-the-record means that the
comments can be recorded, used and attributed to the speaker. Subjects may switch
between ‘off the record’ and ‘on the record’ comments. Attribution means to
whom the comments are attributed. Unattributable means that the source is not
identified. However, agreement may be sought to attribute comments to a ‘former
minister’ to underline the authority of the source. Alternatively, sources may be
anonymised, e.g. ‘a northern Labour MP, A argued that Minister B’s comments had
been reported out of context’.
How should you first approach them to arrange

an interview?
The approach will depend upon the cultural context. Always follow local conven-
tions and protocol. In the UK, the procedure is generally:
1. Select the ‘target’ political elites on the basis of their recorded involvement in your
research topic. Do not rely on any one elite providing you with an interview or all the
information: adopt a ‘multi-track strategy’. Aim high rather than low by approaching the
‘A-list’ elites rather than lesser elites whom you may consider to be more approachable.
2. Access personal details from the Internet or parliamentary etc., directory. Telephone
their PA (personal assistant or secretary) to confirm their full title and preferred form
of address (e.g. Dr, Mr, Mrs, Ms, Minister, etc.). Then write a personal letter on your
Departmental University letterhead. Ask for their ‘help’: ‘help’ is a so-called ‘hook word’ to
which people have become conditioned to respond positively. Describe the importance
of your research and their potential contribution. Ask for an interview. Emphasise that the
research is for academic purposes. Reassure them that their comments can be treated
entirely in confidence, off-the-record or, alternatively, not individually attributed if they
wish. Ask them for an interview preferably before … Add how grateful you will be and end
by saying how much you look forward to meeting them. Offer a contact and reference
(your research supervisor) so that their staff can authenticate your identity and status.
Say that you will telephone their PA on a specific date to agree a time and date for an
interview. Describing your intention to call on a specified date means that the PA – who
will normally screen and ‘triage’ their mailbox into ‘immediate’, acknowledge/pending or
‘acknowledge/bin’ – will have to classify your letter as ‘immediate’ to seek instructions
(or give advice) whether the interview is to be granted or not. Retain the initiative: don’t
expect them to contact you first. When you telephone, you may be told that the elite is
unable to see you. In this case, ask why in a casual way: the PA may let slip information
that may be useful, e.g. ‘Oh, he wants to forget about that: it wasn’t his finest hour’.
Paradoxically, a non-interview can be a good source of useful material. An example of
an initial letter is shown in Box 9.1.
3. When you speak to the PA, try to obtain a late-morning appointment. This will allow your
elite interviewee to extend a ‘good’ interview by inviting you to lunch where you are likely
to be introduced to other contacts. Confirm the appointment in writing (email will suffice).
4. If you are unable to speak to the elite or their PA after several attempts, write again.
Attach a copy of your earlier letter. Be gently persistent. Exploit contacts and leverage.
For example, if the subject is an MP, attend their constituency surgery. Or ask your own
BOX 9.1 Example of introductory letter
UWUK
Department of Politics
University of Watersea
Watersea
WA3 7HU
Rt Hon Tina Atkins MP, PC

House of Commons
London SWIA 0AA
Dear Ms Atkins
I’m writing to ask for your help.
I’m a local government practitioner studying for an MA in Urban Policy at the University
of Watersea. The subject of my dissertation is policy change theory. I’ve chosen ‘public
participation in the National Health Service’ for my case study. My special interest is the
apparent contradiction between the public’s readiness to comment as patients on their own
experiences of NHS services, and unwillingness to become involved as members of the
public in broader consultation exercises. I know that, when you were Minister of State at the
Department of Health (2005–6), this was a specific area of your responsibility. I understand
that, despite the successful introduction of YCTP, your wider ambitions were frustrated by
(what the Guardian described, 22 August) as ‘the opposition to reform from the medical
profession and leading foundation hospital trusts’.
I would therefore welcome an opportunity to meet you in London to learn at first hand
of your experiences and insights of promoting policy change. Your contribution will be very
significant. I can assure you that information gained will be used for academic purposes
only – although I hope that my research can be published in Policy Research Quarterly.Your
contribution can – if you wish – be entirely ‘off the record’ or ‘unattributable’.
I will phone your PA on Wednesday to – hopefully – arrange a date, time and place
where we can meet. Ideally, I would like to see you in June–July and, in any event, before
12 August. I look forward to meeting you.
Yours sincerely
Ted Smith
Email address: ted.smith@watersea.ac.uk Mob: 07790432111

Authentication (if required): Dr Mona Lott (supervisor) Tel: 01903 366 042 (direct line)
MP to arrange an interview. Ask your MP or local councillor to help set up meetings with
reluctant civil servants or local government officers. Remember that serving Ministers,
civil servants and local government officers are likely to be less willing to be interviewed
than others because of the heavy demands on their time and the risk of loss-of-office
from unanticipated disclosure. They will want to ensure deniability of any material that
inadvertently reaches the public domain. So only one researcher will be allowed to attend
the interview. Retired ministers, MPs and officials are likely to have more time on their
hands and may welcome the renewed attention and the opportunity to set the record
straight. But, whilst they may be less discreet – especially about former colleagues –
they will not necessarily be more truthful.
How should you get the best out of the interview?

• Prepare well in advance. Research the subject using the latest edition of Who’s Who in
the university or public reference library (or local equivalents), Internet search or seek
a CV from the subject’s PA. Look for points of common interest (e.g. birthplace, school,
university or support for a particular football team) and publications for prior reading.
Read the opening chapter and conclusion of their latest publication. Also use the Internet
to search for recent press coverage.
• Find a photograph of the elite so that you recognise them at first sight. Having written to
say how much you admire their contribution to your topic, it can be very embarrassing to
be unable to identify them in a crowded room.
• Look and dress the part, i.e. conform to the setting. Observe the setting closely. What
can we learn from the subject’s choice of books or CDs? Which books look well read?
Who are in the photographs? Why does C have photographs on their desk of pets
rather than the customary partner-and-children? What image is the subject trying to
convey by their habitat: gravitas or person-of-the-people? Conspicuous consumption
or thrift?
• Come prepared with a semi-structured interview proforma, an audio-recorder and
clipboard. Politicians generally welcome a recorder as a symbolic acceptance of their
authoritative, recordable voice. It will also enable eye contact – to transfix you with their
gaze – to be maintained. But civil servants and local government officers will decline
(refuse) to be recorded: they want to retain deniability. If the interview is particularly
important, use two recorders.
• Try to avoid an eyeball-to-eyeball arrangement – which is more likely to become an
adversarial interrogation. Instead, seek a more relaxed setting where you and the subject
are seated at an angle (see Figure 9.1 below). Ideally, place the recorder to one side so
that you and the subject can ignore it.
• Accept the hospitality provided. The offer of coffee is a social convention in an office or
home environment. Don’t insist on Fairtrade decaffeinated. Adopt normal good manners:
expect and reply to questions about the weather and travel. ‘How are you?’ is a form
Avoid if possible Seek . . .
Figure 9.1 Alternative interview seating
of greeting to which the expected reply is ‘fine, thank you’ (rather than a recital of your
current medical conditions).
• Follow Elton Mayo’s advice:
1. Give your whole attention to the person interviewed, and make it evident that
you are doing so.
2. Listen – don’t talk.
3. Never argue; never give advice.
4. Listen to:
(a) What he wants to say.
(b) What he does not want to say.
(c) What he cannot say without help.
5. As you listen, plot out tentatively and for subsequent correction the pattern
(personal) that is being set out before you. To test this, from time to time
summarise what has been said and present for comment (e.g. Is this what
you are telling me?) Always do this with the greatest caution, that is, clarify but
do not add or distort.
(Mayo, 1949: 64)5
• Seek to develop and maintain rapport.

• Respect silence. What you are seeking is profound disclosures. Use the psychologist’s
prompts to help the subject cross barriers, e.g. ‘go on . . . ’, ‘can you tell me more?’, ‘how
did you feel?’, etc. Act as a mirror. Don’t answer the subject’s questions, reflect them.
Ask questions like the psychiatrist Professor Anthony Clare did – not a sneering Jeremy
Paxman (BBC2 Newsnight’s feared interrogator). So don’t retort (like Paxman): ‘Surely
you don’t expect me to believe that, do you?’ But don’t be over-eager to ingratiate yourself.
And don’t agree with (and thereby reinforce or lend credence to) irrational prejudices,
fantasies or self-beliefs.
• Try to maintain eye-contact. Minimise note-taking. If you take verbatim notes then you will
be unable to study the subject. They are also likely to slow down their speech so that it
becomes too well-considered.
• Don’t be intimidated. Be gently assertive.
• Watch the subject’s body and look out for differences between what the subject is saying
and what is being communicated in body language. Touching the face around the mouth
or averting gaze can indicate a less-than-truthful response. Conversely, much waving
of open palms – a sign of openness and honesty – may be more evident of a learned
attempt to simulate sincerity (see Chapter 11).
• Retain control (which is difficult with authority figures).You must terminate the interview –
generally ten minutes before the hour likely to have been set aside. This early end is
likely to lead to apparently unimportant comments by the interviewee which may be very
significant. (GP’s [UK family doctors] are told that the most important remarks by patients
are made when they are opening the door to leave the consulting room.) Conclude by
asking if there is anything that you should have asked but missed out and what would
have been the subject’s reply.
• Write up your field-notes as soon as possible after the interview. Accurate recall fades
quickly (which is why wartime bomber crews were ‘de-briefed’ immediately after they
landed, even after very long flights).
The structure of the elite interview

Generally speaking, a successful elite interview will last 50 minutes and consist of
five open questions and five supplementary questions. Your aim is to obtain depth –
to reach the subject’s innermost and private thoughts (of which they may perhaps be
unaware). I call this ‘level 5’. The interview can therefore be likened to a funnel. So
the sequence of questions should ideally begin from the inoffensive ‘level 1’ question
through levels 2, 3, 4 before level 5 is reached. What you are trying to do is to win
the subject’s trust and interest before asking the most intrusive questions. But you
may never reach level 5. It may be necessary to return to a less intrusive question
when you sense resistance. By the same token, you may find that the obvious rapport
established early in your meeting enables you to ask more difficult questions from
the start.
Consider the example of the interview with former minister, Tina Atkins about
her involvement with the government’s programme for greater public participation
in the NHS (National Health Service). What you are really trying to find out is
whether the government’s apparent commitment to greater public participation
is genuine, or (as you suspect) a stratagem to counter the power of the medical
professions and hospital managers who oppose further reform. But, if you ask
that question first, she’s likely to repeat government policy about greater public
involvement in shaping government policy. So a possible sequence of questions and
supplementaries might be:
1. Minister, your (Labour) party founded the NHS 60 years ago. It has been described
as the ‘jewel in the crown’ of your party’s record in government. More recently, your
government has substantially increased the funds provided for health services. What do
you consider has been the NHS’s greatest success?
This is a non-adversarial question which is designed to flatter, to place the government
minister at ease and to give her the opportunity to be enthusiastic and, thus, open.
Her answer may well cite the success of NHS services for children where mortality
rates have been drastically reduced. Your supplementary question may therefore be:
Why do you think that the NHS has been so successful with children’s services?
2. Many GPs and hospital consultants objected to the founding of the NHS. The minister,
Aneurin Bevan famously said that he had obtained the consultants’ agreement by
‘stuffing their mouths with gold’. Despite very generous new contracts, many doctors
remain critical of the NHS. How best can the government win their support? Or – as
Denis Healey said – is acquiescence sufficient?
(Possible supplementary) What lessons have you learned from the ‘new contracts’?
3. Demand by patients for emergency and other health services continues to grow as do
the costs of new drugs and equipment. Commentators have referred to a ‘crunch time’
when either services will have to be rationed, or taxes or charges increased. What’s your
view of the future?
(Possible supplementary) How can your own MPs’ historic opposition to increased
charges be overcome?
4. The government favours greater delegation of decision-making to local NHS trusts and
greater public participation in the process. But this can lead to different practices being
adopted in different areas which the news media describe as a ‘postcode lottery’. The
government also sets national policies in the form of ‘national frameworks’ specifying
standards of care. Given this complexity, what therefore is the real scope for public
participation?
(Possible supplementary) I think I understand. But then why did the government scrap
the Community Health Councils and the replacement Forums for Patient Involvement?
5. A final question: some critics claim that the government’s new support for greater
participation by the public, is merely a cynical ploy to disguise its necessary strategy
of wholesale reform of the NHS. They say that you enlist the support of the public only
when it suits you, and then as a counter-balance in a ‘force-field’ to the opposition by the
medical professions, hospital managers and unions. I understand that you had some
reservations to the high priority given by your colleagues to ‘Patient Choice’. What do
you think?
(Possible supplementary) Your constituency has two district hospitals providing the
full range of services. The government now favours a mix of larger, ‘locality hospitals’
offering full services, and more localised walk-in and GP surgeries. How will you ‘square
the circle’ of defending both government policy and your constituents?
6. Minister, thank you for your time. Are there any questions which you believe that I should
have asked but didn’t. Can you recommend other people I should speak to?
Remember to remain flexible: don’t continue to ask your pre-arranged questions

regardless. Be prepared to use your quick intelligence, charm, humour, guile and
cunning to get the most from the interview. Note that you must show technical
mastery of your subject if you are to be regarded as a credible interviewer and if you
are to achieve any depth. So don’t waste interview time by asking for information
that the elite will have expected you to have obtained before the interview.
One contentious subject is whether you should mention your own hypothesis
during the course of the interview and, if so, when. You must weigh the balance
of advantage of disclosure. My experience is that, where rapport is very good, then
there are advantages in giving the elite the opportunity to dissuade you from your
hypothesis.
Post-interview
• transcribe the recording and apply the format of commentary-transcript-analysis
• separate out ‘quotable quotes’
• summarise the answers to your original questions
• triangulate with other sources
• follow-up leads
• write thanking the subject and forward a copy of your abstract at a later date. You may
want to interview your source again for other research. Use this opportunity to seek
confirmation of unexpected replies to important questions.
Interviews with other people
You are also likely to want to interview other people – ‘non-elites’ – to provide
perspectives from users or practitioners. For example, if you are researching
social exclusion, then you will need to interview a cross-section of socially-
excluded people and those responsible for implementing government policy. They
may include homeless people, unemployed people, single-parents, older people,
JobCentrePlus staff, social workers, voluntary agencies and council officers. Unlike
political elites, they are more likely to be:
• inaccessible or not readily identifiable

• apathetic, wary or hostile.They may suspect that you are a police or council officer, benefit
inspector, journalist, salesman, crook or confidence trickster, or (worse of all) a ‘university
type’ exploiting their dire circumstances to gain a degree or academic reputation
• unused to expressing their innermost beliefs and personal experiences – especially to
outsiders
• not able to fully understand what you are saying – and vice versa – either because you
speak a different language, or because their use of your common language is highly
nuanced.
The issue of communicability is critical to understanding. Schramm’s model of

communication is useful here (Schramm, 1961: 5–6).6 He argued that individuals
have fields of experience which include language, culture, conventions, institutions
and life experience. So people can only communicate effectively in those parts of
their fields which overlapped:
A B
Figure 9.2 Schramm’s ‘Fields of Experience’ (1961: 5–6)
So, to really see the world through the subject’s eyes, you must, first, identify
your cultural and communications overlap and, second, expand your own field of
experience. Living in the locality and carrying out pilot interviews can be very useful
to develop the local vocabulary and to understand local discourses.
You should be able to access the research subjects using either the snowball or
nomination methods described in Chapter 8.
The interview should follow the same semi-structured interview schedule of
open-ended questions and supplementary questions used for elite interviews.
However, the general style will be gentler and softer. You are more likely to engage at
a closer, personal level and gain enduring friendships. Indeed, interviewing members
of the general public can be the most enjoyable and rewarding part of the research
process. Members of the public are more likely to make admissions and confessions
to complete outsiders than to members of their own family. The interview may
prove a liberating experience for them.
BOX 9.2 Twelve commandments for interviewing non-elites
1. Don’t start ‘cold’ (Berg, 2001, 99–100)7

2. Don’t talk too much (Smith, 1993: 40–1)8
3. Use open-ended questions; vary voice and gestures (Smith, 1993)
4. Don’t use stress tactics (but try to avoid feel-good interviews) (Smith, 1993)
5. Don’t emphasise status differences (Smith, 1993)
6. Interview in a comfortable place (comfortable to subject, e.g. their office or home) (Berg,
2001)
7. Use comments (‘active listening’), summaries and transitions (giving opportunities
for interviewee to correct) but do not state initial analysis which will otherwise bias
responses to following questions (Smith, 1993)
8. Use ‘mirroring’ to copy the subject’s body language and, more importantly, use their
words and terminology to build a ‘key vocabulary’ (Bergman, 2003)
9. Be respectful, cordial and appreciative (Berg, 2001)
10. Respect silence: the subject is likely to break-in with important admissions (Bergman,
2003)9
11. Remember the likelihood that subject will also make significant admissions at the end
of the interview in ‘hand on the door’ comments once the tape-recorder is switched off.
When this occurs, seek a follow-up interview (Bergman, 2003)
12. Take notes (or appear to take notes) (Smith, 1993)
Various guidelines are offered by leading commentators. These can be consoli-

dated into ‘twelve commandments’ for researchers (set out in Box 9.2 above).
Payment
In the UK, payment is not expected for interviews. Indeed, its offer may offend.
But, in other countries, some ‘reciprocity’ is expected. You will therefore need to
find out local custom and practice. In the UK and elsewhere, the offer of a gift of
a small memento in the form of a university key-ring or letter-opener at the end of
the interview is likely to be appreciated. Where a subject has been especially helpful,
then sending a relevant paperback book may be appropriate.
Recording interviews
Interviews should generally be audio-recorded which should be supplemented by

notes of key non-verbal events or observations. Ideally, the audio-records should be
Table 9.1 Three-column transcription of interview
Interview A01 Questions, Responses Analysis

Notes and Line Number
Q.001 And what was your
002 reaction to the news?
A. [coughs]
003 I rushed to the House
4 sec pause, rubs 004 immediately to . . wary of expressing real intention?
chin support ? Was he really a strong
supporter of Ken or did he regard him
as most likely winner and wish to see
what advantage he could gain? Or did
he lie low until the outcome became
clearer?
7 sec pause 005 offer my support to . .
looks up to see 006 … Ken
reaction
telephone rings in
outer office
transcribed into text using the three-column format shown in Table 9.1. Note how
you should give each subject and interview a unique reference and add line numbers
to enable accurate references to be given, e.g. (A01: 003).
A 50-minute interview generally takes a whole day to transcribe. Audio-tapes can
be transcribed using a pedal-operated player. Digital recordings can be transcribed
direct to your PC. Voice transcription software is now available (Dragon Naturally
Speaking Professional 9 from Nuance Communications) which claims to be able to
transcribe one voice without extensive calibration. This may be very useful for
transcribing narratives. However, voice-transcription software is not yet able to
automatically transcribe several voices at the same time (although this appears likely
in the future). Full transcription is therefore generally confined to selected parts of
elite interviews where a detailed textual analysis is proposed. A supplementary or
alternative approach (which I prefer) is to write-up an ethnographic account which
combines observations, events and quotable quotes. An example is given in the
box (9.3) overleaf.
Difficult questions and subjects
Every community has its own pariah or difficult topics where special care is required
in introducing, framing and expressing questions. British people have many. One
explanation of why British people discuss the weather at their first meeting is that
it is the only common, safe topic in a repressed society where religion, politics, sex
BOX 9.3 Ethnographic account of an interview
The Landowner (2001)

The ‘big house’ was a modest, weathered, stone-built, seventeenth century long-house
separated and screened from the village below. The only access was a narrow metalled
road which terminated in the shadow of the main building. Access was limited to a single
archway separating the house from its former coach-house.Two general-purpose family cars
stood inside the small yard. A green, woollen sportsman’s tie twisted in the wind from the
long washing-line. The main frontage of the house had been rendered and painted white. It
looked down a private vista of banked, grazing land bordered by copses of birch.
Three, black, Labrador gun-dogs guarded the small child playing outside on the terrace in
the unseasonable February sunshine. They barked furiously until stayed by the neat, small
woman who introduced herself as the landowner’s wife. He was waiting in the office at the
west end of the house which he shared with a teenage secretary.
He was short, stocky, boundlessly energetic and, contrary to local reports, very likeable.
He wore a green woollen jumper, checked-shirt and light-green moleskin trousers. He led the
way into the small, painted dining-room so that he could snatch a lunch whilst he answered
the questions. The whole style of the interior of the house was well-maintained, ordered
simplicity.
He explained that the estate consisted of 7,000 acres of moorland, 300 acres of forestry,
1,500 acres of enclosed land – let as agricultural land to ten tenants – and eight houses … etc.
He’d ‘come into’ the estate 15 years ago on the death of his aunt. Since then he had always
‘sought to square the circle … on a daily basis of traditional obligations and responsibilities
with the present-day needs of modern estate management. It’s a matter of getting the
balance right’.When he took over, he had inherited a workforce consisting of a house-keeper,
gardener, two foresters and a game-keeper. He couldn’t justify a house-keeper or gardener
and the wages paid to the foresters greatly exceeded the net value of timber produced - for
which prices had remained static for 20 years.
Asked whether his style of estate-management was traditional-stewardship-paternalistic
or modern-exploitative-impersonal, he answered: ‘both … getting the estate economically
viable is the overriding consideration … if something isn’t viable, then it must be changed …
putting it off will only let everyone down’. His approach was ‘to make the assets sweat …
to realise their full potential and to provide only good jobs where the men can recognise
their own self-worth and contribution’. Similarly, he granted tenancies, not on the basis of
sentiment, but to whoever offered the highest price – subject to references – to ensure that
‘the land would be kept in good heart’. He was therefore opposed to subsidy or any other
‘distortion of the market’ which was likely to prove unsustainable and would prejudice rather
than protect people’s futures, etc.
and even sport are regarded as essentially difficult and therefore private matters. The
pariah topics in western society have been categorised as:
• illness and disability – especially sexually transmitted diseases and mental illness
• illegal and contra-normative behaviour including crime, tax evasion, drinking and sexual
practices
• financial status including income and savings.
(Foddy, 1993: 115)10
There will also be some people who are unwilling or unable – consciously or
subconsciously – to answer your questions.
In all these circumstances, projective techniques or vignettes may be appropriate.
Projective techniques and vignettes
Projective techniques are designed to prompt answers. The main types are:
• Quick-fire word association, e.g. night/day, tea/biscuits, etc. Blair/?

• Quick–fire ‘yes/yes/yes’ a set of questions followed an inadvertent ‘yes’ which are
designed to undermine the arrogance or self-confidence of hostile interviewees: e.g.
• ‘Is your name XX? yes
• You’re a student at Watersea? yes
• In Politics? yes
• You stole the photocopier? yes, … er, I mean no …
• Well, is it yes or no? It’s no
• You’re sure? [pause] yes
• Well, although you’re unclear about simple matters, let us move onto more complex
issues’. This technique is used by barristers and police.
• ‘Nothing unusual’ questions, e.g. ‘Have you ever smoked cannabis? after all, nothing
unusual, most students do one time or another’.
• Substituting difficult ‘if ’ questions by ‘when’ enquiries. For example, asking the question:
‘Have you ever had unprotected sex?’ implies that unprotected sex is ‘bad’. So the subject
is likely to answer ‘no’. However, a truthful admission is more likely when you re-phrase
the question as: ‘When did you last have unprotected sex?’
• Asking comments on list or photographs, e.g. ‘What sort of person do you think wrote this
list?’
• Using unexpected noises.
• Using film or video, e.g. ‘What do you think happened next? Why?’
• ‘Desert Island Discs’ questions, e.g. ‘If you were to be castaway on a desert island which
eight music recordings or books would you like to take with you? Why?’
You will appreciate that not all types of projective techniques are appropriate to all
subjects and circumstances. Tact is essential. But you will find that many subjects
find some of these types of question to be enjoyable and act as ‘ice-breakers’ in
otherwise tense situations.
A vignette is a more subtle form of question. Vignettes are:
… Short stories generated by the researcher and focussing on hypothetical characters in

particular situations. Interviewees are asked what they would do in these circumstances
or, alternatively, how they think that a third party might react. The latter mode of question
specifically distances the interviewer from the issues being studied and, in this sense,
is impersonal and less threatening.
(Arksey & Knight, 1999: 94–5)11
Vignettes are premised on empiric evidence that what people identify as their
beliefs are often compromised by their actions. For example, when people are asked
whether they are honest, they will generally apply affirmatively. Similarly, they are
likely to denounce racial prejudice. But actions – what people actually do – provide
better evidence of underlying belief systems. Indeed, small changes of behaviour
can prompt fundamental changes of belief. Good vignettes pose some form of moral
dilemma. They may also ask what the subject should do and what they would be most
likely to do. For example, most people, if asked, will say that they strongly oppose
cruelty to children or racial segregation. They will also say that they are honest.
In these circumstances, vignettes may be helpful. Consider the three examples:
1. A family has moved into the terraced house next door to you. After a few months of
heated arguments, the husband quits the house leaving the mother alone to care for
two children under six years old. The boy is a ‘bit of a handful’. But you hear him being
slapped and being told that he’s ‘stupid’ and ‘just like his dad’.
What should you do:
A Speak to the mother?
B Speak to the authorities?
C Do nothing: mind your own business?
What would you actually do? Why?
2. You join the queue at the bus stop. When your bus arrives, it’s nearly full. You’re last on
board. There are only two seats vacant. They’re next to people of your same age and
sex. One is the same race as you. The other is a different race. They both remove their
bags. You have three options:
A Accept the place next to the person of your same race?
B Accept the place next to the person of a different race?
C Stand?
What should you do?

What would you actually do? Why ?
3. You approach the vending machine in the refectory to buy a coffee. You are about to
insert a 50p coin when you notice that there is already one lying in the change slot.
What should you do?
A Hand the 50p coin into the porter’s lodge as ‘lost property’?
B Use the coin to get your drink?
C Put the coin in the charity box in the bar?
What would you actually do? Why ?
Good vignettes are difficult to write and should be pre-tested before finalising. One
useful adjunct to vignettes is to ask respondents – at the end of the interview –
what they actually did in memorable circumstances to calibrate their general responses.
Examples might include whether they’ve always voted for the same political party.
One set of questions that I’ve found useful is to ask older subjects their support (on
a scale of one-to-ten) for the mineworkers in the ‘three-day-week’ (national strike
which led to the fall of the Conservative government) of 1974 and, after they have
answered, their support for the mineworkers in 1982–3.
In many circumstances, partners may ask if they can stay during the interview.
You must always give parents or carers of children this option. This can inhibit the
subject. But it can also provide or deny corroboration. I well remember that, when
a husband replied that he was ‘fully behind the miners in 1983’, his wife added: ‘Oh,
no you weren’t. You said that Arthur Scargill should be strung up!’12
Focus groups
Focus groups were initially developed largely by market researchers as a means of

finding out the potential market for new products and services and, subsequently,
‘blind tasting’ the prototype product against competitors. This type of focus group
is also known as a ‘hall trial’, ‘consumer panel’ or ‘citizen jury’. Focus groups have
now become widely used in a similar way by political parties to learn how potential
leaders, topics and policies are regarded by cross-sections of the general public.
More recently, focus groups have been adopted by academic researchers in
medicine, sociology and Politics to investigate the dynamics of interpersonal
relationships, inter-group conflict and how consensus (or dissensus) emerges. A focus
group is not a ‘group interview’ in which the subjects are asked for their replies to the
same question. Instead, the focus group seeks to emulate the spontaneous discussion
of people who broadly share the same circumstances or identity. In particular,
the focus group should ‘draw upon the respondents’ attitudes, feelings, beliefs,
experiences and reactions in a way that would not be feasible by other methods,
for example, observation, one-to-one interviewing or questionnaire surveys

(Gibbs, 1997).13
Your role as the researcher is essentially as a facilitator who:
• invites the participants

• arranges the venue, refreshments and ‘gifts’
• initiates the session
• moderates the proceedings
• arranges and transposes the recording of the encounter.
The optimum number of participants is usually between five and eight. This
will ensure a variety of views and an opportunity for each person to speak. The
participants should, in the first place, provide a cross-section of people sharing a
particular identity and location. They can be recruited by word-of-mouth, public
notice or via a social gatekeeper intermediary. Where a social gatekeeper is available
then they should be used, first, because they are likely to be able to recruit a willing
cross-section more easily and, second, because if their help is not sought, then they
may be offended and oppose the research. They may be clergy, teachers, councillors
or the secretaries of local groups. They will advise you whether and what kind
of ‘incentives’ (i.e. a small gift) should be offered to the members of the group.
However, you should personally write to or contact the people suggested to you to
invite them to the meeting and to outline your research, the benefits to them, and
the rules-of-the-game.
The venue should ideally be a neutral, central location in which the group can sit
comfortably for as long as two hours. ‘Hospitality’ should be offered in accordance
with local conventions. So, in the UK, tea or coffee and biscuits should be provided.
Occasionally, wine may be offered. The seating should be set out in a circle. But
many researchers find that people feel more confident when they are sitting in a
circle of easy chairs or, if these are not available, around a defensive perimeter of
tables. You should sit down first and allow the members of the group to sit where
they wish. The seating pattern is likely to reflect their assessment of their relative
social position. So, the members considering themselves most important and capable
of speaking out are likely to sit facing you. Conversely, the least confident members
are likely to sit at each side of you.
You should start the session by explaining the nature of the meeting and re-
stating the rules-of-the-game. Where you have a central topic, then you should
specifically invite the views of the group. Alternatively (and often better), you should
ask the group which issues are most important to them. Ideally, the session should
then proceed entirely driven by its members. But, in practice, some intervention –
moderation – by the researcher may be necessary. This may include (tactfully)
silencing dominant members (by a finger to the lips), inviting those who have
remained silent to state their views, stopping rows emerging and moving on when
a topic has been exhausted. So moderating requires good interpersonal, leadership

and managerial skills.
In addition to learning of individuals’ beliefs and experiences and how group
views emerge, the focus group should provide evidence of the language that
they use in everyday conversation and local institutions. Focus groups can also
provide important ‘forums for change’ where the disparate individuals become
empowered by their new collective identity and develop solutions of their own
(Race, 1994: 738).14 In these circumstances, you must be prepared to offer to help
the group by seeking information and writing letters and reports, etc. for them.
Some practitioners strongly argue that the focus group should always be
supplemented by meetings with individual members. This enables the researcher
to learn individual’s views and how these relate to the group consensus, serious
misgivings or changes of belief as a consequence of taking part in the discussions.
But the difficulty here is determining what really were the individual’s initial views
before the focus group was held. So, some researchers argue the case for a split-half
strategy to be adopted in which half the group are interviewed individually before
the focus group, and the others afterwards. In this way, the role of the group in
developing new perceptions or understandings can be identified better. But holding
individual interviews as well as a focus group compromises an important benefit of
focus groups: their great efficiency in terms of your time.
The focus group should be audio-recorded and the recording transcribed.
However, recording a long meeting and transcribing the record can be difficult –
particularly where voices overlap, or where it is difficult to attribute contributions.
It is also very difficult for you to make notes and moderate the group at the same
time. So a further refinement is to use a back-up researcher who will sit outside the
group, organise the recording and make notes of key contributions and any off-tape
events.
The writing-up of the focus group follow the format of an annotated, three-
column transcription ideally supported by an ethnographic account of observations
and ‘quotable quotes’.
1. Draft letters to each of the following seeking an interview to learn their views on
the use of military options to counter terrorism:
George W. Bush
Tony Blair
President Chirac
Martin MacGuiness
2. The (Labour) Leader of the city council in the north of England has agreed to
be interviewed on the subject of corruption in local government. Your research
hypothesis is that ‘standards of public office’ are locally contingent and emergent
rather than absolute. A recent press report alleged that the leader had favoured
a cousin in a construction contract. Draft a semi-structured interview schedule.
Complete and video record an in-class simulation in which two students separately
interview the Leader of the council (who is best played by the course teacher or a
mature postgraduate). Contrast and compare the interviews.
3. A bacon factory in a nearby market town slaughters and packs pork in a continuous,
industrialised process. It employs over 2000 workers. The majority are young
women who are bussed in from villages in the upland areas. Union membership
is low and staff turnover high. The management recognises USDAW for pay (etc.)
negotiations.Your hypothesis is that membership is low because of a combination
of a rural culture of deference, the portrayal by managers of the union as a ‘big
city thing’ and the domination of union offices by men. The HR staff and union
have nominated a cross-section of female workers for you to interview in paid time
as a group and singly. Draft a semi-structured interview schedule to be used in
individual meetings with the workers. How would you organise and structure a
focus group of the workers? Simulate, contrast and compare the interview and
meeting.
4. You have arranged to meet a newly-retired infantry major who took part in the
invasion of Iraq. He strenuously denies that torture is used by British soldiers.
Design a series of vignettes to test your hypothesis that undue pressure was
exerted by some soldiers on prisoners to obtain information with the knowledge
of their officers who turned a ‘blind eye’ to these practices.
FURTHER READING
Arksey, H. and Knight, G. (1999) Interviewing for Social Scientists. London: Sage.
This textbook provides a rich variety of theoretically-based, practical techniques
for all social science researchers.
Berg, B.L. (2001) Quantitative Research Methods for the Social Sciences. Boston:
Allyn & Bacon. pp. 66--99. Chapter 4 adopts a ‘dramaturgical look’ at interviewing
as variously an art, skill, science, game and face-to-face interaction. It examines
the separate roles of the interviewer as an actor, director and choreographer.
This text provides very good advice for the inexperienced interviewer.
Burnham, P., Killand, K., Grant, W., and Layton-Henry, Z. (2004) Research
Methods in Politics. Basingstoke: Macmillan. pp. 205–20. Chapter 9 of this
textbook is devoted to interviewing political elites.
Foddy, W. (1993) Constructing Questions for Interviews and Questionnaires.

Cambridge: Cambridge University Press. This well-established text provides
sound advice on wording and phrasing of questions and sensitive issues.
Harrison, L. (2001) Political Research. London: Routledge. pp. 89–104. Chapter 6
is devoted to conducting interviews in Politics research with particular regard to
practical and ethical concerns.
McEvoy, J. (2006) Elite Interviewing in a Divided society: lessons from Northern
Island. In Politics, 26 (3) 184–91. Joanne McEvoy discusses the particular
difficulties raised when the researcher is interviewing elites from ideological
extremes. She argues that researchers must consider how their identity may
effect the outcome of interviews, bias, trust and how they might seek to ‘probe
beyond ethnic party positions based on mistrust of the other side’.
Neuman, W.L. (2003) Social Research Methods. London: Pearson. pp. 390–401.
The author distinguishes between ‘typical survey interviews’ and ‘typical field
interviews’ where the former are standardised to enable comparison and the
latter tailored to the subject and circumstances. He also discusses the role and
value of ‘informants’ and the contribution of focus groups.
Richards, D. (1996) Elite interviewing: approaches and pitfalls. In Politics, 16(3)
199–204. This journal article provides good analytical and perceptive advice on
the significance of elite interviewing to Politics research.
Notes
1 Note that, in the UK and many other countries, it is illegal to record telephone conversations
without the consent of the respondent.
2 bona fides, Latin, ‘in good faith’.
3 Friedrich, C.J. (1937) Constitutional Government and Politics. New York: Harper & Brothers.
4 An elite’s ‘opposite number’ will normally be the member of the opposition party or group
nominated to lead the opposition to a particular policy area for which a minister holds
responsibility. They may be called the ‘shadow spokesperson’. A ‘parliamentary pair’ is the term
used to describe a grouping of a government and opposition MPs in which, when one is unable to
vote in Parliament for unavoidable personal or parliamentary reasons, the other will abstain.
5 Mayo, E. (1949) Hawthorne and the Western Electric Company. London: Routledge.
6 Wilbur Schramm, How Communication Works. The Process and Effects of Mass Communication
ed. Wilbur Schramm (Urbana, Ill.: The University of Illinois Press, 1961) pp. 5–6.
7 Berg, b.l.(2001) In Quantitative Research Methods for the Social Sciences. Boston: Allyn & Bacon.
p. 99–100.
8 Smith, M. (1993) Personal Interviewing: A Mini-Training Workshop In Breakwell, G.M.; Foot, H.
and Gilmour, R. (eds.) Doing Social Psychology: Laboratory Exercises. Leicester: BPS. pp. 23–31,
40–1.
9 Dr Max Bergman, lecture to Essex Summer School as part of the course ‘Qualitative Data
Analysis: Interpenetrative Methodologies for Analysing Text and Talk’.
10 Foddy, W. (1993) Constructing Questions for Interviews and Questionnaires. Cambridge:

Cambridge University Press. p. 118.
11 Arksey, H. and Knight, G. (1999) Interviewing for Social Scientists. London: Sage. pp. 94–5.
12 Arthur Scargill was the controversial, left-wing General Secretary of the National Union of
Mineworkers during this ‘last, great’ miners’ strike.
13 Gibbs, A. (1997) Focus Groups. In Social Research, 19. Guildford: University of Surrey.
14 Race, K.E., Hotch, D.F. and Parker, T. (1994) Rehabilitation programme evaluation: use of focus
groups to empower clients. In Evaluation Review, 18 (6) 730–40.
Chapter 10
Questionnaire Surveys
Our researchers with Public Opinion are content

That he held the proper opinions for the time of the year;
When there was peace, he was for peace;
When there was war, he went
W.H. Auden (1940) The Unknown Citizen
1. To understand the special characteristics of questionnaires and where they

are best used.
2. To consider how questionnaires should be structured, designed, tested and
coded.
3. To discuss how questionnaires can best be ‘administered’.
4. To discuss issues of personal safety.
Questionnaires
Questionnaires are essentially pre-designed lists of closed questions with pre-

designated, alternative answers. They are used to collect data from a sample of
individual subjects (respondents, interviewees). The data can then be aggregated
(combined) to create a representative profile of the sample and cross-tabulated to
explore the relationships between classifiable variables, e.g. income and education.
You can draw inferences about the research population from questionnaire responses
of representative samples. You can also use them to measure and compare attitudes
(although the accuracy of this data is contested).
The essential differences between questionnaire surveys and interviews (or focus
groups) are that questionnaire surveys:
• use very large samples (e.g., opinion polls use samples of 1,500 people, medical research
can use samples of 60,000 or more)
• use only closed questions
Questionnaire Surveys 141
• use pre-determined answers – including Don’t Know (DK) and No Response (NR)
• take less time for the subject to complete
• seek to use representative samples
• can be administered (carried out) by intermediaries
• similarly, the tasks of questionnaire design, administration, coding and analysis can be
carried out by separate individuals or teams
• can be administered either:
• face-to-face with the subject
• remotely by telephone
• self-administered by the subject in their own home (or workplace, etc) using self-
completion questionnaires delivered and returned by mail
• email.
• are a common tool of quantitative method
• are designed and coded to be readily analysed by computers
• necessarily involve a trade-off between cost, reliability and time.
But another important characteristic of modern questionnaire research is declining

response rates. 50% is now considered high for commercial market research in face-
to-face surveys. Market researchers reconcile low response rates with information
claims by adopting the assumption that non-respondents or non-responses (NRs) –
including ‘don’t knows’ (DKs) – are not inherently different than respondents.
So, the information continues to be typical and representative. However, this
assumption can be challenged on the basis that non-respondents may well be atypical.
Furthermore, they may share some common characteristics which makes them
unwilling to take part or complete the questionnaire.
Good response rates require good access to potential respondents, the ability to motivate
them, and effective questionnaire design to maintain their interest and co-operation.
Access
Access concerns the means by which the subject is contacted. It requires a sample
frame (a list of the research population from which the sample can be drawn),
a means by which a representative sample can be selected and some form of address
or arena where the questionnaire can be administered to them. In the UK, the two
most common forms of lists were the electoral register and the telephone directory.
So, if you wanted to research the general adult population of a city or part of a
city using a 10% sample, then you would use the electoral register to select every
tenth person. The register would also give their address where the researcher could
call on them, post the questionnaire to or, after looking up their number in the
telephone directory, phone them. However, the introduction of the ‘community
charge’ (‘poll tax’) in 1990 led to a widespread avoidance of registration. There

were also problems in urban areas where population turnover was high. Despite
the introduction of a ‘rolling register’, the electoral list remains incomplete. Many
students are not included (although, exceptionally, they are permitted to register
both at the university and home address and to vote in both localities in local
government elections). But, access to the electoral register has become restricted as
a consequence of its misuse by stalkers and criminals to locate victims. Now, the
full register can only be inspected at local authority headquarters. Whilst a register
continues to be published, this is confined to those electors who have agreed
to their names and addresses being published. Although the telephone directory
continues to be published, a third of (mainly higher income) subscribers choose to
be ‘ex-directory’. At the same time, increasing numbers of people – especially young
people – use mobile phones in preference to land-lines. Market researchers have
responded to these changes by buying lists of names and addresses from commercial
agencies, by using ‘random walks’ and by using on-street questionnaire surveys. Even
so, some potential subjects remain difficult to access. They include: housebound
people, people working away from home, and urban nomads who are not registered
with NHS, electoral registrar or national insurance. Homeless people, by definition,
lack a contact address. So accessing the necessary full sample frame is very difficult.
The absence of free, published lists and the limited resources of most Politics
researchers, mean that you are likely to need to use other effective, low-cost means
of accessing subjects appropriate to the research. Where you are researching a small
geographical area, then the simplest way to achieve a cross-section of the population
is to use the nth address technique. In this method, you deliver a questionnaire to
(or seek to conduct a doorstep questionnaire survey at) every, say, tenth, house in
every street. However, rather than starting at No. 1, you should choose a number
between 1 and 9 at random. If that number is 6, then the house numbers will be 6,
16, 26, 36 etc. Where the size of the geographical area is much greater, then you can
use the random walk technique. You begin at the centre calling at every nth house
in the first street, turning left at the first intersection, right at the second and so on.
Where the morphology of the urban area consists of the classic, concentric rings
of inner city, inner and outer suburbs it is likely that a crude representative sample
of households will be achieved. But in much Politics research, questionnaires are
confined to specific groups who may be councillors, trade unionists, party members,
etc. In this case, the best method – if a list of members is not available – is to ask the
‘secretary’ or other social gatekeepers to forward questionnaires to every member or
nth member.
Where a shorter, ‘vox pop’ (‘voice of the people’) questionnaire survey is useful
to find out, say, voting behaviour or support for particular parties, then an on-street,
face-to-face questionnaire survey of a quota sample is probably the most effective
means of accessing a representative sample. Advice on the design of a quota sample
is given in Chapter 7.
Pilots
Having designed the quota sample, then the next stage is to pilot your questionnaire
survey, i.e. test it on a small sample. This is most readily done by trialling the draft
on non-teaching staff at your university. They will tell you if they find the questions
difficult to understand or answer. They may suggest more appropriate (to the local
population) ways to ask your questions. The pilot should also help you identify or
amend your list of potential answers.
Administering the questionnaire
In this context, administering means ‘taking the questionnaire to the people’.

The full survey will generally be carried out in the local shopping centre(s) at the
weekend. You will therefore look for people who appear to correspond to each
of your quota boxes. It is relatively easy to identify potential respondents by sex
and age. The most difficult characteristic to identify is socio-economic class, where
‘working’ and ‘leisure dress’ styles have become more universal. Trial, error and
intuition (learned experience) become important. One street pollster told me that
she had found the quality of shoes to be the best visible indicator. Otherwise, as the
survey proceeds, you will have greater difficulty in ‘filling your quota’.
The general advice is to avoid people who are busy or otherwise fully occupied.
So, avoid parents with young children, people laden with shopping bags or rushing
about. Instead, approach those people who have time on their hands. People sitting
on benches are ideal. They may be waiting for a partner to finish their shopping and
may welcome a conversation with you. You should approach them by ‘catching their
eye’ (making eye contact), smiling gently and greeting them. Your initial remarks
might be:
Hello, what a lovely/cold/awful day.You look as if your enjoying the rest … Can you help?
I’m a student at the university. We’re trying to find out what local people think about the
EU and why …. May I sit down? It won’t take long … Do you live in Watersea?
Motivation
Response rates and co-operation are likely to be much higher when the potential
subjects feel motivated to assist. Motivation means providing the necessary
incentives for respondents to give their time, views and personal details. This involves
appeals to altruism (concern for others’ welfare), public or personal benefits, or
small measures of ‘reciprocity’, e.g. a pen, token, lottery ticket or donation to
charity. Motivation also involves dispelling doubts about your identity, real purpose
and confidentiality. Modern urban relations are characterised by distrust between

strangers. They may fear that you are really trying to get them to sign up for double-
glazing, or trying to steal their identity. So motivating them begins by you identifying
who you are: the importance of the research to them and how the information they
give will be used. You should display your name and university or college on a
photo-ID card and carry a letter of authorisation. Your dress and manner should
not be intimidating. In terms of privacy and confidentiality, you should be able to
reassure them from the start that you do not need or want to know their names or
addresses, and that personal views and details will be treated in confidence. They
don’t have to answer a question if they don’t want to. However, if they make a very
‘quotable quote’ in the course of the interview, you should ask if you can repeat it
as an anonymous comment.
Effective questionnaire design
Ideally, a questionnaire should optimise information collected and time involved to

enable a large sample to be questioned and representativeness maximised. But good
questionnaires are very difficult to design. The most common faults are that they
are too long, too pedantic in their use of language, difficult to understand, or ask
obvious or stupid questions, e.g. are you male/female or don’t know?
The general guidance is:
• make the language simple

• make the language short
• avoid ambiguous or double-barrelled questions
• avoid leading questions: be neutral
• avoid negative questions: especially double-negatives
• avoid questions which assume subjects have perfect knowledge
• always allow for N/R or D/Ks and continue politely
• be aware that many words have different meanings to different people: e.g. ‘cool’, ‘gay’
• distinguish and separate factual from non-factual questions and answers
• avoid ‘prestige bias’ (when you imply that certain forms of behaviour associated with high
status groups are necessarily better)
• clarify the time frame, e.g. weekly or monthly income
• place sensitive questions in context
• don’t ask very detailed or objectionable questions
• don’t ask too many questions: try to keep to 12 or less.
The main problem for students is framing questions in plain conversational language.
There is a tendency to use academic or technical language and to be precise
to the point of pedantry. Excessive qualification of questions tends to produce

gobbledygook (literally, ‘turkey-talk’). Verbal and written questions require very
different styles of question. Preamble to questions or examples can be helpful to
subjects. Remember that, to generate comparable data, each subject must be asked
the same question in the same way: avoid the temptation to interpret or personalise
the question.
The sequence (order) of questions is critical. Begin by checking that the potential
subject is within your sample frame. For example, if you are doing a street-survey of
support for political parties in the city centre, first establish that the person is, in fact,
entitled to vote in UK elections. But it is difficult to ask someone their nationality.
So, instead, begin by asking Do you live locally?. They are likely then to say either
Yes, or No, I come from Leeds, or No, I’m a US citizen etc. Start with questions which
are easy (and, if possible, enjoyable) and which will put the respondent at ease and
lead to good rapport (open, friendly, communicative relationship). Leave sensitive
or contentious questions to the end. Conclude by seeking key classification data for
possible determination of independent variables or cross-tabulation. Classification
data are: gender, age group, socio-economic group, education, housing tenure and
dependants. Some can be guessed by careful observation but should be checked with
the subject.
Sensitive questions should be left until nearer the end of the questionnaire, very
carefully designed and presented or omitted entirely. In the UK, people – especially
middle-aged women – are reluctant to tell you their age. But, older people tend
to be proud of their age. But, paradoxically, all are likely to tell you when they
were born if asked direct: they’re used to being asked the question to confirm their
identity. Income and class can also be sensitive. Political or religious affiliation is now
less problematic on mainland UK. Be wary of asking questions of children without
the consent of parents or teachers, etc. Find out what questions are sensitive to UK
minorities or in countries overseas by speaking to colleagues or cultural groups first.
Sensitive questions can be asked by using:
• proxy-indicators
• show-cards
• post-codes.
Proxy-indicators
One way of asking sensitive questions is to use proxy-indicators where precision is
unimportant. A proxy-indicator is an alternative means of measuring or describing
a variable, e.g. using housing tenure (whether subjects own or rent their homes
and the number of bedrooms) as a proxy measure for income. The indicator is
effectively an approximation. Alternatively, you can use established classification
schemes. The social classification scheme most widely adopted by government and
private organisations is the A, B, C1, C2, D, E. The broad definitions and percentages
of the population of England given in the 2001 census are shown in Box 10.1:
BOX 10.1 Socio-economic classification scheme, UK

Census 2001
A Higher managerial/administrative/professional
22%
B Intermediate managerial/ administrative/professional
C1 Supervisory, clerical, junior managerial/administrative/professional 30%
C2 Skilled manual workers 15%
D Semi-skilled and unskilled manual workers 17%
E On state benefit, unemployed, lowest grade workers 16%
This ‘ABC’ classification is a composite of income and status developed to reflect

purchasing power. University lecturers are included in A whereas students are
included in E together with others who are ‘not economically active’. Adults are
classified either by their present or last employment or occupation, or collectively
as ‘pensioners’. Children are classified by their parents’ occupation.
It is, of course, difficult to ask people which social class they belong to. Asking
them whether they are middle class or working class can be difficult and offers only
a crude twofold classification. In these circumstances, you can ask questions about
subjects’ occupation, newspaper readership, education and housing tenure. Typical
questions might be:
• ‘What do you do for a living?’ (If they reply that they are retired or not working, then ask:
‘What did you do last’?)
• ‘Which daily newspaper do you normally read’?
• ‘When did you leave school: at 16 or 18?’ (if they reply ‘18’, you can then ask: ‘Did you go
on to university?’)
• ‘Do you rent or own your home?’
Note how the questions are phrased using conversational English. Note also how
they avoid the prestige bias of implying that the respondents should go to university
and should own their own home.
The issue of newspaper readership is interesting. Currently, UK newspapers rely
on commercial advertising for over three-quarters of their income. To survive,
they must therefore sell their readerships to advertisers. The National Readership
Survey audits sales and readership using the ABC classification. Newspaper groups
seek to ensure maximum market penetration by publishing a range of ‘titles’
(newspapers) which are designed to appeal to all the major market segments.
Examples are:
Times, Guardian, Independent, Telegraph (AB)

Mail, Express (C1,C2)
Sun, Mirror, Star (C2/D)
So, asking people which newspaper they read can be an easy, inoffensive way of
gaining information into their lifestyle group and social class. In the UK, there is also
a relationship between newspaper readership and political affiliation. The Telegraph,
Times, Mail and Express are traditionally associated with the Conservative Party and
The Mirror with Labour. The Guardian and Independent are more sympathetic to
Labour and the Lib-Dems whilst the (best-selling) Sun is regarded as supporting
whichever party is most likely to win the next election. Therefore, if a subject
tells you that they read The Telegraph, then, rather than asking which party they
support, you can ask (with a slight smile) whether they support the Conservative
Party.
Show-cards
Another ways of seeking information on class, socio-economic group or other
sensitive issues is to use show-cards. Here you ask the subject to look at a card and
to say ‘which letter corresponds to your income’. The card might read:
BOX 10.2 Sample show-card
Q.11 Income per year (before tax)

A £0 5 000
B £5 000 10 000
C £10 000 15 000
D £15 000 20 000
E £20 000 25 000
etc
J £100 000 or more
L Don’t know
M Won’t say
N No response
Postcodes
A third method is to ask subjects for their (home) postcode (similar to the US
zipcode). The housing market in the UK tends to bring together people of similar
age, income and socio-economic class. Groups of around 50 homes are classified as
enumeration districts (EDs) which form the basic building block of population census
data. The firm CACI has matched small area census data and postcodes to create
ACORN : a classification of residential neighbourhoods.
ACORN categorises all 1.9 million UK postcodes, which have been described using
over 125 demographic statistics within England, Scotland, Wales and Northern Ireland,
and 287 lifestyle variables, making it the most powerful discriminator, giving a clearer
understanding of clients and prospects.
(www.caci.co.uk)
So, a simple way of finding out social class, etc. is to ask people their postcode.
The weakness of this approach is twofold: first, its assumption that all people sharing
the same postcode are very similar; and, second, the cost of subscribing to CACI.
However, you can avoid the subscription fee by using a map of post-codes (likely
to be held in your university library) to cross-refer to a map of enumeration areas
and the census data.
Scaling attitudes
A major claim of opinion pollsters is the ability of questionnaire surveys to find

out people’s attitudes and beliefs by measuring these using scales. Scales are linear
representations of measurement. Three techniques are widely used: the semantic
differential, Likert scale and rating-scale.
Semantic differential
The semantic differential was developed by Good in 1957. It uses a five- or
seven-point scale to assess respondents’ assessment of conditions or abilities, etc.
between pairs of words with diametrically opposite meanings.
For example:
‘[In your opinion] how honest is the Prime Minster?’

Extremely Very Quite Neither/Nor Quite Very Extremely
Honest ✷ ✷ ✷ ✷ ✷ ✷ ✷ Dishonest
Alternatively, you can give the respondent a list of semantic differentials for
comment, e.g. trustworthy/untrustworthy, sincere/insincere. The difficulty is to
find generally-acceptable differentials for some words like ‘aloof’ etc.
Likert scale
The Likert scale (Likert, 1932) puts the question in the form of a statement with
which the respondent is asked to scale their agreement or otherwise.
For example:
‘How strongly do you agree or disagree with the following statements: the Prime Minister is
essentially honest’, etc.
Strongly agree agree neither agree nor disagree disagree strongly disagree
✷ ✷ ✷ ✷ ✷
One variation of the Likert scale is to use wingdings to indicate agreement or

disagreement. For example:
☺ ☺ ☺ " # # #
Guttman scale
A weakness of both the semantic differential and Likert scale is the marked tendency
for many respondents to select the central, neither/nor answer. The Guttman
scale seeks to overcome this problem by asking a series of related questions
which will determine the intensity of the respondent’s attitudes to a specific issue.
It can therefore be likened to limbo-dancing or high-jumping. It seeks to build a
continuum in which the respondent’s answer to one question is likely to be repeated.
The process begins with you drafting a series of statements referring to the chosen
topic. Ideally, you should draft as many as 100. You should ask your research
colleagues to separately consider the drafts and to say whether or not each statement
directly relates to the topic or not. The ten statements receiving the highest ‘yes’
vote are selected. A pilot study of the sample population is carried out in which
the respondents are asked whether they agree or not with the statements. The
answers are then tabulated into a scalogram in which the questions are ranked by
the highest number of positive responses received. The ranking constitutes the scale.
An example of the Guttman scale is the Bogardus Social Distance Scale which seeks
to measure respondents’ attitudes towards social groups in terms of the degrees of
social acceptance. For example, a Guttman scale question might ask each subject:
Would you accept Afro-Caribbean immigrants as:
• close relatives by marriage (score 7.00)

• personal friends (6.00)
• neighbours (5.00)
• work-colleagues (4.00)
• UK citizens (3.00)
• visitors (2.00)
• would exclude from my country (1.00).
Note that agreement with the first statement indicates likely agreement with the
following statements until the last. So respondents most favouring immigrants would
achieve the highest score, 28 (i.e. 7 + 6 + 5 + 4 + 3 + 2) whereas those most hostile
would score only 1.
Another example would be attitudes to capital punishment, e.g.
Tell me whether you agree or not with the following statements:
• There can never be any justification for the death penalty (score 0.00)
• The death penalty can only be justified in exceptional circumstances following a trial
where evidence had shown beyond a shadow of doubt that the accused are guilty of:
• crimes against humanity including genocide (1.00)
• a murder following a previous conviction for murder (2.00)
• the murder of two or more children (3.00)
• murder and rape (4.00)
• murder in the course of other criminal activity (5.00)
• any murder (6.00)
• lethal injection is not painful enough (7.00)
• executions should be carried out in public (8.00)
In this case, abolitionists would achieve a score of 0.00 whilst ‘hangers and floggers’
would score 36.00. Note that, except for the first statement, agreement with other
statements implies agreement with all preceding statements. The responses can then
be displayed in a frequency table and chart which show the distribution and intensity
of opposition and support for the capital punishment (see Fig 10.1).
Ideally, the statements should be scrambled before they are presented to the
respondent. But, in practice, it is much easier for the researcher and subject if the
questions are asked in accordance with the scale.
Scale-rating
Scale-rating is a new scale developed largely by the entertainment media and widely
used by younger people. It asks people to ‘rate’ a particular individual or other entity
on a scale of 1–10 (where 10 is ‘well cool’).
Table 10.1 Guttman scale: summary of respondents and scores
Proportion of Population (%) Support for Death Penalty (Maximum 36)

33 0
18 1
14 4
10 8
9 13
7 19
5 26
4 35
40
Intensity of support for
35
30
death penalty
25 Support for
20 death penalty
15 (max 36)
10
5
0
0 10 20 30 40
Respondents (%)
Figure 10.1 Chart of Guttman responses
So, a question might be:
Q. ‘On a scale of 1 to 10, how did you rate John Kerry as a

Democratic Presidential Candidate?’
But whilst a question might be simple and straightforward for ‘media-savvy youngsters’,
it might not be readily understandable to older people.
Ranking
Subjects can also be asked to rank the importance or significance they give to
particular characteristics, e.g.
Q. Which of these characteristics do you consider the most

important for a councillor?
A honesty
B intelligence
C local knowledge
D party loyalty
E shrewdness
F business experience
G personal wealth
H involvement in the voluntary sector
Q. Which do you consider to be the least important?
Whilst this technique has certain advantages, it can irritate subjects when they are asked to
rank a complete list of criteria. It is therefore best used sparingly and with a show-card and
pencil to hand where conditions or the short-term memories of respondents prove difficult.
Personal safety
Face-to-face interviewing poses potential risks to you. These can be reduced by
observing the following do’s and don’t’s:
DO
• carry ID and a letter of authorisation from your course leader or tutor written on
departmental letterhead and providing a series of contact telephone numbers
• carry a mobile phone
• always carry out your questionnaire surveys within sight and earshot of another
student
• carry out your surveys in daylight in a public place
• consider borrowing or buying a personal alarm
• comply with any instructions from a police officer to ‘move on’
• where the survey is being carried out as part of an assignment, ask the course
leader to write to the local police commander and local authority notifying them
that you will be carrying out the survey, when and where.
DON’T
• don’t harass or hinder members of the public

• don’t stand in doorways
• don’t get involved in arguments.
Overseas students often worry that they will have greater difficulty in recruiting
members of the public to help with their surveys. However, experience in York has
been that overseas students have rather less difficulty than ‘home’ students because
the public are curious and want to find out what they think of the UK.
Coding
An important feature of questionnaire design is the coding of questions and
alternative responses to enable manipulation of data and analysis by computers.
A code is the representation of any group of words (or numbers) by a unique
set of numbers or letters. The initial pilot questionnaire should establish the range of
most frequent responses (scoping). Advances in software have enabled the coding
to be simplified. A simple generic form of coding can be used which combines the
number of the question and a number allocated to each potential response.
For example, if Question 7 asks whether the respondent left school at 16 or 18
then the coding (see brackets) might be:
Left at 16 (07.1) :
Left at 18 (07.2) :
Don’t know (07.3) :
No reply (07.4) :
In this example, the coding is entirely nominal: it has no mathematical significance.

However, giving a numeric code to a semantic differential or Likert scale has
mathematical significance and allows the ‘average attitude’ to a particular topic to be
calculated simply. For example, the Likert ‘strongly agree’ can be coded 2, ‘agree’
Table 10.2 Schedule of coded responses to questionnaire survey
Respondent Ref. Q01 Q02 Q03 Q04 Q05 etc Notes

A00001 1 4 2 1 3 interrupted
A00002 2 1 5 2 4
A00003 1 3 4 1 4
A00004 3 2 3 2 3
A00005 1 2 2 3 2
A00006 1 2 2 2 rushed
A00007 2 1 3 2 5
A00008 3 4 5 2 2
A00009 4 2 2 1 2
coded 1, ‘neither agree nor disagree’ 0, ‘disagree’ −1, and ‘strongly disagree’ −2.
So, if the average is 0.25 then the average view of the public is only marginally
in agreement with the statement. But there is a flaw: the assumption that levels of
agreement etc, are linear and equal, so that one person’s ‘strongly agree’ is equal to
two people’s ‘agree’. This assumption is highly questionable.
The coding will enable the answers received to be recorded directly into a table
or spreadsheet: you do not need a separate questionnaire for each face-to-face interview. It
is usual to list the respondents in the left-hand column and the questions in the
top row. Each cell in the table will contain the sub-code for the answer given (see
Table 10.2). Each respondent should be given a unique reference which, where
a team of researchers is used, should identify who administered the questionnaire.
This can be extended to denote where and when the survey was carried out.
The best way to learn how to design a questionnaire is to complete one. An
example of a questionnaire is therefore given below as a demonstration.
Questionnaire
You have been set the assignment of answering the research question:
‘Why are the British so anti-EU’?
Examination of the Eurobarometer site has shown that support by the UK public
for the EU has fluctuated. But support is greater in Scotland and Wales. Overall
UK support reached its lowest following French and German refusal to join the
invasion of Iraq and widespread opposition to the proposed new constitution.
The UK government has deferred any decision on ‘joining the Euro’ (European
Monetary System) and any new constitution which would involve majority voting.
However, the UK government continues to support further ‘enlargement’ – possibly

to reduce Franco-German dominance. Your theoretical and literature review
leads you to argue that UK people are indifferent towards the EU and adopt a
rational choice perspective of economic self-interest. So, when the UK economy is
performing better than continental Europe, people become less enthusiastic about
the EU. You therefore reject the argument that hostility to the EU reflects an
increasing nationalism shaped by the anti-EU Murdoch press (News International
Corporation) whose front page headline was ‘Up yours Delors’ (when Jacques
Delors was head of the EU commission and favouring greater EU integration at
sub-national level to create a ‘Europe of the Regions’).
Your assignment requires you to design a questionnaire to administer to a quota
sample of 100 adults in Watersea (which happens to offer a microcosm of the
UK population).
Your initial design is:
Hello, can you help me? I’m a student at the university. We’ve been asked to find out why
British people are so anti-EU. I’d like to ask you ten questions. It will take five minutes.
I will not ask you your name or address, so whatever you tell me can’t be traced back
to you.
Q1. Do you live locally?
• Yes Code 1.1

• No 1.2
• Don’t know 1.3
• No response 1.4
[This is a filter question. If the answer is no or no response, then you should

thank them for their time and move away. If they reply that they don’t know, then
show them a map of the Watersea administrative area to discuss where they live and
whether or not it forms part of the city.]
Q2. How would you describe your overall attitude to the EU?
• Strongly supportive 2.1

• Supportive 2.2
• Opposed 2.3
• Strongly opposed 2.4
• Neither support nor oppose 2.5

• No response 2.7
Q3. So what would you prefer the British government to do in

the future?
• Become more involved with the EU, for example,

by joining the Euro at a date in the future 3.1
• Leave the EU 3.2
• Remain the same 3.3
• No response 3.5
Q4. Supporters of the EU claim the following benefits of

membership:
• A larger ‘common market’ for goods 4.1

• Greater protection for workers’ rights 4.2
• Freedom for British people to work abroad 4.3
• Greater anti-pollution controls 4.4
• Protection from ‘dumping’ by other countries 4.5
• Counter-balance to USA 4.6
[Pass show-card]
Which do you consider the two most important of these benefits?
• No response 4.8
Q5. On the other hand, critics of the EU say that the costs are
considerable. They argue that the disadvantages are:
• Loss of UK sovereignty 5.1

• Harm to Third World caused by tariff walls 5.2
• High price of EU food 5.3
• Common Agricultural Policy, wine lakes and butter mountains 5.4
• Freedom for Poles and other EU people to enter UK to seek work 5.5
• The ‘Brussels bureaucracy’ 5.6
[Pass show-card]
Which do you think are the two biggest disadvantages on this list?
• No response 5.8
Q6. I’d like to ask you how patriotic you feel. How proud are you to
be a resident of
A lot A little Not at all

The city (Watersea) ✷6.1 ✷6.2 ✷6.3
The county (Watershire) ✷6.4 ✷6.5 ✷6.6
The region (Yorkshire and Humberside) ✷6.7 ✷6.8 ✷9.9
England ✷6.10 ✷6.11 ✷6.12
United Kingdom ✷6.13 ✷6.14 ✷6.15
EU ✷6.16 ✷6.17 ✷6.18
Don’t know 6.19
No response 6.20
[Classifier data]
Some people claim that age, sex and background affect people’s attitude to the EU.
Q7. When were you born? Or, how old are you?
• Before 1939 68+ 7.1

• 1940–49 58−68 7.2
• 1950–59 48−57 7.3
• 1960–69 38−47 7.4
• 1970–79 28−37 7.5
• 1980–89 18−27 7.6
• No response 7.7
Q8. Which daily newspaper do you normally read?
• The Times 8.1

• The Telegraph 8.2
• The Independent 8.3

• The Guardian 8.4
• The Mail 8.5
• The Daily Express 8.6
• The Sun 8.7
• The Mirror 8.8
• The Star 8.9
• The Yorkshire Post 8.10
• Other 8.11
• None 8.12
• No response 8.14
Q9. What do/did you do for a living? [ask them for their occupation
which you allocate to the most appropriate group below]
A Higher managerial/administrative/professional 9.1

B Intermediate managerial/administrative/professional 9.2
C1 Supervisory, clerical, junior managerial/administrative/ 9.3
professional
C2 Skilled manual workers 9.4
D Semi-skilled and unskilled manual workers 9.5
E On state benefit, unemployed, lowest grade workers 9.6
(and students)
G Not classifiable 9.7
G Don’t know 9.8
G No response 9.9
Male 10.1
Female 10.2
Thank you for your help. Have you any questions for me or additional comments about
the EU that you’d like to make?
Note how the temptation has been avoided of asking the respondents’ knowledge
of the research office by, for example, asking: ‘who is Head of the European
Commission?’ A questionnaire should avoid being a quiz or asking questions which
seek to verify answers: don’t be too clever.
1. Pilot the questionnaire on at least two members of the public (not students). What
improvements would you suggest and why?
2. You are asked to test the hypothesis that age and education strongly influence
the likelihood of electors voting in general elections. You must use a ques-
tionnaire. Identify the sample frame. Devise a quota sample for an on-street
survey. Design, test and revise a questionnaire of no more than 12 questions.
Administer the questionnaire to ten members of the public in the main shopping
street. Comment critically on your experiences as a questionnaire writer and
questioner.
3. Discuss critically the claim that questionnaires can measure people’s attitudes
and beliefs.
4. Most of the national opinion polls published in the lead-up to the 1992 UK general
election wrongly forecast the outcome. Critically assess the various explanations
given by analysts.
FURTHER READING
Burnham, P., Grillard, K., Grant, W. and Layton-Henry, Z. (2004) Research

Methods in Politics. Basingstoke: Palgrave Macmillan. pp. 92–105. This
text provides sound, practical advice on questionnaire design and coding.
It includes memorable examples of poorly-worded or ambiguous questions
from national surveys.
De Vaus, D.A. (2001) Surveys in Social Research. London: Routledge.
pp. 80–105. This textbook differs from most others by emphasising the
importance of designing the questionnaire to measure the dependent and
independent variables, test variables and background measures. It also offers
additional guidance on telephone interviews and approaches to pre-testing
questionnaires.
Gilbert, N. (2003) Researching Social Life. London: Sage. pp. 85–104, 227–51.
Gilbert provides more pages of advice and examples than most other texts
and offers extensive guidance on pre-coding data.
pp. 46–55. Harrison discusses the methodological issues and provides rich
case studies including analysis of the errors made in the national opinion
polls leading up to 1992 when most of the pollsters’ forecasts (that the
Conservatives would lose) proved wide of the mark.
Yates, S.J. (2004) Doing Social Research. London: Sage. pp. 31–64. Whilst
coverage of the topic may seem slight, this book excels in the rich case studies
(including the 1992 UK polls) and the discussion of model answers to the
questions asked of readers.
Chapter 11
Observation
‘A politician can never [lie] flawlessly because their body language will always give
them away, psychologists say’.
(BBC News 5 September 2006)1
1. To consider the relevance of observation to Politics research.

2. To critically assess the claims made for a new ‘science of body language’.
3. To consider other claims which consider the origins of body language and
argue that, rather than being instinctive, body language is culturally specific
and often learned.
4. To consider how to systematically observe, record and interpret behaviour.
Introduction
A senior nurse pursuing doctoral research in Politics at York argued that too
much attention is paid by Politics researchers to what people say and write and
too little to how they behave and what they do. She described the importance
given in medicine to the systematic observation of patients as part of the initial
process of analysis and diagnosis and, subsequently, in assessing the efficacy of
any ‘intervention’ (treatment) and ‘management’ of illnesses (which are no longer
regarded as ‘cured’). Particular attention is paid to how the patients initially ‘present’
themselves: degrees of consciousness; whether they exhibit cheerfulness, sadness,
pain or anxiety; whether their faces are pale, perspiring or ‘florid’; whether they
walk with or without difficulty; whether they are obese, well-nourished or thin; the
attention they pay to their surroundings; hearing or other sensory difficulties; how
they treated their carers and medical staff, and so on. These initial observations
are noted and followed by measuring the ‘vital signs’ of blood pressure, pulse,
temperature, breathing and heart rate. This researcher argues that Politics researchers
should similarly observe, record and analyse how subjects present themselves as an
integral part of the data collection and analysis process.
Chapter 9 has already commented on the value of recording the appearance,

manner and surroundings of political elite or other subjects during elite interviews
or focus groups. But the main purpose of recording this information is to provide
richer material for your readers so that they can better share your experience.
The claim that this information can provide a basis for analysis is contestable.
The patient who ‘presents’ to the medical team does so because they or others
believe that their condition is abnormal and heightened. In most cases, hopefully,
the political elites who you interview will not normally be displaying noticeably
unusual behaviour. But when they do, it is quite right and proper that you note
this and consider its significance later. This is particularly important when you find
yourself or others affected by their charisma (or lack of it) or loathfulness or feel
intimidated by them. My own experience is that personality does count in arenas of
power. Dominant personalities are often able to drive through policy change which
their peers recognise as disastrous but find themselves – like rabbits frozen in the
headlights of a vehicle – unable to oppose. Mrs Thatcher and community charge
(poll tax) is an example of this phenomenon.
Historically, lobby correspondents (Parliamentary journalists) and political oppo-
nents have suggested that observation of the mannerisms of political elites provides
important clues to sincerity, truthfulness and real intentions. More recently, body
language has been developed and promoted as a scientific technique for the analysis
of intentions. Its disciples repeat the assertion that:
over 80% of effective communication is non-verbal.

(anon)
But what are the claims of body language? How can they be assessed? What is its
value to Politics research?
Body language: the claims
There is a large literature – especially in management science – which presents body

language as a science. It claims that mastery is essential for success as a negotiator,
salesperson or analyst. Duly equipped, the manager can detect liars and, as an added
bonus, simplify courtship by ‘reading the signs’ (or tells – from ‘tell-tales’). There
are a large number of books on the theme of Learning Body Language in a Week etc.
Body language has also attracted the attention of the news media who can use
their back catalogue of video-tapes to illustrate its application. Claims made are that:
Tony Blair unconsciously fiddles with his little finger whenever an opponent makes him
anxious … George Bush bites the inside of his cheek at anxious moments … Bill Clinton
Observation 163
tended to bite his lip when he wanted to appear emotional … George Bush walks like
a bodybuilder, hanging his palms to the rear as though laden by huge muscle, to imply
that the is larger than he is’ … [However] no amount of coaching or media training can
co-ordinate the hand gestures and facial expressions to fully cover up what a person
knows not to be true.2
You can’t fake [a genuine smile]. It’s too difficult.3
Expert opinion
Leading British commentators include Dr Desmond Morris: Oxford Fellow and

former Keeper of Mammals at London Zoo. His publication The Naked Ape (1967)
stimulated interest in body language and was the first of many publications. However,
his reputation may have suffered from later over-exposure in the news media and
the popularisation of his literature. However, unlike other ‘lite’ body language
publications, he presents well-argued explanations and encourages caution in the
interpretation of human behaviour.
Morris’ Manwatching (1977) begins with the claim that our everyday actions –
postures and movements – are performed unconsciously.4 In most cases, they are
a response to a perceived stimulus. He argues that the claims that all actions are
natural (innate) or all actions are nurtured (learned) are false. Blind, deaf and dumb
children will smile and cry. However, empiric evidence shows that the meanings
of other actions (e.g. crossed-fingers which historically reflect the crucifixion in
western societies) are unique to different cultures. He therefore begins by drawing
a distinction between genetic inheritance (innate) and discovered, absorbed, trained and
mixed actions.
Discovered actions are those which people develop as they mature. An example is
the folded arms posture which Morris suggests is derived from the ready movement
and hinging of the arms, and the comfort and protection provided.
Absorbed actions are those that are unconsciously copied. Copying is more likely
of higher than lower status groups. He cites the example of the camp posture
adopted by some adolescent homosexuals. But technology and fashion can play
a part. He attributes the informal posture of teenagers to the jeans that they have
adopted. So he argues that the same people wearing suits will adopt more formal
postures to avoid damaging the less-forgiving material. But he also attributes the
new informality to changes in philosophy and teaching which emphasise more
open styles and less overt deference, and the desexualisation and sexualisation of
dress.
Trained actions are those that require some teaching, e.g. winking, clicking fingers
and hand-shaking (although the distinction between absorbed and trained actions
seems perhaps over-stated).
Mixed actions have, as the term implies, a combination of origins. For example,
innate actions like crying are modified by social pressure to an occasional tear or
sniff in western society and wailing in other cultures. Another mixed action is the
realisation of the comfort derived from sitting cross-legged. But US culture regards
cross-leggedness as effeminate: hence, perhaps, the US practice of men sitting behind
the backs of the chairs to achieve manly postures. Social etiquette also affects the
precise form of actions.
Gestures are a sub-group of actions: they are adopted to visually transmit a signal
to the observer. Morris introduces a convenient classification of gestures as either
incidental or primary. Primary gestures are those which you perform alone. They
are non-social and convey no meaning. You rub an itch because it relieves the
discomfort. Winking, on the other hand, is an incidental gesture which provides mood
signals (in this case, collusion). Sometimes you deliberately use primary gestures as
stylised incidental gestures to convey mood, e.g. when you deliberately slump or yawn
to communicate your boredom and disrespect for the teacher.
There are six primary gestures of which five are unique to Homo erectus (‘Man’).
The exception is the expressive gesture. This is made by the face or hands. But, as we
mature, they can become modulated by cultural and gender influences.
Incidental – purposeful, communicating – gestures include mimic gestures. They
can be social mimicry (putting on a good face), theatrical mimicry (which imitates specific
actions or seeks to emulate the imagined mood of the part), partial mimicry (where we
imitate an entirely different species like a bird or cat) and what Morris terms vacuum
mimicry – which today you would call air mimicry like air-guitar-playing – where,
for example, you emulate drinking to indicate your wish to drink. Other types of
signal gestures include schematic, symbolic, technical and coded gestures. Schematic gestures
are abbreviated (e.g. using two fingers to emulate cigarette smoking). Symbolic signals
are essentially local social conventions: for example, tapping the head and pointing
in western society can mean mad, whereas in Saudi Arabia this signal is given by
lowering the eyelid. Technical gestures are made by specialist minorities where voice
communication is not reliable in potentially dangerous situations. They include,
for example, the hand-signals given by construction workers to crane operators,
media directors to presenters, and those given by bat-men to pilots of aircraft on
the taxiway. Although specialised, these gestures are intuitive. Coded gestures are
the formal language of sign language (of which there are several types), semaphore
and the bookmaker’s assistant (tick-tack man) at race courses. Some gestures may
be relics where they imitate an action outdated by technical change (e.g. indicating
a bad smell by holding the nose and imitating the action of pulling-the-chain to
the historic overhead cistern). Morris argues that the near-universal head-shake –
meaning ‘no’ – is derived from babies’ actions when they have had enough milk and
push the breast way. Similarly, the cigarette is an adult baby-comforter (dummy).
Baton signals have been called: ‘beating time to the rhythm of words’ (1977:56).
They are widely used by political elites. They include the open palm, the air grasp,
Observation 165
the hand-chop, the air punch, the raised fist and the palm back. The raised fist is widely
used to add vehemence (Martin Luther King) or aggressive intent (Mohammed Ali),
whereas the lowered, tight fist transmits unshakeable determination (Nixon). Guide
signs are used to indicate direction. They can indicate ‘go there’ (extended hand) or
‘come here’ (beckoning, hinged hand).
Gaze behaviour (eyeballing) is complex:
when two people meet and make eye contact, they find themselves in an immediate
state of conflict. They want to look at each other and at the same time they want to look
away. The result is a complicated series of eye movements … (1977: 71).
Gaze can also show dominance or submission. Passive dominance or passive

submission will involve looking away whereas active aggression or fear involves
intense looking (even glaring) at the other person. You are also likely to look away or
down when speaking to prevent other people from catching your eye and disrupting
you. So you indicate turn-taking (the alternation of dialogue between speakers) by
looking up and at the other person to show that you have finished and to give
them the opportunity to speak. But people can also lower their gaze when they are
concentrating. So a parent’s appeal to a child ‘to look at me when I’m talking to
you’ may be seeking the child’s submission rather than their attention.
Postural echo refers to the social mimicry that is demonstrated by people at ease
with one another. You will probably deduce from two people whose body language
is mirroring each other’s that they are developing a relationship. Body-contact tie-signs
refer to the degree of exaggeration of normal greetings like shaking hands. The
bone-crunching hand shake, the grasping of the other person’s hands with two
hands or gripping their arms can all be signs of special friendship or dominance
(but do close friends shake hands?). In contrast, auto-contact behaviour is the term
used to describe our actions when we touch ourselves. Common forms include
the jaw-support, chin support, hair clasp, cheek support, mouth touch and temple
support. Some – like touching the mouth – are popularly believed to be signs of
telling lies.
Morris includes lying as a type of non-verbal leakage: clues that you give away
unwittingly your real feelings or intentions. Like everyone else, you will lie in
your everyday life: you say that you are OK when you may be unhappy. Morris
argues that you best lie with those parts of your body with which you are most
familiar. So, because of your knowledge of your face from mirrors, you can lie more
easily by adopting a combination of words and facial expressions of, say, happiness
(putting on a brave face), etc. But you are less aware of the movements of our
lower limbs. So Morris advises that: ‘if you have to lie, do it over the telephone or
peering over a wall’. He disparages research that seeks to show that, when people lie,
they make more frequent hand movements, hand-to-face contacts, mouth-covers
and nose touches. Instead, he argues that these are physiological auto-responses to
pressure. These include the increase in the sensitivity of the nasal lining which nose-
touching reduces. So, these tells should never be regarded as conclusive. Similarly,
barrier signals – for example, touching shirt-cuffs or crossing limbs across the body –
are also more likely to be responses to a sense of nervousness. Examples of activities
stimulated by nervousness – displacement activities – include: jingling coins, clicking
biros and re-checking boarding passes (reportedly ten times more common among
air than rail passengers).
The most reliable indicator of mood appears to be eye-pupils. These react
spontaneously to changes of light and emotion. They cannot be controlled. Your
pupils will automatically narrow in brighter light or when you see an unpleasant
object. They will widen when you see something you like, for example, a baby or
a kitten. Pupil signals are both unconsciously given and received. (Hence, perhaps,
the appeal of dark glasses to some insecure political elites and poker players.) But
the difficulty remains actually being able to detect the width of pupils beyond very
close range and without appearing over-curious (which presumably would affect the
pupils under our observation). In any event, medical sources point out that pupil
dilation can also be a symptom of illicit drugs or amphetamines, drug overdose,
certain medications, eye drops, brain injury, stroke, brain tumour or, in extreme
cases, death.5
Morris concludes by discussing threat signals, dress, body adornments, infantile signals,
sexual signals and play patterns, etc. He also discusses (and appears to support) the
claim that, as human beings are the only female primates to have both hemispherical
breasts and buttocks, the female breast shape has evolved as a mimic of the other
primates’ buttocks which are the principal body parts used for sexual display In two
later books, Morris explores the distribution of signs and gestures in a large sample
of European and North Africa states (Morris et al, 1979).6 They show that, whilst
the same signs are common throughout Europe, they may have entirely different
and opposed meanings. For example, the ring sign means OK in the UK, Ireland,
Spain, Germany and southern Italy. But in France and Tunisia, the ring sign means
zero. Yet in Greece, Malta and Sardinia, it means orifice. In Tunisia, it means threat
(1979: 99–103). The influence of trade and cultural trade links is obvious from the
mapping.
Morris extends his study of body language worldwide by trying to identify signs
and gestures which are universal and thus more probably innate (Morris, 1994).7
Nearly 600 informal gestures are studied. Only 50 are worldwide. The gestures (G)
(and their meanings) are:
• arms akimbo: (keep away from me!)

• arms reach: (I offer you my embrace)
• belly rub: (hunger)
Observation 167
• body leant forward: (I am paying attention)

• body leant forward with hands gripping chair: (I am about to leave)
• chest tap: (me!)
• chin jut: (threat)
• chin lift: (I am above such things!)
• chin rub: (I don’t believe you: subconscious)
• chin withdraw: (fear)
• ear cup: (speak up!)
• ear rub: (I don’t wish to hear this: subconscious)
• ear scratch: (I am confused)
• ears block or cover: (stop the noise!)
• eyebrows flash: (greeting)
• eyes side-glance: (bold shyness or coyness)
• eyes stare: (threat)
• eyes weep: (distress)
• fist punch: (forceful emphasis)
• fist shake: (threat)
• footlock: (discomfort)
• finger beat: (moderate threat)
• forefinger point: (threat)
• hand chop: (I cut through argument)
• hand jab: (I insist)
• hand purse: (I am hungry)
• hands scissor: (that is finished)
• hands wring: (please, help me)
• head nod: (yes!)
• head support: (boredom)
• knees clasp: (I am about to leave)
• knees crossed: (I am very relaxed)
• ankles crossed: (I am politely relaxed)
• lips kiss: (love: possible relic of food passing)
• mouth smile: (pleasure)
• neck clamp: (I am angry)
• nose flare: (anger)
• nose tap: (I am – unconsciously – hiding something)
• nose up: (superiority)
• nose wrinkle: (disgust)
• palm up cupped: (please give)
• palm up flat: (pay up!)
• palms back: (I embrace you)
• palms down: (calm down)

• pupils dilate: (widen) (I like what I see)
• shoulders shrug: (I don’t know)
• stomach clasp: (I am hungry)
• tongue protrude: (insult)
• tongue protrude – slight: (I am concentrating)
Note that, in this book, Morris contradicts his earlier argument that nose-touching
could better be explained as a response to changes in the nasal lining caused by
nervousness. Conversely, some gestures that are acceptable in some cultures may
be unacceptable and insulting in others. For example, in western society, the
knees-cross is an indication of relaxation and confidence. But the extended knees-
cross will inevitably show the sole of the raised foot to onlookers (observers).
In Saudi Arabia, Egypt, Singapore and Thailand, this gesture is interpreted by
onlookers as showing that that they are considered to be the lowest of the low and
dirt-ridden. It is therefore especially offensive and can provoke extreme violence
(Morris, 1994: 77).
Recording and analysing body language
You will find many public meetings of many councils and health trusts, etc. and
court proceedings very dull. A useful way of spending your time is to record the body
language, and frequency and tenor of speeches. You can do this by drawing either
‘matchstick figures’ or, if you are more ambitious, sketching profiles. (You are not
allowed to photograph or record meetings of UK public bodies.) But drawing in this
way enables you to analyse what you would otherwise overlook in a photograph.
So it uses your time productively. Pay particular attention to posture and the attitude
of heads and hands. My experience is that people who sit close to the table and
remain still are primarily lending support rather than taking part in the deliberations.
They are usually backbenchers or officials. Most of the main protagonists will
adopt – literally – a laid-back approach resting back in their seats. They may appear
bored. Those leaning forward are likely to be aspiring councillors or others keen to
demonstrate their attack and defence skills to their leaders. Note how – in a new
tradition derived from the US – the most senior party member present will often
take off their jackets and sit in shirt-sleeves. Note the exaggerated, theatrical gestures
and how they appear to lose their tempers. Watch the young male councillor who
leans back with his hands behind his head indicating his self-satisfaction. Note how,
when party leaders say something even slightly funny, the others laugh loudly. This
courtier-like behaviour is a classic sign of dominant power. Compare this to when
a junior councillor says something really amusing and no one laughs until the leader
Observation 169
Figure 11.1(a) Simple matchstick drawing
leads or halts the laughter. Look at the mutual dependency between the senior
councillors and the senior officers. Record who speaks to whom and the tenor
(volume, pitch, rhetorical style, etc.) of their voices. Show these as thick or thin
arrows in your record. Note that, despite the apparent intense arguments between
the parties, they readily enjoy one another’s company after the meeting. What you
have seen is the elaborate role play of the ritual of democracy: the real decisions are
often taken privately elsewhere.
Examples of matchstick figures and more representative drawings and annotation
are in Figure 11.1(a)-(d). These drawings and analysis were made from sketches made
at a meeting of a local council meeting. The same techniques can also be applied to
Figure 11.1(b) Simple representative drawing
Figure 11.1(c) Detailed representative drawing

Observation 171
Figure 11.1(d) Communication chart
focus groups although the initial sketches should best be made from memory after
the meeting.
Conclusions
The principal conclusion must be that – perversely – body language is an inconclusive

indicator of underlying behaviour. As a famous US movie star remarked: ‘there is
less to this than meets the eye’.8 There are relatively few innate gestures. But those
which are innate can be consciously emulated for effect. Similarly, many of the
indicators of insincerity – like nose- or mouth-touching – may be more readily
attributed to nervousness. They can also be restrained. It can also be argued that, in
our televisual age, body language has become learned and exploited by political elites
to enable them to communicate their own ‘spin’ on events more readily. Similarly,
the news media are able to extract seconds of tape from hours of film to illustrate
any point. So the body language of non-elites and others untrained in the artistry
of body language may be more revealing. In any event, body language is a classic
case where the observations should be recorded separately and before any analysis
is attempted or conclusions claimed. Perhaps the most important conclusion that
we should draw is of the potential for our own repertoire of gestures to antagonise
people from other cultures during fieldwork. So, in the final analysis, self-awareness
and restraint of your own body language during fieldwork may be more important
than your observation of others’ behaviour.
1. What significance should be attached to the observation and interpretation of body

language in Politics research.
2. Review the outcome of a search on the Internet for ‘body language pictures’. Using
a data projector, display and discuss your interpretation of the various catalogues
of pictures from sources including the BBC.
3. Record, project, discuss and analyse the non-verbal communications techniques
used in the most recent wave of party political broadcasts.
4. Attend a meeting of your local council or NHS trust. Note and comment on the
body language displayed by the main participants when they are speaking and
listening to others.
5. ‘A politician can never [lie] flawlessly because their body language will always give
them away’ (Collett, 2006). Discuss using examples and illustrations.
FURTHER READING
Morris, D. (1977) Manwatching; A Field Guide to Human Behaviour. London:

Jonathan Cape.
Morris, D., Collett, P., Marsh, P. and O’ Shaughnessey, M. (1979) Gestures: Their
Origins and Distribution. London: Jonathan Cape.
Morris, D. (1994) Bodytalk: A World Guide to Gesture. London: Jonathan Cape.
Notes
1 http://news.bbc.co.uk/1/health/5316916.stm (10 October 2006).

2 http://news.bbc.co.uk/1/health/5316916.stm (10 October 2006) Quotations attributed to Dr Peter
Collett, formerly of the University of Oxford speaking to the British Association’s Science Festival in
Norwich. He co-authored with Desmond Morris the book Gestures: their origins and distribution
(1979). See end-note 5 below.
3 http://news.bbc.co.uk/1/health/5316916.stm (10 October 2006) quotation attributed to Prof.
G. Beattie, University of Manchester.
4 Morris, D. (1977) Manwatching; A Field Guide to Human Behaviour. London: Jonathan Cape.
5 www.wrongdiagnosis.com
6 Morris, D., Collett, P., Marsh, P. and O’Shaughnessey, M. (1979) Gesture: Their Origins and
Distribution. London: Jonathan Cape. Collett and Marsh were members of the Department of
Experimental Psychology at Oxford. O’Shaughnessey is a language graduate.
7 Morris, D. (1994) Bodytalk: A World Guide to Gestures. London. Jonathan Cape.
8 www.centralsquare.com/quotations.html
Part IV
Data Analysis
Chapter 12
Analysing Research Data: The
Process
‘A single swallow doth not a summer make.’

(Greek proverb cited by Aristotle in Nichomachean Ethics, I, vii, 16, fourth
century BC)
‘One swallow doth not a summer make. But two swallows might attract a research
grant, and three are enough to work up a paper in a marginal journal of climatology …’
(James Meek, The Guardian, 9 November, 2000)
‘Politics researchers tend to draw summers of conclusions from single swallows of data.’
(Anon)
1. To understand the weaknesses and limitations of earlier methods of analysis.

2. To consider the recommendations made by authorities.
3. To develop a general hierarchy of analysis for Politics research.
Introduction
Traditionally, the analysis by Politics researchers of the information collected –

especially qualitative information – has been weak if not altogether non-existent.
Previously, the research has been characterised by:
• a loose, implied methodology

• the absence of supporting evidence (i.e. critical supporting information) in the form of:
• transcripts of interviews or meetings
• other accounts of meetings
• records of other empiric observations
• instead, generalisations are made which are sexed-up (exaggerated) with quotable
quotes (which are not cross-referenced to sources and could be entirely fictional)
• analysis and discussion are given without cross-reference to the information on which
they rely
• evidence is not compelling: acceptance is based almost entirely on trust: no opportunity
or means are given to corroborate the findings
• much statistical analysis assumes an in-depth knowledge of the mathematical routines,
formulae and symbols used, and may be impenetrable to most UK students; the rationale
and meanings of the techniques adopted go without comment, explanation or indication
of confidence limits
• quantitative and qualitative methods of analysis have been seen as entirely different and
lacking any common process
• research findings are rarely persuasive to those readers not otherwise subscribing to the
implicit assumptions adopted
• consequently, much Politics research tends to confirm or qualify existing explanatory
theory in an ever-decreasing circle
You will be familiar with the axiom: ‘rubbish in: rubbish out’.
Five guidelines for effective analysis are suggested:
1. The data never speaks for itself: ‘all knowledge is theoretically impregnated’ (Silverman,
2001: 23)1 : your analysis and interpretation are always required.
2. Similarly, you must make the process of analysis transparent if the critical reader is to
share your findings: trust and belief in your scholarship and honesty are not enough.
3. Ensure that the analysis and interpretation are proportionate in depth and extent to the
information to which they are applied.
4. The means of analysis you use must be relevant and appropriate to the research
question and to the information collected; you must justify using them.
5. You must identify, consider and justifiably discount (discard) other explanations before
you claim firm conclusions.
You may have gathered the data in quantitative (numeric) or qualitative (non-
numeric) form. However, the former distinction between quantitative and quali-
tative analysis has been overtaken by technological development. Whilst numeric
data can only be analysed by quantitative means, you can now analyse qualitative
data by qualitative or quantitative methods.
The process of analysis
What do we mean by analysis? What does it involve? Oddly enough, the authorities
are largely silent – presumably, on the assumption that, after all, we all know what
Analysing Research Data: The Process 177
analysis means. However, the following characteristics of analysis can be inferred

from the discussion they provide:
• analysis is an objective-seeking, rational process

• in academic research, the objective is to answer the research question either by
confirming or infirming the explanations set out in the hypothesis, or, in inductive research,
the relationship, if any, between the key variables
• analysis essentially involves deconstruction – breaking down the whole into its
constituent parts
• analysis involves a systematic, deliberate, planned examination of these parts, and of
the relationship between them
• analysis invariably involves some comparison between variables
• analysis seeks to identify patterns or rhythms of relationships
• analysis also involves creative interpretation
• testing is used to eliminate any spurious relationships
• the analysis concludes by inferences being drawn. These are generalisations about the
research population drawn from samples or other limited information. For example, if your
study of a sample of South American states has shown a causal, contributory relationship
between, say, levels of interpersonal trust and economic performance, then how likely are
you to find this same relationship in other countries? These inferences must be followed
by discussion before firm conclusions can be drawn.
The greatly increased interest in analysis has spawned a very good literature,
including Miles and Huberman (1994), and Ritchie and Lewis (2003).2,3 However,
this emerging literature tends to concentrate exclusively on either qualitative
or quantitative data analysis and implies that they adopt divergent approaches.
Exceptionally, Neuman (2003) provides a valuable review of similarities and
differences which have been tabulated in Table 12.1.
A hierarchy of analysis
Despite the differences, it is possible and useful to suggest a general, generic model
of analysis (which draws in part from Ritchie and Lewis Analytical Hierarchy, 2003:
212, Box 8.1). The model has five, iterative, component phases: raw data assembly,
validation and reduction; classification, coding and sorting; testing; inference-drawing; and
theory development. The process can be illustrated as a pyramid (after Maslow)4 which
reflects the successive reduction and refinement of the volume of raw data into
information and evidence.
Table 12.1 Qualitative and quantitative analysis compared
Similarities between Qualitative and Quantitative Analysis

1 Both involve making inferences
2 Both methods of analysis involve ‘a public [transparent] method or process’
3 Comparison is central to all analysis
4 Researchers ‘strive to avoid errors, false conclusions, and misleading inferences’
Differences between Qualitative and Quantitative Analysis
Quantitative Analysis Qualitative Analysis

Uses standardised, statistical techniques Less standardised
Delays analysis until all data collected Begins earlier before all data collected
Tends to test hypotheses Can create new concepts and theories
Low level of abstraction High level of abstraction
Source: (abstracted from Neuman, 2003: 439–40)
Δ Theory development
Inference drawing from evidence
Testing
Information classification, coding and sorting
Raw data assembly, validation and reduction
Figure 12.1 Hierarchy of analysis
Raw data assembly, validation and reduction
This first phase involves you bringing all the raw data together at the same place and
time. The objective here is to reduce the mass and mess of data in sheets of text,
photocopies, emails, scoping documents, transcripts, letters, memos, questionnaires,
etc., audio and video-records and photographs. You should selectively discard all
the non-relevant and duplicate data to a ‘recycle bin’, transposing the relevant
data into a common format (e.g. Microsoft Office) and organising it into folders
and files. The files are then validated. This involves checking the data for errors
Analysing Research Data: The Process 179
or omissions. It may also involve you anonomysing people and places (hiding their
real identities). You should also identify and flag (mark) outlier data. Outliers are
extreme and exceptional data which lie outside the expected range. For example,
where a new leader has shown an increase in support of, say, 5–15% in most opinion
polls, then you might classify an increase in one poll of, say, 25% as an outlier.
But, you should check first whether it is exceptional because of some recording
or transposition error, or may represent a significant finding. Where outliers are
considered dubious, then you can discard them. By the end of this initial process,
you will have reduced and organised the mass of raw data into useful information
ready for coding.
When this first stage is complete, you will be able to begin the processes of
classification, coding, testing and inference drawing. These processes are very different for
quantitative and qualitative information and are therefore described separately in the
following chapters.
1. What do you understand by ‘analysis’?

2. The suggested hierarchy of analysis seeks to reduce (raw) data to (reliable)
information, and this information to evidence. An operational distinction between
these is suggested in Chapter 7:
Commentators and politicians tend to use the word ‘evidence’ to describe

what they would wish us to regard as ‘conclusive, compelling information’
which either proves or, in its absence, disproves allegation. But is there a real
difference between data, information and evidence? Certainly, the dictionary
meanings are similar. But some distinction is useful. Researchers tend to
speak of data as the mass of disordered, raw material from which information
(knowledge) is abstracted to provide evidence to support argument and
conclusions. Information informs. Evidence supports conclusions. So it is
helpful to conceive of research of involving three stages. First, the raw data is
gathered. Second, the data is organised and distilled into information. Third,
evidence is abstracted for or by analysis from the information.
Is the suggested distinction between data, information and evidence useful,

exaggerated or dangerous in the context of analysis?
FURTHER READING
Spencer, L., Ritchie, J. and O’Connor, W. (2003) Chapter 8: Analysis: Practices,

Principles and Processes. In Ritchie, J. and Lewis, J. (eds.) Qualitative
Research Practice: A Guide for Social Science Students and Researchers.
London: Sage. pp. 199–218. Whilst this textbook is written for a readership of
social scientists undertaking qualitative research, this chapter provides a very
succinct account of the principles underlying analysis which is of value to
Politics researchers pursuing quantitative, qualitative or mixed methods.
Notes
1 Silverman, D. (2001) Interpreting Qualitative Data: Methods of Analysing Talk, Text and Interaction.
London: Sage.
2 Miles, M.B. and Huberman, A.M. (1994) Qualitative Analysis: An Expanded Sourcebook. London:
Sage.
3 Ritchie, J. and Lewis, J. (2003) Qualitative Research Practice: A Guide for Social Science Students
and Researchers. London: Sage.
4 Abraham Maslow’s Hierarchy of Needs (1943), see Maslow, A.H. (1970) Motivation and
Personality. New York: Harper and Row.
Part IV
A: Quantitative Analysis
Chapter 13
Calculating and Interpreting
Descriptive Statistics
‘Lies, damned lies and statistics’, attributed to Benjamin Disraeli by Mark Twain.
1. To understand what is meant by descriptive statistics.

2. To learn the separate functions of the mean, mode and median, range,
interquartile range, variance and standard deviation.
3. To learn how to calculate these using MS Excel.
Introduction
Descriptive statistics resolve complexity by summarising and compressing data

to identify their essential characteristics to create a brief but relatively accurate
impression to the observer. You already do this in everyday use by describing
the total number of, say, your fellow students’ range of ages, their average age,
male/female split and ethnic composition.
Consider this example. In reply to a question from your parents, you tell them
that your university class consists of about 20 students, half are women and six come
from the Indian sub-continent. You add that most students are about the same age.
(Twenty years old but there are three older students.) The oldest is 48. This creates
a reasonable picture in their minds. You use this rough description or numeric
shorthand to avoid having to describe each student individually.
However, more accurate statistics may be required – by say, by the registrar’s
staff – giving the precise average age. The ages of the class are (in ascending order):
18,18,18,18,19,19,19,19,19,19,19,19,19,19,19,19,20,20,20,36,45,48.
There are actually 22 students. To calculate the average, you must, of course,
add all the ages together and divide the sum by the number of students. In this
case, the total ages are 489. There are 22 students. So the average is 489/22 =
22.23 years.
So what? Well, in this example, none of the students is aged 22. So the average
is misleading. The reason for this apparent discrepancy is because the class contains
much older students. This is more obvious when the students and their ages are
individually plotted on a type of graph or chart termed a scattergram:
In this simple example, the average has been distorted by the inclusion of three
much older students. But, suppose that one of the 18-year-old students were to
‘drop out’ and an 85 year-old admitted. In this case, then the average mean would
increase to 25.3 and the graph to:
60
50
40
Ages
30
20
10
0
0 10 20 30
Individual students
Figure 13.1a Scattergram of students and their ages
90
80
70
60
50
Ages
40
30
20
10
0
0 10 20 30
Individual students
Figure 13.1b Scattergram of students and their ages (revised class)

Calculating and Interpreting Descriptive Statistics 185
In this second case, the calculation of the average has been grossly distorted by the
replacement of a more typical student by a very much older one. This is called the
outlier effect – the effect of those highest or lowest terms to skew – distort – your
everyday, ‘average’ mathematical description of complex information. (Incidentally,
some quantitative researchers argue that qualitative researchers rely on outliers; the
exceptional case is more attention-grabbing than more typical information.)
To overcome this weakness, statisticians use a set of simple statistical measures to
better describe, say, a student class. These are the arithmetic mean, median, mode, range,
variance and the standard deviation. These are termed measures of central tendency
(or measures of dispersion). As the terms imply, they describe numerically the extent
to which the individual terms cluster around the ‘centre’. The class is termed the
population: the total group of people or events being described or under study. The
population is represented by the Roman capital, N. The individual measurements
of data from the population are called terms. A group of terms which measure the
same characteristic at the same moment in time is called a series.
The arithmetic mean is what is generally called the average, i.e. the sum of the
terms divided by the number of terms. In statistics, the terms are called X and the
number of terms N . ‘Sum of’ is represented by the Greek symbol " (pronounced
sigma). The arithmetic mean is called X bar and represented nowadays by the symbol
X . So X = "X N
The median is the middle term when the series is ranked (normally in ascending
order), e.g. 4,10,2,8,6 becomes ranked into 2,4,6,8,10. The median is the (N+1)th
term/2. So, the series 2,4,6,8,10 has five terms. Therefore the median is the
(5+1)/2th term = 3rd term. The third term in the series is 6. If the series were reduced
to four terms – 2,4,6,8 – then the median would be the (4+1)/2th term = 2.5. In this
case, the median is calculated as lying midway between the second term (4) and the
third term (6). So the median is 5.
The mode is the most frequent term in the series, e.g. in the series 1,3,6,5,1,2 then
the mode is 1. This is rarely used.
Returning to the example of the initial class of 22 students, then, whilst the
arithmetic mean may be 22.3, the median is the (22+1)/2th term, i.e. the 11.5th
term or half-way between the 11th and 12th term. In this case, the 11th and 12th
terms are both 19 so the median is 19. The mode (the most common term) is also 19.
So, in this example, the median and the mode provide better descriptions than the
mean (of 22.23). In the revised class, the median and mode remain 19 and the effect
of the (85 year old) outlier is effectively discounted. In cases like these where most
terms lie below the mean, then the data – or distribution – is termed positively
skewed. In many universities, the final degree classification uses the median exam
mark rather than average mark. In this way, exceptionally good or bad exam marks
are discounted.
Another statistical measure of the data is the range – the difference between the
highest and lowest term. However, as noted earlier, the range can be distorted by
exceptionally high or low outliers, e.g. the very mature student. In the revised class,
the range would be 67 (i.e. 85–18). So statisticians developed the interquartile
range. This is the difference between the first and third quartiles when the terms of
the series are placed in ranked order. So in a series of 99 terms, the interquartile
range is the difference between the 25th and 75th term. The first and third quartiles
are identified in a similar way to the median by calculating
(N + 1) 3(N + 1)
and
4 4
For example, in our initial class of 22 students ranked by age, then the interquartile
range is the difference between the ages of the 5.75th and 17.25th students, i.e. 19
and 20 = 1.
The greatest use of median measurements probably lies in representing unequal
distributions especially in terms of resources. Income is a prime example where a
small number of people may have vast wealth and, at the other end of the scale, a
large number virtually nothing. In these cases, the inequality is shown by calculating
and comparing the income of the 10th and 90th ‘percentiles’. So, in a fair society
where incomes are equal, then the 10th and 90th percentiles will be the same
whereas, in less equal countries, the ratio of 10th to 90th percentiles may be as high
as 100. A more sophisticated descriptive statistic for measuring unequal distribution
of income and wealth is provided by the Gini coefficient. This is the ratio between
the areas above and below the Lorenz curve (of cumulative incomes) and the area
of equal distribution. The coefficient varies between less than 0.25 (Greenland)
and 0.6 (Namibia). No coefficients are calculable for Sub-Saharan Africa. The Gini
coefficient for the UK is 0.35–0.39.
In the UK, the official (upper) level of poverty is calculated as 60% of median
household income. ‘Deep poverty’ is calculated as 40% of the median household
income. The definitions reflect a rejection of the historic concept of poverty being
calculable as an absolute level in favour of relative measures.
Lorenz curve for a typical Complete income equality

income distribution
Cumulative Cumulative
income income
share (%) share (%)
0 0
0 100 0 100
A Population share (%) B Population share (%)
Figure 13.2 Charts of income distribution (from www.statistics.gov.uk)

However, the median itself can be misleading. Take for example, data obtained
from applying the Likert scale to the question:
Q. To what extent do you agree with the following statement:
The Anglo-American invasion of Iraq has actually increased the threat of terrorism
in both countries. Do you strongly agree, agree, neither agree nor disagree, disagree,
or strongly disagree? You code the possible answers 1–5 (where 1 is strongly
agree, etc.). Let us say that you ask three samples of ten people of three different
age groups (in practice, you would use much greater samples). The results (ranked
in descending order) are:
Young people 5 5 5 4 3 2 1 1 1
Middle-aged people 4 4 3 3 3 2 2 1 1
Older people 4 3 3 3 3 3 3 3 2
In each of these three groups, the median is 3 (neither agree nor disagree) although
the intensity of agreement or disagreement is significantly higher among the sample
of younger people. To overcome this problem, mathematicians developed the
variance.
The variance
The variance is a descriptive statistic which measures the degree of concentration

or dispersal of the terms around the mean. This is found by calculating the difference
between each term and the mean, i.e. X − Xi . In some cases, the difference will
be +, in others, –. (Indeed, if added together they should cancel each other out,
leaving an answer of 0.) So, to overcome this difficulty, the differences are squared –
thus eliminating the minus quantities. The variance is then found by adding all
these squared differences and dividing by the number of terms. The formula for the
variance is:
"(X − Xj )2
where j is the number of terms in the series
N
This sounds more complicated than it is. Consider the example above of the Likert
scores for the sample of ten, younger people:
Young people, X 5 5 5 4 3 2 1 1 1
Arithmetic mean, X = 3 3 3 3 3 3 3 3 3
Difference between term (5−3) (5−3) (5−3) (4−3) (3−3) (2−3) (1−3) (1−3) (1−3)
and mean (X −X ) = +2 +2 +2 +1 0 −1 −2 −2 −2
Difference 2 (X −X )2 = 4 4 4 1 0 1 4 4 4
Sum of difference2 "(X − X )2 = 26
"(X − Xj )2 26
Variance = = 2.6
N 10
Standard deviation
A further development of the variance is the standard deviation, SD or σ which

is simply the square root of the variance. So the formula is:
!
"(X − Xj )2
where the value of σ is ±
N
N and n
You will note that, in most scientific research publications, it is the number of terms
and standard deviation which is normally given to describe experimental data, e.g.
n = 1600, σ n−1 = ±12.1.
However, the symbol n rather than N is given. Why? The answer is that the
symbol N is used to denote the number of terms – people or events, etc – in the
population: n is used to denote the number of terms in the sample. Similarly individual
terms in the sample are denoted by x rather than X .
The symbol for the standard deviation of the sample is also written in lower case
as σ n−1 . But why is the n − 1 notation added? The reason is that, where the series
is a sample of the population, then the number of terms may be comparatively
small. To compensate for the small number of terms and the potential experimental
error, the number of terms is notionally reduced by subtracting 1 to give a larger
standard deviation. To distinguish a sample from a population, the symbols x, n
and σ n−1 are used. So the formula for the standard deviation of a sample of n
terms is:
!
"(x − xj )2
(n − 1)
Grouped frequency distribution
Data – whether populations or samples – may contain a large number of terms.

To simplify their representation and analysis, the terms are translated into what is
called a grouped frequency distribution. For example, a small sample of eight
families may show the following numbers of children in each: 2,1,2,3,2,1,0,4, i.e. a
total of 15 children. These can then be tabulated in a grouped distribution as:
Table 13.1 Grouped frequency distribution table
Number of Children in Number of Families Total Children

Each Family
x f fx
0 1 0
1 2 2
2 3 6
3 1 3
Total families, n = 8 Total children = 15
The measure in which data is grouped is called an interval. In the example above,
the interval used is the number of children. The most commonly used age intervals
in Politics research are years, i.e. 1998, 1999, etc., age groups, e.g. 0–17, 18–28,
29–38, 39–48, 49–58, 59 +. , and incomes, e.g. $0– $999, $1000 – $1999, etc.
Grouping data by frequency distribution also enables the data to be displayed as
a graph, e.g.
3.5
3
Number of families
2.5
(Frequency)
2
1.5
1
0.5
0
0 1 2 3 4 5
Number of children
Figure 13.3 Chart of Grouped frequency distribution table: Frequency of children in

families
In nature, charts of grouped frequency distributions of most phenomena (e.g. weight

of adults, their heights, intelligence, shoe sizes, etc.) form a normal distribution.
Examine Fig 13.4. In this case, the mean (average) number of clubs etc. to which
members of the sample (of 1,000) were members is 12. Almost 100 of the sample
belonged to 12 clubs whilst less than 10 belonged to only one club.
Note how the normal distribution (or Gaussian distribution) is characterised by a
symmetrical, bell-shaped curve in which the mean, median and mode coincide at the
most frequent term. The normal distribution is also asymptotic, i.e. the frequencies
approach but never reach zero.
100
Frequency (memberships)
80
60
40
20
0
0 10 20 30
Number of clubs joined
Figure 13.4 A normal distribution: Membership of clubs
How to calculate descriptive statistics
Most descriptive statistics can now be calculated using a good calculator, for example,
the Casio fx-83 series (which can be bought for as little as £6.99). However, most
students are more likely to have a PC or laptop readily available than a calculator.
Unlike a calculator, a computer can save the data and calculations as a file and enable
them to be pasted (copied) into a report. Whilst a wide variety of statistical software
is available, most students use the MS Office software package which includes the
spreadsheet Excel. Most students are also familiar with using Excel which is quite
adequate to perform descriptive statistics.
Before using Excel to perform statistical calculations, it is necessary to add the
Analysis ToolPak – which contains a large number of specialist statistical formulae –
bundled into the software. This is relatively straightforward. Open an Excel spread-
sheet. On the menu bar, click on Tools. Then click on Add-Ins. Click on Analysis
ToolPak and Analysis ToolPak VBA at the head of the dialogue box and click OK.
You can now enter your data. Insert the term Income in cell A1. Then insert the
following data in cells A2:A11 in Column A. This is a small sample from a northern
benefits office of the annual household income of current single-parent claimants
having three children for income support. In practice, of course, we would seek to
use a much larger sample of 35 terms or more for greater accuracy.
Claimants’ household income:
Cell A1: Income £

Cell A2: 12,317
Cell A3: 14,861
Cell A4: 8,003
Cell A5: 11,370
Cell A6: 12,561
Cell A7: 13,889

CellA8: 15,270
Cell A9: 18,241
Cell A10: 25,279
Cell A11: 19,671
Save the data as a file titled Claimants’ Household Income. We want to calculate
the mean, median, tenth and ninetieth percentile, range, variance and standard deviation.
The first step is to copy the data onto a nearby column, E using the Copy/Paste
routine. This safeguards your original data. Your next step is to critically examine
the data. This is best done by sorting the data into ascending order and by drawing
a chart (graph).
To do this, first highlight the data (including the title ‘income’), then click on
Data and Sort. In the sort dialogue box, click OK. The data will now appear in
ascending order. Leave it highlighted. Then click on Insert and Chart. Click on X–Y
(Scattergram). Click on Next until the graph appears.
30,000
25,000
20,000
Income
15,000
10,000
5,000
0
0 2 4 6 8 10 12
Benefit claimants
Figure 13.5 Chart of distribution of claimant’s household income
This graph shows that the distribution of incomes is relatively straightforward and
that there are no ‘outliers’ or other unexceptional terms.
To calculate the descriptive statistics, click on Tools, then Data Analysis, then
click on Descriptive Statistics in the dialogue box and OK. The Descriptive Statistics
menu will appear. In the box marked Input Range, insert E2 : E11. (Whilst there
is a short-cut alternative to inserting the cell numbers, it can complicate routines at
this initial stage.) Click on Grouped by columns. The Output Range is the cell where
you want the statistics to appear: insert G1. Then place a tick in the box marked
Summary Statistics. Finally, click OK. The table of statistics below will appear:
By reading this table and rounding up to one place of decimals, we can report that:
mean income = £15,146.2

median income = £14,375
Table 13.2 Descriptive statistics output of claimants’ household income data
Income £ Column1
8,003
11,370 Mean 15146.2
12,317 Standard error 1543.355
12,561 Median 14375
13,889 Mode #N/A
14,861 Standard deviation 4880.518
15,270 Sample variance 23819453
18,241 Kurtosis 1.035046
19,671 Skewness 0.839166
25,279 Range 17276
Minimum 8003
Maximum 25279
Sum 151462
Count 10
standard deviation = ± £4,880.5

variance = 23819453
range = £17,246
In this case, the summary statistic for the standard deviation calculated is the standard
deviation of the sample. If the data were a population, then you would have to use
a different formula =STDEVP( ). Click on the cell where you want the standard
deviation of the population to appear. Insert =STDEVP(A2:A11) on the task bar
next to the fx notation. Press the Enter key. The answer 4630.066 will appear.
To record the meaning of this data you will have to enter standard deviation of the
population in the cell above or to the right. (If you use a cell to the left then the title
will be partially obscured unless you increase the column width.) Note that the
standard deviation of the population is smaller than the standard deviation of the
sample.
To calculate, the 10th and 90th percentile, click on Insert and then Function. The
Insert Function box will appear. In the first box headed Search for a function, insert
percentile and then click on Go. PERCENTILE will appear highlighted blue on
the larger box below. Click on OK. A box marked Function Arguments will appear.
In Array, insert E2 : E11. In k, insert 0.1 (meaning 10%). As soon as you insert this
data, the answer will appear to the right of Formula result = 11,033. Change the k
value to 0.9 (i.e. 90%) and the formula result will be 20,232. So we can report that:
10th Percentile = £11,033

90th Percentile = £20,232
You will appreciate that it makes no difference to the practical ease of completing
the calculation whether the spreadsheet contains – as in our example – 10, or 10,000
terms.
Types of numbers
In the example above, numbers have been used to represent incomes. However,
there are two different groups and four different types of number that you can use.
The two groups of numbers are categorical and cardinal.
Categorical numbers
Categorical numbers involve nominal numbers and ordinal numbers (rankings).
Your university number is an example of a nominal number. Your university number
gives you a unique identity and enables your name to be replaced by a number. If
your university number is lower than a friend’s then it does not mean that you
are better or worse in any respect. Nominal numbers have no mathematical
significance: they are entirely nominal. In contrast, ordinal numbers are used to
indicate rankings. But their significance is confined to some ranking relative to a list
of criteria, e.g. 1 may indicate the person with the highest mark.
Cardinal numbers
Cardinal numbers involve interval and ratio scales. Ratio scales have an absolute
zero, e.g. income, time, degrees Kelvin (degrees centigrade of Fahrenheit are
interval scales). Interval scales measure the variance above or below a particular
benchmark. Ratio and interval scales can be negative. For example, an annual
income of −£500 means that you owe someone else £500. Accountants express
this as (£500).
Wherever possible, quantitative researchers seek ratio scales.
Where variables are expressed as categorical or cardinal numbers they can be termed
categorical variables or cardinal variables. Variables may also be discrete or continuous.
Discrete variables are those where there is only a limited number of intervals in
the range. For example, the numbers of children in a household are discrete because
you can only have zero, one, two, etc children: you cannot have 1.4 children
(although that may be the arithmetic mean of children in all households). In contrast,
continuous variables are those where there are an infinite number of intervals in
the range. So, for example, the ages of the children in a household are continuous:
even twins will have different ages when measured in terms of minutes or seconds.
Generally speaking, categorical variables are discrete whilst continuous variables are
cardinal.
Whilst the distinctions between categorical and cardinal numbers and discrete
and continuous variables may appear academic, they are important. For example,
the descriptive statistics in the example above can only be calculated using cardinal
numbers. As you will see in Chapter 15, the type of number will dictate which
advanced statistical formulae can be used.
Questions for discussion, workshop tasks or assignments
1. Access a UN, World Bank or EU data bank on the Internet, e.g. http://unstats.un.
org/unsd/cdb/cdb or http://mdgs.un.org/unsd/unsd/ Choose a table of interest to
you. Copy it to Excel. Calculate the summary descriptive statistics, 10th and 90th
percentile. Compare the mean and median and explain the difference.
2. Create your own file of 1,000 random numbers.To do this click in turn on Tools in the
menu bar then Random Number Generation. In the Random Number Generation
Dialogue box, enter:
Number of Variables (columns) = 10

Number of random numbers (rows) = 100
Distribution = normal
Mean = 50
Standard deviation = 15
Random number seed =5
Output Range = A1
Then click ‘OK’
Save as (file) Normal distribution random numbers. Then calculate the summary
statistics for the entire data set. Retain the file for future use as your own bank of
random numbers.
3. Now analyse the random numbers above into a grouped frequency distribution.
First, decide what class interval you wish to adopt. Your data extends from 0 to
100.You may decide to use a class interval of 5.Then enter 5 in cell K1 and continue
to enter 10, 15, 20 etc until you reach 100 in cell K20. (You can, of course, use
the Autofil routine.) Then click on Tools and Data Analysis. In the Data Analysis
dialogue box, click on Histogram and OK. In the Histogram dialogue box, insert
Input Range A1:J100

Bin Range K1:K20
Output Range L1:L20
Then click on Chart Output and click OK.

The output and chart will appear:
Table 13.3 Output and chart
5 Bin Frequency
10 5 12
15 10 12
20 15 20
25 20 32
30 25 42
35 30 48
40 35 76
45 40 78
50 45 79
55 50 101
60 55 92
65 60 89
70 65 83
75 70 77
80 75 53
85 80 41
90 85 32
95 90 12
100 95 9
100 7
120
100
80
Frequency
60
40
20
0
5
20
35
50
65
80
95
Bin
Figure 13.6 Histogram

You will note that the histogram apparently takes the shape of a uniform
distribution. However, as we constructed our 1,000 cell bank of random numbers
from a normal distribution then the chart is reassuring rather than surprising.
But note how, despite having been calculated from 1,000 random numbers, the
histogram does not follow a perfect uniform distribution: there will always be some
variations from the perfect normal distribution when bins are used.
FURTHER READING
Clegg, F. (2005) Simple Statistics: A Course Book for the Social Sciences.
Cambridge: Cambridge University Press, pp. 1–28. This is an excellent, readily
readable, authoritative account for students new to or lacking confidence in
statistics.
Burnham, P., Gilland, K., Grant, W., and Layton-Henry, Z. (2004) Chapter 5:
Descriptive Statistics. In Research Methods in Politics. Basingstoke: Macmillan.
pp. 114–42. This account provides good advice on the ‘charting’ of descriptive
statistics using examples from Politics.
Pennings, P., Keman, H., and Kelinnijhuis, J. (2006) Chapter 5: Explorative and
descriptive statistics. In Research Methods in Politics. London: Sage.
pp. 88–131. This is an advanced text better suited to experienced researchers
wishing to develop their mathematical skills.
Chapter 14
Using and Understanding
Inferential Statistics
1. To understand what is meant by inferential statistics.

2. To learn how the normal distribution enables generalisations to be drawn
from the descriptive statistics of representative samples.
3. To learn how to calculate the confidence limits which can be attached to
opinion polls.
4. To understand the meaning of the null hypothesis and how it can be applied
to sample data.
Introduction
Consider the example of a survey of a random sample of 100 Labour party members
which reveals an average weekly net disposable income of £84.12 with a standard
deviation of £21.07. Twenty-six of the sample (i.e. 26%) regularly read The Guardian
newspaper. In terms of other (non-print) news media, 72% rely mainly on BBC TV
and radio services for national and international news.
Given this information from a small sample of people, then how can you calculate
the extent to which it is shared by the general population? Furthermore, how can
you tell whether or not your sample is representative in the first place?
The answer is that you can use inferential statistics. These are:
a branch of applied mathematics or statistics based on a random sample. They let

the researcher make precise statements of the levels of confidence they have in the
results of a sample being equal to the population parameter (where a parameter is a
characteristic of the entire population drawn from a sample).
(Neuman, 2003: 539, 541)
The solution to drawing inferences from sample data lies in the attributes of the
normal distribution and the central limit theorem.
The normal distribution
In the previous chapter, you learned how to describe a series of data by the number
of terms, N , the arithmetic mean, X and the standard deviation, σ . The standard
deviation probably appeared contrived and of no real practical use. However,
consider the normal distribution in Figure 14.1
An important characteristic of the normal distribution is that 95% of the terms
will lie within two standard deviations (actually 1.96 SD) either side of the mean;
conversely, only 5% will lie outside two standard deviations. Of the terms, 99%
will lie within 2.97 SD. These characteristics can be expressed as probabilities.
A probability is the likelihood that an event will occur. For example, if an event
is certain – like death albeit in the distant future – then the probability is 100% or
1.00. Alternatively, a remote possibility – like winning the UK national lottery next
week – may be 0.001% or 0.00001. A one-in-four likelihood would be expressed
as a probability of 25% or 0.25. A bookmaker would give odds of 4–1 on an
event having a 25% probability. So, returning to the normal distribution, there is
100
80
Number of men
60
40
20
0
0 50 55 60 65 70 75 80 85 90 95 100 105 110 115
x ± 1 SD
x ± 2 SD
x ± 3 SD
Diastolic blood pressure (mmHg)
Figure 14.1 The normal distribution

Using and Understanding Inferential Statistics 199
a probability that of 95% or 0.95 any individual term will lie within 1.96 standard
deviations of the mean. This means that it is likely that 95 out of every 100 terms
will lie within 1.96 standard deviations of the mean. We can increase the likelihood
of any term lying in the distribution to 99% by increasing the span to 2.97 standard
deviations. There is therefore a trade-off between probability and accuracy.
Central limit theorem
The means of samples randomly taken from a population will always differ. This
difference is termed the sampling error. The central limit theorem (where a
theorem is a mathematical deduction accepted as a truth) effectively proposes that the
arithmetic means of large samples (greater than 30 terms) will approach a normal
distribution – even where the population is not normally distributed. Therefore, by
referring to the distribution of terms in the normal distribution curve above, we
can say that 95% of all the sample means lies within 1.96 standard deviations of the
‘mean of the means’.
(For samples of less than thirty terms, the sample means will follow the Student’s
T distribution. This is symmetrical and looks very similar to the normal distribution.
However, the spread of the distribution reflects both the standard deviation and the
sample size. The accuracy of inferences from the T distribution is therefore lower
than from a normal distribution and decreases as the sample size falls.)
Standard error of the sample mean
The standard error of the sample mean, SE, provides the statistical method of
generalising from samples to populations. It estimates the likely variance and therefore
accuracy between the sample mean and population mean. Its formula is:
sd
SE = √
n
where sd is the standard deviation of the sample and n the number of terms in the
sample (sample size)
The formula which brings together the characteristics of the normal distribution,
central limit theory and standard error of the sample mean is that, at 95% probability,
the population mean X , lies within 1.96 standard errors SE, of the sample
mean x, i.e:
X = x ± 1.96SE
Consider again the initial example of a random sample of 100 Labour party
members that reveals an average weekly net disposable income of £84.12 with
a standard deviation of £21.07. So we can say that:
X = 84.12
Sd = 21.0
n = 100
√ √
The standard error, SE = sd/ n = 21.07/ 100 = 21.07/± 10 = ± 2.107
You can calculate the income of the population of all Labour party members at
a probability of 95% by substituting in the formula:
& = x̄ ± 1.96SEP
X
&
X = 84.12 ± (1.96)(2.107) = 84.12 ± 4.12972 = 79.9902 to 88.24972
Alternatively, you can complete the calculation using Microsoft Excel. Open an Excel
spreadsheet. Then click on Insert and fx function. In the Insert Function dialogue
box, select the category statistical and scroll down the functions until you reach
CONFIDENCE. Click on OK. The Functions Argument dialogue box will appear
asking you to insert values for Alpha, Standard deviation and Size. Alpha is the
significance level for which 95% is the default for Politics. For a significance level of
95%, the Alpha is 0.05. The standard deviation is £21.07. The (sample) size is 100.
Insert these figures. The answer (formula result) of 4.129 will appear automatically at
the foot of the dialogue box. Click on OK and the answer will be entered in cell A1
(or any other cell that you’ve highlighted in advance).
What the manual or Excel calculations mean is that you can infer that 95 of
every 100 members of the Labour party overall will have a net weekly disposable
income of between £79.99 and £88.25. These two figures are termed the upper and
lower confidence limits. The variance – difference – between these two figures is
termed the confidence interval or margin of error. The margin of error can either
be expressed as the difference between the confidence limits or their variance from
the sample mean. So, in the example above, we could report that the margin of error
was either 8.26 or ± 4.13. Note how the numbers calculated have been ‘rounded
up’ to two places of decimals only in the final answer: this avoids accumulated errors
which can occur where initial calculations are rounded up. If you wanted to learn the
likely spread of income of 99% of all party members, then we would use the figure
of 2.97SEs or an Alpha of 0.01. The result will be a confidence interval of ±5.42
This initial example of sample data also showed that, of our random sample of
100 Labour Party members, twenty-six of the sample (i.e. 26%) regularly read The
Guardian newspaper. In terms of other (non-print) news media, 72% rely mainly on
BBC TV and radio services for national international news. How can you calculate
the confidence limits of likely Guardian readership and BBC news of other party
members from the sample?
In this case, you can use the standard error of the proportion, SEp where
√
SEp = {p(1 − p)/n}
Where p is the standardised proportion of the sample sharing the characteristic. So the value
p for 26% is 0.26; 100% would be 1.0.
To calculate the population proportion P, you can use a formula similar to that
using the standard error. In this case, at the 95% confidence level:
P = p ± 1.96SEp
In our initial sample (% of Guardian readers):
p = 0.26 (i.e. 26% expressed as a proportion of one)

n = 100
Therefore, the standard error of the proportion:

√
SEp = √{p(1 − p)/n} √
= {0.26(1 − 0.26)/100} = {0.26(0.74)/100}
= 0.043863244 (do not round up!)
Therefore, substituting in the equation:
P = p ± 1.96SEp
P = 0.26 ± (1.96)(0.043863244) = 0.26 ± 0.0859723118
P = 0.2161365756 to 0.348597231182
Now, you can round up to say that P = 0.216 to 0.349 or, in percentage terms,
you can say that, at 95% confidence limits, between 21.6 and 34.9% of all Labour
party members will read The Guardian.
Similarly, you can calculate that, if 72% of the sample relied on BBC news
services, then the percentage of the national members who also rely on the BBC
at 95% confidence limits would lie between 63.2% and 80.8%. In this case, the
confidence interval is 17.6%. This confidence limit is obviously too large – because
of the small sample size – for useful inferences to be drawn. Greater predictive
accuracy requires much larger samples.
National opinion polls
Opinion pollsters – NOP, Gallup, etc – generally use a random sample of 1600.
Consider, therefore, the example of opinion polls which show that, by February,
support for the government is slowly rising, whilst support for the main opposition
party is falling. The very latest poll (of 1600 electors) shows that 39% of the electorate
say that they will vote for the government ‘if an election were held tomorrow’ and
33% for the main opposition. This is the highest level of support given for the
government for two years. Should the prime minister risk an election in May, or
wait another year? What is the margin of error?
In this case, the standard error of the proportion, SE p is 0.01219. There would
be a 95% probability that support for the government could lie between 36.6 and
41.4%. Conversely, opposition support could lie between 30.7% and 35.3%. At the
99% probability level, support for the opposition could conceivably be as great as
36.6% and therefore greater than the lowest level of support for the government at
35.5%. So, despite a 6% lead over the opposition and a national random sample of
1600, the government could lose the election.
The margins of error are therefore much larger than newspapers – who commission
polls as a low-cost source of ‘news’ – admit. This is important because of the potential
effect of opinion polls in manipulating attitudes and discouraging people from voting
where they believe that the outcome is a ‘foregone conclusion’, e.g. UK general
elections in 2001 and 2005 (Heath and Taylor, 1999: 180).1 However, where the
proportion of the sample sharing the same characteristic rises to nearly 100% then
the margin of error falls considerably. So, if 90% of the sample of 1600 support the
government’s policy on, say, ‘tougher sentences’, then, at the 95% confidence level,
the margin of error falls to 1.5% ‘either way’, i.e. between 88.5 and 91.5%.
The null hypothesis
The null hypothesis is effectively the practical application of the test of falsifiability
or fallibalism added by Popper (see Chapter 3). It is the assumption – similar to the
assumption of innocence in UK and US law – that the data is untrustworthy until
shown conclusively otherwise. It is generally explained as the concept of ‘no differ-
ence.’ Its general application is to the control and experimental samples. Any difference
in the experimental group must first be attributed entirely to chance. It can only
attributed to the effect of some stimulus (cause) when chance has been eliminated.
The null hypothesis is termed H0 and the alternative hypothesis – our hunch – is
termed H1 . H0 must be disproved – shown to be false – before H1 can be accepted.
In the case of samples and populations where only one variable is being studied –
so-called univariate statistics – the null hypothesis is that the sample has not been
drawn randomly from the population and is therefore not representative.
Is the sample representative?

To accept a sample as representative and therefore dismiss the null hypothesis, two
questions must be satisfactorily answered. First, is the sample genuinely drawn from
and statistically representative of the population? And, second, could the findings
from the sample be entirely by chance?
To answer the first question, you must verify that the sample is really drawn
from the population. In your initial example, the population was the national
membership of the Labour party. To be statistically reliable, the sample must be
randomly drawn from that population. So, if the sample were gained from only
one region or constituency of the UK, it would not be statistically representative.
A very small sample would be unlikely to be representative. You can also check
the sample against known characteristics of the population – e.g. age or sex. So, if
the male/female ratio of the population is, say, 60:40 then you would expect that
of the sample to be similar. You can test this mathematically. The average age of
all party members is 59 (source: Labour party, London, January 2007) whereas the
average age of the sample is 57 with a standard deviation of, say, 3. You can use the
formula:
X = x ± 1.96SEp
Therefore, in your example:

√
X = 59 ± 1.96(3/ 100) = 59 ± 8.88
= 50.12 or 67.88
As you know that X is 59 and therefore between the upper and lower levels predicted
above, you can say that you are 95% confident that the sample could indeed have been drawn
from the population.
You can also investigate the significance (the likelihood that any similarity between
sample and population means occurred entirely by chance) of the difference between
a population mean X , and the sample mean x, by using the t-test for a population
mean. Where the variance of the population is unknown, then the formula to
calculate the test statistic, t is
√
t = (x − X)/(sd/ n)
In the case of the ages of the sample and population of the Labour party
X = 59
X = 57
sd = 3
n = 100
You will calculate that t = −6.67. By referring to the t-distribution tables published in
most statistical tables, you can see that, for degrees of freedom, df of 99 (i.e. n − 1) and
a probability level of 0.05 (i.e. 95% confidence levels), the critical value of t is ±1.65.
As your calculated t statistic is greater than the critical value, then you can conclude
with 95% confidence that the data from the sample has not arisen by chance. The
null hypothesis can therefore be rejected and your alternative hypothesis – that the
sample is statistically representative of the population – accepted.
1. Work your way through each of the worked examples shown above at your own
pace to test and improve your understanding of the procedures.
2. Discuss what is meant by confidence level, confidence interval and confidence
limits. Where and why should they be used?
3. Develop your own table for calculating the confidence limits of opinion polls. In
column A enter sizes of sample from 10, 20, 30, …100, 200, 300, … 3,000, 4,000,
5,000, 6,000, … 10,000, 15,000, 20,000 … 100,000.You can use the Autofil routine
to simplify this process. In row 1, enter the percentages of the sample sharing the
same characteristic or attitude from 01, 02, 03 ,04, 05,… 99%. Write your own
formula for calculating the confidence limits at 95% confidence levels, i.e. values
√
of 1.96( p(p − 1)/n. Copy your formula across the spreadsheet and save.
FURTHER READING
Cambridge: Cambridge University Press. pp. 31–124. This is an excellent text
for researchers who lack confidence or experience of statistical methods. The
text is supported by operation schedules which demonstrate the use of the
statistical techniques. A major strength is that the demonstration calculations
are undertaken manually rather than computed elsewhere.
Levine, D. and Stephan, D. (2005) Statistics: Even You Can Learn. London:
Pearson Prentice Hall. pp. 103–23. This is a very readable account which
includes step-by-step instructions for using Microsoft Excel to complete
calculations and downloadable practice files.
De Vaus, D. (2002) Analysing Social Science Data: 50 Key Problems in Data
Analysis. London: Sage. pp. 147–234. This is an advanced text which uses
SPSS for calculations. Its discussion of the principles involved in sample sizes,
significance and confidence intervals is good.
Kanji, G. K. (1999) 100 Statistical Tests. London: Sage. p. 27 t -test for a population
mean (variance unknown). This is another advanced text. It shows clearly which
tests should be carried out and includes simple worked examples and statistics
table.
Note
1 Heath, A., and Taylor, B. (1999) Chapter 9: New Sources of Abstention. In Evans, G. and Norris, P.
(eds.) Critical Elections. London: Sage. pp. 164–80.
Chapter 15
Testing for Association
1. To consider the concept of association between variables.

2. To learn how to test for association using correlation analysis and applying
tests of significance to the results.
3. To learn how association between variables can be expressed as equations.
4. To learn what is meant by regression and how to produce explanatory
equations between two or more variables by applying linear regression
analysis and multiple linear regression analysis.
Introduction
In the previous chapter (Chapter 14), you considered how individual measures of
characteristics of a sample could be generalised into statements about the population
from which they were drawn. However, only one characteristic was measured for
each member of the sample: whether they voted or belonged to a union, income and
strengths of attitude, etc. These are called univariate statistics, i.e. one number. But
a greater ambition of Politics research is to identify causes and effects – the independent
variables and dependent variables. Because they involve two (paired) variables, they
are referred to as bivariate statistics. Where two or more independent variables
cause a particular outcome or effect they are termed multivariate statistics. By
convention, the independent variable (cause) is termed x, and the dependent variable
(effect or outcome), y. This relationship is expressed mathematically as:
y = f(x) i.e. y is a function of x.
The relationships between bivariate or multivariate variables may be linear (i.e. a

straight-line relationship), exponential, logarithmic or curvilinear. For example, empiric
research has shown that, given perfect choice between comparable locations, then
demand is a function of the ratio of their perceived utility and the perceived
Testing for Association 207
costs – expressed as the inverse of travel time to the power 1.8. This is known
as the transport equation which can be expressed as:
"
Di = µ{"i (u/C 1.8 ) "i−n (u/C 1.8 )}
What this means is that the likelihood of a decision-maker choosing to travel to

location i rather than any other location is a function of its perceived benefits u, and
perceived costs C, by comparison to the perceived benefits and costs of travelling
to other locations. This equation belongs to a family of so-called gravity models.
Politics research concentrates instead on linear relations between cause and effect.
Consider the data in Table 15.1 that shows paired data for relative pay and union density
(% of workers in a particular industry belonging to a trade union) in the ten highest
paid and ten lowest paid industries and services.
The issue of pay and union density has occupied politicians and Politics
researchers – particularly labour historians – for some time. The subject was of
particular importance in the 1970s when the Conservatives – especially under
Mrs Thatcher – believed that the unions were to blame for UK national decline
Table 15.1 Pay and union density.
MLH Industry Pay as % of all industry Union density

and services %
485 Newspaper printing 157.8 94
262 Mineral oil refining 143.2 59
101 Coal mining 133.6 97
Underground workers 145.9 99
707 Air transport 130.8 85
601–2 Electricity and gas 122.9 95
489 Other printing 120.7 94
706 Port and inland Water 119.8 83
271 General chemicals 117.5 59
383 Aerospace eng. 117.3 80
812 Wholesale distribution 86.7 15
XIII Textiles 86.2 99
894 Motor repairs and distribution 85.5 60
831–2 Industrial materials 85.3 15
441–9 Clothing 84.0 42
820–1 Retail distribution 83.6 15
414 Woollen worsted 82.3 47
872 Educational services 80.6 78
884–8 Catering 77.9 8
1 Agriculture 72.4 23
Source: Department of Employment, 1981: Table 54 where MLH is Minimum List Heading
because of their power to inflate wages under threat of industrial action (strikes) and
their influence on Labour governments. On the other hand, the trade unions argued
that they had been responsible for securing real improvements to living conditions
for workers and their families through ‘free collective bargaining’ and responsible
partnership with Labour governments. (Few argued that increases in pay caused
union membership to rise.)
You can chart the data in Excel as an X–Y scatter graph:
Pay as % of all other industries
180
160
140
and services
120
100
80
60
40
20
0
0 50 100 150
Union density %
Figure 15.1 X–Y graph of Table 15.1, Union density and pay, 1981
You will see that, as union density rises, pay also appears to rise. The graph would
appear to support the belief – shared by Left and Right – that higher union density
can be associated (related to or connected with) with higher pay. However, note
that there are some sectors where high union density is not associated with high
pay – especially in textiles and education services. But generally the relationship
could be represented as a straight line and, therefore, described as linear. This linear
relationship might be expressed as an equation:
Pay = po + (g) Union density
Where po is (basic) pay levels where unionisation is nil (zero) and g is the gradient
(steepness, slope) of the straight line.
Measuring association
Apparent association can be measured by calculating the coefficient of

correlation, r.
This coefficient was first devised by Sir Francis Galton (1822–1911), a gentleman
mathematician, brother-in-law of Charles Darwin. He was also the father of eugenics,
a discipline founded on the scientific belief that the physical and mental abilities of
society should be improved by selective breeding. Galton found that, when he

compared the heights and weights, etc, of people, shorter people produced taller
children and vice versa and underweight people produced heavier children etc. He
described this tendency as regression – literally going back. Furthermore, the tendency
of pairs of characteristics to regress could be measured by the coefficient of correlation
and expressed as a linear regression equation.
The statistical formulae used in regression analysis are complex. They essentially
perform ANOVA: the analysis of variances between variables. Simply put, this involves
calculating the means and seeing whether the variances of the paired variables show
commonality. There are various statistical formulae to calculate the coefficient of
correlation, r. Pearson’s r is the most widely used.
Correlation is expressed on the scale between (+1.00) and (−1.00) where +1.00
indicates perfect positive correlation, 0.00 means no correlation and −1.00 indicates
perfect, negative correlation. Their respective graphs are shown in Figures 15.2a and
15.2b. Figure 15.2c shows a distribution with a low or nil correlation.
12 12 9
10 8
10
7
8 8 6
5
6 Series1 6 Series1 Series1
4
4 4 3
2
2 2
1
0 0 0
0 5 10 15 0 5 10 15 0 5 10 15
A B C
Figure 15.2 a: Perfect positive correlation, b: Perfect negative correlation, c: little or

no correlation
A correlation coefficient r of greater than ±0.3 is statistically important. However,

in the social world, a correlation of greater than ± 0.7 may indicate collinearity, i.e.
both variables may be the product of the same, unseen variable. Positive correlations
are recorded where the variables both increase (or decrease) together, e.g. income
and expenditure. Negative correlations are recorded when one variable increases
and the other decreases, e.g. car-ownership and the likelihood of owners supporting
the Labour Party. The general rule-of-thumb is that where:
r = 0.1: the association is termed of small importance

r = 0.3: the association is termed of medium importance
r = 0.5: the association is termed of large importance
This interpretation is derived from the coefficient of determination, R 2 . This is

the square of the coefficient of correlation, r. It is the proportion of the variance in
one variable that can be explained or attributed to the other. So, where:
r = 0.1, R 2 = 0.01, this means that 1% of the association can be attributed

Correlation is a measure of association. But evidence of association does not necessarily

mean causation: theoretical exposition and explanation are required.
Calculating regression statistics
The equations for calculating coefficients of correlation is complicated. Fortunately,

regression statistics can readily be performed using Excel.
Calculation of regression statistics should only be carried out after the data has
been plotted as a chart to show whether or not there is prima facie (at first sight) signs
of apparent association. The data must be entered in a spreadsheet in which the
paired data are recorded side-by-side in columns (see Table 15.1). The next steps
involve clicking on Tools and then Data Analysis. When the Data Analysis dialogue
box opens, click on Regression and OK. The Regression dialogue box will appear.
By way of example, use the data from Table 15.1 (Union density and pay). In
the Input Y Range, enter the addresses of the first and last cells of the columns of pay,
i.e. C8:C27. In the Input X Range, enter the addresses of the first and last cells in the
column of union density, ie, D8:D27. Click on New Worksheet Ply and OK. The
following results will appear:
The most important statistics have been highlighted. They are:
Multiple R 0.688452
Intercept 71.48414
X Variable 1 0.564809
What do they mean? They mean that the coefficient of correlation (termed
‘Multiple R’ in Excel) is +0.69, that the point where the straight line crosses the
Pay axis is £71.50 and the gradient of the straight line is 0.56 (which means that,
for every change in union density of 1%, pay will increase by 0.56%). The linear
equation between pay as an outcome of union density is therefore:
Pay = 71.5 + 0.56 Union Density

Table 15.2 Summary output of regression analysis of data in Table 15.1
Summary Output
Regression Statistics
Multiple R 0.688452
R square 0.473966
Adjusted R square 0.444742
Standard error 20.01117
Observations 20
ANOVA
df SS MS F Significance F
Regression 1 6494.574 6494.574 16.21831 0.00079
Residual 18 7208.046 400.447
Total 19 13702.62
Coefficients Standard t-value P-value Lower 95% Upper 95%

Error
Intercept 71.48414 9.822873 7.277314 9.18E-07 50.84703 92.12124
X Variable 1 0.564809 0.140249 4.027196 0.00079 0.270157 0.859461
Other results are also useful. As previously explained, R Square (written R 2 ) is the
proportion of Y (outcome) that can be attributed to X (cause). In this case, it is 0.47.
So 47% of pay can be attributed to union density. So there must be other drivers.
What might they be? The Adjusted R Square is a more sophisticated and reliable
estimate than R Square. In this case, Adjusted R Square is 0.45.
However, before you can accept relatively high coefficient of correlation of +0.69
as evidence of association, you must be able to refute the null hypothesis. If your
starting hypothesis, H1 was that union density affects pay, then the alternative
hypothesis, H0 must be that there is no causal relation and that the association
between data has arisen entirely by chance. In other words, you must be able to
show that the data and results are statistically significant
There are various statistical tests of significance for paired data. They include the
Pearson product-moment correlation, Z-test and t-test. Perhaps the simplest way, is to
use the confidence interval estimate of the slope. This is calculated by multiplying the
t-statistic by the standard error of the slope and then adding and subtracting this product
to the gradient. This sounds complicated. But, fortunately, Excel has already done
this. The results are at the foot of the summary output of Table 15.2:
This shows that, at 95% confidence levels, the upper and lower levels of the
gradient are 0.270 and 0.859. Because these figures are greater than 0, then you
can be satisfied that a significant relationship exists between the union density and
pay data. The null hypothesis can therefore be refuted. Alternatively, you can refer
in statistical tables to the Critical Values of Pearson’s r. These will tell you critical
Table 15.3 Extract from summary output in Table 15.2
Coefficients Standard t-value P-value Lower 95% Upper 95%

Error
Intercept 71.48414 9.822873 7.277314 9.18E-07 50.84703 92.12124
X Variable 1 0.564809 0.140249 4.027196 0.00079 0.270157 0.859461
values of r for a given sample size and levels of significance. You will note that, for a
level of significance of 0.05 (95%) for a one-tailed test for a sample size of 20, the
critical value of r is 0.378. Because your calculated r (of 0.69) is greater than the
critical value, you can conclude that the association between union density and pay
is statistically significant at 95% confidence limits.
Another simple alternative is to use the P-value automatically calculated by Excel
and included in Table 15.3. The P-value is the observed level of significance. Basically,
you can reject the null hypothesis Ho if the P-value is less than the level of significance
you have chosen (typically 0.05 in Politics research) and vice versa: ‘if the P-value is
low, then Ho can go’. In the example, the P-values are well below 0.05 and the null
hypothesis can be rejected.
You will note that statistics tables distinguish between one-tailed and two-tailed tests.
The default position in politics research is the one-tailed hypothesis: that, as one
variable increases (say, union density), the other (pay) will also increase. However,
there may be circumstances where, although you hypothesise that, as one variable
changes, the other will also change but you can’t predict whether it will go up or
down. In that case, your hypothesis will be two-tailed. What this means is that,
for the same confidence limits, the critical value of r will have to be much higher if
significance is to be attributed.
Regression analysis also enables predictions to be made of variables for which data
is not available. You will recall that the linear regression example for pay/union
density was:
Pay % = 71.5 + 0.56 Union density
So you can calculate that, for a union density of, say, 50%, pay will be:
= 71.5 + (0.56)(50) = 98.5%
The predicted value of a term is denoted by a ‘hat’ sign. So, a predicted value of y
is denoted by ŷ and called ‘y hat’. In the example above, the predicted value of pay
is called ‘pay hat’. However, if we substitute data in the formula from our original
paired data, there is a difference between predicted pay (pay hat) and recorded pay.
For example, in Table 15.1, the pay and union density for distribution are 85.5 and
60%. But if we substitute a union density of 60 in the formula:
Predicted pay (pay hat) = 71.5 + (0.56)(60) = 71.5 + 33.6 = 105.1
This difference between predicted value (105.1) and actual data (85.5) is called the
residual error. It is represented by the Greek letter ϵ (pronounced ‘epsilon’).
The general equation for linear regression is therefore:
y = a + bx + ϵ
where:
y is the outcome (effect)

a is the y-intercept (value of y when x is 0)
b is the gradient
x is the cause
ϵ is the residual error.
The use of regression analysis to calculate y can only be used between the range of
values for x for which data is available.
Multiple linear regression analysis
The example of pay/union density showed that, despite a high coefficient of

correlation of 0.69, less than half (0.47) of the change of pay could be attributed
to changes of union density. Inevitably, in a complex social and political world in
which everything seems related to everything else, other variables will be involved.
So most regression analysis undertaken by Politics researchers involves multiple
linear regression analysis whose general equation is:
y = A + B1 x1 + B2 x2 + B3 x3 + . . . + Bn xn
where x1 , x2 , x3 , … xn are the various independent variables.

Returning to the example of pay/ union density, you can anticipate that
other, subordinate independent variables are involved. What might they be? Your
theoretical and literature review is likely to have identified some of these other
variables. They may include the physical concentration of workers, the proportion
of male workers (whose pay and conditions have been historically much greater than
women), and the extent to which the industries have to compete globally (thereby
creating downward pressures on pay), and whether they are protected by tariffs or
enjoy other advantages which give them a near monopoly in the home market.
Calculating multiple linear regression equations

using Excel
Assume that you have found other, comparable, additional data from government
sources for the proportion of the workforce employed in plants of 500 or more, the
male proportion of the workforce, and UK market share. These are combined with
your initial data in Table 15.4.
Table 15.4 Pay/union density, workplace size, males and

UK market share, 1981
Pay Union % of Workers % Men % UK

density in Plants of Market
500+ Share
Newspaper printing 157.8 94 40 95 95
Mineral oil refining 143.2 59 65 92 90
Coal mining 133.6 97 70 99 85
Underground workers 145.9 99 10 60 100
Air transport 130.8 85 35 55 45
Electricity and gas 122.9 95 22 85 100
Other printing 120.7 94 24 73 60
Port and inland Water 119.8 83 65 97 100
General chemicals 117.5 59 78 96 37
Aerospace eng. 117.3 80 85 88 38
Wholesale distribution 86.7 15 34 82 91
Textiles 86.2 99 71 22 36
Motor repairs and distribution 85.5 60 18 77 87
Industrial materials 85.3 15 69 88 33
Clothing 84.0 42 47 28 47
Retail distribution 83.6 15 26 62 82
Woollen worsted 82.3 47 73 26 31
Educational services 80.6 78 12 45 91
Catering 77.9 8 11 13 77
Agriculture 72.4 23 0 79 65
Your first step is, once again, to produce an X–Y graph of the data:
You will note in the chart that there is prime facie evidence of association
between pay and the proportion of male workers and proportion of the domestic
market. However, there is little evidence of a strong relationship between pay
and the proportion of workers employed in plants of more than five hundred
workers.
You can complete the multiple linear regression analysis in entirely the same
way as the linear regression analysis. The summary output is shown in
Table 15.5.
Pay/other variables
200
and UK market share
worplace size, men
Pay, union density,
Pay
150
Union density
100
% of workers in
50 plants of 500+
% Men
0
0 5 10 15 20 25 % UK market share
Industrial and service sectors
Figure 15.3 X–Y chart of variables in Table 15.4
Table 15.5 Multiple regression output summary from Table 15.4
Regression Statistics
Multiple R 0.813484
R Square 0.661756
Adjusted R square 0.571558
Standard error 17.57806
Observations 20
ANOVA
df SS MS F Significance F
Regression 4 9067.796 2266.949 7.336683 0.001757
Residual 15 4634.824 308.9883
Total 19 13702.62
Coefficients Standard t Stat P-value Lower 95% Upper 95%

Error
Intercept 35.2295 18.32924 1.922038 0.073805 −3.83837 74.29738
X Variable 1 0.463829 0.130775 3.546763 0.002929 0.185088 0.74257
X Variable 2 0.144154 0.206786 0.697117 0.496393 −0.2966 0.584909
X Variable 3 0.315062 0.174371 1.806843 0.090885 −0.0566 0.686726
X Variable 4 0.214856 0.221124 0.97165 0.346633 −0.25646 0.686171
By default, Excel has automatically designated the variables of union density,

workers in plants of 500 staff or more, etc., as X variables 1, 2, 3, and 4. You will
note that the overall coefficient of correlation for the entire range of variables has
risen to 0.81 (high, positive correlation) and that 66% of changes in relative pay can
be attributed to the four, potential causal factors.
The multiple regression equation is therefore:
y = 35.2 + 0.46x1 + 0.14x2 + 0.31x3 + 0.21x4 + ε

That is:
Relative Pay % = 35.2 + 0.46 [Union Density %] + 0.14 [%Workers in plants
of 500+] + 0.31 [% male workers] + 0.18 [% UK market
share] + residual error
However, when you examine the values for the Lower and Upper 95%
confidence limits, you will note that these extend across zero (e.g. Variable 2
from −0.32 to + 0.49). The null hypothesis cannot therefore be rejected. Similarly,
the P-values for X2, X3 and X4 are higher than 0.05 which also means that the null
hypothesis cannot be rejected.
The regression summary does not include the coefficients of correlation between
pay and each of the four individual variables. To calculate these, click once again
on Tools and Data Analysis. In the Data Analysis dialogue box, scroll upwards to
correlation and click. In the Correlation dialogue box, insert the first and last cells of the
five columns of data (B3:F22) in the Input Range box and click OK. The output
appears in Table 15.6.
Table 15.6 Correlation coefficients of all variables
Column 1 Column 2 Column 3 Column 4 Column 5

Column 1 1
Column 2 0.688452 1
Column 3 0.199587 0.158195 1
Column 4 0.537148 0.199638 0.225699 1
Column 5 0.32949 0.168877 −0.52341 0.314504 1
Where Column 1 is pay, Column 2 is union density, Column 3 is workers

in larger plants, Column 4 is male workforce and Column 5 is UK market share.
The coefficients of correlation between pay and union density is (as previously
calculated) 0.69, workers in large plants 0.20, male workforce 0.54 and UK market
share 0.32. Note also how Excel calculates the coefficients of correlation between
all the variables, e.g. the coefficient of correlation between UK market share
and workers in large plants is −0.49. This additional data indicates that there
may be some previously overlooked relationship between apparently independent
variables.
The display in Table 15.6 is called a correlation matrix – also termed the
R-matrix. It is the building block from which factor and cluster analysis are developed.
These are the subject of the next chapter (16).
Pitfalls of linear regression analysis

Excel and other software enable linear regression analysis to be completed quickly.
Paradoxically, the greater simplicity increases the likelihood that errors will be made
in the process and interpretation of the summary output. The greatest errors arise
from the pitfalls of:
1. applying the formulae before first considering the likelihood of relationship – if any – in
theoretical terms and without creating X–Y graphs of variables
2. failing to identify collinear variables
3. failing to interpret the summary output correctly
4. misapplying or misinterpreting the tests of significance
5. consequently, wrongly rejecting or accepting the null hypothesis
6. predicting values of ŷ outside the range of known x data.
1. ‘Correlation does not necessarily mean causation’. Discuss. Explain how you would
investigate a high correlation for causation.
2. The table shows paired data for the total number of UK workers registered as
unemployed and membership of the British Communist party, 1929–39.
Table 15.7 Paired data:1929–39
Year UK Unemployed (000s) CP Membership
1929 1,216 3,200

1930 1,917 2,555
1931 2,630 6,279
1932 2,745 5,600
1933 2,521 5,700
1934 2,159 5,800
1935 2,036 7,700
1936 1,755 11,500
1937 1,484 12,250
1938 1,791 15,570
1939 1,514 17,756
Is there any evidence of association between unemployment and party member-

ship between 1929 and 1939?
Using Excel, draw an X –Y graph of unemployment on the x-axis and CP
membership on the y-axis.
Calculate the coefficient of correlation.
Calculate the linear regression equation.
Calculate the contribution made by unemployment to CP membership.

Test the statistical significance of the calculation. Can the null hypothesis be
dismissed?
The data shows that, after 1934, CP membership increased whilst unemploy-
ment fell. What other causes of increasing CP membership can you suggest
and why?
3. Read carefully the extract from the publication by Rallings, C., and Thrasher, M.,
(1997) Local Government Elections in Britain, London, Routledge, pp. 46–63.
Examine the relevant 2001 census data for the wards of a UK city of your choice
from web-site www.neighbourhood.statistics.gov.uk.
On the basis of the information given in the selected text and the census data
available, set out an hypothesis to the research question: which three factors are
most likely to have affected electoral turnout in local elections in your chosen city?
Find and save the most recent headline election data for ward turnout for all of
the wards in your selected city from its web-site www.[selected city].gov.uk
Select three independent/collinear variables from the census information for
testing your hypothesis.
Transpose the data for ward names, turnout and your three selected census
variables into a single spreadsheet. Produce X –Y charts of the data.
Using Microsoft Excel spreadsheet software, create a spreadsheet consisting
of all wards, the turnout data and the three variables you have selected from the
census data.
Calculate the coefficients of correlation between turnout and selected census
characteristics. What inferences can you draw from the association of turnout and
selected census data? Are the data significant? What limitations do you attach
to these inferences? Using the appropriate formula within Excel, calculate the
multiple linear regression equation between ward turnout (Y ) and the independent
variables (X1 . . .Xn ).
Your submission should be no less than 2,000 words in report form. It must
critically review the text by Rallings and Thrasher and clearly justify your choice
of potential independent variables. You must then explicitly describe, justify
and explain the analytic techniques you have adopted, the ‘results’, the linear
regression equation calculated and the limitations attached to the output.
FURTHER READING
Clegg, F. (2005) Simple Statistics. Cambridge: Cambridge University Press.

pp. 124–40. Clegg provides a readable and authoritative introduction to
correlation and provides worked examples in pp. 167–87.
Levine, D. and Stephan, D. (2005) Chapter 10: Regression Analysis. In Statistics:

Even You Can Learn. London: Pearson Prentice Hall. pp. 182–208. This is a
very readable account which includes step-by-step instructions for using
Microsoft Excel to complete calculations and downloadable practice files.
Kanji, G. K. (1999) 100 Statistical Tests. London: Sage. This is another advanced
text. It shows clearly which tests of significance should be carried out with
simple worked examples and statistics tables.
Lewis-Beck, M. S. (1995) Data Analysis: An Introduction. London: Sage. This
relatively short book (77 pages) provides a good introduction for students with
experience of higher mathematics. It also provides additional, valuable
guidance on tests for association for nominal data (using Goodman and
Kruskals’ Lambda) and ordinal data (Kendall’s tau) which will be especially
helpful to researchers intending to use comparative method (pp. 22–8).
Pennings, P., Keman, H. and Kelinnijenhuis, J. (2006) Chapter 6: Multivariate
Analysis and Causal Inference. In Doing Research in Political Science. London:
Sage. pp. 132–82. This is a very advanced text which demonstrates the
application of regression analysis to Politics case studies.
Chapter 16
Applying Factor Analysis and
other Advanced Techniques
1. To understand the meaning and application of factor analysis and cluster

analysis.
2. To learn how to carry out factor analysis.
3. To introduce other advanced functions and techniques including the binomial
and Poisson distributions and time series analysis.
Introduction
You may have noticed that UK research methods textbooks rarely go beyond linear
regression analysis because of space and complexity. This is unfortunate for two
reasons. First, the data and software now available to researchers make more advanced
statistics much easier to apply. And, second, the more advanced techniques are,
paradoxically, more useful they are than the basic descriptive or inferential statistics
with which you are more likely to be familiar.
The underlying purpose of this chapter is, therefore, to widen your appreciation
of the new possibilities in quantitative data analysis readily available to you to achieve
more practical and adventurous research.
Factors and factor analysis
Chapter 15 ended with an example of a correlation matrix which showed how

the independent variables shared some small degree of association. However, no
causal connection was inferred from the analysis. This demonstrates the old
adage (saying) that, in the social world, everything seems related to everything
else in complex ways and patterns. Factor analysis enables you to simplify that
complexity.
Applying Factor Analysis and other Advanced Techniques 221
Consider, for example, the physical and social character of many of the inner
city areas of western cities located between the city centre (US ‘central business
district’, CBD) and the outer rings of ‘better suburban housing’ (first identified
by Park, 1925 and Wirth, 1928 of the ‘Chicago School’).1 Physically, the inner
city remains characterised by decaying, predominantly older, high-density, terraced
housing. Most is privately rented. The larger houses have been sub-divided into
‘houses in multiple occupation’ (HIMOs). The schools, parks and shopping areas
are dilapidated. The roads are congested with traffic moving between the suburbs
and the city centre. Most of the mills, factories and other industrial buildings have
closed. Air quality is poor. Socially, the area has become the entry point for successive
‘waves of immigrants’. Other temporary migrants include students and other single
people or single parents seeking lowest-cost housing. Turnover is high. There is
also a large number of single, older pensioners living on low-incomes. Many suffer
chronic illnesses. There is a relatively high level of drug and alcohol abuse, and
street crime. Economically, job opportunities and wages are very low. Most of the
residents rely on some form of income support from the state. In other words, these
physical, social and economic characteristics come together – cluster – in the inner
cities of the western world. The geographical clusters are termed areas of multiple
deprivation.
Governments and local authorities have developed statistical indices of deprivation.
They combine census data on: population age, household size, employment, health,
and housing conditions, with other data on: incomes, benefits and health treatments.
The UK government uses the indices as part of the funding formulae (developed by
multiple linear regression) to distribute central government grants in an objective
(but contested) framework to local authorities, police and health trusts. Local
authorities also use the indices to identify the localities in greatest need and to
lobby government for additional resources. Both use indices to measure changes
over time and the success or otherwise of intervention.
Why these characteristics co-locate is disputed. That dispute is fundamental to
political divisions in society. The Right explains the inner city as the manifestation
of conditions which are created where individuals choose a lifestyle of welfare
dependency: they get what they deserve. The policy solution therefore lies in welfare
cuts to force them to work and become self-reliant. (My grandfather, a self-made
businessman, alderman and church warden subscribed to this view despite both his
grandmothers having died in workhouses.) On the Left, inner city residents are seen
as the victims of a capitalist economy in which those unable to work or find work and
therefore with the least economic power are consigned to the worst areas by market
forces. They are prescribed meagre handouts and a large police presence to prevent
them from challenging the system. The solutions advocated include redistribution of
income, greater income support and empowerment of local institutions. The Centre
takes a slightly ambivalent view in which the inner city is seen as a cost to the national
economy, in which the residents are underused resources. The solutions therefore
rely on education, training, capacity building and a combination of incentives and

sanctions to encourage participation.
So, what’s all this to do with factor analysis?
Well, the underlying principle of factor analysis is that, where characteristics (like
poor health, housing and incomes) occur together, they may be the observable
symptoms of underlying common factors also called latent variables. Where the
symptoms can be measured quantitatively, then the number and strength of factors
can be statistically identified. These factors point to underlying causes, for example,
the role of institutions and those whose interests are served by them. But, whilst
factor analysis can identify the factors, the labels – social exclusion or underclass –
must be assigned by the researcher.
Factor analysis serves three main purposes. First, the analysis enables the underlying
structure of a complex set of variables to be identified. Second, it can reduce
and compress the raw data to more manageable evidence. And, third, it enables
researchers to develop more effective and efficient follow-up questionnaires or other
surveys which directly target the key underlying factors.
Factor analysis
Factor analysis involves four stages:
1. First, an R-matrix is produced. This enables the variables with the highest coefficients
of correlation to be identified. It also enables the clusters of variables to be detected by
visual analysis. Consider for example, the (simplified) R-matrix in Table 16.1. Note that
the R-matrix is called a ‘square matrix’ because there are as many rows as there are
columns:
Table 16.1a R-matrix
Variable 1 Variable 2 Variable 3 Variable 4 Variable 5

Variable 1 1
Variable 2 0.5 1
Variable 3 0.4 0.3 1
Variable 4 0.1 0.1 0.1 1
Variable 5 0.1 0.1 0.2 0.7 1
2. Second, the underlying factors are extracted from the R-matrix. The simplest method
is called principal factor extraction. When you shade or colour the cells containing
coefficients of correlation with a value of 0.3 or more, then two clusters become apparent:
You will note from Table 16.1a that variables 1, 2 and 3, and variables 4 and 5 show
high degrees of association. The two separate clusters they form are termed the two
Table 16.1b R-matrix : higher correlations shaded
Variable 1 Variable 2 Variable 3 Variable 4 Variable 5

Variable 1 1
Variable 2 0.5 1
Variable 3 0.4 0.3 1
Variable 4 0.1 0.1 0.1 1
Variable 5 0.1 0.1 0.2 0.7 1
principal factors. These factors can be named A and B. In this stage, the eigenvalues
are calculated. The eigenvalue is the measure of the variance in the original variables
that can be attributed to a particular factor.
3. Thirdly, the principal factors are rotated using the varimax procedure (which maintains
the independence of the factors) in which the factor axes remain at right angles to one
another. What this means is that the variables are plotted along the axes of the factors.
This is shown diagrammatically below:
0.8
0.7
0.6
0.5
Factor B
0.4 Factor B
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Factor A
Figure 16.1 Factor chart
4. Lastly, the factor scores are calculated. These are estimates of the people’s individual
standing with regard to the factors identified.
This procedure sounds complicated. But advanced computer software enables

the process to be completed in a small number of commands. However, whilst the
R-matrix can be calculated using Microsoft Excel, the subsequent steps are currently
beyond its capacity. You have two alternative courses of action. First, you can
use add-ons to Excel such as WinStat taking advantage of the 30-day free trial
period. These add-ons can be bought for around £100. Or second, you can use
one of the advanced statistical software packages such as SPSS, STATA or Minitab.
You are likely to find that your university already supports one or more of these
on its network. Alternatively, you may find that you can buy your own copy from
your computer services under special annual licences for the cost of the CD-ROM,
i.e. £5. But you will only be able to renew the licence as long as you are a student or
employee of the university. After that you will have to pay to purchase the software
(£1200 for SPSS) and pay the annual licence renewal (around) £200. But you can
download a copy for a 30-day trial period.
SPSS
SPSS will be used for factor analysis. SPSS was originally the acronym for Statistical
Package for the Social Sciences. More recently, the acronym has changed to Statistical
Product and Service Solutions as the company has extended its range of products
and customers beyond its original academic base. The web-site is www.spss.com.
Remember that learning to use a different type of software is like learning a new
language: it is useful and develops your intellect.
The software can appear initially perplexing but it is very easy and has become
more intuitive to use. SPSS uses a full Windows environment. Like Excel, all the
data is held on a single spreadsheet called the Data Editor. This has two alternative
screens: a Variable View and a Data View.
The Variable View is the spreadsheet where you enter the variables as separate
rows and nominate the name, type, width, decimals label, values etc. For example, these
might be: name (age of respondent), type (numeric), width (4), decimals (2), label (age), etc.
The Data View is the spreadsheet where you enter your data. In this spreadsheet,
the variables occupy the columns whilst each case is entered as a new row.
How to carry out factor analysis using SPSS
Consider the example where you want to find out what the public thinks should be
the most important characteristics and qualities of candidates for the post of prime
minister.
Following a pilot study, you reduce the characteristics to twenty-one attributes:
(their) age (youthfulness); sex (male or female); appearance; religion; origins
(class); peccadilloes (known or alleged personal misbehaviour); family (partner
and children); wealth; identity (notably English, Scots, etc); speech (speaking and
debating abilities); modernity (not old-fashioned); honesty; sincerity; integrity; guile
(cunning); principle; loyalty; courage; patriotism; and, Euroscepticism (anti-EU).
You enter these variables in the opening Variable View shown in Table 16.2.
Note how an additional variable, ‘cases’ has been added to the Variable View. The
‘cases’ are the individual members of the public whom you have asked to rate the
importance they attach to each of these short-listed characteristics on a scale of 0 to
10 (where 0 is least importance and 10 is highest importance). The members of the
Table 16.2 Variable view
Name Type Width Decimals Label Values Missing Columns Align Measure
1 Age Numeric 3 0 Age None None 4 Right Scale
2 Sex Numeric 3 0 Gender None None 3 Right Scale
3 Appearance Numeric 3 0 Looks None None 4 Right Scale
4 Religion Numeric 3 0 Religion None None 4 Right Scale
5 Origins Numeric 3 0 Background None None 3 Right Scale
6 Peccadillos Numeric 3 0 Peccadillos None None 3 Right Scale
7 Family Numeric 3 0 Private/family life None None 4 Right Scale
8 Wealth Numeric 3 0 Wealth None None 4 Right Scale
9 Identity Numeric 3 0 Regional identity None None 3 Right Scale
10 Speech Numeric 3 0 Speaking skills None None 4 Right Scale
11 Modernity Numeric 3 0 Modernity None None 4 Right Scale
12 Honesty Numeric 3 0 Truthfulness None None 4 Right Scale
13 Sincerity Numeric 3 0 Sincerity None None 4 Right Scale
14 Integrity Numeric 3 0 Integrity None None 4 Right Scale
15 Guile Numeric 3 0 Guile None None 4 Right Scale
16 Intellect Numeric 3 0 Intellect None None 4 Right Scale
17 Principles Numeric 3 0 Principles None None 4 Right Scale
18 Loyalty Numeric 3 0 Loyalty None None 4 Right Scale
19 Courage Numeric 3 0 Courage None None 4 Right Scale
20 Patriotism Numeric 3 0 Patriotism None None 4 Right Scale
21 Eurosceptic Numeric 3 0 Euroscepticism None None 4 Right Scale
22 Cases Numeric 8 0 Case number {1, case number} None 8 Right Scale
general public will have been randomly selected. The sample size will be 630, i.e.
calculated at 30 people per variable.
You will enter the data collected in the Data View. The Data View for the first
ten cases is shown in Table 16.3.
Starting the factor analysis procedure is relatively simple. On the Data View, click
in turn on:
Analyze
Data Reduction
Factor Analysis
On the Factor Analysis dialogue box, highlight all the variables listed in the left-hand
pane except ‘cases’. Then, click in turn on:
The higher ‘radio button’ this will transfer the list of variables to the pane marked
‘Variables’
Descriptives
In the Factor Analysis: Descriptives dialogue box, click in turn on:
Univariate descriptives
Initial solution
Coefficients (Correlation Matrix)
Reproduced
Continue
You will return to the Factor Analysis dialogue box. Now click in turn on Extraction.
In the Factor Analysis: Extraction dialogue box, click in turn on:
Principal Components (Method)

Correlation matrix
Unrotated factor solution
Scree plot
Eigenvalues over 1
Continue
You will return once again to the Factor Analysis dialogue box. Click on Rotation.
In theFactor Analysis: Rotation dialogue box, click in turn on:
Varimax
Rotated solution
Continue
Table 16.3 SPSS Data View (cases 1–10 only of 630 cases)
Age Sex Appear- Religion Origins Peccadi- Family Wealth Identity Speech Modernity Honesty Sincerity Integrity Guile Intellect Principles Loyalty Cour- Patriot- Euroscepti- Cases
ances llos age ism cism
1 7 6 8 0 5 3 5 2 4 7 6 6 7 7 7 8 9 7 7 8 5 1
2 4 8 8 4 8 8 8 8 9 8 2 7 8 6 8 5 7 8 8 9 9 2
3 9 1 8 2 1 2 4 2 5 8 9 7 7 7 8 6 7 7 8 7 3 3
4 1 1 4 1 1 1 3 1 2 6 4 9 9 9 5 7 8 7 6 7 2 4
5 2 2 8 0 0 0 1 0 5 9 7 7 7 6 8 8 6 7 6 4 0 5
6 0 9 9 1 7 1 3 5 5 7 2 6 7 8 9 4 6 7 8 8 9 6
7 3 9 8 0 0 0 1 0 9 8 8 4 4 4 8 7 5 8 9 9 9 7
8 0 0 2 0 0 0 0 0 0 9 9 5 5 5 8 9 8 8 8 5 5 8
9 7 9 8 5 6 7 8 5 8 9 2 6 6 6 8 5 7 7 8 9 9 9
10 2 9 8 7 5 8 7 8 9 8 2 5 5 5 8 7 7 7 9 9 9 10
You will finally return to the Factor Analysis dialogue box where you will
click on OK.
The output will appear immediately on a new screen. This will consist of nine
tables and one chart. The most important for your purposes are the Total Variance
Explained, the Scree Plot and the Rotated Component Matrix
Part of the Total Variance Explained is shown below in Table 16.4. You will note
that Component 1 (i.e. Factor 1) has an eigenvalue of 8.7 and explains 41.6% of the
variance. Component 2 explains 24.6% of the variance. Five components (factors) have
eigenvalues greater than 1.0 and explain 88.9% of the variances overall.
The table is illustrated in the Scree Plot which shows the components having the
highest eigenvalues.
Table 16.4 SPSS output: total variance explained
Total Variance Explained

Component Initial Eigenvalues
Total % of Variance Cumulative %
1 8.7 41.6 41.6
2 5.2 24.6 66.2
3 1.9 9.0 75.2
4 1.7 8.0 83.1
5 1.2 5.7 88.9
6 0.8 3.9 92.7
7 0.8 3.7 96.4
8 0.5 2.4 98.8
9 0.2 1.2 100.0
8
Eigenvalue
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Component number
Figure 16.2 SPSS output: scree plot

Table 16.5 SPSS factor analysis: rotated component matrix
Component
1 2 3 4 5
Wealth .961 .168
Peccadillos .948 −.114 .245
Private/family life .917 .376
Religion .883 .115 −.205 .151
Modernity −.867 .235 −.195 −.194 .237
Background (class) .862 .228 .196
Patriotism .664 .335 .200 .566 .266
Euroscepticism .641 .606 .294 .301 −.105
Gender .620 .395 .522 .313
Regional identity .560 .354 .539 .312
Truthfulness −.922 −.205
Sincerity .136 −.911 −.126 .180
Integrity −.883 .387
Courage .350 .819 .235
Loyalty −.104 .664 −.151 −.320
Principles .143 −.291 −.835 .106 .168
Looks .324 .784 .445
Guile .149 .444 .641 −.408
Intellect −.570 .211 −.619 −.247
Speaking skills .442 .108 −.846
Age .125 .948
Notes:
Extraction method: principal component analysis.
Rotation method: varimax with Kaiser normalization.
A rotation converged in seven iterations.
The rotated component matrix is shown in Table 16.5. This matrix identifies the
relationship between the original 21 variables and the five principal components
into which the data can be reduced. Components 1 and 2 are clearly the most
important. Component 1 shows very high values for wealth, peccadilloes, private/family
life, religion, background (class), patriotism, euroscepticism and gender. You must now
interpret this result to identify what is the family of characteristics incorporated
in Component 1 interpretation as ‘traditional values’. Indeed, these are the
characteristics associated with the Conservative party. The strongest characteristics
of Component 2 are courage, loyalty and guile. (Sincerity and truthfulness are negative
aspects here.) One interpretation of these characteristics might be ‘street fighting
qualities’.
The analysis shows that two factors dominate the characteristics that the British
public wish to see in candidates for prime minister. These can be labelled
‘traditional values’ and ‘street fighting qualities’. Age and intellect are not important.
Margaret Thatcher in the period 1975–90 appears to fit these preferred characteristics
well and may well explain her popular appeal.
The additional advantage of reducing the variables to two key factors is that a
previous battery of potential questions in questionnaires can be reduced to just two
thereby reducing the number of questions, costs and time required and allowing
additional alternative questions to be added.
Time series analysis
Consider the data in Table 16.6 of civilian casualties in Iraq from March 2003 and
October 2006.
Table 16.6 Iraqi civilian casualties 2003–6
2003 2004 2005 2006

Jan 542 946 1186
Feb 561 1111 1236
Mar 2778 917 644 1423
Apr 2859 833 1200
May 507 541 1024 1658
Jun 541 747 1060 1897
Jul 593 682 1336 3590
Aug 735 724 1941 3009
Sep 525 805 1117 3345
Oct 461 827 953 3709
Nov 433 1323 1053
Dec 504 802 828
Source: Iraqbodycount.org
You will be aware that coalition spokespersons (especially for US forces) continued
to claim for some time that, despite occasional ‘blips’, the level of violence was
actually declining. Using Excel, you can chart the table:
In Figure 16.3, you can see that the general movement for monthly casualties is
to increase over time. You can identify the underlying trend by using the Moving
Average function in Excel (click in turn Tools, Data Analysis, Moving Average).
Essentially, this calculates the average of the months March–May, then April–June,
May–July and so on. The calculated moving averages for the first year are shown in
bold in Table 16.7
The moving averages can be charted to show the trend (forecast):
You will note that there is a variance between the actual data (Y) for casualties
and the trend.
4000
3500
Civilian casualties
3000
2500
Month no.
2000
Casualties
1500
1000
500
0
0 10 20 30 40 50
Time (months)
Figure 16.3 Chart of Iraqi civilian casualties 2003–6
Table 16.7 Calculation of moving averages
Year Month Month No. Casualties #N/A

2003 Mar 3 2778 #N/A
Apr 4 2859 2048
May 5 507 1302
Jun 6 541 547
Jul 7 593 623
Aug 8 735 618
Sep 9 525 574
Oct 10 461 473
Nov 11 433 466
Dec 12 504 493
Moving average
4000
3500
3000
2500
Actual
Value
2000 Forecast
1500
1000
500
0
1 6 11 16 21 26 31 36 41
Data point
Figure 16.4 Trend line

Table 16.8 In-migration to UK from other EU States (000s)
Year/Quarter Q1 Q2 Q3 Q4
2003 78 62 56 71
2004 84 64 61 82
2005 92 70 63 85
2006 100 81 72 96
This variance has four potential components:
Trend Component, T
Cyclical Component, C
Seasonal Component, S
Residual Component, R
Where: Y = T + C + S + R
The Trend Component is the underlying movement calculated by the method of

moving averages. The Cyclical Component is the variation that is generated by cyclical
movement of trade or business cycles. In the case of Iraqi war casualties, there may
be no cyclical component. The Seasonal Component is the variation attributable to
the months or seasons of the year. You may recall that there was reference during
the early years of the occupation to a ‘fighting season’. When you examine the
charts in Figures 16.3 and 16.4, you will see that there is some evidence of seasonal
variations with casualties reducing in December–February.
You can carry out sophisticated time series analysis using SPSS. However, the data
output is complex and difficult to interpret. Excel is unable to go beyond calculating
the moving average. The simplest method is to adopt the manual Method of Quarterly
Deviations using your own formulae created in Excel.
Method of quarterly deviations
Consider the example of in-migration to the UK from other (‘pre-enlargement’)

EU states:
The trend and the deviation from the trend may be calculated easily by calculating
in turn:
1. the four-quarter moving total

2. the eight-quarter moving total
3. dividing the eight-quarter moving total by 8 to calculate the trend
4. finally, subtracting the trend from the actual data to calculate the deviation between the
trend and the actual data.
Table 16.9 Calculation of trend and deviation
Year Quarter In-migrants (Y) Four- Eight- Trend (T) Deviation

quarter quarter from
moving moving trend
total total (Y−T)
2003 1 78
2 62
3 56 267 540 67.5 −11.5
4 71 273 548 68.5 2.5
2004 1 84 275 555 69.375 14.625
2 64 280 571 71.375 −7.375
3 61 291 590 73.75 −12.75
4 82 299 604 75.5 6.5
2005 1 92 305 612 76.5 15.5
2 70 307 617 77.125 −7.125
3 63 310 628 78.5 −15.5
4 85 318 647 80.875 4.125
2006 1 100 329 667 83.375 16.625
2 81 338 687 85.875 −4.875
3 72 349
4 96
The completed calculation is shown in Table 16.9

The next step is to calculate the seasonal variation. This is calculated by finding the
average of the deviations for each quarter. The deviations for the third quarter of
each year are: −11.5, −12.75 and −15.50 which total −39.75 which, divided by
three, give an average of −13.25. When you add the average deviations for the four
quarters, you will find that these add up to +0.25. As the sum should be zero, then
this error figure is divided by four and applied as a correction factor to each seasonal
deviation.
The residual component or irregular movement (R) can then be calculated by subtracting
the seasonal deviation (S) and trend (T) from the actual data (Y) as shown in
Table 16.10.
You will note that the highest residual variations – irregular movements − are +2.188 in
the final quarter of 2004 and −2.187 in the third quarter of 2005. These are relatively
small. Your task as a Politics researcher is to explain these irregular movements. One
explanation might be the policy announcement of tightening controls made in 2004,
which prompted more in-migration in the final quarter before the controls came
into effect in the third quarter of 2005. You will appreciate that changes in public
policy often create a j-lag effect where variables continue to get ‘worse’ until the
new policy begins to ‘improve’ the situation.
Table 16.10 Calculation of residual variation (shown bold)
Year Quarter In-migrants (Y) 000s Trend (T) Seasonal Residual

Variation (S) Variation (R)
2003 1 78
2 62
3 56 67.5 −13.313 1.813
4 71 68.5 4.312 −1.812
2004 1 84 69.375 15.52 −0.895
2 64 71.375 −6.521 −0.854
3 61 73.75 −13.313 0.563
4 82 75.5 4.312 2.188
2005 1 92 76.5 15.52 −0.02
2 70 77.125 −6.521 −0.604
3 63 78.5 −13.313 −2.187
4 85 80.875 4.312 −0.187
2006 1 100 83.375 15.52 1.105
2 81 85.875 −6.521 1.646
3 72
4 96
Time series analysis is a very underused but useful research method in Politics.
Try to use it. It will pay dividends by identifying those irregular movements which
may be the outcome of policy change. Time series analysis is therefore a very
powerful tool for monitoring public policy and assessing the effectiveness of policy
change.
The binomial distribution
In Politics research, population characteristics are assumed to follow the normal

distribution (see Chapter 14). However, there will be circumstances where the data
are non-numeric and involve only two potential answers, for example, the replies
YES or NO to a question, or heads or tails in respect of coins being ‘tossed’. Yes/no
or heads/tails are termed dichotomous nominal data. In these circumstances,
the outcomes are termed success p, or failure q. The probability of success p, can
be expressed mathematically. For example, the probability of a tossed coin landing
‘heads’ is 0.5. The probability of a thrown die ‘giving’ a six is 0.167 (i.e. one divided
by six). Each separate toss of the coin or throw of the die is termed an event.
Tests in which dichotomous outcomes are measured and where the outcomes are
independent of each other are termed Bernoulli trials (after their originator, Jacques
Bernoulli, 1654–1705). In Bernoulli trials, the null hypothesis is that the probability of
success in each event is the theoretical probability, e.g. for each toss of the coin, 0.5.
Take the example of two food parcels, A and B, being prepared by an aid agency
for use overseas to sustain IDPs 2 in a specific war-torn state. Both parcels contain
different food of equal nutritional value. The parcels have transparent packaging
so that the people can choose A or B. A two-parcel strategy has been adopted to
optimise cost, local production capacity and variety. The theoretical probability of
a person choosing parcel type A is therefore 0.5. An initial trial of a random sample
of 100 of the target population shows that 61 preferred parcel type A. The research
problem is therefore: does this higher preference represent the choice likely to be made by the
population as a whole or has it occurred entirely by chance?
The problem can be solved using SPSS. In the Variable View, enter ‘parcel’ in
the first row and ‘frequency’ in the second row. In the Data View, enter ‘1’ (for A)
and ‘2’ (for B) in the column marked ‘parcel’ and ‘61’ and ‘39’ in the column marked
‘frequency’. Now click in turn on Data View and Weight Cases. In the Weight Cases
dialogue box, click on Weight Cases and, using the radio button, transfer ‘frequency’
to the Frequency Variable. Then click on OK. You will return to the Data View.
Click in turn on Analyze, Nonparametric Tests and Binomial. In the Binomial Test
dialogue box, use the radio button to transfer‘parcel’ to the Test Variable List. Note
theTest Proportion box. This is where the theoretical probability is inserted. The
default proportion is 0.5 so no new entry is required. Finally, click on OK. The
Output is:
Table 16.11 Binomial test
Parcel Category N Observed Test Prop. Asymp.

Prop. Sig.
(2-tailed)
Parcel 1A Group 1 1 61 .61 .50 .035(a)
Parcel 1B Group 2 2 39 .39
Total 100 1.00
Note:
Based on Z Approximation.
The key calculation is the Asymp. Sig. (2-tailed) of 0.35. This is less then the test
proportion (theoretical probability) of 0.5. This means that the null hypothesis can
be rejected. So the preference of parcel A cannot be attributed to chance. The
implications are that the aid agency should either change the ratio of production
from 50:50 to nearer 61:39 or improve the appeal of parcel B and re-test using a
new sample.
A special type of Binomial distribution occurs where the population is very large
and the incidence of events is very small. This special type is termed the Poisson
distribution (named after Siméon Denis Poisson, 1781–1840 who developed the
distribution to forecast the likelihood of any of his fellow cavalrymen being kicked
Table 16.12 Poisson distribution: data
Description Data events per 100,000 children

Illness events in city C,x 5
Illness mean worldwide 3
by a horse). It is widely used to forecast the likelihood of a rare event occurring

at any time or, alternatively, several rare events happening at the same time. It is
therefore widely used by fire and other emergency services to plan staffing levels to
ensure adequate cover. A special application is the analysis of illness and disease. For
example, assume that the world wide incidence of a rare, fatal disease is 3 in every
100,000 children a year. Data for city C shows that the rate of the disease is 5 per
100,000. Does this indicate a special cluster or could the higher figure have arisen
entirely by chance?
In this case, Excel provides a simpler solution than SPSS. Enter the data into a
spreadsheet:
To calculate the probability of rate of illness in city C exceeding worldwide
rate, highlight a cell and enter = POISSON. Excel will automatically complete the
formula for you as:
= POISSON (x, mean, cumulative)
In this example, x = 5 and the mean = 3. The Help? Button will tell you that:
‘cumulative is a logical function that determines the form of distribution returned’.
You can enter either TRUE or FALSE. When you enter TRUE, Excel will calculate
the ‘cumulative Poisson probability that the number of events occurring will be
between zero and x’. If you enter FALSE, then Excel will calculate the ‘Poisson
mass probability function that the number of events will be x exactly’. First, complete
the formula as:
= POISSON (5, 3, TRUE)
Press the Enter key. The answer will appear immediately in the highlighted cell:
0.916.
Repeat the process, inserting FALSE. The answer is 0.1008. This answer tells
you that there is 10.1% probability (0.1008 − 0.916 ×100) of the number of illnesses
being 5 where the mean is 3. So the illnesses in city C could have arisen entirely by
chance and are therefore not exceptional. However, if the incidence was ten children
a year, then you will calculate that the probability is 0.0008. In other words, the
incidence could not have arisen by chance. So emergency intervention is required
to identify and tackle a local cause or source.
Questions for discussion in class or seminars or for assignments
1. Work through each of the examples given in this chapter. Seek help when you
encounter difficulties.
2. Collect new data from the class population of their assessment of candidates
for UK prime ministers using the criteria listed in the chapter. Add four additional
criteria of your own choice. Carry out a factor analysis and comment on the results.
3. Obtain unadjusted data for registered unemployment in the UK since May 1979.
Apply time series analysis. Identify the size of irregular movements and offer
cogent, potential explanations.
4. Two Departments of Politics, having equal and good RAE and TQA ratings in ‘new
universities’ A and B both have 30 places available for similar Politics degrees.
A attracts 56 applicants whilst B attracts 73. Could the variation have arisen entirely
by chance or does it indicate that B’s department or university are significantly more
attractive than A? If so, what should A do?
5. The Conservative party has a working majority of 11 over all other parties
combined in the House of Commons following the general election. On aver-
age, five MPs die each year causing by-elections to be held. The new prime
minister believes that harsh economic policies are essential for the first three
years. He assumes that half the Tory-held seats will be lost in by-elections.
Calculate the probability that the party can survive in office for three years.
(Remember that each time a government loses one seat, their majority reduces
by two.)
FURTHER READING
A notable feature of SPSS is the number of very good textbooks that are available
which combine excellent explanations of the statistical concepts and
instructions how to carry them out using SPSS software. Two titles can be
recommended:
Kinnear, P. R. and Gray, C.D. (2004) SPSS for Windows Made Simple. Hove:
Psychology Press Ltd. This offers a very good introductory and intermediate
text. Earlier editions are also available.
Field, A. (2005) Discovering Statistics Using SPSS (and Sex, Drugs and Rock ‘n’
roll). 2nd edn. London: Sage. This more extensive and expensive book is very
readable and provides a comprehensive range of material for new and more
experienced researchers.
Notes
1 Chicago University’s School of Sociology was the first established at a US university. As the
‘Chicago School’ it developed a reputation for ‘urban ecology’ which identified the concept of the
‘zone of transition’ and, later ‘inner city’ that provided the initial theoretical underpinning for US and
UK urban programmes and urban regeneration. See Park, R.E. and Burgess, E.W. (1925) The City
and Wirth, L. (1928) The Ghetto.
2 IDPs: acronym used by UN to distinguish ‘internally displaced persons’ from other refugees.
Part IV
B: Qualitative Analysis
Chapter 17
Analysing Qualitative
Information: Classifying, Coding
and Interpreting Information
1. To learn how to develop and apply your own codes.

2. To learn how to interpret coded information using tables and diagrams.
3. To understand how to use the techniques of ‘memoing’, case analysis
meetings and interim case summaries.
4. To learn how to construct your own typology.
Introduction
In Chapter 12, you were introduced to the hierarchy of analysis which begins by
assembling all the raw data, validating it and reducing it to ordered information.
At the end of this essential first stage, you will have uncluttered the chaos of your
desk and reduced the paperwork etc. to the ‘nuggets of gold’ on which you can
begin your analysis.
There are three principal methods for analysing qualitative information. Most
qualitative information takes the form of talk-and-text - spoken or written words.
The methods are:
• coding
• content analysis
• discourse analysis.
Content analysis is probably the oldest form of analysis. It is essentially a

quantitative process in which you count the frequency of key words individually
and, in more advanced methods, their concurrence. Its use has been recently revived
thanks in part to developments in communications science, semiotics, the Internet
and new computer software. Content analysis is considered in Chapter 18.
Discourse analysis is considered in Chapter 19. It is founded on the belief that

words are never neutral or merely descriptive, but part of vocabularies which are
written by others and laden with meanings. The three methods are not necessarily
mutually exclusive.
Coding
This has become the most widely used method of qualitative analysis. Coding
is essentially the process of replacing or substituting groups of words, phrases or
sentences by letters or numbers (or a combination). Unlike a cipher, a code is not
designed to hide the original meaning of the word or phrase. Indeed, an essential
property of an effective code is that it should be readily understandable to users and
readers. A word or phrase can be given several different codes.
Codes can be theoretical, descriptive or inferential.
Theoretical codes are derived from the independent and dependent variables
identified in your theoretical framework. A descriptive code merely describes what
it seeks to represent. An inferential code is used when you draw an inference usually
from repeated patterns or clusters of juxtapositions of different codes. One special
type of inferential code is a factor code. A factor is a latent, underlying variable
like ‘power’. Chapter 16 showed you how to identify underlying factors from
quantitative data. There will be occasions when you identify important variables
which fall outside your initial range. In this case, you can develop new codes as you
continue. These are called in vivo codes (‘code-as-you-go’).
Codes are usually designed generically and given alphabetical references to help
identify and distinguish them intuitively. For example, where ‘education’ is the
theoretical variable, then a family of codes could be, say:
Education E
Pre-school Ep
Early Years Epe
Nursery Epn
Infants Ei
Junior Ej
Secondary Es
Sixth-Form E6
Teacher Et
Further Education Ef
Higher Education Eh
University Ehu
Uni-Tutor Ehut
Uni-Friend Ehuf
Analysing Qualitative Information 243
The process of designing and applying theoretical, descriptive or in vivo codes to text is
termed first-level coding. They are applied by underlining or making bold all the
key passages and recording the codes in the right-hand margin. Alternatively, you
can use Nud*st, ATLAS.ti or other (‘CAQDAS’) software to code the information
electronically. In this way, passages can be multi-coded. Authorities suggest that you
will be able to remember and use as many as 60 generic codes without difficulty
(Miles and Huberman, 1994: 58).1
Tagging and coding
You must have already reduced your raw data to organised information before
you begin tagging and coding. This will have involved transcribing your interviews
into a three column format and having omitted all extraneous clutter and ‘noise’.
However, you don’t need to have collected all your data before you begin the
process of reducing available data and coding. Indeed, there are great advantages in
designing your codes and applying them at an early stage of your fieldwork to test
their appropriateness and the value of your initial data and information.
The initial part of the process – identifying key, codable words or phrases before
they are coded – is termed tagging. You are likely to be already familiar with
the most basic tagging and coding of text. This involves highlighting or underlining
words, phrases or sentences with coloured pens. Each different type of text is
marked with a different colour. So the text becomes colour-coded. However, you
will encounter two or more problems. First, you are likely to run out of colours.
And, second, many words or phrases may be coloured two or more times so
the coding becomes unreadable. Yet, words or phrases which attract many codes
are particularly important because they demonstrate the links between different
characteristics and concepts.
The case study below has been developed to demonstrate how to design and
apply simple codes and to analyse the coded information.
Case study
The research question is: why are UK leading political elites first drawn into politics and
particular parties?
You have chosen a comparative approach which relies, in the first instance on
primary, published sources. Your starting, theoretical hypothesis is that choices of
party and political careers are caused by a combination of environmental factors.
You believe that they will include: family, education, religion, class, occupation and
circumstances. You have chosen a modified form of grounded research in which you
start with a small sample of two cases. You will analyse these before seeking further
cases until you achieve theoretical saturation.
You have been fortunate to obtain transcripts of interviews by the same

interviewer of two UK party leaders and prime ministers in the late 1960s early
1970s. They are Harold Wilson (Labour) and Ted Heath (Conservative). They
are both described as ‘self-made’ men. Wilson was the first Labour prime minister
to be born into the Labour party. You could argue that his career trajectory was
determined by the set of institutions into which he was born. Heath was the first
Tory prime minister since Disraeli to be born outside the circle of rich, Anglican,
landed, aristocrats or wealthy businessmen. His acceptance by the Tories can be
attributed to his Oxford education, wartime service and the need for the party to rid
itself of its archaic, grouse-moor image of previous premiers, Macmillan and Lord
Douglas-Home.
Harold Wilson, 1963 (1916–95), UK Prime Minister, 1964–70, 197–76
(Transcript of interview first published in The Observer from Harris, K. (1967) Conversations.
Hodder and Stoughton. pp. 266–86.)
Harris: Some men go into politics almost as a matter of course and

some form of personal Damascus. Why are you in politics?
Wilson: I suppose the short answer is because politics are in me, as far as I can
remember. Farther than that: they were in my family for generations before me, as they
were in the families of dozens of members of the Labour party. The first time that I
can remember thinking systematically about politics was when I was seven. I was in
hospital with appendicitis. My parents came in to see me the night after my operation
and I told them not to stay too long or they’d be late to vote – for Philip Snowden.
Then when I was 10 I went to West Australia where my uncle was an Australian
MP and later President of the Upper House. That was my first visit to a politician.
Harris: Why was your family so politically minded?
Wilson: They were non-conformist by religion and radical by temperament. The Lib-Lab
tradition. The day after the 1906 election results came out in Manchester, my grandfather –
he was a Sunday school superintendent – chose the hymn: ‘Sound the loud tumbrel
o’er Egypt’s dark sea! Jehovah hath triumphed, his people are free’. It’s the old story of the
pursuit of religious freedom and indignation with social injustice combining to conflict with the
established social order. My other grandfather, too, was a deeply religious man who believed
that politics represented the nation’s application of religious principles.
In my childhood, it was chapel and the scout movement, that kind of pattern. My wife’s the
daughter of a Congregationalist minister – I met her at a sports club. So I was impregnated
with nonconformity. It was the soil out of which the Labour party grew. Incidentally, there are
plenty of second-generation nonconformist radicals in the Labour party, men and women
whose approach to politics stems from the religious values their parents planted in them.
There are a lot of first generations, too. I don’t think that you can understand the Labour party
if you don’t bear that in mind. Or the trade union movement. Or the Co-op[erative Movement].
Harris: Does the religious side of the inheritance mean anything

to you today? Are you a religious man?
Wilson: I have religious beliefs, yes, and they have much affected my political views. But I’m
no theologian…
Harris: You talk of the nonconformist radical influence on

the development of your political views. How much has Marx
influenced you?
Wilson: Not at all. I’ve studied the subject as history: you can’t understand the Russians
without it. But, quite honestly, I’ve never read Das Kapital. I only got as far as page 2 – that’s
where the footnote is nearly a page long. I felt that two sentences of main text and a page
of footnote were too much.
Harris: Who has influenced you?
Wilson: Well, as I have said, my parents and the Scout movement … but there was another
indirect influence. The two men who influenced me most directly were schoolmasters.
One of them is still alive … a lifelong socialist. Not by argument, but by example. He was
a great teacher – unselfish and unstinting. When I was fourteen I caught typhoid, and missed
two terms at school.When I got back to school, I was miles behind – miles. It looked hopeless.
I reckoned I could catch up in things like history. But not in mathematics. The second day of
term he called me up to his desk. He said: ‘if you put in an extra hour in the afternoon after
school closes, you should be able to make it all up this term. If you’re willing, I am’. I did.
Harris: Was the other teacher a socialist?
Wilson: Yes, a much younger man. He died very young …
Harris: You seem to have a high regard for teachers.
Wilson: Coming from my kind of background, the teachers were the most important adults in
your life. And my mother and my sister were teachers, and two uncles. Politics is education.
Harris: I’ve taken you off the point. I was asking who
influenced you?
Wilson: At Oxford, my tutors: for example, my tutor in politics, R B McCallum, who is Master of
Pembroke now. A Liberal politically. He taught me so much about the mechanisms of politics,
such as parliamentary standing orders, the works of the public accounts committee and so
on. When I became a don at University College, there was G D H Cole, a socialist. Both had
a lot of influence on my ideas, and on the training of my mind.
Harris: Who had the most personal influence on you at Oxford?
Wilson: Beveridge. I was his assistant – his research assistant – for a couple of years before
the [1939–45] war … Then there was Attlee. A great prime minister. The story of Attlee
has yet to be told, though I think that the public has been getting a fuller notion of his real
contribution to history in the last four or five years.
Harris: Is it true that as leader of the party you are modelling

yourself on him?
Wilson: I would like to think that I was. I learnt a great deal from him. I think that I’m
driving the party at a faster rate than Clem [Attlee] did – at the moment. His method
was to sit and listen and then say: ‘We all seem to want it this way’. I talk more, far too
much in fact, put more cases, am more positive. But the circumstances are different …
Harris: Did anybody else influence you personally? How?
Wilson: Nye [Aneurin Bevan]. It’s very difficult to abstract from Nye’s influence on me. It was
the whole man, you see. He taught me the power of the public platform. I wish I had a tenth
of his power on the public platform. And he corrected my interest in detail. Nye always took
the broad view. In many ways he was lazy, and avoided detail, if he could, anyway, but he
had the gift of instinctively seeing the horizon. He never failed to see the wood for the trees.
I regard him as the best-educated man I have ever met, particularly in philosophy. He could
see everyday political issues in the deep human perspective and he could communicate
them simply and instantly to anybody.
Edward Heath, 1966 (1916–2005), UK Prime Minister, 1970–74.
(Transcript of interview first published in The Observer from Harris, K. (1967) Conversations.
Hodder and Stoughton. pp. 255–65.)
Harris: What made you a Conservative in the first place? Are you
a Conservative because your father was a Conservative?
Heath: No, in more ways than one. The other day my father said that he had been a Liberal.
No, I certainly didn’t become a Conservative because my family were Conservative. I’m not
sure what they were. We hardly ever talked about politics in the family because as a family
we weren’t particularly interested in politics.
Harris: What were your parents like?
Heath: My father was a very likeable and good man, and a good craftsman – a carpenter. I
was born in Broadstairs [Kent] but during the First World War he went to work at Vickers –
woodwork for aircraft – and he came back to Broadstairs when I was seven or so, and the
family with him. I still have a sense of everything beginning when I came back to Broadstairs
at the age of seven.
My mother was a fine character with a strong personality, a high sense of morality and public
responsibility, but none of which she projected on anybody, including the family. She had the
spiritual sense of her Christian faith, not that she was all ethereal about it – she kept her
feet on the ground. I think that she was rather a remarkable woman, because she had high
standards and strong views and yet she was a wonderfully forbearing mother towards her
children. For instance, when I was thirteen I went to Europe on my own, or with one or two
other boys, and I only discovered later how much she worried about me without letting me
ever know.
Harris: Even if your parents weren’t politically minded, do you

think that the home background had any effect on you becoming
a Conservative?
Heath: Oh yes. I am what I am politically because of a combination of home background

and the times I was living in as a boy and a certain amount of thinking about both. Living in
Broadstairs [Kent], we didn’t see the mass unemployment they suffered in S Wales, Scotland,
N Ireland and parts of England. We knew about it, and felt about it, but it didn’t stamp
itself upon our minds as it did on some of the people who lived in the middle of it. But
conditions were tough for nearly everyone nearly everywhere, and I saw my father working
hard, showing enterprise, taking risks, and gradually developing his capacity in spite of very
difficult circumstances.
I had the feeling then – and I have it more clearly now – that father could not have achieved
what he did – I’m thinking about personality, and capacity, not income or status – if he had
been left free to do so. My mother’s influence – it follows from what I’ve said already – had
to do with freedom too. She left us boys free – to go one’s own way, and because she did
that, I had the sense of moving forward, growing up, without the feeling that I was doing it
against my parents. On the contrary, I was doing it with them, because of them, and that
one’s parents were behind one – which makes such a difference.
As I got to 17, 18 years old, the country as a whole was moving out of the period in
which it was preoccupied with the suffering of the unemployed – things were looking up –
and moving into the threat of dictators. This was the moral dimension – of my youthful
associations with the attractions of freedom. By the time I went up to Oxford, if there was
one political idea in the broadest sense of the term which governed me it was a deep
feeling about and for the idea of freedom. I was very conscious of this. I joined all the
three parties on my first term at Oxford, to see what they were up to, but by then I had
accepted that I was a Conservative as part of accepting, so to speak, the facts of life.
Harris: Who influenced you most at Oxford?
Heath: Just after I got to Balliol – I was an organ scholar there – they were putting a new organ
into the chapel, and, partly because of this, and partly because I later became President
of the Junior Common Room, I saw a great deal of the Master of College, A D Lindsay.
(Harris: A socialist).
Heath; yes, and in a way because of it he strengthened my own Conservatism. He was

completely non-dogmatic and a non-doctrinaire thinker and doer. Any way, what influenced
me most about him was that he was a great believer in the democratic expression as central
to the whole system of political democracy, which in term, he thought the only political system
which enabled men to apply the freedom which is the prerequisite for doing the only thing in
life that really matters – the chance for each man to live a full life – not a prescribed full life,
but his own full life …
Harris: Oxford, then, was a very formative experience for you?
Heath: Yes, formative, inspirational. But, as I say, it shaped and gave a clearer meaning to
what … But it also started processes for me. It wasn’t merely that I went up to Oxford but
that I went to Balliol. Balliol is a college of high academic standards and strong democratic
traditions where old Etonian sons of dukes or millionaires are expected – and expect – to
meet on equal terms, well, the son of a carpenter. Balliol opened all the doors to me …
Harris: You were quite a prominent Christian at Oxford, weren’t

you? How much of that remains with you? Does Christianity mean
something fundamental to you?
Heath: Good lord, yes. It’s been the same all along. I was brought up in a Christian home,
as they say in the obituaries, and at Oxford – well, I just went on as I do today … I believe in
God, very much the kind of God, I would say, that ministers of the Church of England would
believe in. I believe in Christ and in the divinity of Christ. I believe in an after-life …
Harris: How far does your religion come into your politics?
I don’t think that there should be too much talk about religion and politics. And I think that
Temple [war-time Archbishop of Canterbury, 1942–44] was right: the Christian should not aim
at programmes of action. It’s the declaration of values and principles and inward personal
attitudes that come from faith …
Harris: Apart from your religion, has any system of ideas, or any
particular person’s ideals, had much influence on your views?
Heath: Yes, a great many. I don’t get much time to read now outside of what I have to read,
but I’m a reader by temperament and when I’m free I read a lot. But it isn’t the people I’ve
read most of that come to mind. It’s more what I’ve read about them. Disraeli, for instance.
His one-nation approach, his refusal to be distracted from the search for the national interest
by the existence of class and sectional interests. I found that profoundly sympathetic. And
I admired him because he didn’t come from the top men of politics of its day, and he was
opposed by them in his early political life and had to struggle to preserve his confidence in
himself, let alone succeed in getting the confidence in others.
Locke also comes to mind too. Locke’s ideas I found gave me a base on which to think about
my own – his emphasis on liberty enshrined in political institutions based upon experience
and judgement, not on doctrinaire theories and utopias.‘No man’s knowledge can go beyond
his experience.’ And his view of the state. It exists for men, not men for the state. Indeed,
it’s only a convenience, the state, to deal with the problems you’d have if each man tried to
be a judge in his own case. And I admire the cast of his mind, Christian, calm, temperate,
good-humoured.
Codes
You have chosen and designed the following generic, theoretical codes:
Variable Education E
Sunday school Ess
prep school Ep
secondary school Es
sixth-Form Es6
university Eu
teachers and tutors Et
variable Family F
Father Ff
Mother Fm
Sibling Fs
Grandparents Fg
Uncles and Aunts Fua
Wife Fw
Children Fc
Others Fo
variable Religion R
Non-conformism Rn
Methodism Rnm
Church of England Rcoe
Ministers Rp
Lay preachers Rlp
variable Politics P
Labour Pl
Conservative Pc
Leading Members Pm
MP Pmp
variable Life Stage L

Child Lc
Student Ls
Young adult Lya
Middle-age Lm
Old age Lo
variable Events/Experiences EX
Depression EXd
World War EXww
Overseas visits EXo
variable Voluntary Bodies V
Trade Union Vtu
Co-op Vcoop
Scouts Vsc
Sports clubs Vsp
The codes can now be applied to the text. This is shown in Table 17.1. Note how
landscape format has been used to transcribe the text. This effectively condenses the
text and provides greater continuity from page to page in both its printed form and
on-screen display. Note how line numbers have been added to your transcript to
enable you to cross-reference quotations later to their original source.
You will quickly find out that this first-time (first-pass) coding is time-
consuming – especially if you code to screen. But it will save you time later.
When you apply the codes, you will learn whether they capture all the independent
variables that you identify in the text. You may decide that additional, in vivo codes
would be useful. Two variables for additional coding might be ‘significant others’
and ‘music’.
Examine your coded sheet. You will see that substantial parts of the transcript
have been coded. This shows that the material is an information-rich source.
Note also how many passages have been coded many times. For example, the
part of the sentence lines 012 and 013 attracts three codes (Fg, Ess, Rlxp). These
codes show the concurrence of family (grandfather), education (Sunday school) and
religion (superintendent). This appears to confirm your original hypothesis of family,
religion and education as causal variables of political consciousness. However, what
is missing from this and other parts of the Wilson interview is any mention of
contemporary events, despite his childhood in the late 1920s in a northern UK
mill town. He makes no reference to the Depression, the rise of Fascism and the
appeal of Communism. His politics appear to reflect an entirely historical rather than
experiential basis. Compare this with Heath’s account. Heath describes the influence
of experiencing his own family’s hardship, reading of unemployment in the depressed
areas, the growth of Fascism and joining all the political parties at university to
experience their appeal. Perhaps this comparison may explain the greater sense of
Table 17.1 Coding of Wilson interview (paras 1 and 2 only)
Remarks Line Coding

Number
001 Harris: Some men go into politics almost as a matter of course and some form of personal Damascus.1 Why
002 are you in politics?
His political 003 Wilson: I suppose the short answer is because politics are in me, as far as I can remember. Farther than that: P
career seems 004 they were in my family; for generations before me, as they were in the families of dozens of members of F
‘written for 005 the Labour party. The first time that I can remember thinking systematically about politics was when I P , Lc
him’. Didn’t 006 was seven. I was in hospital with appendicitis. My parents came in to see me the night after my operation Ef , Fm
he ever rebel 007 and I told them not to stay too long or they’d be late to vote – for Philip Snowden. Then when I was ten I Pm, Lc
against the 008 went to West Australia where my uncle was an Australian MP and later President of the Upper House. EXo, Fa, Pmp
orthodoxy or 009 That was my first visit to a politician.
look at
alternatives? 010 Harris: Why was your family so politically minded?
No mention of
Damascene 011 Wilson: They were non-conformist by religion and radical by temperament. The Lib-Lab tradition. The Rn
events 012 day after the 1906 election results came out in Manchester, my grandfather – he was a Sunday school Fg, Ess, Rlp
013 superintendent chose the hymn; ‘Sound the loud tumbrel o’er Egypt’s dark sea! Jehovah hath triumphed,
014 his people are free’. It’s the old story of the pursuit of religious freedom and indignation with social injustice
Social and 015 combining to conflict with the established social order. My other grandfather, too, was a deeply religious
political 016 man who believed that politics represented the nation’s application of religious principles. // In my Fg, R
engineering? 017 childhood, it was chapel and the scout movement, that kind of pattern. My wife’s the daughter of a Lc, Rn, Vsc
018 Congregationalist minister – I met her at a sports club. So I was impregnated with nonconformity. It Fw , Rlp, Vsp
019 was the soil out of which the Labour party grew. Incidentally, there are plenty of second-generation Rn, Pl
Is Wilson 020 nonconformist radicals in the Labour party, men and women whose approach to politics stems from Pl , F , Lc
being entirely 021 their religious values their parents planted in them. There are a lot of first generations, too. I don’t think
truthful or is 022 that you can understand the Labour party if you don’t bear that in mind. Or the trade union P , Vtu, Vcoop
this post-facto 023 movement. Or the Co-op.
justification to
appeal to
readers?
political conviction and obduracy displayed by Heath when he was prime minister
in 1970–74. However, you must beware that these observations are being made
from only two, relatively short interviews and with the benefit of hindsight.
The next stage is to abstract the coded passages into a table whose columns are
formed of the primary codes used. In your research, the outcome is membership
of a particular political party. So place this dependent variable – politics - in the
first (left-hand) column. The process of abstracting highlighted passages is relatively
easily done by copying and pasting the highlighted passages and their codes. The line
numbers are added in brackets to enable cross-referencing to their source. Where
passages are multiply-coded, then they should be copied in each of the relevant
columns and the relevant sub-sections highlighted. This is shown in Table 17.2.
The next task is to look for patterns and clusters. In this context, a pattern is a
repeated set of relationships. One of the simplest methods of identifying these is to
compress the analysis table and to colour-code or grey-scale the contents. This is
shown in Table 17.3.
This tabulation enables patterns and clusters to be seen to emerge. For example,
religion and family can be seen to be more strongly associated with politics.
However, that represents only the first two paragraphs of the interview. Suppose
that the tabulated coding of the full interview produced Table 17.4.
You can see how education, events and significant others become associated with
politics. If you wish, you can calculate the number of passages.
Second-level pattern coding
Second-level pattern coding builds on the original coding by developing meta-

codes which group various interviews, etc. into sets, themes or constructs (analogous
to cluster or factor analysis in quantitative analysis). This technique can be used to
develop maps and flow charts illustrating the linkages interpreted between variables.
For example, in the Wilson case, you could begin to sketch out the variables, overlaps
and links.
In this way, you can explore and demonstrate overlaps using a combination of
text boxes, circles and arrows. The preceding text boxes demonstrate the sequence
of influencing variables and how they relate to one another. What emerges from
this analysis of the Wilson interview is the enormous overlaps between family,
religion, voluntary organisation and early education. This suggests that they are
visible manifestations of an underlying latent variable or factor. What might that
be? One strong contender would be ‘class’ – northern, industrial working class
(although Wilson described his parents as lower, middle-class). And what (from a
Marxist perspective) is the determinant of class: the mode of production. You have
therefore revived the explanatory power of class which was central to post-war UK
political science.
Table 17.2 Wilson interview: coded passages tabulated
Politics Family Religion Education Life Stages Events Vol.

Organisations
Politics… were in Politics … were in my
my family; for family; for generations
generations (004)
Remember thinking Remember
about politics thinking about
when I was seven politics when I
was seven (005)
My uncle was an I was ten when I went to I was ten when I went to
MP Australia where my went to Australia Australia
uncle was an MP (008)
Non-conformist by Non-conformist by
religion (011) religion
My grandfather was a My grandfather was My grandfather was
Sunday school a Sunday school a Sunday school
superintendent (012) superintendent superintendent
Other grandfather Other grandfather was a Other grandfather
was a deeply deeply religious man was a deeply
religious man who believed that religious man
who believed that politics represented who believed that
politics the nation’s politics
represented the application of religious represented the
nation’s principles (015) nation’s
application of application of
religious religious
principles principles
(Continued)
Table 17.2 Cont’d
Politics Family Religion Education Life Stages Events Vol.

Organisations
In my childhood, In my childhood, it In my childhood, it
it was chapel and was chapel and was chapel and
the scout the scout the scout
movement, that movement, that movement, that
kind of pattern kind of pattern kind of pattern
My wife’s the daughter of My wife’s the My wife’s the
a Congregationalist daughter of a daughter of a
minister – I met her at Congregational- Congregationalist
a sports club (017) ist minister – minister – I met
I met her at a her at a sports
sports club club
So I was So I was
impregnated with impregnated
nonconformity. It with
was the soil out of nonconformity.
which the Labour It was the soil
Party grew. out of which the
Incidentally, there Labour party
are plenty of grew (018)
second-
generation
nonconformist Second-
radicals in the generation
Labour party, nonconformist
men and women radicals in the
whose approach Labour party,
to politics stems men and
from their women whose
religious values approach to
their parents politics stems
planted in them. from their
religious
values their
parents planted
in them. (019)
Table 17.3 Wilson interview: coded passages grey-scaled
Politics Family Religion Education Life Stages Events Vol. Orgs.
The next task is to apply this potential explanation to the Heath interview. Heath
attributes the appeal of Conservative politics to a ‘combination of home background
and the times’. He extends the influences to ideas of freedom. But perhaps the
most telling passage is his description of his father: ‘. . . working hard, showing
enterprise, taking risks, and gradually developing his capacity in spite of very difficult
circumstances’. His father was a self-employed carpenter working in a predominantly
rural county. In short, the Heaths were petty-bourgeoisie and church – rather than
chapel – goers. However, the evidence is not wholly conclusive. But both appear
to have conformed to the political norms of their social class.
Another step would be to prepare a comparative table drawing together material
from the texts.
End stage
The last stage in your analysis is to construct a typology – an explanatory, conceptual

framework which represents graphically all the cases within two, unrelated variables
or, preferably, factors. It therefore provides a very powerful display of your data
and analysis and provides a basis from which you can develop or extend theoretical
explanations.
In terms of explaining the causal factors of political identity, you might construct
a typology using the factors of class-of-birth and conformity where wider experience
was likely to lead to individuals rejecting the party of the class into which they had
been born.
Footnote
However, there are no easy fixes in coding and its analysis. You will have to explore
and prepare different ways of ‘differentiating and integrating’ your information until
Table 17.4 Wilson interview: tabulation of complete interview
Politics Family Religion Education Life Stages Events Vol. Orgs. Sig. Others Music
Table 17.5 Calculated number of passages
Politics Family Religion Education Life Stages Events Vol. Orgs. Sig. Others Music
29 16 15 21 3 8 2 14 0
Family
Religion
Voluntary
Education organisations
Significant
others
Figure 17.1 Graphical representation of relationship between variables
some signs of clarity appear. Be prepared to invest considerable time in the process:
as much as (if not more than) you spent collecting and transcribing the data.
Other methods of analysis
These include memoing, case analysis meetings, interim case studies and pre-structured cases
(Miles and Huberman, 1994: 73–85).1
Memoing
Writing a memo generally means writing an informal note to yourself or colleagues
of new information. In the context of research, memoing has a more specific
meaning:
the theorising write-up of ideas about codes and their relationships as they strike the
analyst while coding … it can be a sentence, a paragraph or a few pages … it exhausts
the analyst’s momentary ideation based on data with perhaps a little conceptual
elaboration.
(Glaser, 1978: 83)2
Table 17.6 Coding: Wilson and Heath compared
Code Wilson Heath

Rn Nonconformism Allied to social justice; soil for
Labour party
Rcoe Church of Mother’s spiritual sense of Chr faith
England
*[his] belief in God that ministers of
CoE share
Rnm Chapel Essential part of childhood
Rp Ministers Father-in-law
Er School (Implied)
(Religious music) Chapel hymns reflected political Organ scholar
events
F f Father (No specific mention) A good man, craftsman, self-made

F m Mother Teacher Fine character, strong personality
F w Wife (never married)
F o Other relatives Uncle, grandfathers
E t School teachers Two: most important adults in

your life
E t Tutors, etc. R B McCallum A D Lindsey
G D H Cole
Beveridge
So Significant others Attlee Disraeli ‘one-nation’
Bevan (Nye)
Locke on liberty and the state
In other words, when you have a Eureka moment, insight or brainwave during your
coding and analysing of text, write it down and continue with your activity. Copy
the memo to your colleagues or supervisor for their views. Return to your memos
when you have finished your coding. Many of your bright ideas will prove false
dawns. But you will not have forgotten them and have evidence of all the leads
that you explored and discounted.
Table 17.7 An analytical typology
Conforming Non-conforming
Conforming working class (Wilson) Nonconforming working-class (D. Davis)
Class Conforming middle class (Heath) Nonconforming middle class (Attlee)
Conforming upper class (Macmillan) Nonconforming upper class (Mosley)
Case analysis meetings

Case study meetings are similar to supervision meetings or peer-reviews. They are
designed to seek formalised discussions on your analysis so far. They are particularly
valuable to review progress against timetables, or to talk-through problems, new
opportunities or unexpected findings.
Interim case study

Miles and Huberman strongly recommend researchers to write a 25-page interim
case study report to: ‘prevent the nightmare of bad or opaque data, systematic
error, or blindingly obvious or trivial conclusions’. You should complete this when
you are one-third the way through the research (comparable to the ‘up-grading’
review procedure through which postgraduates must pass before the initial MPhil
registration can be raised to DPhil/PhD). This interim report should demonstrate
the type of data and information that you have collected and the effectiveness of
your analytical methods. If you are unable to demonstrate a functional relationship
between the data and its analysis, then clearly you must consider collecting different
types of data or adopting other methods of analysis – or both.
1. Consider and discuss the transcripts of the interviews with Harold Wilson and
Edward Heath. On the basis of your theoretical understanding gained from other
courses, what causal influences on their political choices would you suggest? Copy
them from www.sage.co.uk/pierce. Develop your own generic codes and apply
these. What additional, in vivo codes would you suggest? Tabulate the codes and
extracts for Wilson and Heath. Tabulate a comparison of the transcripts. Develop
a typology of your own. What inferences would you suggest?
2. Visit www.bbc.co.uk On the home page, enter ‘Paxman Interviews’. Select an
interview with a UK political elite. If in doubt, use interview with Tony Blair on
27 April 2005. Click on ‘printable version’. Download, transcribe and code. Note
that you can also watch video recordings of many BBC interviews.
FURTHER READING
Gubrium, J. F. and Holstein, J. A. (1997) The New Language of Qualitative Method.

Oxford: Oxford University Press. p. 244. This concise textbook explores the ‘new
language’ of analysis within the discipline of sociology and the perspectives of
naturalism, emotionalism, ethnomethodology and postmodernism.
Miles, M. B. and Huberman, A. M. (1994) Qualitative data Analysis. London: Sage.

p. 338. This US textbook provides detailed guidance on coding of text and its
analysis. The authors pay particular attention to the value of matrices, tables
and charts in data analysis. The book draws its examples from US public policy
practice, especially education reform.
Ritchie, J. and Lewis, J. (eds.) (2003) Qualitative Research practice: A guide for
social science students and researchers. London: Sage. p. 336. This is a
well-written and readable textbook which is written for a readership of all social
science disciplines. It draws on examples from UK practice. However, only two
chapters describe the analysis of qualitative data. The analytic hierarchy shown
on p. 212 is invaluable to understand the process of analysis.
Silverman, D. (2001) Interpreting Qualitative Data: Methods for analysing talk, text
and interaction. London: Sage. p. 325. This is an excellent monograph by a
leading UK academic. He promoted three models for interpreting interview
data: positivism, emotionalism and constructionism. Methods used include
content and discourse analysis. Exceptionally, the textbook includes a chapter
(Chapter 7) on the analysis of visual images.
Notes
1 Miles, M. B., and Huberman, A. M. (1994) Qualitative Data Analysis. London: Sage.
2 Glaser, B. G. (1978) Theoretical Sensitivity: Advances in the Methodology of Grounded Theory. Mill
Valley: Sociology Press.
Chapter 18
Using Content Analysis
1. To consider alternative definitions of ‘content analysis’.

2. To compare qualitative and quantitative content analysis.
3. To consider the sub-types of content analysis.
4. To learn how to carry out content analysis using latest available software.
Introduction
Content analysis is also called textual analysis and, somewhat pejoratively,

text mining. In this context, content is words, texts, transcripts of speeches or
conversations, pictures, ideas, themes or messages. Content analysis is concerned
with the bits and pieces – words – of communication rather than the generality.
It is a long-standing tradition and was widely used by church authorities to search
out heresies and to prosecute heretics. Max Weber encouraged its use, in Weimar
Germany, to monitor and compare press reports – in terms of coverage and bias –
of his Social Democratic Party. During the Cold War, content analysis was widely
used by Kremlinologists and Sinologists to monitor the stream of radio traffic, official
reports and statements from USSR and Chinese sources. Its use has become even
more widespread as a result of new software designed to automatically scan all
telephone and emails etc. to detect key words as part of the ‘war on terrorism’.
In a sense, content analysis has become part of the apparatus of the vigilant – or
oppressive – state and a source of increasing paranoia among political activists. It
is a ‘key concept’ of the new popular discipline of media studies (O’Sullivan et al,
1997: 62).1 In this context, text is considered to have an independence from its sender
or intended receiver. It can be any message written – visual, spoken or sung – as a
medium for communication including books, newspapers, advertisements, speeches,
official documents, films of video, musical lyrics, photographs, clothing (especially
T-shirts), graffiti, works of art and national anthems.
The advantages to researchers of using content analysis include:
• ‘there’s plenty of it’: widespread availability of texts – especially on the Internet

• greater ease of scanning or downloading text for analysis
• relatively low cost
• simple, covert research method which minimises the research effect: (Hawthorne, etc.)
• distance from subjects
• quantifiability
• consequently, claims to objectivity.
On the other hand, the criticisms that can be made of content analysis include:
• researcher bias in the selection of texts

• conceptual assumptions are highly contestable
• application of Boolean and other mathematical software is bad science
• problem of identifying the population of texts and, therefore, of obtaining a random,
representative sample of texts
• heavy promotion by commercial interests
• benefits are exaggerated by users.
I regard content analysis as a potentially useful but supplementary tool of analysis.
Qualitative and quantitative content analysis
Content analysis may be qualitative or quantitative. Both can use the same body
of textual and other data. Qualitative content analysis is highly interpretive.
It essentially involves the reading of texts, etc. to determine the extent of bias in
terms of supportive, critical or (more or less) neutral accounts of organisations,
institutions, concepts or figures. In York, it was used by students over many years
to analyse the bias of British print media towards the Loyalists and Nationalists in
Northern Ireland. The approach here, as elsewhere, is to select a sample of texts
over a long period of time for independent reading and analysis by a number of
researchers whose own biases are recorded at the beginning of the exercise. The
assessments are then compared.
Alternatively, qualitative content analysis can be used to compare different
perspectives on the same topic by different speakers. In 2004, as an exercise
at the Essex Summer School, I compared the interpretations of ‘democracy’ of
world leaders in their speeches to the UN following 9/11. My analysis is shown
below. You will note the differences in the meanings of democracy adopted and
promoted by the US President and European leaders. Note also how Putin adopts
an instrumental view of democracy.
BLAIR
CHIRAC
cooperation
liberty
human rights
responsibility
peace and prosperity
BUSH fear freedom and dignity
peace and
tyranny multilateralism
development
censorship state sovereignty
peace and prosperity
from taxes rule of law under UN
freedom hope rule of law
equal dignity of all cultures
to choose respect for others
respect for diversity
raise families liberty rather than
dialogue
vote freedom
assemble stability partnership SCHROEDER
worship uniting disparate institutions to guarantee human
groups tolerance rights
God + Democracy + Liberty justice popular participation
hard work shared trade
opposite of responsibility to defend
offering asylum to sustainable development
communism greater vigilance,
security refugees disarmament and non-proliferation
military action merit full state sovereignty
sacrifice worth social and material security
intervention minority rights
response BURLESCONI
pre-emption freedom and democracy PUTIN
can't be neutral defence of human rights market economies
peace and development decent standard of living
freedom from totalitarianism choice
route from poverty European humanism
anti-totalitarianism
myth of partnership (NATO)
caliphate
Figure 18.1 Interpretation of world leaders of ‘democracy’ in speeches to the UN following 9/11
This rough and ready approach has been replaced by the development of discourse
and narrative analysis as the principal method for the qualitative analysis of text.
Quantitative analysis
Most content analysis is now entirely quantitative. It counts the incidence and
frequency of words. Its theoretical underpinning is provided by Zipf’s Law –
an experimental law developed by George Zipf (1902–50), Harvard professor of
linguistics. Zipf’s Law is, technically, a power-law distribution. The law states that
the frequency of occurrence of some event (P), as a function of the rank (i) when the
rank is determined by the above frequency of occurrence, is a power-law function
Pi ∼ 1/i a with the exponent a close to unity. What this means is that, in any body of
text, the frequency of any word is inversely proportional to its rank in the frequency
table (of most commonly-used words). In English texts, the word ‘the’ is the highest
ranking word. So, in any text, it is likely to be the most frequently used word and,
indeed, is used twice as often as the next, highest-ranking word.
What Zipf means in plain English is that, where uncommon words (or phrases)
are used often in a text, then they express and reflect the greatest concerns of the
communicator. These are termed key words. For example, a Zipfian analysis of
G. W. Bush’s presidential inaugural address of 20 January 2005 shows that, in a
speech of only 2,083 words, the key words used were:
Key word Frequency

freedom 27
liberty 15
hope 8
history 7
tyranny 5
God 4
However, the word ‘Iraq’ was not mentioned at all. This example readily
demonstrates the great weakness of content analysis as a tool of Politics research: that
the key words used by political elites (and their speech-writers) do not necessarily
express what may be their greatest concerns. Hence the criticism made of Margaret
Thatcher by a cabinet colleague that: ‘The trouble with Margaret is that, when she
speaks without thinking, she says what she thinks’.2 Indeed, the key words may be
used deliberately to conceal or divert attention from real policy concerns. Political
speeches have become carefully-planned, ‘communications events’ whose contents
will have been analysed to ensure that the ‘right words’ are used the ‘right’ number
of times and ‘resonate’ (appeal) to the target audiences. However, paradoxically,
content analysis by others can enable omissions to be identified.
Using Content Analysis 267
You can carry out your own simple content analysis by copying texts of speeches
from the Internet and then using the Edit/Replace function in Microsoft Word.
First, identify your key words. Then ask Word to replace all your first key word by,
say, xyz. Word will tell you how many occurrences have been replaced.
Quantitative content analysis can be either structural or substantive. Whichever you
adopt, you must firstly clearly identify the population of texts and sample frame and
justify fully the (random or non-random) sampling method you adopt.
Structural content analysis
Structural content analysis is primarily concerned with how the text is presented
and reported rather than the frequency of key words used. It seeks to measure:
• space (or time) devoted to the text

• volume: headline type and font size
• position or prominence of the text
• use of accompanying illustrations or photographs
• indications of direction or bias (for example, by stereotyping)
• persistence over time (for example, Daily Mail ’s campaign over ‘asylum seekers’ and
Daily Express’ pre-occupation with Princess Diana)
Structural content analysis is widely used as a comparative technique by media

watchers. It is also employed by political parties to make complaints of bias against
public broadcasters and thus to secure better, corrective coverage. But the public
broadcasters also use structural content analysis in the design of their programmes
to rebut complaints. One very useful, simple and instructive exercise that you can
perform is to track and compare the front page editions of all the national newspapers
over one week. Note how the choice of front-page stories varies between the
‘quality’ (Times, Guardian, Independent, Telegraph) and popular, ‘redtop’ newspapers.
Then compare them with the headlines on the BBC, ITN and Sky Internet news
sites. For example, the headlines for 13 January 2007 are:
BOX 18.1 National Newspapers, circulation (000s) and

headlines, 13.01.2006: TV News headlines, 13.01.2007
The Times 635 NHS faces treatment rationing

The Guardian 365 Revealed: the 11 government ministers fighting NHS cuts
Daily Telegraph 899 Brown’s manifesto for Britishnesss
(Continued)
The Independent 238 Shoot the messenger: Blair blames press for anti-war mood
Daily Express 773 Big new tax on house owners
Daily Mail 2,311 Now a school bans crucifix
The Mirror 1,540 Big brother shot my dad dead
The Sun 3,026 Shoot him, mother’s anger at beast who raped girl, 3
The Star 750 Big Bro on skids
TV news (audience for post-watershed news)
BBC News 5,000 Reid moves to quell offender row
ITN news 3,000 Boys killed by Tube train
Sky News 2,000 Men killed in Tube tragedy
Source: Audit Bureau of Circulation www.abc.org.uk, BBC, ITN and Sky News web-sites
Substantive content analysis
Substantive content analysis analyses selected texts by counting the frequency

and distribution of key words. More advanced, software-based techniques use
Boolean search parameters which enable you to refine your research by counting
the frequency and distribution of combinations of key words.
A very good example of traditional, substantive content analysis is provided by
Budge’s use of the method to analyse changes of UK party policy and ideology in
the period 1945–97 (Budge, 1999: 1–21).3 He uses the party manifestos published
by the major parties as election programmes which he argues are:
not widely read by the British public.Their importance is that they are read by the political
and media elite and reported intensively in newspapers, TV and radio. Thus their textual
emphases set the tone and themes of campaign discussion (1999: 2).
He explains that the task of analysing the data is simplified by the coding of all
Western European manifestos since 1945 by the Manifesto Research Group of the
European Consortium for Political Research (ECPR). The Group uses a base of 46,
major, policy areas which are coded generically. They include: 101 Foreign special
relationships: positive; 102 Foreign special relationships: negative; 103 Decolonisation; …
706 Non-economic demographic groups. Each sentence is coded and counted under
one policy area. The percentage of sentences devoted to each policy area is then
calculated. In this way, the coverage given to any policy area can be plotted over
time by extending the analysis over all of the parties’ manifestos in chronological
order.
Budge extended the research by analysing the movement of manifestos across the
political spectrum to establish how far (and often) the parties have changed their
ideological positions and appeals to the electorate. He adopted the left-right coding
scale developed by Klingermann et al (1994).4 This is shown in Table 18.1.
Table 18.1 Left-right coding scale
Left-Wing Emphases (Sum of %s for) Right-Wing Emphases (Sum of %s for)

Decolonisation Pro-military
Anti-military Freedom, human rights
Peace Constitutionalism
Internationalism Effective authority
Democracy Free enterprise
Regulate capitalism Economic incentives
Economic planning Anti-protectionism
Pro-protectionism Economic orthodoxy
Controlled economy Social services limitation
Nationalisation National way of life
Social services expansion Traditional morality
Educational expansion Law and order
Pro-labour Social harmony
Source: Klingermann, et al 1994: 40
Budge applied this scale by adding all the sentences in Klingermann’s ‘left list’ and
subtracting them from all the sentences in the ‘right list’ to provide a scale between
+100% (manifesto wholly ‘right wing’) to –100% (all ‘left wing’). The scores are
charted in Figure 18.2.
40
Left wing (−) / Right wing (+) scores
30
20
10
for each party
0 Conservative
−10 0 20 40 60 Liberal
−20 Labour
−30
−40
−50
−60
General Election years (where 1945 is 0)
Figure 18.2 British parties ideological movement on a left-right scale, 1945–97

Source: Budge, 1999: 5, Figure 11.1
Budge interpreted the chart to argue that:
Labour moves sharply rightwards from 1992 and for the first time in post-war history
shows a preponderance of right wing positions over left ones positions (+5%) … Labour
moved rightwards and ‘leapfrogged’ over the [Liberal–Democrats] … in relative terms,
Labour became the most centrist party.
(Budge, 1999: 6)
You may criticise Budge’s method for its reliance on assumptions that you regard
as very problematic (e.g. the categorisation of ‘human rights’ as a right-wing
phenomenon). However, you must accept that his analysis does appear to chart
accurately the ideological emergence and drift of New Labour to the (more electable)
centre-ground. However, this very good example of effective content analysis also
demonstrates that the method is not a quick fix. In this research project, it has
involved reading and coding 42 election manifestos. So structural content analysis
can be very laborious. And users can also be criticised for substituting hard labour
for the less-laborious but more intellectually demanding methods such as discourse
and narrative analysis (see Chapter 19).
Content analysis software
New software removes the hard labour of traditional content analysis but compounds
criticisms of its use as a principal research method. There are now a large number of
English language software programs for quantitative content analysis. They include:
AutoMap; CatPac; General Inquirer (Harvard); Hamlet II; Leximancer; TACT; Textpack
(Cologne); Texstat (Berlin); VBPro; Wordsmith; WinMax. You can find reviews of
and information on many of these programs at: http://lboro.ac.uk/research/methods/
research/software/stats.html
Each of these types of content analysis software perform a wide range of functions.5
The core functions include:
• calculating word frequencies

• excluding stopwords (‘the’, ‘and’, ‘in’ etc.) automatically
• adopting lemmitisation to combine words with the same stem, e.g. go, going, gone
• using synonyms to categorise as one word all others having the same meaning, e.g.
gone, quit, departed, etc.
• recording concordance by showing each word in its context (termed KWIC: key words
in context)
• using cluster analysis to group together words used in similar contexts
• using co-word citation to identify the concurrence of key words. This is used by the US
government’s Echelon Project to scan emails for terrorist activity. So, emailers using a
combination of, say, ‘bomb’ and ‘Islam’ are likely to automatically attract the attention of
the security services.
The software is designed to analyse relatively short texts of up to 10,000 words.

However, no coding is required. All you have to do is to enter the text you wish to
analyse and the key words. Your selection of text and appropriate key words is therefore
critical to the effectiveness and value of the analysis. Many other programs designed
primarily for coding and analysing texts can perform simple content analysis. They
include ATLAS.ti, Nud*st and SPSS for text.
Hamlet II
Hamlet II can be recommended for a number of reasons. First, it was developed

uniquely by a political scientist, Dr Alan Brier of Southampton in conjunction
with the foremost continental researchers, Ekkehard Mochmann and Bruno Hopp
of the Central Archive for Empirical Social Research in Cologne. Second, the
software provides supplementary graphical displays of completed analysis in the
form of dendograms and three-dimensional (Minissa) displays (which can be rotated).
These dendograms and displays can illustrate what might otherwise be uninteresting
tables and bring welcome, additonal interest to your research report. Third, you
can download a free, 30-day trial copy of Hamlet II from www.apb.cwc.net. A free
tutorial guide is also available.
Hamlet II offers a number of analytical procedures including:
Joint Frequency Analysis

simple cluster analysis
multi-dimensional scaling
PINDIS (Procustean Individual Difference Scaling)
KWIC (key words in context)
Wordlist
Compare (to compare two or more texts)
Profile (which displays the distribution of words and sentence lengths)
To provide an example, Hamlet II has been used to complete a content analysis of the
full text of President Bush’s State of the Union Address 2005. This was downloaded
from the White House’s Internet site at www.whitehouse.gov/news/release/
20050202 as a plain text file. A list of key words was selected from a quick skim of
the speech. They were (in alphabetical order) better world, budget, democracy, economy,
free*, freedom, God, history, hope, justice, liberty, sacrifice, tax, terror* and tyranny. (The
wild card * is used to enable all words beginning with the chosen stem, e.g. free*,
to be counted together.) No synonyms were used. However, the word ‘Applause’
(included in the transcript as ‘(Applause)’) was added to find out which key words
generated the greatest support from Members of Congress. A demonstration copy
of Hamlet II was installed and the instructions followed. The analysis was saved and
copied and is shown in Boxes 18.2a, b, and c:
BOX 18.2a HAMLET II analysis: word search
HAMLET II analysis: word search

HAMLET - Computer-assisted Text Analysis - 17/01/2007 17:32:57
=================================================================
The text is read from the file: State of the Union Address 2005.txt
Counting collocations within a span of 50 words -
WARNING : No collocates for "freedom"
WARNING: some characters were not recognised when reading this file!
WORD-SEARCHING IS INSENSITIVE TO CASE.
There are 17 main entries in the search list.

No synonyms / related items are recognised:
CATEGORY/WORD COUNTS
VOC.LST. FREQUENCY % VOC.LST. % TEXT CONTEXT UNITS
Applause 67 36.61 1.30 67

better world 6 3.28 0.12 6
budget 3 1.64 0.06 3
democracy 8 4.37 0.15 8
economy 11 6.01 0.21 11
free* 27 14.75 0.52 27
God 1 0.55 0.02 1
History 5 2.73 0.10 5
Hope 3 1.64 0.06 3
Justice 4 2.19 0.08 4
Liberty 7 3.83 0.14 7
Sacrifice 1 0.55 0.02 1
Society 5 2.73 0.10 5
Tax 6 3.28 0.12 6
terror* 27 14.75 0.52 27
tyranny 2 1.09 0.04 2
You will note that the most frequently used key words were free* (27), terror* (27),
democracy (8), liberty (7), ‘better world’ (6), tax (6) and history (5). ‘Applause’ was recorded
in the transcript on 67 occasions
BOX 18.2b
5162 words were read from the text file.

183 of these were in the search list, and
collocations within up to 50 words were counted.
JOINT FREQUENCIES ......................................
COLLOCATIONS within up to 50 words:
i
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16
+-----------------------------------------------------
Applause 1 |
better world 2 | 4
budget 3 | 7 0
democracy 4 | 9 0 0
economy 5 | 20 3 2 0
free* 6 | 25 0 0 4 2
God 8 | 1 0 0 0 0 1
History 9 | 6 0 0 1 2 6
Hope 10 | 4 1 0 1 0 3
| 0
justice 11 | 7 0 0 1 2 1
| 0 0
liberty 12 | 6 0 0 3 1 9
| 1 1 0
sacrifice 13 | 1 0 0 0 0 0
| 0 0 0 0
society 14 | 4 0 0 0 0 1
| 0 0 0 0 0
tax 15 | 11 1 2 0 6 0
| 1 0 0 0 0 0
terror* 16 | 26 0 0 6 0 23
| 0 6 0 3 0 0
tyranny 17 | 2 0 0 2 0 3
| 0 1 0 0 0 0 4
BOX 18.2c HAMLET II analysis: Joint Frequency
The joint frequency table clearly shows that ‘Applause’ was most strongly associated with
terror* (26 times), free* (25 times) and economy (20 times). Other frequent associations
are; free* with terror* (23 times); liberty with free* (9 times); history and free* (6 times);
and, hope and free* (3 times).
The probability of a specific pair of words being present in any pair of words is measured
by the Jaccard coefficient. This is shown calculated in Box 18.3c:
HAMLET II analysis: Jaccard coefficients
STANDARDISED JOINT INDEX VALUES

Jaccard coefficient - ignores joint non-occurrence
i
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16
+---------------------------------------------------
-----
Applause 1|
better world 2| 0.06
budget 3| 0.11 0
democracy 4| 0.14 0 0
economy 5| 0.34 0.21 0.17 0
free* 6| 0.36 0 0 0.13 0.06
God 8| 0.01 0 0 0 0 0.04 0
History 9| 0.09 0 0 0.08 0.14 0.23 0 0.20
Hope 10| 0.06 0.13 0 0.10 0 0.11 0 0
| 0
justice 11| 0.11 0 0 0.09 0.15 0.03 0 0
| 0 0
liberty 12| 0.09 0 0 0.25 0.06 0.36 0 0
| 0.09 0.11 0
sacrifice 13| 0.01 0 0 0 0 0 0 0
| 0 0 0 0
society 14| 0.06 0 0 0 0 0.03 0 0
| 0 0 0 0 0
tax 15| 0.18 0.09 0.29 0 0.55 0 0 0
| 0.10 0 0 0 0 0
terror* 16| 0.38 0 0 0.21 0 0.74 0 0
| 0 0.25 0 0.10 0 0 0
tyranny 17| 0.03 0 0 0.25 0 0.12 0 0
| 0 0.25 0 0 0 0 0 0.16
Figure 18.3 Dendogram
The dendogram and minissa display are shown in Figures 18.3 and 18.4. They
demonstrate the functional relationship between the key words used and their
relative proximity.
The example demonstrates the main strengths and weaknesses of using content
analysis software. The main strengths are:
• the ready availability of texts

• the scope for comparative analysis
• easy-to-use (with practice) analytical software
• good tabulations of results
• eye-catching illustrations.
The main weaknesses are:
• concurrence of key words does not necessarily indicate causal links

Figure 18.4 Minissa 3-D display
• the meanings of key words change over time and can be deliberately changed by political
parties who exploit ‘feel good’ or ‘feel bad’ words to make their policies more attractive
(e.g. ‘community’, ‘homeland’)
• overall, the analysis is not really conclusive enough to provide sufficiently reliable
evidence to confirm your research hypothesis.
Content analysis can therefore arguably be recommended only as a supplementary

research method.
1. Where and when should content analysis best be used?

2. Compare and contrast two or more of the available software packages.
3. Working in teams, complete a structural content analysis of last week’s national TV

and print news media accounts of national and international news from Monday
to Sunday. What inferences and conclusions can you draw?
4. Complete a content analysis of all President George W. Bush’s State of the Union
Addresses. Compare the analyses. What are your main conclusions? What are
their limitations?
5. Select two quality newspapers. Using CD-ROMs or Internet sources, evaluate their
(emergent) bias for or against the Republican and Loyalist movements in N. Ireland
between the Good Friday Agreement (1998) and the St Andrew’s Agreement
(2006). What conclusions can you draw? What reasons can you suggest for any
changes of support?
6. A comparison of the key words used by President Bush in his Inaugural Address
(2,083 words) and State of the Union Address in 2005 (5,162 words) shows a
variation in their frequency:
Key word Inaugural Address State of Union Address

Freedom 27 27
Liberty 15 7
Hope 8 3
History 7 5
Tyranny 5 2
God 4 1
Are these differences significant? What alternative explanations can you give?
FURTHER READING
Most textbooks of research methods in Politics mention content analysis but few
give more than a few pages’ coverage.
Krippendorff, K. (1980) Content Analysis: An Introduction to its Methodology.
London: Sage. Despite its age, this remains an essential text for students
contemplating the use of content analysis. Its early publication (1980) means
that it pre-dates PCs and the availability of software. Hence, Krippendorff
provides a very good, detailed account of the assumptions and processes
adopted, and shows how these can be applied manually.
pp. 113–20, provides an authoritative account of the scope and application of
content analysis. She provides a case study of the portrayal by UK news media
of political leadership qualities.
Burnham, P., Gilland, K., Grant, W. and Layton-Henry, Z. (2004) Research

Methods in Politics. Basingstoke: Palgrave Macmillan. pp. 236–42, provides
a well-referenced introduction to content analysis including two case studies.
Budge, I. (1999) Chapter 1: Party Policy and Ideology: reversing the 1950’s?
In Evans, G. and Norris, P. (eds.) Critical Elections: British Parties and Voters in
Long-Term Perspective. London: Sage. pp. 1–21, provides an excellent example
of the application of content analysis. The book also shows how content
analysis can be used to scope and foreground other methods of analysis.
Klingermann, H-D., Hofferbert, R. and Budge, I. (1994) Parties, Policies and
Democracy. Boulder: Westview Press.
Notes
1 O’Sullivan, T., Hartley, J., Saunders, D., Montgomery, M. and Fiske, J. (1997) Key Concepts in
Communication and Cultural Studies. London: Routledge.
2 Matthew Parris, The Times, 13 January 2007.
3 Budge, I. (1999) Chapter 1: Party policy and ideology: reversing the 1950s? In Evans, G. and Norris,
P. (eds.) Critical Elections: British Parties and Voters in Long-Term Perspective. London: Sage.
4 Klingermann, H-D., Hofferbert, R. and Budge, I. (1994) Parties, Policies and Democracy. Boulder:
Westview Press.
5 Brier, A.P. and Hopp, B. HAMLET a Multidimensional scaling approach to text-oriented policy
analysis. In Journal of Diplomatic Language 2(1) (2005) (online at http://www.jdlonline.org//
IIbrier.html), illustrates the comparison of a number of sources.
Chapter 19
Understanding and Adopting
Discourse and Narrative Analysis
‘… the world is structured by discourse’

(Foucault)
‘We cannot get iffy over other people’s power-games with language, and then pretend
we are not players in the game too’
(Curt, 1994:19)1
1. To understand some of the origins of the discipline of discourse.

2. To consider and compare various schools of discourse analysis.
3. To consider in greater detail what is meant by critical discourse analysis.
4. To learn how critical discourse analysis can be applied.
5. To discuss what is meant by narrative.
6. To learn how to apply narrative analysis.
Introduction
Discourse analysis is characterised, at one hand, by growing enthusiasm for its use by
ambitious students and researchers and, at the other, by a variety of interpretations
and advice from leading authorities. The paradox is that discourse analysis has many
interpretations: there is a discourse on discourse. Some regard it as a method.
Others profoundly disagree: they conceptualise discourse analysis as a discipline.
It is, therefore, another ‘essentially contested concept’ (Gallie, 1956).2 Indeed, there
are times when, faced with the wide range of differing authorities, you may be
forgiven to likening discourse analysis to Churchill’s description of Russia as:
a riddle, wrapped in a mystery, inside an enigma.3

In everyday speech, discourse is generally used to describe a discussion, conver-

sation, talk or text. Discourse is essentially a ‘mode of communication’. However,
within the social sciences, a variety of explanations are given. The most readily
understandable and succinct are:
• language as social practice determined by social structures (Fairclough, 1989)4

• systems of meaning, including all types of social and political practice, as well as
institutions and organisations (Howarth, 1995)5
• a representation of what we want the world to be like, rather than a representation of how
the world is, the correctness of which [can] be tested (Carver, 2002; 5)6
• a specific ensemble of ideas, concepts and categorisations that are produced, repro-
duced, and transformed in a particular set of practices and through which meaning is
given to physical and social realities (Hajer, 1995: 44).7
In other words, ‘language is political’.

Similarly, discourse analysis has many explanations including:
• a description for studies focusing only on linguistic units above the level of the sentence
(Stubbs, 1983)8
• analysing the way systems of meaning or ‘discourses’ shape the way that people
understand their roles in society and influence their political activity (Howarth, 1995)9
• an examination of the relationships between units of talk, writing, or other representational
forms, and of the significance of these relationships for our subjective experience (Squire,
1995)10
• discourse analysis does not look for truth – but rather at who claims to have truth (Carver,
2002: 53).11
Essentially, discourse involves language (rather than individual words). But the
modes of discourse also include other sounds (a referee’s whistle), non-verbal
communication (gestures, holding a child’s hand) and signs (for example, traffic
lights). The primary uses of language are:
• ideational (the expression of ideas)

• referential
• propositional.
Additionally, language is used as a means of:
• emotional expression
• social interaction
• an instrument of thought
Understanding and Adopting Discourse and Narrative Analysis 281
• expression of identity
• co-ordination and control of others.
Examples of the use of language for co-ordination and control include parade-
ground commands, children’s nursery rhymes and the litany of religious rituals. But
the meaning of the language will be determined by the context and societal norms.
Consider the example of a brick. A brick on a building site is a component of
construction. A brick as part of a wall is a means of enclosure. A brick to an artist
is a highly textured and variegated surface within a regular and repeated shape. But
if you carry a brick through a crowded shopping street, then – especially if you are
wearing a ‘hoodie’ – it may well be regarded as a potential weapon and you, the
bearer, as a would-be criminal.12
Origins of discourse theory and discourse analysis
Discourse theory is an ever-evolving discipline. Its roots lie in linguistics, semiotics,

philosophy, psychology, and social theory. Politics departments – especially at the
University of Essex – have played an increasing role in its more recent development.
Linguistics
Linguistics is ‘the science of language’ (Crystal, 1995: 425).13 Its original focus
was grammar and the rules of language. In the Enlightenment period, the rationalist
scholars investigated what appeared to be universal features of language as evidence
of universal rationality. Their nineteenth-century successors studied the evolution
of language and grammar. In the twentieth-century, the focus moved to psychology
and social theory. A paradigmatic shift was provided by the Course in General
Linguistics (1916) by Ferdinand de Saussure (1857–1913). He distinguished between
language (facility of speech), langue (language system) and parole (speech act). Langue
is essentially a system of signs which has two parts: the signifier (what signifies) and
the signified (what is signified or meant). He also distinguished between diachrony
(study of the historical development of language) and synchrony (study of the ‘lived’
language of the present).
The synchronic approach of concentrating study on the structural forms of
language – rather than diachronic study – became termed structuralism. Its name
derived from the belief that that individual agency (personal autonomy) is superficial:
behaviour is determined by underlying structures of society. Individuals are not the
architect of their universes but their products. This approach was extended by the
French scholars: Claude Levi-Strauss (1908–),who was particularly concerned with
the role of myths in society); cultural seminologist, Roland Barthes (1915–80),
and psychoanalyst; Jacques Lacan (1901–81) and others (Jary, 1995: 660–2).14
Anthropologist, Franz Boas (1858–1939) argued that language is essential to the

acquisition and transfer of knowledge. Later, post-structuralism was developed by
Jacques Derrida (1930–2004) and Jean Baudrillard (1929–2007) to focus on the
nature of surface knowledge. Special attention was given to the rhetorical devices of
which language was constructed. Their principal method of analysis was therefore,
deconstruction.
A paradigmatic shift was achieved by the MIT Professor of Linguistics and
subsequent political activist, Noam Chomsky (1928–). He distinguished between
competence (the individual’s knowledge of the rules of grammar) and performance (their
use of the language). Linguistics should concentrate more on linguistic competence as
part of the individual’s psychological capacity:
How [is] it that human beings, whose contacts with the world are brief and personal and
limited, are nevertheless able to know as much as they do?
(Chomsky, 1986: xxvi)15
Semiotics
Semiotics is the:
study of the properties of signs and signalling systems especially as found in all forms
of human communication.
(Crystal, 1995: 430)16
More recently, the signifiers have been extended to include consumer durables,
gadgets, gifts, music and architecture. For example, archaeological research has
shown that the great fortified gatehouses built by the Romans had little military
value: their role was to express to the native population the overarching power
of Rome.
Philosophy
The philosophers principally associated with the development of discourse theory are
the Austrian-born philosopher, Ludwig Wittgenstein (1889–1951) and the French
philosopher, Michel Foucault (1926–84).
Wittgenstein argued that language pictured the world. Furthermore, there could
never be a ‘private language’. Language was bounded by language games where each
game had its own rules and was grounded in its own ‘form of life’.
Foucault was, in part, a poststructuralist and iconoclast (a person who challenges
accepted beliefs). He re-examined topics and prevailing beliefs ‘through the
other end of the telescope’. He argued that there was no absolute truth or
objective knowledge. The maxim that knowledge is power was a myth. In practice,
power is knowledge. Elites use language to promote their interpretations of knowledge
through fabricated truth regimes. Discourse was a system of representation in the form
of epistemes and discursive formations. In particular, language is a vehicle for the
production of new knowledge and systems of classification of society. Knowledge
did not liberate individuals. Instead, it has become part of the apparatus of control by
the state. He was particularly concerned (and personally affected) by the emergent
discourse of scientific and medical knowledge by which individuals became labelled
as mad or homosexual, and panoptically complicit in their own exclusion and
confinement. Madness and sexuality were entirely social constructs. In this way,
individuals were objectified and structured by discourse. His approach has been widely
adopted by Foucauldian critical discourse analysts.
Other significant contributions were made by Canadian-born sociologist Erving
Goffman (1922–82), US sociologist, Harold Garfinkel (1917–) and Canadian media
guru, Marshall McLuhan (1911–80). Goffman studied face-to-face interaction
through the lens of social interactionism where the meanings of language were never
static or fixed but negotiated by the participants. Garfinkel was the founder of
ethnomethology which studied the everyday methods used by members of different
societies (ethnomethods) in their conduct and interaction. McLuhan contributed the
maxim that: ‘the medium is the message’ (1962).17 So ‘television is more significant
than the content of its programmes’ (O’Sullivan, 1997: 176–7).18
The schools of discourse theory and discourse analysis
The leading UK authority on discourse in Politics, David Howarth, identifies five

different theoretical schools of discourse and, therefore, discourse analysis: positivists;
realists; Marxists; critical discourse analysis; and post-structuralists (Howarth, 2000:
2–5).19
Positivists (and empiricists) see discourses as ‘frames’ made by groups to interpret
their worlds in the same way and for specific purposes. These frames legitimate (and,
therefore, justify or excuse) collective action. This is:
essentially a political definition, focusing on attempts by groups to impose their own

assumptions and values on others in order to promote their own interests.
(Burnham, 2004: 243)20
Members of this school see the task of discourse analysis as, therefore, to identify
these frames, their effectiveness and the consequences of their use.
Realists argue that objects have an existence independent of society’s conception or
perception. The objects have their own properties and causal capacity. So discourses
are objects in their own right which contribute to events and social development.
The task of discourse analysis is to identify the discourses, show how they contribute
to processes of development and to expose the ‘underlying material resources which
make discourses possible’ (Parker, 1992: 1).21
Marxists see discourses as ‘ideological systems of meaning’ which legitimate the
unequal distribution of resources and power. The discourses are economic. They
are promoted by capitalists via their political parties, state education system and news
media. The task of discourse analysis is: first, to expose the false consciousness sustained
by discourse; second, to identify the mechanisms and means adopted: and third, to
liberate those people whom it imprisons.
Critical discourse analysis is essentially similar to the Marxist conception of discourse.
However, it privileges social over economic influences. This is described in greater
detail later in the chapter where it is adopted to analyse a case study.
Post-structuralists (and Post-Marxists) regard:
social structures as inherently ambiguous, incomplete and contingent systems of

meaning … discourses constitute symbolic systems and social orders … the task
of discourse analysis is to examine their historical and political construction and
functioning.
(Howarth, 2000: 4–5)
The proponents of this view include Derrida, Foucault, and, more recently,
Howarth’s colleagues, Laclau and Mouffé.22 Laclau and Mouffé set out ten
propositions. These include the argument that a political discourse can never
dominate a discursive field to the exclusion of other discourse: it is defined by
a subordinate one. Their propositions are very sophisticated and probably too
advanced for first-time discourse analysts.
Confused by discourse theory? This brief review demonstrates one essential element of
discourse analysis: don’t think about using discourse analysis unless you are willing to
confront a very challenging literature. If your research project is merely a means to an
end outside academia, then choose another method of analysis. Go for an easy option.
But if you regard your research as a unique opportunity to develop your intellect by
tackling new concepts ranging beyond the conventional boundaries of political science, then
consider using discourse as both the theoretical framework and method of analysis. On the
practical side, an understanding of discourse can help you to compete better in the growing
communications and media sectors of the economy.
You may well feel intimidated by Laclau and Mouffé’s post-structuralist approach.
In that case, you are likely to find that critical discourse analysis provides a
better starting point for your journey into discourse analysis. It offers three great
advantages to the political scientist. First, critical discourse analysis sees its primary
role as ‘emancipatory’:
to expose the way in which language and meaning are used by the powerful to deceive
and oppress the dominated.
(Howarth, 2000: 4).
Second, critical discourse theory draws on wide-ranging authorities with whom you
are likely to be already familiar: Gramsci, Althusser, Foucault, Giddens, Habermas
and Bakhtin (1895–1975). And third, a leading school is led by Norman Fairclough,
formerly Professor of Language in Social Life at the University of Lancaster, whose
very readable, authoritative texts draw examples from Politics – especially the
discourses of Thatcherism and New Labour.
Fairclough’s critical discourse analysis
In the introduction to his major text, Language and Power (2000), Fairclough
explains that his language study is critical (CLS) because it seeks to expose
otherwise hidden connections between language, power and ideology (2000: 4).23
He differentiates this approach from others by concentrating on how the audience
interprets the text (rather than the intentions of the ‘text-producer’) drawing on their
members’ resources (MR, i.e. accumulated experience). He also includes within
the scope of discourse analysis the turn-taking of everyday conversations (which
others tend to treat as a specialist sub-field).
Critical language study adopts a concept of discourse which Fairclough defines
succinctly as: ‘language as social practice determined by social structure’ (2000: 14).
So language is integral (rather than external) to society and a socially-conditioned
process. The relevant conditions are the social conditions of production and the social
conditions of interpretation. These social conditions have three levels of organisation: the
immediate social situation; the social institution; and the level of society. Fairclough
develops three arguments about the relationships between language, discourse and
social practice. First, the discourse used is socially determined by orders of discourse -
conventions set out in social institutions. Second, orders of discourse are ideologically
shaped by power relations of class and power in capitalist society. Third, (following
Giddens’ structuration theory) discourse shapes social structures and vice versa.
So this dialectic of structures and practices achieves both social continuity and
change.
Fairclough’s critical language study develops Gramsci’s concept of hegemonic
power. Discourse is used by the state to win consent and acquiescence. However,
this non-coercive power can be in or behind discourse. Power in discourse is used

by dominant groups to control or constrain the contributions of others in terms
of what is said (contents), the social relations of the discourse (e.g. teacher and pupil)
and the subject positions they occupy. In particular, news media limit the contents
of bulletins and representation of minorities. This constraint is an example of
the hidden power of discourse. But the power behind discourse is more subtle. One
example is the promotion of an elite dialect as the national language. Standard
English became the essential qualification for advancement in UK professions
(early phonographs record Gladstone’s Worcestershire accent). Whilst ‘Estuarial’
(or ‘Mockney’) English may be the dialect of popular television, it is not used in
news bulletins – especially of serious news. Standard English remains authoritative.
It remains authoritative by mutating (as notably chronicled by Raymond Williams,
1976).24 Access to discourse is subtly restricted. Free speech is a myth. Professions
monopolise technical language and develop their own narrow jargon. Similarly,
turn-taking is dictated by social class especially in formal situations like court
proceedings.
Fairclough is particularly concerned with the relationships between discourse,
common sense and ideology. He argues that ideology prevails when it becomes
elevated to and accepted as common sense. As such, it is natural and therefore
not only acceptable but desirable. Many older people believe that the unequal
distribution of wealth and income is common sense and ‘for the best’ (Panglossian).
The meaning of words is also common sense. For example, the word ideology is
equated with totalitarianism, fascism, communism and Marxism. So political leaders will
argue that their views are, basically, common sense, whilst their opponents are entirely
ideological. Similarly, each will preface their arguments by stating that ‘the truth of
the matter is …’. Another device is to use scare quotes. For example, the popular
press undermined the scientific discourse of GM foods by calling them Frankenstein
foods.
Fairclough’s exposition of the concept and components of language as discourse
is relatively straightforward. Where his particular discourse becomes more difficult is
when it becomes operationalised as a complex system of discourse analysis. He argues
that you can’t just cite a speech by President Bush and say that this is a good example
of hegemonic discourse: formal analysis is necessary.
The conceptual framework for Fairclough’s critical discourse analysis is the
relationship between text, interactions and contexts, illustrated in Figure 19.1.
Fairclough develops a three-stage system of critical discourse analysis:
1. description: identifying the formal properties of the text

2. interpretation: identifying the relationship between the text and interaction
3. explanation: identifying the relationship between the interaction and the social context.
Description answers ten questions in three sections.

Social conditions of production
Social conditions of
interpretation Process of production
Context
Process of interpretation
Interaction
Text
Figure 19.1 Discourse as text, interaction and context

Source: (after Fairclough, 2000: 21, Figure 2.1)
BOX 19.1 Description: ten questions (Fairclough, 2000:

92–3)
Vocabulary
1. What experiential values do the words have?

1. What classification schemes are being used?
2. Are there words which are ideologically contested?
3. Is there rewording or over-wording?
4. What ideologically significant meaning relations are there between words?
2. What relational values do words have?
1. Are there euphemistic expressions?
2. Are there markedly formal or informal words?
3. What expressive values do the words have?
4. What metaphors are used?
Grammar
5. What experiential features do grammatical features have?

1. What types of process and participant predominate?
2. Is agency unclear?
3. Are processes what they seem?
4. Are nominalisations used?
5. Are sentences active or passive?
6. Are sentences positive or negative?
Continued
6. What relational value do grammatical features have?

1. What modes (declarative, grammatical question, imperative) are used?
2. Are there important features of relational modality?
3. Are the pronouns we or you used, and, if so, how?
7. What expressive values do grammatical features have?
1. Are there important features of expressive modality ?
8. How are simple sentences linked together?
1. What logical connectors are used?
2. Are complex sentences characterised by co-ordination or subordination?
3. What means are made for referring inside and outside the text?
Textual structures
9. What interactional conventions are used?

1. Are there ways in which one participant controls the turn of others?
10. What larger-scale structures does the text have?
These terms need explanation. Experiential value refers to traces of how the
text producers’ own experiences of the world – social or otherwise – are represented.
By way of example, Fairclough cites the difference between the use of the terms
‘solitary confinement’ and ‘seclusion’ by different groups of psychiatrists. The two
terms reflect two opposing ideologies. The replacement of ‘solitary confinement’
by ‘seclusion’ is an example of rewording. Relational value refers to traces of
the social relationship in the discourse. In other words, how the words used create
social relations between the participants (for example, addressing each other by title,
family name, first name or nickname). Expressive value refers to the text producer’s
view of reality (for example, ‘globalisation’ and ‘internationalisation’). Euphemisms
provide traces of expressive values, for example, ‘defence forces’.
In terms of grammatical description, Fairclough analyses the grammar of sentences
into subject (S), verb (V), object (O), complement (C) and adjunct (A). Their
combination and order express three types of process: actions (SVO, e.g. Thatcher
attacked the welfare state), events (SV, e.g. Thatcher was speaking), or attributions
(SVC, e.g. the welfare state was attacked by Thatcher). Attributions identify causes
(or allocate credit or blame to agents). In contrast, where agency is unclear,
then causes are concealed – perhaps for legal reasons. Nominalisation is used
to convert a process into a noun. In this way, meaning and attribution become
(conveniently) ambiguous. A verb requires a subject and may cite an object. For
example, using the word ‘development’ rather than the verb ‘to develop’ enables
the speaker to be much less specific about who is going to develop what, how
and when. ‘Community development’ is a wonderful example of nominalisation in
practice.
Fairclough distinguishes between modes of sentence, modality and pronouns. Modes
can be declarative ( SV), grammatical question (no O) or imperative (no S). Each treats the
subject in a different way. Using grammatical questions or imperatives underlines
the power of the subject. Modality refers to the authority of the speaker. Relational
modality is the social authority of the speaker, whereas expressive modality is concerned
with the speaker’s authority of truth or reality, and can be recognised by their
use of ‘is’, ‘may’ or ‘can’. Similarly, using ‘we’ or ‘you’ can evidence power
relations.
Aspects of Fairclough’s textual analysis are sophisticated and, at first sight,
complicated, He can perhaps be criticised for appearing to omit other key
contributions to the discourse of particular importance to the Politics researcher.
These are primarily contextual. They include the when, where, to whom (the audience)
and how (the medium), the role of the speaker or writer (the ‘text-producers’) and
the position of the discourse in preceding or contemporary events. For example,
you may consider that contextual factors like a formal speech by the British prime
minister to the UN in the wake of 9/11 may be more significant in the first instance
than whether individual sentences have objects or not.
Interpretation
Fairclough distinguishes between interpretation and explanation. Interpretation is the

process of mediation between text and social structures in which the audience draws
on the common-sense assumptions embedded in the resources (MR). Explanation
concerns: ‘the relationship of discourses to processes of struggle and power relations’
(2000: 117).
Interpretation boils down to answering three questions of the text: first, what
interpretation are the speaker and audience giving to the situation and ‘intertextual
context’? Second, what discourse type is being used? And, third, are participants
likely to interpret the context and discourse type in different ways? (And does this
affect the course of the interaction?)
Explanation
Fairclough argues that, whilst the processes of production and interpretation

reproduce members’ resources (MR), explanation seeks to locate the discourse as social
practice, to show how it is affected by social structures and to identify how these
structures are reproduced or modified. Explanation therefore has two analytical
dimensions: first, to reveal the role of discourses as elements of social struggles in

which structures are contested, sustained or re-shaped; and, second, to identify how
and which power relations determine discourses. Explanation can be achieved by
answering three questions of the text: first, what power relations at situational,
institutional and societal levels shaped the discourse? Second, what aspects of the
members’ resources used are ideological? And third, does the discourse sustain or
change previous power relations? (2000: 138)
Fairclough concludes by arguing that the researcher must draw on their own MR
to explain how the participants use theirs. The analysis must be an ‘inside job’. So,
the analyst must be reflexive and seek to develop their own MR by developing their
own theoretical understandings and experience of society and politics. The critical
discourse analyst can therefore be likened to an ‘expert forensic witness’.
Fairclough demonstrates the application of critical discourse analysis by a case
study of a radio interview of Margaret Thatcher in 1985 (2000: 143–5). You will
note that his analysis is eight times the length of the transcript! A small extract
demonstrates the level of analysis:
‘Question 1: What relational values do textual features have? Are

there inconsistencies in relational values which could indicate a
new articulation of discourse type?
1. We. Mrs Thatcher uses the pronoun we mainly in lines (11–9) and (79–81), both
inclusively and exclusively … The inclusive use (e.g., ‘ now we do enjoy a standard
of living which was undreamed of then’) is rationally significant in that it represents
her, her audience and every one else in the same boat. It assimilates the leader
‘to the people’ … (Fairclough, 2000: 148)
He also subjects New Labour to discourse analysis in a separate text (Fairclough,

2000a).25 His essential argument is that the re-naming of Labour as New Labour
signals ideological change and manipulates language to control public perception.
Great emphasis is given to ‘getting the language right’ by, for example, substituting
the ‘privatisation’ of Thatcherism to ‘public-private partnership’. (This is effectively
the opposite of Thatcherism in which radical social and economic policy was offset
by appeals to the traditional Conservative party discourse of the virtues of patriotism,
strong pound, strong defence, law and order, and the ‘3Rs’.)26 New Labour was
wholly committed to the neo-liberal global economy. It portrayed globalisation as
a ‘natural’ and ‘inevitable’ process in the face of which government was forced to
respond by ‘allowing’ greater ‘labour flexibility’, etc.
Fairclough identifies a discourse of New Labour characterised by lists – of

effectively motherhood objectives – in which the connections and means of
implementation go undisclosed. So, by speaking of, say, ‘not only economic
dynamism but social justice’, the contradictions, antagonisms and unpleasant
consequences of the duality are omitted. But he argues that New Labour portrays
its portfolio of platitudes with an overriding consensus formed from the ‘old’ clash
of (outdated) socialism and capitalism. The New Labour keywords are: we, Britain,
welfare, partnership, new, schools, people, crime, reform, deliver, promote, business, deal,
tough and young (2000a: 17). He also pinpoints how the moral discourse becomes
authoritarian when expressed by statements like: ‘if you can work, you should work’
and ‘if you can save, you have a duty to do so’ (2000a: 42). The rhetoric and
presentation of Tony Blair are also honed by the discourse. ‘Tony’ is an everyday
man who uses glottal stops, says ‘yeah’ rather than ‘yes’ and speaks of ‘Cherie and
the kids’.27 Yet he combines ‘moral righteousness with toughness’ (2000a: 148).
Fairclough ends the case study by re-asserting that:
… (a) that politics and government are social practices in which language is salient –
this is a durable feature of these social practices in comparison with others [he cites
agriculture as a case where language is less salient], (b) language is becoming more
salient within these practices.
(Fairclough, 2000a: 156).
Case Study: President George W Bush’s State of the Union Address, 2007
The annual State of the Union Address is a constitutional requirement of the

US President. It is made to a joint meeting of Congress. It effectively combines the
UK Parliament’s Queen’s Speech and the Chancellor of the Exchequer’s Budget statement.
Read the introduction, passages on Iraq and Afghanistan and the conclusion from
www.whitehouse.gov/news/releases/2005/02/20050202-11.html. Note that you may also be
able to watch the video of the speech at the same web-site.
The task is to apply discourse analysis to the section of President Bush’s speech dealing
with Iraq and the Middle East. However, before beginning the detailed analysis, you should
ponder its context. This is highly significant. You will note that:
• The State of the Union Address is a formal speech

• whilst the President occupies centre-stage, he is overlooked by Vice-President Cheney
and Democrat Nancy Pelosi, Speaker of the House of Representatives elected on
Continued
4 January as the first woman speaker and the most senior elected politician in the country.
She is ‘two heartbeats away from the Presidency’
• The President is effectively reading a prepared speech written by speech-writers and
honed to perfection by rehearsals
• the speech is being made to the Members of Congress face-to-face and, via the news
media, to the citizens of the US
• the speech will also be considered carefully by the US allies, supporters, critics and
enemies
• by tradition, the speech is made unchallenged by opposition; it is punctuated only by
applause which is triggered by pauses in the President’s delivery
• President Bush is ‘beleaguered’: the Iraq project has become the dominant political issue
in the US and, in consequence, the President’s Republican Party has lost control of the
Senate and House of Representatives
• The President’s approval ratings are at an all-time low
• the ‘separation of powers’ set out in the US constitution means that, if bills are not to
be vetoed by the President or Congress, then a new ‘bi-partisan’ approach will have to
be negotiated by the two institutions if progress is to be made on new policy on social
security, Medicare, taxation, home security and defence
• big business, the Pentagon, the ‘neocons’, and potential candidates for the presidential
election will be watching for signals of support for their own constituencies
• The President has just rejected the recommendations of the bi-partisan Iraq Study Group
to involve Syria and Iran in the resolution of Iraq
• a predecessor, President Reagan was called the ‘Great Communicator’; George W. Bush
is no great communicator
• US commentators tell us that, whatever the public’s view of any individual President, they
retain great respect for the Presidency.
A comparative content analysis of key words used in all President Bush’s State of the Union
Addresses shows the changing emphasis given to key issues over the years:
The table is illustrated in Figure 19.2
Analysing the case study
The analysis will adopt Fairclough’s three-stage system of description, interpretation and
explanation.
Table 19.1 State of the Union Addresses 2001–7
Topic 2001 2002 2003 2004 2005 2006 2007

Iraq − 2 21 24 27 16 34
Afghanistan − 13 3 5 3 2 4
Economy 6 7 13 17 14 23 8
Iran − 2 3 1 3 6 5
Oil − 1 − − − 2 9
Source: New York Times: 21/01/07
40
Number of mentions of
35
Iraq
30
keywords
25 Afghanistan
20 Economy
15 Iran
10
Oil
5
0
0 2 4 6 8
Years (2000 onwards)
Figure 19.2 President Bush’s State of The Union Addresses 2001–07: frequency of
key words
Source: Table 19.1 from New York Times: 21/01/07
Explanation
You will recall that the questions to be asked of the text are: what are the experiential,
relational and expressive values and metaphors used?
Experiential values refer to the traces of ‘text-producer’s’ experience of the natural and social
world included in the text. Aspects include the classification schemes used, whether words
are ideologically contested and over-wording used. In this case, the classificatory schema
adopted by the President is primarily to reduce the world to the US, ‘Americans’ and US
interests. Ideologically contested words are minimised in favour of the language of consensus
and learning-through-experience. This is also achieved by portraying the enemy in over-
worded, graphic terms:
… possessed by hatred and commanded by a harsh and narrow ideology. They preach
with threats, instruct with bullets and bombs, and promise paradise for the murder of
the innocent.
Continued
Note also the use of alliteration (the rhetorical style which uses the repetition of opening
consonants to achieve emphasis): ‘hatred … harsh, bullets and bombs …, promise
paradise…’. Notice how ‘our American values’ are contrasted with their ‘extremism, ideology,
and totalitarianism’. By contrast, ‘our government’ is using ‘lawful and proper use of
intelligence, etc.,’ in a list in which ‘military action’ is mentioned last, implying a ‘last resort’.
What they (the terrorists) ‘fear most is human freedom’. There is a superabundance of purple
prose:
free societies where men and women make their own choices, answer to their own
consciences and live by their hopes rather than resentments.
Relational values refer to traces of social relationships enacted via the discourse. In this
case, the social relationship sought with his immediate audience of Congressmen is that of
friendship in adversity: e.g. ‘Like many before us, we can work through our differences, and
achieve big things for the American people’ and: ‘our success in the war’. You will note
that, in the video recording, emphasis is given by the President to we. In terms of the
wider audience of the American people, President Bush is seeking to project himself as
their resolute protector. But the social imagery induced is more like the ‘strong father’ than
‘benevolent mother’ or ‘kindly grandfather’ (which President Reagan was able to evoke). But
this President is no wise old man sitting out on the porch: he is a military scout and fire-fighter.
Are there any euphemistic expressions? Well, the whole speech seems to avoid saying either
‘my disastrous invasion of Iraq, my failure to secure peace there, the worsening situation,
and 3,000 US dead’. Instead, he asserts that ‘our success in this war is often measured
by the things that do not happen’. Are there any markedly formal or informal words? The
speech is ‘US formal’ which seeks to underline the dignity and authority of the Presidency
and the seriousness and fitness-for-office of the current President.
Expressive values refer to traces of the speaker’s evaluation of reality and their ideological
significance. In this case, the President ‘congratulate[s] the Democrat majority’. He appears to
avoid terms ideologically confined to the Republican Party and uses instead terms common
to the Republicans and Democrats: ‘extending the nation’s prosperity, to spend the people’s
money wisely … to guard America against evil’. The ideological struggle is between America
and its enemies. There is no accommodation possible which does not compromise US
freedom: ‘the security of our nation is in the balance’.
What metaphors are used? There appears to be very few metaphors used. The President
avoids using the popular metaphors of baseball or American football or the ‘tough talk’ (e.g.
‘I say, bring it on’ for which he apologised later). He’s ‘talking Presidential’, using measured
words.
What experiential value do grammatical features have? The main grammatical features are
short, grammatically correct sentences. Most sentences are given in the active (rather than
passive) tense in which the subject–verb–object predominates. So does the present tense.
The subject is ‘we’ rather than ‘I’ or ‘you’. Agency is clear: terrorists. The sentences are mainly
positive and confident. There are few nominalisations.
The grammatical mode is almost wholly declarative (rather than grammatical questioning
or imperative). The relational modality is President to Congress and ‘our citizens’. The
expressive modality is characterised by the use of ‘can’, ‘will’ and ‘know’ rather than ‘may’
or ‘might’. The President is confident and certain. The main logical connectors employed
are hyphens. There are no lists. Applause is cultivated to maintain control and to enable
topics to be changed. Indeed, applause is the main interactional element between speaker
and audience. The President speaks direct to his audience rather than to the ceiling and
notes. There are no asides or quips. He is reading from the transparent video-screens at
each side.
Interpretation
There will be many interpretations of a Presidential address. The interpreters will
include: Congressmen, the television audience, the newspaper readers and the ‘expert
commentators’ who mediate between the text and the wider audience. Each group is
diverse. They will all draw upon very different members’ resources (MR). Then there will
be your own member’s resources as the Politics researcher. Knowing how the State of the
Union Address will ‘play in Peoria’ will be difficult for a British or Australian researcher, and
doubly-difficult for a researcher who is not wholly fluent in American-English. So, ideally
you should be able to identify which interpretations are likely to be given by the main
constituents to whom the President is seeking to appeal and influence. There are many.
You must identify the principal constituencies and suggest how they might interpret the
speech.
Take, for example, the US military who, despite forces of over a million men and women and
a budget of $600 billion, appear unable to stem the rising violence in Iraq. What’s in it for
them? Well, their Commander-in-Chief declares that:
on this day, at this hour, it is still within our power to shape the outcome of this battle. Let
us find the resolve, and turn events towards victory’. Victory will be achieved by sending
‘more than 20,000 additional soldiers and Marines to Iraq.
Continued
An additional 92,000 troops will be recruited, and civilians hired for rear echelon duties.
Our country is pursuing a new strategy in Iraq, and I ask you to give it a chance to work.
And I ask you to support our troops in the field, and those on their way.
This last statement evokes memories of Vietnam when President Nixon argued that it was
primarily the lack of support for US forces at home that undermined the war effort over there.
You may conclude that US soldiers based in Baghdad would interpret the speech to mean
‘more of the same’ with the options of withdrawal or ‘taking the fight to the real enemy in
Iran and Syria’ ignored. They may be disappointed or even angry that the death of so many
soldiers goes unstated.
Explanation
Fairclough’s critical discourse analysis concludes by identifying the power relations,
ideologies and effects. He is particularly concerned to identify how the audience’s members’
resources are reproduced unconsciously through their interpretation of the text. In this case,
the President achieves – or seeks to achieve – reproduction of his discourse by minimising
the differences between Republicans and Democrats, by wrapping himself in the flag of
consensual American values and by painting his enemies as wholly evil and hell-bent on
destroying our America. His discourse is therefore of integration and differentiation. American
values are being defined by the enemy. It is their ideology which is defined. He continues to
link the war in Iraq with 9/11 and seek popular support for a war of righteous retribution against
evil which threatens ‘the security of [our] nation’. So the war is essential to defend American
values and way of life. He raises the sword of ‘human freedom’ which ‘every terrorist fears
most’. The appeal is to emotions rather than intellect. But – despite the stentorian rhetoric –
the speech is comparatively uninspiring and lacks the power of John Kennedy or Churchill
to reinforce the determination of the people.
Postscript
In his State of the Union Address, the President spoke of the budget that would ‘shortly
be sent to Congress’. His budget was announced on 15 February. Despite his pledges to
balance the budget and to protect Medicare provision, the budget proposed to increase
military spending by nearly 12%. This would be achieved by reducing federal spending
on Medicare by $66 billion over five years and cutting back $12 billion from the Medicare
healthcare scheme for lowest-income citizens. However, he argued that these reductions
could be achieved by attacking costs rather than by reducing the level and volume of
services.
Conclusion:
How useful is critical discourse analysis?
The case study demonstrates the strengths and weaknesses of applying Fairclough’s
approach to critical discourse analysis. The highly-structured framework of analysis enables
you to offer a wealth of critical assessments. The framework provides a good starting
point for the ambitious researcher keen to adopt discourse analysis. However, there
are a number of weaknesses which arise in part from its roots in linguistics. Too
little emphasis is given to the context in which the discourse is produced. It is better
suited to the analysis of interactive discourse than the speeches and monologues which
characterise politics. It emphasises the interpretations given by the audience rather than
the intention of the ‘text-producer’. Identifying the interpretations given by the audience is
highly speculative. This requires understanding of their members’ resources. Furthermore,
the audience is likely to consist of a number of constituencies with very different
resources
You may therefore wish to modify the approach by including other methods, such as:
• randomly selecting texts from a sample frame

• identifying the contexts
• considering the intentions of the text-producer by asking: what is the text seeking to
achieve?
• using content analysis to calculate the actual frequency of key words like ‘will’ and ‘may’
and sentence length
• critically using interviews, focus groups and questionnaire surveys to identify both the
members’ resources drawn upon by the audience to interpret the text, and how they are
fixed or changed by the text.
Narrative analysis
Narrative analysis is a sub-type of discourse analysis concerned with the analysis of

narratives. A narrative is essentially an account of a past event. It can be verbal,
written, visual or aural. Often discourse is a combination of media. The narrators
mediate between the subjects and the audience. They can be a politician, journalist,
commentator, teacher or other authoritative narrator. They will seek to locate the
narrative in the audience’s members’ resources and to strengthen or modify them.

Modern politics has been likened to a battle of narratives in which the victor is
the most successful in providing the most popular, enduring and damning narrative
of their opponents. For example, the Labour leader, Harold Wilson succeeded in
devastating the Conservative government in 1964 by repeating the narrative of
‘thirteen wasted years of Tory misrule’.
Various definitions and explanations have been given of narrative in the social
sciences. They include (in chronological order):
• any form of communication (Barthes, 1966)28

• the main mode of human knowledge (Bruner, 1986)29
• a socially-symbolic act in the double sense that it (a) takes on a meaning only in a social
context, and (b) plays a role in the construction of that context as a site of meaning within
which social actors are implicated (Mumby, 1993)30
• stories that take place in time (Berger, 1997);31
• a story with a beginning, middle and end that reveals someone’s experiences (Manning
and Cullum-Swan, 1998).32
Origins
The origins of narrative analysis include linguistics, social and literary theory, history
and discourse theory. The classical theory underpinning narrative analysis is the
Morphology of Fairy Tales (Propp, 1927). Vladimir Propp (1985–1970) examined
the main Russian fairy tales. He found that the underlying narrative structure was
remarkably common. Each fairy tale could be deconstructed into 5 categories of
elements and 31 narratemes. The five categories were: the dramatis personae (‘cast’);
the conjunctive elements of misfortune and good fortune; motivations; forms of
appearance of the cast (e.g. the flying fairy godmother); and, lastly, attributive
elements (e.g. Jack’s beanstalk, the old lady’s shoe). The narratemes began with
α, the initial situation when the hero and members of family are introduced,
followed by " when one of the members of the family goes missing. The
stories all end with the last narrateme, ', when the hero marries and ascends the
throne. Western fairy tales also followed this pattern. Propp therefore argued that
fairy tales constituted pre-knowledge – comparable to Fairclough’s earliest members’
resource – by which people became conditioned to believe that their own lives
would follow a similar fairy tale course. These expectations enabled them to
accept that early misfortune was inevitable and part of the essential tapestry of
life that – ultimately – would end in happiness. Parents and grandparents become
reconditioned and reproduced by their role as household narrators. The concept of
pre-knowledge can explain how readily the British public came to see Princess
Diana as the ‘fairy princess’ and Camilla Parker-Bowles as the ‘wicked witch’
and ‘step-mother’ to the young princes. Indeed, the western antipathy to step-
mothers and mothers-in-law can be attributed to the bad press they receive in fairy
tales.
Whilst Propp’s work has been hugely influential, he has been criticised for
overlooking the nuances of context and mood. He can also be criticised for
overlooking the competing pre-conditioning role of religion and Biblical narrative
on western society. The narrative of Christ is the antithesis of the fairy tale. Through
this combination of fairy tales, religious tracts and local folk stories (King Arthur,
etc.), the structures of stories become both historically embedded and conservative
(retrospective). The participant or observer is conditioned to believe that there is
little opportunity for – or point in – agency. This historicist model can be likened
to Heidigger’s ‘horizons of meaning’.
History – especially national history – is grand narrative. History shapes national
identity. Identity shapes and reinforces the boundaries of the state. History is shaped
by dominant power: ‘it is the victors who write history’ (Carr, 1990).33 So whose
history prevails is critical. Hence the continuing argument in the UK about the
contents of the unified, national curriculum for history in which – as one insider
told me – the central question was: ‘how much can we afford to tell the working
class about their past?’ Some concentration camp survivors argue that narratives of
the Holocaust were suppressed in Western Europe during the Cold War to facilitate
reconciliation and military coalitions against the ‘Soviet threat’. In the same way, the
historical narrative of Japan was re-written by the British government in the 1980s
to air-brush out Japanese wartime atrocities and therefore to reduce opposition to
Japanese inward investment.
Narrative analysis
Definitions of narrative analysis include:
• the analysis of act, scene, agent, agency and purpose (Burke, 1966)34
• how respondents in interviews impose order on the flow of experience to make sense of
events and actions in their lives (Kohler Reissman, 1993)35
• the analysis of how stories mark out identities, mark out differences, differences define
‘the other’, and the other helps structure the moral life of culture, group and individual
(Plummer, 1995)36
• the analysis of the production of narratives: the process of association, building and ‘the
and, and and’ connections between actions and events and negotiating them with readers
(Czarniawska, 1998)37
• the study of lives from the narrator’s experience as a shared production with social
scientists (Social Anthropology, Feminist Research, McRobbie, 1982).38
Narrative analysis essentially asks:
• Who is ‘writing’ the story?

• Who is telling the story? How?
• Who is the ‘target audience’?
• What is the story trying to achieve?
• What are its effects?
The approaches to narrative analysis include:
• identifying the formal structures (properties) of stories in terms of the plot, setting,
characterisation, and temporal ordering
• identifying the social roles of stories in terms of:
• the ways they are reported
• the ways they are read
• how they change
• their role in the political process
• stories as speech acts.
Speech Act Theory is central to narrative analysis in explaining how words ‘work’. The
underlying concept is that words ‘do action’. The speech acts can be distinguished
and classified as (Austin, 1962):39
• locutionary speech acts: constative, descriptive statements, e.g. ‘it’s hot today’
• illocutionary speech acts that achieve action ‘in’ saying (directly): e.g. ‘in saying it I was
warning him’
• perlocutionary speech acts that ‘do something’ indirectly. They use the language of a
constative statement: e.g. ‘students who do not attend classes are more likely to fail’.
Often these categories of speech act are linked. Not all speech acts are felicitous (suc-
cessful): some are infelicitous. Narratives can also compete for support. For example,
consider the famous case of the Congressional hearings in 1991 when President
Bush’s nominee for the US Supreme Court was contested. The nominee was Judge
Clarence Thomas, a 43-year old, conservative, African-American from Georgia.
Critics argued that, whilst Thomas would maintain the racial composition of the
Supreme Court, he would replace a black liberal judge by a conservative who would
oppose reform in decisions involving Affirmative Action and abortion. The hearing
considered the objection by Anita Hill (a former colleague) of sexual harassment.
Both were high-achieving, black Republicans from poor backgrounds whose
successful careers were assisted by early ‘affirmative action’ programmes. The hearing
was held in the context of the imminent presidential elections in which Republicans
were arguing that affirmative action was discriminatory. The full transcript of
the hearing is available on http://chnm.gmu.edu/courses/122/hill/hillframe.htm.
When you read them, you will note how both Thomas and Hill adopted
‘narrative strategies’. She sought to argue on the basis of the ‘facts’ (detailed
descriptions of diarised events). He presented himself as a ‘victim’ citing his deprived
background and the ‘high-tech lynching for uppity Blacks’ by the liberal news media.
Furthermore, he criticised his sister, Mae as a ‘welfare queen’ and portrayed Hill as
a ‘black lady’. Thomas’ narrative prevailed by appealing to embedded US narratives
of the self-made man and ‘victimhood’.40 You may note how the contemporary
narrative metaphor of ‘black’ - which had recently replaced the more pejorative
‘negro’ - has now been superseded by ‘African-American’ in the same period in
which ‘Red Indian’ has been superseded by ‘Native American’ (cf., Canadian ‘First
Nations’).
Narrative constituents
Narrative analysts adopt a scheme of terms to classify the components of narrative
discourse. They are:
• story : the raw, temporarily sequenced or causal narrative of life

• plot : emerges from unexpected twists in the narrative that draw attention to differences
from the conventional story
• narrative of the self : the story or stories by which self-identity is reflexively understood
by the individual and others
• reflexive projects of the self : the process by which self-identity is constituted by the
reflexive ordering of self-narratives (Giddens, 1991)
• discursive register : the style of narrative, especially stylistic narratives ‘borrowed’ from
other narratives, e.g. ‘ball-park figure’ from US sport; and ‘take out’ from US military texts
• epiphany : a crucial turning-point (Damascene conversion) which stimulates a new, radical
consciousness
• text : a finite, structured whole composed of language signs
• actors: passive subjects, objects
• agents: active subjects who affect others by their actions
• events: transitions from one state to another
• fabula: the series of logically and chronologically related events that are caused by or
experienced by actors or agents
• kernels: narrative moments that give rise to cross-roads or branches in the direction taken
by events. They are nodes or hinges in the structure which form cross-roads or branching
points which force a movement into one or more possible paths
• satellites: relatively minor events which embellish the kernels, add detail or ‘flesh
them out’
The narrative-of-the-self is critical. Narrative analysts argue that ‘the self is

nothing but stories’. And it is also unstable. The self and its narratives are heavily
contextualised. So you add, change, edit and embellish stories of yourself. In
particular, you avoid recounting the tedium of boring, everyday life in favour of tales
of rare, interesting moments. So the self is in a constant state of flux. To paraphrase
Shakespeare:
all the world’s a stage in which you act the part dictated by the combination of the set
and your interests at the time.
(This view contradicts the argument of Sartre and others that, despite peripheral
flux, people have solid, stable, cores of authentic self.)
Case Study
Consider the example of the famous interview of Princess Diana by Martin Bashir
broadcast on BBC Panorama programme in 1995. You can watch the video-recording
and inspect the transcript by inserting ‘Diana, Bashir’ in the search box on the web-site
www.bbc.co.uk
You can identify the narrative of the self employed by the Princess:
• innocent foal/Bambi eyes/ fairy tale princess

• brave, bright-eyed innocent (virgin)
• let down by advisers
• trapped in media spotlight
• frequent use of ‘I’
• inference of brutal, uncaring husband, isolated, alone
• epithanies of engagement, marriage, Alice Springs walkabout
• other people are disappointing
• her support for other victims: drug addicts, battered people
• her ill-health, depression, self-disgust
• abused but loyal wife and mother
• relations with Charles are damned with ‘false praise’
• intuitive knowledge of Charles’ relationship
• un-loved
• tactics adopted by Charles’ ‘friends’ to portray her as ill, unbalanced
• embarrassment, isolation and desolation

• ‘Three of us in the marriage’
• ‘The [Royal] Family’
• the news media – abusive relations
• denial of self-pity
• ambassadorial aspirations … to do good
• Mother Theresa
• Diana as victim
• I will survive … the fairy tale is not yet ended
You can also identify the many discursive registers adopted:
• psychological language register

• confessional register
• packed with language of therapy
• marketing language register
• romantic love register (seduced by Hewitt)
Her victim strategy was successful in her narrative struggle with Prince Charles for the ‘heart
of the nation’. Her narrative was defined and reinforced by Prince Charles’ own TV interview
later (with Jonathan Dimbleby) in which he admitted his adultery.
1. Working in small groups, discuss your understanding or misunderstandings of

discourse, discourse analysis, narrative and narrative analysis.
2. Critically assess Fairclough’s process of discourse analysis.
3. Select and obtain transcripts of two or more of the State of the Union Addresses
by President George W. Bush. Complete a comparative discourse analysis of the
transcripts. What conclusions can you draw?
4. Contrast and compare the narratives of the most recent general election man-
ifestos of the British Conservative, Liberal-Democrat and Labour parties and
US Republic and Democrat presidential elections. How do they compare to the
narratives of French presidential rivals?
FURTHER READING
Howarth, D. (2000) Discourse. Buckingham: Open University Press. pp. 166. This
excellent textbook is ‘essential reading’ for students contemplating the adoption
of discourse analysis in their research. The author critically examines and
compares the various concepts of discourse.
Fairclough, N. (2000) Language and Power. 2nd edn. Harlow: Longman. pp. 224.
This text is also ‘essential reading’ for students contemplating discourse
analysis and wishing to learn how critical discourse analysis can be applied.
Fairclough, N. (2000a) New Labour: New Language? London: Routledge. pp. 178.
This is a very readable and wry demonstration of the application of critical
discourse analysis. It is less technical than his previous text (Fairclough, 2000).
It might have been strengthened by a comparison with the discourse of the
earlier, new Labour (of the ‘white heat of technology’ and economic
restructuring) successfully forged by Harold Wilson in the 1960s (which proved
equally disappointing in the long term).
Austin, J. L. (1962) How to Do Things with Words. Oxford: Oxford University Press.
Williams, R. (1988/ 1976) Keywords. London: Fontana. Raymond Williams
(1921–90) describes and analyses how the meanings of ‘keywords’ like culture,
class, community and democracy have evolved through adaptation and
struggle. He defines keywords as: ‘significant, binding words in certain activities
and their interpretation; they are significant, indicative words in certain forms of
thought. Certain uses [bind] together certain ways of seeing culture and
society …’ (1988: 15).
Notes
1 Curt, Beryl C. (1994) Textuality and Tectonics; Troubling Social and Psychological Science.
Buckingham: Open University Press, cited by Bergman, M. at Essex Summer School, 2003.
2 Gallie, W.B. (1956a), Essentially Contested Concepts. In Proceedings of the Aristotelian Society,
56: 67–198.
3 Churchill’s quote is widely attributed to his (post-war) assessment of China. However, he first
coined it in 1939 about the Soviet Union in the wake of the Germany-Soviet Union
Non-Aggression Pact.
4 Fairclough, N. (1989) Language and Power. London: Longman.
5 Howarth, D. (1995) Discourse Theory. In Marsh, D. and Stoker, G. (eds.) (1995) Theory and
Methods in Political Science. Basingstoke: Macmillan.
6 Carver, T. (2002) Discourse analysis and the ‘linguistic turn’. In European Political Science,
Autumn 2002, cited by Bergman, M. at Essex Summer School, 2003.
7 Hajer, M. (1995) The Politics of Environmental Discourse. Oxford: Oxford University Press, cited
by Bergman, M. at Essex Summer School, 2003.
8 Stubbs, M. (1983) Discourse Analysis. Oxford: Blackwell, cited by Bergman, M. at Essex Summer
School, 2003.
9 Howarth, D. (1995) Discourse Theory. In Marsh, D. and Stoker, G. (eds.) (1995) Theory and
Methods in Political Science. Basingstoke: Macmillan, cited by Bergman, M. at Essex Summer
School, 2003.
10 Squire, C. (1995) Discourse Analytical Psychology. In Wilkinson, S. and Kitzinger, C. (eds.) (1995).
Feminism and Discourse: Psychological Perspectives. London: Sage, cited by Bergman, M. at
Essex Summer School, 2003.
11 Carver, T. (2002) Discourse analysis and the ‘linguistic turn’. In European Political Science,
Autumn 2002, cited by Bergman, M. at Essex Summer School, 2003.
12 A ‘hoodie’ is the slang term popularly given in the UK to the hoods of jackets worn over the head
by young men. It is interpreted both as a means of concealing their identity from CCTV cameras
and as a self-proclaimed sign of the wearer’s separate and alienated identity.
13 Crystal, D. (1995) The Cambridge Encyclopaedia of Language. Cambridge: Cambridge University
Press.
14 Jary, D. and Jary, J. (1995) Sociology. London: HarperCollins.
15 Chomsky, N. (1986) Knowledge of Language. New York: Praeger, cited by Crystal, D. (1995) The
Cambridge Encyclopaedia of Language. Cambridge: Cambridge University Press. p. 409.
16 Crystal, D. (1995) The Cambridge Encyclopaedia of Language, Cambridge, Cambridge University
Press.
17 McLuhan, M. (1962) The Gutenberg Galaxy. London: Routledge and Kegan Paul.
18 O’Sullivan, T., Hartley, J., Saunders, D., Montgomery, M. and Fiske, J. (1997) Key Concepts in
Communications and Cultural Studies. London: Routledge. pp. 176–7.
19 Howarth, D. (2000) Discourse. Buckingham: Open University Press. pp. 2–5.
20 Burnham, P., Gilland, K., Grant, W. and Layton-Henry, Z. (2004) Research Methods in Politics.
Basingstoke: Palgrave Macmillan.
21 Parker, I. (1992) Discourse Dynamics. London: Routledge, cited by Howarth, D. (2000) Discourse,
Buckingham, Open University Press.
22 Laclau, E. and Mouffé, C. (1985) Hegemony and Socialist Strategy. London: Verso. Laclau, E. and
Mouffé, C. (1987) Post-Marxism without Apologies, New Left Review, 166: 79–106.
23 Fairclough, N. (2000) Language and Power. 2nd edn Harlow: Longman.
24 Williams, R. (1976) Keywords. London: Fontana.
25 Fairclough, N. (2000a) New Labour: New language? London: Routledge.
26 The ‘3Rs’ is popular shorthand for ‘reading, rewriting and arithmetic’ widely regarded by older
people as the core curriculum of education which had been marginalised by trendy, socialist,
schools reform which had allegedly ‘left our children ill-qualified for the world of work’.
27 A ‘glottal stop’ occurs when air from the larynx suppresses the ‘t’ in words. In its extreme form,
‘bottle’ becomes ‘bo’ol’. A milder form is very popular with young middle-class people who are
trying to play down their accents. Blair suppresses his consonants in interviews.
28 Barthes, R. (1966) Introduction to the Structural Analysis of the Narrative. Birmingham: University
of Birmingham, cited by Bergman, M. at Essex Summer School, 2003.
29 Bruner, J.S. (1986) Actual Minds, Possible Words. Harvard, Mass.: Harvard University Press, cited
by Bergman, M. at Essex Summer School, 2003.
30 Mumby, D.K. (1993) Narrative and Social Control. Newbury Park, Calif.: Sage, cited by Bergman,
M. at Essex Summer School, 2003.
31 Berger, A.A. (1997) Narratives in Popular Culture, Media and Everyday Life. Thousand Oaks:
Sage, cited by Bergman, M. at Essex Summer School, 2003.
32 Manning, P. K., and Cullum-Swan, B. (1998). Narrative, content, and semiotic analysis. In Denzin,
N.K. and Lincoln, Y.S. (eds.), Collecting and Interpreting Qualitative Materials. pp. 246–73,
Thousand Oaks, CA: Sage Publications, cited by Bergman, M. at Essex Summer School, 2003.
33 Carr, E.H. (1990/1961) What is History? London: Penguin.
34 Burke, K. (1966) Language as Symbolic Action. Berkeley: University of California Press, cited by
Bergman, M. at Essex Summer School, 2003.
35 Kohler Reissman, C. (1993) Narrative Analysis. London: Sage, cited by Bergman, M. at Essex
Summer School, 2003.
36 Plummer, K. (1995) Telling Sexual Stories: Power, Change and Social Worlds. London: Routledge,
cited by Bergman, M. at Essex Summer School, 2003.
37 Czarniawska, B. (1998) A Narrative Approach to Organization Studies. London: Sage
Publications, cited by Bergman, M. at Essex Summer School, 2003.
38 McRobbie, A. (1982) The Politics of Feminist Research: Between Talk, Text and Action. In Feminist
Review, 12, pp. 46–58, cited by Bergman, M. at Essex Summer School, 2003.
39 Austin, J. L. (1962) How to Do Things with Words. Oxford: Oxford University Press.
40 Lubiano, W. (1992) Black Ladies, Welfare Queens and State Minstrels. In Morrison, T. (ed.)
Race-ing Justice, En-Gendering Power: Essays on Anita Hill, Clarence Thomas and Social
Reality. New York: Pantheon. Compare with Carver, T.(1997) Identity and Narrative in Prime-Time
Politics: The Hill-Thomas Hearings. In Carver, T. and Hyvarinen, M. (eds.) Interpreting the Political:
New Methodologies. London: Routledge.
Part V
Communicating Research
Chapter 20
Writing-up
‘… and in the end was the beginning’

(Chapter XIII, The Last Time)
1. To consider when to begin writing up your research.

2. To identify essential contents.
3. To learn how to communicate your research so that it stands out favourably
from other similar works.
4. To consider how best to maintain the interest of your readers.
Introduction
Paradoxically, most research methods textbooks – and this is no exception –

end by reviewing the process of writing the research report. This is termed –
ungrammatically – writing-up. This implies that writing-up is the last part of the
research process. This was true many years ago when academic research was divided
into three separate stages of reading-up, fieldwork and writing-up. But, nowadays,
writing-up is the first and last stages of the research process:
Report-writing is not, or should not be, a frantic activity carried out at the end of the
project. It is a process of varied stages all of which need to be recorded at the time
they are completed. Your first drafts will certainly need to be revised and in some cases
completely re-written but the foundations of the report should have been established at
the planning stage.
(Bell, 1997:152)1
Writing-up research
Clegg memorably describes writing-up as the response to four questions

(1998: 141):2
Why?
How?
What?
So what?
Where ‘so what’ means: why is this research really so important?

However, I believe that there are two additional, more significant questions, to
be asked at the outset:
To whom? That is, who or what is the readership?

Where? That is, in what publication? What are their rules?
Readership
The readership is all important. This will determine the language used. The reader’s
prior knowledge or experience will – or should – also determine how concise or
expansive the theoretical and conceptual exposition will need to be and, critically,
how the conclusions are expressed. Conclusions must be relevant to the reader.
Remember that your initial readers are likely to be your external examiners or
the referees for journal articles. They have been selected because they are leading
authorities in your research field.
Readership will also affect the style of writing-up. The discipline of Poli-
tics contests objectivity. There are few facts but many interpretations. Political
researchers are interpreters. So the clinical style of the natural scientist is rarely
adopted. Instead, Politics researchers admit their own subjectivity, communi-
cate their interest in the research topics and conclude with almost jaw-jutting
assertions. Most researchers seek, especially in case study research, to tell a
good story. However, the danger in this enthusiastic, assertive style of writing-
up is that the empiric experience – the facts such as they are – can become
so inseparable from the interpretation as to appear highly selective or fabri-
cated. Similarly, the great emphasis given to theoretical exposition can lead to
springs of conclusions being drawn from a few swallows of evidence. A good
research report should lead your readers stage-by-stage and paragraph-by-paragraph
through the research process and enable them to fully share (or reject) your
conclusions.
Writing-up 311
Publication rules
The readership – or their publisher – will also have their own rules to maintain
comparability and scholarship. They may seem anachronistic or petty, e.g. margin
size and line spacing. But they cannot be ignored – especially with regard to
word length – if your dissertation is to be approved or your research published.
Remember, words count: illustrations, tables and graphs rarely do so. When
you are preparing an MA or PhD dissertation, check the format required by
the university regulations. These will include printing margins. In many cases,
the double-spaced, single-sided pages of ink-and-parchment days have been
superseded by more recent formats which include 1.5 line spacing and verso
(printing on both sides). These new formats reduce the volume and weight of
your research report and, arguably, make your dissertation easier to read. But
note that regulations allowing double-sided submissions set specific additional
requirements for gutter widths. You may be tempted to exceed the maximum
word limit on the grounds that your department or university won’t count them.
That’s correct. However, departments receiving a large number of dissertations
of 20,000 words length know that these should be about 95 double-spaced
pages. So they will recognise when the word limit is likely to have been
exceeded. New regulations also require that an electronic copy is also submitted
for interrogation by plagiarism software. The software will automatically calculate
the word length.
Structure
An important issue is the degree of structure (chapters, sub-headings, paragraphs) that

you include in your writing-up. Following a rigid structure can help to order and
organise material. But a structure of paragraph headings and sub-headings – especially
where generic paragraph numbers and insets are used – can compartmentalise
your material and arrest the flow and development of argument. Similarly, the
incorporation of extensive tabulated data and statistical tests can cut across what
you are trying to say. You can avoid this by using graphs and diagrams to simplify
complex data (which can be annexed as technical appendices). Similarly, you can
give ancillary evidence or explanation without compromising the continuity of
the text by using extensive footnotes. Similarly, many researchers prefer to give
full citation of texts in footnotes rather than using the Harvard system. Where
footnotes (or endnotes) are used, they should be annotated at the end of the sentence
and after the full stop. Do not use the Latinism op. cit. in the footnote to refer
to a previously cited reference: repeat it in full. (Regrettably, the software used
by Sage and other publishers to scan texts for publication precludes the use of
footnotes.)
When to write-up
As already noted, writing-up has traditionally been regarded as the last stage of the
research process. However, the practice of leaving it all until the end is neither
encouraged nor accepted. Instead, you will be expected to ‘write-up-as-you-
proceed’ in parallel with the research process. This is particularly important in the
initial phases of contextualisation, theoretical and literature review, and hypothesis
setting. Indeed, you will begin the research process by writing your outline and
detailed dissertation proposals which will provide the foundation for your first
chapter of the research report.
Writing is an exploratory and creative process. Ideas can emerge in this process
which might have eluded mere abstract contemplation. Writing-up-as-you-go provides
a substantive record of progress for circulation to, and discussion with, supervisors
and colleagues and as back-up in the event of some disaster befalling you. Most
sponsors demand this parallel writing-up as a basis for monitoring – and controlling –
the research process and budget. Most supervisors require their supervisees to submit
draft chapters before any meeting as part of the teacher-student ‘supervision contract’
so that there is a substantive document to discuss.
You will have already received expert advice on essay-writing in your earlier
module guides. Most of that advice applies to research reports and need not be
re-stated. Whilst there is no right way to write-up research in politics, the following
conventions are recommended:
• write generally in the third person; e.g. ‘it was considered that …’
• however, you can use the pronoun ‘I’ to emphasise your own position – especially
where you disagree with accepted theory: but use this sparingly so that it retains its
impact
• avoid writing ‘one’: use ‘you’ to share the text with the reader, e.g. ‘as you have seen’, etc.
• avoid gender problems of ‘his/her’ by referring to subjects as ‘they’ if possible
• using the present tense to describe historical events may be a good literary device to
create excitement: however, it can become very confusing in academic reports. Instead,
use the past tense to describe the methodology and field-work; reserve the present tense
for your analysis and conclusions
• qualify your assertions – but not excessively
• always state your assumptions clearly
• strike an appropriate balance between the coverage and erudition given to theoretical
discussion, methodology, data, interpretations and conclusions
• don’t write the data speaks for itself : you must interpret the data
• try to avoid excessive and spurious accuracy in the text, e.g. don’t say ‘48.73%’,
write ‘nearly half’: where you think that precise figures are helpful, give them in
footnotes
Writing-up 313
• follow normal grammatical conventions

• avoid excessive acronyms: after a time, the reader may forget what they stand for and their
significance. Avoid inventing acronyms which are the same or similar to well-established
ones, e.g. don’t label independent research analysis as IRA
• seek to maintain pace and interest
• ensure that your report is read by another person before you submit it: we are all very poor
at spotting our own errors. If English is not your first language, then seek the assistance
of a proof-reader
• be positive and up-beat: if you don’t sound confident of your research, then your readers
are even less likely to have confidence in your report.
Structure and contents
A suggested structure for research reports is (after Denscombe) threefold:3
BOX 20.1 Research report contents

I
Title:
Short, interest-grabbing, classifiable. A two-part title can be useful, consisting of a snappy,

headline question and brief technical description (which can be readily identified by a search
machine), e.g.:
Left-right position matters, but does social class?

An investigation into the outcome of 1992 British general election
Note that capital letters are now much less used in titles than previously
Abstract:
This is a synopsis of the research of 200–300 words. The abstract is a statement of what
you have done – not of your starting intentions.
Dedications:
UK academics generally find dedications somewhat embarrassing especially when they

adopt the emotional style and length of Hollywood Oscar-winners’ acceptance speeches. If
you believe that the sacrifices made by your family and friends should be made public, then
include them in the copies which you have made after your degree has been awarded.
(Continued)
Preface
A personal statement explaining why you, as the researcher, chose the topic, your
starting biases and prejudices, and what the research means to you, i.e. the reflexive
relations between you and your social world. Often students speak of carrying out research
which enables them to shed fresh light on specific experiences and to make sense of
their past.
Acknowledgements
This provides thanks to academic supervisors, sponsors, respondents and research

gatekeepers. The real purpose of the acknowledgements is to show off the many,
many people who you have consulted or interviewed. You will normally end your
acknowledgements by stating: ‘However, any errors are entirely my own’.
Declaration
A statement to the effect that: ‘The dissertation is entirely my own, original work with the
exception of those academic and other sources which are separately identified in the text,
fully attributed in the footnotes and listed in the bibliography’.
In many scientific research papers, a statement of conflicts of interest is normally given
here – especially in medical research. Such admissions or declarations would be a useful
addition to Politics research papers.
Contents
List of Figures
List of Tables
Acronyms and Abbreviations
II
Chapter One
Introduction
The context, the aims of the research, research question and hypothesis, overview and
(chapter by chapter) contents.
Chapter Two
Theory and literature review
Critical review of theory, contemporary research and literature review leading to the
identification of the gap in the literature.
Writing-up 315
Chapter Three
Method of data collection
• concept/indicator link
• what methods were used
• why
• when did the research take place
• where
• how was access to the data and subjects obtained
• who was involved (population, sample, case studies)
• how many were involved
• how were they selected
• why the procedures were chosen
• limitations of methods employed
Chapter Four
Data: e.g. the case study
Chapter Five
Analysis and Discussion
Chapter Six
Conclusions
Has the hypothesis been confirmed or not? What contributions to theory has the research
made? What are the implications and recommendations for follow-on research (and
grants)? Provide a retrospective, critical evaluation of your research and proposals for its
improvement. This shows that you are fully aware of the limitations and know how these
might be overcome.
III
Appendices
• organised tabulated data

• questionnaires
• transcripts
• technical specification
• statistical methods
References and Bibliography

Index
Bell provides a very good check-list for reviewing the final draft report
(1997: 162):4
• Is the meaning clear?

• Is the report well written? (tenses, grammar, spelling, punctuation)
• Is the referencing well done?
• Does the abstract really give the reader a clear idea of what is in the report?
• Does the title indicate the nature of the study?
• Are the objectives of the study stated clearly?
• Are the objectives fulfilled?
• If hypotheses were postulated, are they proved or not proved?
• Has a sufficient amount of relevant literature been studied?
• Does the literature review provide an indication of the state of knowledge in the subject?
Is your topic placed in the context of the area of study as such?
• Are all the terms clearly defined?
• Are the selected methods of data collection accurately described? Are they suitable for
the task? Why were they chosen?
• Are any limitations of the study clearly presented?
• Have any statistical techniques been used? If so, are they appropriate?
• Is the data analysed and interpreted or merely described?
• Are the results clearly presented? Are tables, diagrams and figures well drawn?
• Are conclusions based on evidence? Have any claims been made that cannot be
substantiated?
• Is there any evidence of bias? Any emotive terms or intemperate language?
• Is the data likely to be reliable? Would another researcher get the same results?
• Are recommendations feasible?
• Are there any unnecessary items in the appendices?
• Would you give the report a good grade of you were the examiner? If not, an overhaul is
necessary.
Re-writing
You are likely to find that your chapters are too long. Trimming or editing is unlikely
to be sufficient. You must expect to completely rewrite every chapter at least once
and plan your work programme accordingly.
Word distribution
There are no rules for the breakdown of words to individual chapters. However,
following the advisory model distribution given below should ensure that proper
Writing-up 317
coverage is given to the theory and literature review and to your analysis and
discussion of your data.
Chapter 1 Introduction 5%
Chapter 2 Theory and Literature Review 30%
Chapter 3 Method of Data Collection 5%
Chapter 4 The Data, e.g., Case Study 30%
Chapter 5 Analysis and Discussion 20%
Chapter 6 Conclusions and Implications 10%
Avoid the temptation to reduce the analysis and discussion. Adequate coverage
is essential if your conclusions are to be shared.
1. Select three research reports from approved journals on the same topics, e.g.
election turnout. Contrast and compare these. Which do you regard as being
most successful in attracting your interest? Why? What improvements would
you suggest to the other two?
2. Select a dissertation previously approved by your department for BA, MA or
PhD purposes appropriate to your own degree. Critically evaluate the writing-up.
What improvements would you make? Give specific examples.
3. Prepare a programme for your research project on a week-by-week basis. Integrate
your proposed supervision meetings and writing, and re-writing of chapters.
FURTHER READING
Similar dissertations by former students.

Bell, J. (1997) Doing your own Research Project; A Guide for First-Time
Researchers in Education and Social Sciences. Buckingham: Open University
Press. pp. 244. Judith Bell gives very good advice on the research process.
Chapter 11, pp. 127–50, Interpretation and Presentation of the Evidence is
especially valuable as is Chapter 12, pp. 151–64, Writing the Report.
Cambridge: Cambridge University Press. pp. 200. This relatively short textbook
focuses on the application of statistical techniques. However, Chapter 16 ‘… In
the last analysis’, pp. 141–52 provides a very good and amusing account of the
writing-up process.
Denscombe, M. (1998) The Good Research Guide for Small-Scale Social

Research Projects. Buckingham: Open University Press. pp. 247. This
authoritative textbook provides complementary advice on writing-up –
Chapter 12, pp. 225–37 – and an excellent checklist on p. 238.
Notes
1 Bell, J. (1997) Doing your own Research Project; A Guide for First-Time Researchers in Education
and Social Sciences. Buckingham: Open University Press.
2 Clegg, F. (1998) Simple Statistics: A Course Book for the Social Sciences. Cambridge: Cambridge
University Press.
3 Denscombe, M. (1998) The Good Research Guide for Small-Scale Social Research Projects.
Buckingham: Open University Press.
4 Bell, J. (1997) Doing your own Research Project; A Guide for First-Time Researchers in Education
and Social Sciences. Buckingham: Open University Press.
Glossary of Terms
α data: new data generated in the course of the research.

β data: data abstracted from other sources.
Abscissa: the horizontal or x-axis on a graph. (Vertical or y-axis is termed the Ordinate.)
Abstract: a synopsis of 200–300 words which follows the title page of the research report.
The abstract is a summary of the research done – not of intent. An abstract should clearly identify
the research topic, research question, main theoretical framework, methodology, conclusions
and assess its contribution to the topic and research field.
Access: means by which the subject is approached and contacted.
Acknowledgements: statement in the research report which names and thanks the
academic supervisors, sponsors, respondents and research gatekeepers who helped the
researcher to complete their research project.
Action research (method): (generally) obtaining data by direct participation by the
researcher in a group of actors experiencing particular conditions in order to understand better
the institutional and other barriers they face in the struggle towards transformation.
Analysis: an objective-seeking, rational process which essentially involves breaking down –
deconstructing – the whole into its constituent parts to identify their relationships and structure.
Anecdotal evidence: evidence in story form based on unique, personal experience. It is
often used by quantitative researchers as a pejorative term to describe qualitative researchers’
evidence.
Antinaturalism: belief that using methods developed in the natural sciences is inappropriate
to social science research, i.e. opposed to Positivism.
Attitude: learned and persisting positive or negative evaluation or predisposition to people,
groups, policies, etc.
Attribute: qualitative variable, e.g. gender.
Baseline or Benchmark: starting level of dependent variable before the independent variable
is changed.
Behaviouralism: specific to Politics: advocacy of research based on empirical observation of
behaviour.
Behaviourism: the school of psychology associated with Harvard professor B.F. Skinner
(1904–90), architect of the controllable ‘Skinner box’ and ‘programmed learning’. Its key tenet
is that only observable behaviour may be scientifically studied.
Bias: prejudice: pre-disposition towards or against subject, movements and interpretations:
a source of systematic errors in research findings.
Case study: a ‘sample of one’ event, instance, state or sub-unit at one point in time.
320 Glossary of Terms
Case study method: gathering and analysing data from an individual example as a means
of making broader generalisations.
Causality (‘cause and effect’ or causation): the relationship between two variables when
one causes a change in the character of the other.
Census: count of 100% of population.
Central limit theorem: this states that, the larger the sample, the greater the tendency for
the sampling distribution of the means to follow a normal distribution.
Central tendency: concentration (cf. dispersion) of values statistically represented by mean,
mode, median, variance and standard deviation.
Classification data: information sought in questionnaire surveys which relate to the circum-
stances of the subject rather than their beliefs and attitudes. They include: sex, age group,
socio-economic group, education, housing tenure and dependants. Some can be obtained by
proxy (e.g. newspaper readership as an indicator of social class). Others can be guessed by
careful observation but should be checked with the subject. These variables may well include
important independent variables. For example, research has shown that older people are more
likely to participate in political activity than others.
Class interval: groupings of data selected by the researcher to enable a frequency distribution
to be constructed to represent and analyse data, e.g. ages 36–45, etc.
Closed questions: questions which seek – or receive – closed answers. Closed answers are
generally short and confined to yes, no or don’t know or specific answers, e.g. date of birth.
Coding: essentially the process of replacing or substituting groups of words, phrases or
sentences by letters or numbers (or a combination). Unlike a cipher, a code is not designed to
hide the original meaning of the word or phrase. Indeed, an essential property of an effective
code is that it should be readily understandable to users and readers. A word or phrase can be
given several different codes.
Cohort: groups sharing characteristic(s), e.g. males 18–29.
Comparative method: obtaining data from a population or sample of equivalent states or
sub-units, sectors or groups at the same points in time. It is the principal method used in
comparative politics and international relations.
Concept: basic idea.
Content analysis: set of techniques to systematically analyse contents of books, reports and
transcripts of speeches, etc. by measuring repetitions and bias.
Contingency: conditional relationship between variables.
Control Group: sample not subjected to changes in independent variable.
Correlation: statistically proven relation between two variables. But correlation does not
necessarily indicate causation. It is measured statistically by the correlation coefficient which
extends from +1 (perfect positive correlation) through 0 (no correlation) to −1 (perfect negative
correlation, i.e inverse relationship).
Counterfactual: what you would expect to happen but didn’t (e.g. Sherlock Holmes’ ‘dog
that didn’t bark in the night’). Counterfactuals may be possible evidence of unseen, dominant
power.
Covariance: phenomenon where two variables appear to change at the same time.
Data: raw material which may be sorted and processed into information. Data has no
meaning by itself. Note that although data is plural, writing data are is now considered pedantic.
Decile: a tenth part.
Declaration: a statement that follows the Acknowledgements page of the research report
which states that ‘This dissertation is entirely my own, original work with the exception of those
academic and other sources which are separately identified in the text, fully attributed in the
footnotes and listed in the bibliography. Signed ... Date ...’.
Deductive research: the research approach which develops specific hypotheses from
general principles or theories and then seeks to confirm or infirm the hypotheses: ’moving
from the general to the specific’ (cf. deductive research: moving from specific to the
general).
Demography: the study of population and influences. Demographic data therefore includes
population size and density, age and male/female distribution, household size, income, health
and life expectancy, socio-economic distribution, ethnicity and migration, etc.
Dependent variable: consequent event; outcome; variable which changes as a consequence
of change in the independent variable; in social research, this is generally represented by y.
Descriptive statistics: quantitative, numeric information which describes or summarises
primary data.
Diachrony: the study of the historical development of language.
Discourse: ‘a specific ensemble of ideas, concepts and categorisations that are produced,
reproduced, and transformed in a particular set of practices and through which meaning is
given to physical and social realities’ (Hajer, 1995: 44).
Discourse analysis: ‘analysing the way systems of meaning or discourses shape the way that
people understand their roles in society and influence their political activity’ (Howarth, 1995).
Directional hypothesis: (one-tailed test): an experimental hypothesis which states a partic-
ular direction of outcome, e.g. that the dependent variable will increase as the independent
variable changes. A non-directional hypothesis – where a causal relationship is hypothesised
but what happens to the dependent variable when the independent variable changes is unclear
– is called two-tailed.
Ecological fallacy: false reasoning which allocates attributes to individuals from information
about groups.
Eigenvalue: (Eigenwerte) used in Principal Component and Factor Analysis as a measure of
the variation in a group of variables attributable to one factor.
Empirical: data collected by systematic observation or experience (cf. theoretical).
Enumeration District (ED): the basic building block of the UK census consisting of around
200 households. They are the lowest level of census data publicly available until individual
household forms are released after 100 years.
Epistemology: study of knowledge.
Ethnography: study of social and cultural life of social or ethnic groups often by participant
observation.
Experimental method: a ‘sample of two’; a control sample and an experimental sample

drawn (ideally randomly) from the population under study.
Experimental sample: the sample subjected to the independent variable.
Facts: the available information. They present a picture of events. Truth is the reality behind
the facts. Sometimes the facts may obscure the truth – perhaps deliberately so.
Fallacy of composition: false reasoning that accords individual’s attributes to groups.
Falsification: after Popper, the doctrine that theories can never be confirmed: they can only
be refuted. So verification by repeated experiment can never be conclusive.
Field research: research undertaken in and of the real, social world (cf. laboratory research).
Focus group: group of subjects recruited by researchers to emulate the spontaneous
discussion of people who broadly share the same circumstances or identity. In particular, the
focus group should draw upon the respondents’ attitudes, feelings, beliefs, experiences and
reactions in a way that would not be feasible by other methods, for example, observation,
one-to-one interviewing or questionnaire surveys’ (Gibbs, 1997).
Gambler’s fallacy: prediction based on continuation of or non-continuation of coincidence
of random events, e.g. the belief that the prior incidence of heads or tails will influence the
outcome of the next spin of the coin.
Generalisability: the facility to draw reliable inferences about a population from a sample.
Grounded research: approach first developed by Glaser and Strauss where the findings are
entirely grounded in the data collected.
Guttman scale: a scale used in questionnaire surveys as a means of determining the
respondent’s intensity and consistency of attitudes to a specific topic by asking them a series
of related questions.
Hermeneutics: the study of interpretation.
Hypothesis: a specific theoretical proposition presented for testing by research. It is the
researcher’s initial answer to their research question.
Independent variable: causal event or driver; the variable causing change to the dependent
variable.
Inductive research: the approach which develops general hypotheses from observation of
specific phenomena (cf. deductive research: moving from general to the specific).
Inferences: generalisations about the research population drawn from samples or other
limited information
Inferential statistics: ‘a branch of applied mathematics or statistics based on a random
sample. They let the researcher make precise statements of the levels of confidence they
have in the results of a sample being equal to the population parameter’ (where a parameter
is a characteristic of the entire population drawn from a sample’) (Neuman, 2003: 539,
541); a statistical method of drawing reliable generalisations about a population from sample
data.
Information: data which has been sorted and classified to inform the research; information
has meaning.
Integer: whole number, e.g. 1, 2, 113.
Iteration: repeated procedure.

Language: the facility of speech.
Law of averages: doctrine that variations and errors in numeric data will be reduced by
repeated experiments.
Law of large numbers: doctrine that, as the sample size increases, it will assume the general
character of the population from which it has been drawn. In particular, the sample mean will
approach the population mean.
Likert scale: a scale used in questionnaire surveys to measure the intensity of attitudes of
subjects. The subject is asked whether they strongly agree, agree, neither agree nor disagree,
disagree or strongly disagree with a statement.
Linguistics: ‘the science of language’ (Crystal, 1995: 425).
Longitudinal method: gathering data from the observation of a cohort – a group of
people, states or organisations, etc. sharing one or more common characteristics – over an
extensive period of time. The common characteristics may include age, education, place or
specific condition.
Mean: a measure of central tendency. There are arithmetic, geometric and harmonic means.
The arithmetic mean, X is generally called the average and is the sum of terms X of a
population, divided by the number of terms in the population N, i.e. X = "X/N. In a sample,
the (arithmetic) mean is denoted by x and the number of terms by n and is given by the formula
x = "x/n.
Median: another measure of central tendency. It is the middle term in a ranked series, e.g.
in the series 2,3,4,5,6, then 4 is the median. In the series 2,3,4,5, then the median is 3.5.
Mode: another measure of central tendency. It is the most common or frequent term, e.g. in
the series 2,2,3,3,3,4 then 3 is the mode.
Method of agreement: term used by Mill to describe the comparative method adopted
in which apparently different cases having the same outcomes are compared to identify the
common – potentially causal – elements.
Method of difference: term used by Mill to describe the comparative method adopted in
which apparently similar cases having different outcomes are compared to identify the absent
or common – potentially causal – elements.
Method of concomitant variations: comparative method used when outcomes of different
magnitude occur to identify the potential causal variable by comparing their magnitude.
Moving average: a technique used in Time series analysis to smooth out seasonal and
irregular movements to identify the underlying trend.
Naturalism: this has two, almost uniquely, contradictory conceptualisations. The first –
the minority view held by Giddens – is that naturalism refers to the adoption in the social
sciences of models of inquiry from the natural sciences. So naturalism is effectively a new, non-
pejorative name for positivism. But the dominant view is that naturalism rejects positivism in
favour of a free-standing method of inquiry which is essentially humanistic and hermeneutic
(after Hermes, interpretative) and concerned with social meanings (actor’s beliefs, motives,
purposes, reasons etc. which lead to social actions) rather than frequency. Whilst positivism
emphasises scientific, controlled, replicable experimentation, naturalism seeks to study everyday
life in naturally-occurring situations. Naturalism is experiental: researchers seek to see the world
through the eyes of the subject and to understand what people feel, interpret and do. Moods,
ideas, identity and beliefs are the stuff of naturalistic research.
Nomination: method of recruiting a non-probability sample in which a social gatekeeper or
other intermediary is asked to nominate – name – potential recruits.
Non-probability samples: samples where members of the population do not have an equal
chance of being selected. They are not statistically reliable. They cannot generate generalisable
data.
Null hypothesis: ‘the statistical proposition of no difference’ (Clegg, 1998: 61). The
proposition – which must be refuted – that data favourable to our experimental hypotheses
occurred by chance (rather than any causal relationship between the variables). In sampling
research, the null hypothesis is that the sample has not been drawn randomly from or
is representative of the population and that therefore no generalisable inferences can
be made.
Ockham’s razor: ‘the principle of parsimony’: the simplest theory is the best.
Odds: ratio of probable outcome in sequence of events, e.g. odds of a coin falling heads is
1:2. Also expressed as a probability of 0.5 (i.e., 5 out of 10) or 50%.
Operationalisation: the process of transposing abstract theoretical concepts into observable,
measurable variables.
Outliers: exceptional lowest or highest terms (or first or last ranked terms) which lie outside
the general distribution and which can skew measures of central tendency and distort
interpretation.
Panel: a sample of individuals from a population who are repeatedly surveyed over time to
provide longitudinal or tracking data. They do not meet.
Poisson distribution: a special distribution used when the sample size is very large and the
probability of incidents, e.g. illness, is very small. (cf. Binomial Distribution used where there are
only two potential outcomes, e.g. heads or tails).
Political elites: individuals who exercise disproportionately high influence on the outcome
of events or policies in the research topic.
Population: the entirety of the group – of people or objects – under study and from which
the sample is drawn.
Positivism: After Comte, the doctrine that the only true knowledge is scientific knowledge.
Positivism emphasises scientific, controlled, replicable experimentation. Auguste Comte (1798–
1859) argued that all societies were fated to move from a theological stage of ‘fictitious
knowledge’ (in which all otherwise inexplicable phenomena were attributed to ‘spiritual’ forces)
via an intermediate metaphysical stage to a ‘positive’ stage. The underlying beliefs of the
theological stage were medieval faith and custom in which the family was the social base.
Those of the metaphysical stage were Enlightenment philosophy, the ‘scientific revolution’ and
the nation state. The state of knowledge of the positivist stage was scientific when rational
logic and humanity prevailed.
Power of a test: sensitivity of a test to identify relationships between variables.
Preface: a personal statement explaining why the researcher has chosen the topic, their
starting biases and prejudices and what the research means to them, i.e. the reflexive relations
between them and their social world.
Primary: data or information is original, unedited and first-hand (cf. secondary information).
Probability: the likelihood that an event will occur. For example, if an event is certain – like
death albeit in the distant future – then the probability is 100% or 1.00. Alternatively, a remote
possibility – like winning the UK national lottery next week – may be 0.001% or 0.00001. A
one-in-four likelihood would be expressed as a probability of 25% or 0.25. A bookmaker would
give odds of 4-1 on an event having a 25% probability.
Proposition: an unproven, generalised, theoretical explanation.
R2 : used in multivariate analysis as a measure of the extent to which the dependent variable
can be attributable to two or more independent variables.
Random Walk: method adopted by interviewers to access a sample of households. The
researcher begins at the centre of the town or city calling at every nth house in the first street,
turning left at the first intersection, right at the second and so on until the boundary of the
urban area is reached.
Rapport: French word used to describe a desirable state of mutual confidence, empathy, trust
and liking between the researcher and their subject(s).
Reflexivity: intellectual self-awareness gained through self-examination to identify deep-
seated biases.
Regression analysis: analysis carried out to calculate the relationship between the depen-
dent variable (y) and one (linear regression) or more (multivariate regression) independent
variables (x, x 1 . . .) represented in the formula y = a + bx where b is the regression coefficient
and gradient (or slope) of the best fit line.
Reliability: literally, the extent to which we can rely on the source of the information and,
therefore, the information. Reliable information is dependable, trustworthy, unfailing, sure,
authentic, genuine, reputable. Consistency is the main measure of reliability. Note that a
measure may be reliable but not accurate and vice-versa, e.g. a reliable weighing machine
may give an inaccurate but consistent recording of the same weight.
Research component analysis: method used to evaluate published research in which the
researcher dissects the published research into 15 or more components which are assessed
separately.
Research effect: the tendency for the population being studied to modify its behaviour and,
therefore, provide unrepresentative data.
Residual error: the difference between predicted and actual data.
Sample: a group of subjects chosen randomly or non-randomly from the population and
from whom it is intended that generalisations can be made.
Sampling error: potential variance between an attribute of the sample and the population.
Sample frame: a list or schedule of the population from which the sample will be drawn. It
may be a membership list, a directory of engineering firms or electoral roll.
Secondary data or information which is second-hand and may be previously edited or

interpreted material.
Semantic differential: a five or seven-point scale used in questionnaire surveys to assess
respondents’ beliefs, etc., between pairs of words with diametrically opposite meanings, e.g.
honest and dishonest.
Semi-structured interview: the most widely-used type of interview – especially in elite
interviews. The researcher uses a pre-designed schedule of a limited number of topic-related
questions and, pre-determined, alternative supplementary questions which question aspects
of the answer received. The format of the semi-structured interview is essentially one of
question-and-discussion.
Semiotics: ‘study of the properties of signs and signalling systems especially as found in all
forms of human communication’ (Crystal, 1995: 430).
Show-cards: printed lists used by researchers administering questionnaire surveys to ask
sensitive questions. For example, a card might contain a list of weekly incomes marked A,
B, C, etc. The subject can then be asked to give the letter corresponding to their own income
range.
Significance: characteristic of importance of data or calculated result.
Statistical Significance: the likelihood that any similarity between sample and population
means occurred entirely by chance.
Snowball sampling: non-random sampling in which subjects identify others similar to
themselves for further research, e.g. drug-users.
Split-half strategy: an approach adopted by some researchers towards focus groups. Half
the group are interviewed individually before the focus group and the others afterwards.
In this way, the role of the group in developing or imposing new collective perceptions or
understandings can be identified better.
Spurious relation: where variables appear statistically related or correlated but bear no
dependent or causal relationship to one another.
Standard deviation (SD or σ ): another measure of central tendency expressing the average
by which individual terms are less or more than the mean. It is calculated as the square root of
the variance.
Standard error of the mean (SE): a statistical measure of the variance between a sample
and the population: the standard deviation of a sampling distribution of the mean.
Structural content analysis: assesses how the text is presented and reported rather than
the frequency of key words used.
Substantive content analysis: analyses selected texts by counting the frequency and
distribution of key words.
Subject: the smallest unit of population adopted for research, e.g. individual people,
households, firms. The term is most widely used by qualitative researchers to de-objectify the
respondent or interviewee.
Synchrony: the study of the ‘lived’ language of the present.
t-Test: statistical test of the significance of the variance between two groups.
Tells: term given by body language analysts to the inadvertent signs given by subjects.
Theoretical saturation: in grounded research, term used by Glaser and Strauss to describe
the stage reached when additional data fail to reveal any further significant findings.
Theory: simply stated, a statement of general principles of the underlying relationship in
phenomena or events. Theory may be expressed as laws, propositions, arguments or hypotheses.
Historically, theory has generally been descriptive in terms of describing and, therefore,
explaining relationships. Some descriptive theories may be called laws, for example, Newton’s
law of universal gravitation, where they are accepted as having been universally verified by
observation. Newton’s law was subsequently superseded by Einstein’s theory of relativity. The
title law is therefore rarely given to theories. Alternatively, theory may be normative where it
proposes what ought to be the relationship. For example, the theory of egalitarianism is that
all people should be treated equally regardless of origin or circumstances.
‘thick/thin’: case study and a quarter; term used to describe the research method used when
the researcher seeks to extend the findings of their single case study by quickly testing them
against another comparable case.
Triangulation: term derived from navigation and surveying; a method of corroborating data
by seeking accounts from three or more perspectives and media.
Truth: the reality behind the facts (the available information). Sometimes the facts may
obscure the truth – perhaps deliberately so.
Type I Error: wrongly rejecting the null hypothesis.
Type II Error: wrongly accepting the null hypothesis.
Universe: population.
Validity: accuracy and appropriateness of measurement; ‘the extent to which a measure,
indicator or method of data collection possesses the quality of being sound or true as far as
can be judged. . . . in the social sciences generally, the relationship between indicators and
measures and the underlying concepts they are taken to measure is often contested’ (Jary &
Jary, 1995: 714).
Variance: a measure of central tendency calculated as the sum of the squares of the
differences between each term and the mean divided by the number of terms. Squaring the
difference makes negative differences positive. In a sample, the sum of the squared differences
is divided by the number of terms minus one, i.e. n − 1.
Verstehen: term adopted by Max Weber (1864–1920) to describe the necessary empathetic
understanding by researcher of their subject; ‘trying to see the world through the subjects’
eyes’.
Vignette: ‘Short story generated by the researcher and focusing on hypothetical characters
in particular situations. Interviewees are asked what they would do in these circumstances
or, alternatively, how they think that a third party might react. The latter mode of question
specifically distances the interviewer from the issues being studied and, in this sense, is
impersonal and less threatening’ (Arksey & Knight, 1999: 94–5).
Wave: term used in longitudinal research to describe each phase of cross-sectional survey of
the research cohort or panel.
For a more comprehensive and technical glossary of terms with examples and illustrations, see:
Jary, D. and Jary, J. (1995) Sociology Dictionary. Glasgow: HarperCollins.
Vogt W. P. (1999) Dictionary of Statistics & Methodology. London: Sage Publications.
Key Formulae and Symbols
Population Sample
Number of cases or terms N n
Individual values X x
Arithmetic mean X x
Variance σ2 s2
Standard deviation σ or SD s or sd
"(X − Xi )2 "(x − xi )2
Variance
√ N √(n − 1) 2
"(X − Xi )2 "(x − xi )
Standard deviation
N (n − 1)
Standard error of the mean, SE n/a sd
√
n
Confidence limits 95% ±(1.96 × SE )
i.e. in 95 out of 100 samples, the population mean will lie between
x + (1.96 × SE ) and x − (1.96×SE)
Standard error of the proportion, SEp
√
(or percentage) (pq/n) (where q = 1 − p)
Bibliography
Amis, K. (1997) The King’s English – A Guide to Modern Usage, cited by Mount, H. (2006) Amo,
Amas, Amat and All That: How to Become a Latin Lover. London: Short Books. p. 40.
Arksey, H. and Knight, G. (1999) Interviewing for Social Scientists. London: Sage.
Austin, J. L. (1962) How to Do Things with Words. Oxford: Oxford University Press.
Bachrach, P. and Baratz, M. (1962) Two faces of power. In American Political Science Review,
56: 947–52.
Barthes, R. (1966) Introduction to the Structural Analysis of the Narrative. Birmingham:
University of Birmingham.
Bell, J. (1997) Doing your own Research Project; A Guide for First-Time Researchers in
Education and Social Science. Buckingham: Open University Press.
Beresford, P. and Hoban, M. (2005) Effective Participation in Anti-poverty and Regeneration
Work and Research. York: Joseph Rowntree Foundation.
Berg, B. (2001) Qualitative Research for the Social Sciences. London: Allyn & Bacon.
Berger, A.A. (1997) Narratives in Popular Culture, Media and Everyday Life. Thousand
Oaks: Sage.
Brier, A.P. and Hopp, B. (2005) HAMLET a multidimensional scaling approach to text-oriented
policy analysis. In Journal of Diplomatic Language, 2 (1).
Bruner, J.S. (1986) Actual Minds, Possible Words. Harvard, Mass: Harvard University Press.
Bryman, A. (1988/96) Quantity and Quality in Social Research. London: Routledge.
Budge, I. (1997) Chapter 1: Party policy and ideology: reversing the 1950s? In Evans, G., and
Norris, P. (1999) Critical Elections. London: Sage
Burke, K. (1966) Language as Symbolic Action. Berkeley: University of California Press.
Burnham, P., Gilland, K., Grant, W. and Layton-Henry, Z. (2004) Research Methods in Politics.
Carr, E.H. (1990/1961) What is History? London: Penguin.
Carver, T. (2002) Discourse analysis and the ‘linguistic turn’: In European Political Science,
Autumn 2002.
Chomsky, N. (1986) Knowledge of Language. New York: Praeger.
Clegg, F. (1998) Simple Statistics: A Course Book for the Social Sciences. Cambridge:
Cole, G.D.H. (1938/66) The Common People. London: Methuen.
Cole, J. (1996) As it Seemed to Me: Political Memoirs. London. Weidenfeld & Nicolson. p. 64.
Crossman, R.H. (1976) The Diaries of a Cabinet Minister 1 Minister of Housing 1964–66.
London: Hamish Hamilton and Jonathan Cape. p. 590.
Crouch, C. (2004) Post-Democracy. Cambridge: Polity.
Crystal, D. (1995) The Cambridge Encyclopaedia of Language. Cambridge: Cambridge:
University Press.
Curt, Beryl C. (1994) Textuality and Tectonics; Troubling Social and Psychological Science.
Bibliography 331
Czarniawska, B. (1998) A Narrative Approach to Organization Studies. London: Sage

Publications.
Dahl, R. (1969) Who Governs? Democracy and Power in an American City. New Haven: Yale
University Press. A new edition was published in 2005, ISBN 03001 03921.
Denscombe, M. (1998) The Good Research Guide for Small-Scale Social Research Projects.
De Vaus, D.A. (2001) Surveys in Social Research. London: Routledge.
De Vaus, D. (2002) Analysing Social Science Data: 50 Key Problems in Data Analysis. London:
Sage.
Downs, A. (1957) An Economic Theory of Democracy. New York: Harper & Brothers.
Engels, F. (1845/1993) The Condition of the Working Class in England. Oxford: Oxford University
Press.
Evans, G. and Norris, P. (1999) Critical Elections. London: Sage.
Fairclough, N. (2nd edn 2000) Language and Power. Harlow: Longman.
Fairclough, N. (2000a) New Labour: New Language? London: Routledge.
Field, A. (2005) Discovering Statistics Using SPSS (and sex, drugs and rock ‘n’ roll) 2nd edn.
London: Sage.
Filstead, W.J. (1979) Qualitative methods: a needed perspective in evaluation research.
In Cooke, T.D. and Reichardt, C.S. (eds) Qualitative and Quantitative Methods in Evaluation
Research. Beverly Hills: Sage.
Foddy, W. (1993) Constructing Questions for Interviews and Questionnaires. Cambridge:
Foot, M. (1973) Aneurin Bevan: A Biography, 2. 1945–60, London: Davis-Poynter.
Franke, R.H. and Kaul, J.D. (1978) The Hawthorne experiments: first statistical interpretation,
American Sociological Review, 43: 623–43.
Friedrich, C.J. (1937) Constitutional Government and Politics. New York: Harper & Brothers.
Galbraith, J.K. (1992) The Culture of Contentment. London: Penguin.
Gallie, W.B. (1956a), Essentially contested concepts. In Proceedings of the Aristotelian Society,
56: 167–98.
Gibbs, A. (1997) Focus groups. In Social Research, 19. Guildford: University of Surrey.
Gilbert, N. (2003) Researching Social Life. London: Sage.
Glaser, B.G. and Strauss, A.L. (1967/1995) The Discovery of Grounded Theory: Strategies for
Qualitative Research. New York: Aldine de Gruyter.
Glaser, B.G. (1978) Theoretical Sensitivity: Advances in the Methodology of Grounded Theory.
Mill Valley: Sociology Press.
Greenfield, S., Singh, S., Tallack, P. et al. (2001) The Science Book. London: Cassell & Co.
p. 302.
Grigg, J. (1978) Lloyd-George: The People’s Champion. London. Eyre Methuen.
Gubrium, J. F. and Holstein, J. A. (1997) The New Language of Qualitative Method. Oxford:
Oxford University Press.
Hajer, M. (1995) The Politics of Environmental Discourse. Oxford: Oxford University Press.
Harris, K. (1967) Conversations. London: Hodder and Stoughton.
Hart, C. (1998) Doing a Literature Review. London: Sage.
Harvey, L. (1990) Critical Social Research. London: Unwin Hyman. pp. 19–32.
Heath, A. and Taylor, (1999) New sources of abstention. In Evans, G. and Norris, P. (eds) Critical
Elections: British Parties and Voters in Long-Term Perspective. London: Sage.
332 Bibliography
Held, D. and Leftwich, A. (1984) Chapter 8: A discipline of politics? In Leftwich, A. (ed.) What is
Politics? Its Activity and Study. Oxford: Blackwell.
Heywood, A. (2000) Key Concepts in Politics. Basingstoke: Palgrave Macmillan.
Hoinville, G., Jowell, R. et al. (1977) Survey Research Practice. London: Heinemann. p. 61 cited
in De Vaus, D.A. (2001) Surveys in Social Research. London: Routledge. p. 73
Honderich, T. (ed.) (1995) The Oxford Companion to Philosophy. Oxford: Clarendon Press.
pp. 405–6.
Hopkin, J. (2002) Chapter 12: Comparative methods. In Marsh, D. and Stoker, J. (eds) Theory
and Methods in Political Science. Basingstoke: Palgrave Macmillan.
Howarth, D. (2000) Discourse. Buckingham: Open University Press.
Hume, D. (1739) A Treatise on Human Nature. 1: 467–8.
Huxley, A. (1932/1994) Brave New World. London: HarperCollins.
Jary, D. and Jary, J. (1995) Sociology. Glasgow: HarperCollins.
Jenkins, R. (2001) Churchill: A Biography. London: Macmillan.
Kanji, G. K. (1999) 100 Statistical Tests. London: Sage.
Kinnear, P. R. and Gray, C.D. (2004) SPSS for Windows Made Simple. Hove: Psychology
Press Ltd.
Klingermann, H-D., Hofferbert, R. and Budge, I. (1994) Parties, Policies and Democracy.
Boulder: Westview Press.
Kohler Reissman, C. (1993) Narrative Analysis. London: Sage.
Krippendorff, K. (1980) Content Analysis: An Introduction to its Methodology. London: Sage.
Lacey, A. (1995) In Honderich, T., The Oxford Companion to Philosophy. Oxford: Oxford
University Press. pp. 145, 705–6.
Laclau, E. and Mouffé, C. (1985) Hegemony and Socialist Strategy. London: Verso.
Laclau, E. and Mouffé, C. (1987) Post-Marxism without apologies, New Left Review, 166: 79–106.
Landman, T. (2003) Issues and Methods in Comparative Politics: An Introduction. London:
Routledge.
Lasswell, H. (1958) Who gets What, When, How. New York: Meridian.
Lee, R. L. (1995) Dangerous Fieldwork. London: Sage.
Leftwich, A. (ed.) (1984) What is Politics? Its Activity and Study. Oxford: Blackwell.
Leftwich, A. (2000) States of Development: On the Primacy of Politics in Development.
Cambridge: Polity.
Leftwich, A. (ed.) (2004) What is Politics? Cambridge: Polity Press.
Levine, D. and Stephan, D. (2005) Statistics: Even you can Learn. London: Pearson
Prentice Hall.
Lewis-Beck, M. S. (1995) Data Analysis: An Introduction. London: Sage.
Lubiano, W. (1992) Black ladies, welfare queens and state minstrels. In Morrison, T. (ed.)
Race-ing Justice, En-Gendering Power: Essays on Anita Hill, Clarence Thomas and Social
Reality. New York: Pantheon.
Lukes, S. (1974) Power: A Radical View. London: Macmillan.
Mace, C.A. (1962) The Psychology of Study. Harmondsworth: Penguin.
Manning, P. K. and Cullum-Swan, B. (1998). Narrative, content, and semiotic analysis.
In Denzin, N.K. and Lincoln, Y.S. (eds), Collecting and Interpreting Qualitative Materials,
pp. 246–73. Thousand Oaks, CA: Sage.
Marsh, D. and Stoker, G. (eds) (1995) Theory and Methods in Political Science. Basingstoke:
Macmillan.
Bibliography 333
Marsh, D. and Furlong, D. (2002) Chapter 1: A skin is not a sweater: ontology and epistemology
in Political Science. In Marsh, D. and Stoker, G. (eds) Theory and Methods in Political Science.
Marsh, D. and Stoker, G. (eds) (2002) Theory and Methods in Political Science. Basingstoke:
Palgrave. pp. 1–16.
Maslow, A.H. (1970) Motivation and Personality. New York: Harper and Row.
Mayer, L.C. (1989) Redefining Comparative Politics: Promise versus Performance. London: Sage.
Mayo, E. (1949) Hawthorne and the Western Electric Company. London: Routledge.
McEvoy, J. (2006) Elite interviewing in a divided society: lessons from Northern Ireland.
In Politics 26 (3): 184–91.
McLuhan, M. (1962) The Gutenberg Galaxy. London: Routledge and Kegan Paul.
Miles, M.B. and Huberman, A.M. (1994) Qualitative Analysis: An Expanded Sourcebook.
London: Sage.
McRobbie, A. (1982) The Politics of feminist research: between talk, text and action. In Feminist
Review, 12, pp. 46–58.
Mill, J.S. (1843) A System of Logic. London: Longman.
Moore, B. (1966) Social Origins of Dictatorship and Democracy. Boston: Beacon Press.
Morris, D. (1977) Manwatching; A Field Guide to Human Behaviour. London: Jonathan Cape.
Morris, D. (1994) Bodytalk: A World Guide to Gestures. London: Jonathan Cape.
Morris, D., Collett, P., Marsh, P., and O’Shaughnessey, M. (1979) Gestures: Their Origins and
Distribution. London: Jonathan Cape.
Mumby, D.K. (1993) Narrative and Social Control. Newbury Park, Calif.: Sage.
Neuman, W.L. (2003) Social Research Methods: Qualitative and Quantitative Approaches.
London: Allyn and Bacon.
Nicolson, N. (1953) King George V. London: Routledge. Chapter XXVII.
Norris, P. and Lovenduski, J. (1995) Political Recruitment: Gender, Race and Class in the British
Parliament. Cambridge: Cambridge University Press.
Oakley, A. (2000) Experiments in Knowing: Gender and Methods in the Social Sciences.
Cambridge: Polity Press. pp. 37–8.
O’Sullivan, T., Hartley, J., Saunders, D., Montgomery, M. and Fiske, J. (1997) Key Concepts in
Communications and Cultural Studies. London: Routledge.
Parker, I. (1992) Discourse Dynamics. London: Routledge.
Pennings, P., Keman, H. and Kleinnijenhuis, J. (2006) Doing Research in Political Science: An
Introduction to Comparative Methods and Statistics. London: Sage.
Plummer, K. (1995) Telling Sexual Stories: Power, Change and Social Worlds. London:
Routledge.
Przeworski, A., Alvarez, M., Cheibub, J.A. and Limongi, F. (1996) What makes democracies
endure? In Journal of Democracy, 7 (1): 35–55.
Race, K.E., Hotch, D.F. and Parker, T. (1994) Rehabilitation programme evaluation: use of focus
groups to empower clients. In Evaluation Review, 18 (6): 730–40.
Reed, M. and Marsh, D. (2002) Chapter 11: Combining quantitative and qualitative method. In
Marsh, D. and Stoker, G. (eds) Theory and Methods in Political Science. Basingstoke:
Macmillan, pp. 231–48.
Richards, D. (1996) Elite interviewing: approaches and pitfalls. In Politics, 16 (3): 199–204.
Ritchie, J. and Lewis, J. (2003) Qualitative Research Practice: A Guide for Social Science
Students and Researchers. London: Sage.
Robertson, D. (1993) Politics. London: Penguin.
334 Bibliography
Robson, C. (1993) Real World Research. Oxford: Blackwell.

Rose, G. (1982) Deciphering Sociological Research. London: Macmillan.
Russell, B. (1971) A Liberal Decalogue. In The Autobiography of Bertrand Russell, 3. 1944–67.
London: George Allen & Unwin.
Sanders, D. (2002) Chapter 2: Behaviouralism. In Marsh, D. and Stoker, G. (eds) Theory and
Methods in Political Science. Basingstoke: Palgrave Macmillan. p. 52.
Schattschneider, E.E. (1961) The Semisovereign People: A Realists’s View of Democracy in
America. New York: Holt, Rinehart. In The Process and Effects of Mass Communication.
Urbana, Ill.: The University of Illinois Press.
Silverman, D. (2001) Interpreting Qualitative Data: Methods of Analysing Talk, Text and
Interaction. London: Sage.
Smith, M. (1993) Personal interviewing: a mini-training workshop. In Breakwell, G.M., Foot, H.
and Gilmour, R. (eds) Doing Social Psychology: Laboratory Exercises. Leicester: BPS.
Smith, M.J. (2003) Social Science in Question. London: Sage.
Spencer, L., Ritchie, J. and O’Connor, W. (2003) Chapter 8: Analysis: practices, principles and
processes. In Ritchie, J. and Lewis, J. (eds) Qualitative Research Practice: A Guide for Social
Science Students and Researchers. London: Sage. pp. 199–218.
Squire, C. (1995) Discourse analytical psychology. In Wilkinson, S. and Kitzinger, C. (eds) (1995).
Feminism and Discourse: Psychological Perspectives. London: Sage.
Stringer, E. (1996) Action Research: A Handbook for Practitioners. London: Sage.
Stubbs, M. (1983) Discourse Analysis. Oxford: Blackwell.
Webb, S(ydney) and B(eatrice) (1975/1932) Methods of Social Study. LSE: Cambridge
University Press.
Williams, R. (1976) Keywords. London: Fontana.
Woodruff, W. (2003) Beyond Nab End. London: Abacus.
Yates, S.J. (2004) Doing Social Research. London: Sage.
Index
3Rs, 305 n.26 Bergman, M., 129 (box) Cardinal numbers, 193–194
Absorbed actions, 163 Bernoulli, Jacques, 234 Carr, E.H., 299
Accessibility, 52 Bernoulli’s trials, 234–235 Carver, T., 280
ACORN, 148 Bevan, Aneurin, 84 case studies:
Action research method, 52, 62–63 Bias, 17–18 case study meetings, 260
All-postal voting, 54–55 Biblical narrative, 299 Evaluating information 86–88
Amis, Kingsley, 80 Bibliographies, 107 (box) Fairclough’s critical discourse
ANOVA, 209 Binomial distribution, 234–236 analysis, 290–291, 291–297
Archives, 104–105 Biography, 84 interim case study, 260
Aristotle, 175 Blair, Tony, 291 narrative analysis, 302–303
Arithmetic mean, 185 Boas, Franz, 282 qualitative information
Arksey, H. and Knight, G., Body language, 161–162 analysis case study, 243–259
133, 137 claims made for, 162–163 single case study, 53–54
Assessment letter, sample, 77 (box) Morris on, 163–168 Categorical numbers, 193
Association, 31 recording and analysing, 168–171 Causality, 29–32
measuring, 208–210 Bogardus Social Distance Scale, Census returns, 103
multiple linear regression 149–150 Central limit theorem, 199
analysis, 213–217 Brier, A.P. and Hopp, B., 278 n.4 Chicago School, 221, 238 n.1
regression statistics, calculating, British Election Study (BES), 44, Chomsky, N., 282
210–213 49, 61 Churchill, Winston S., 279, 304 n.3
testing for, 206–208 British Household Panel Survey Clegg, F., 196, 204, 218, 310, 317
Association of Colleges on data, 14 (BHPS), 61 Coding, 153–154, 242–243
Auden, W.H., 140 Bruner, J.S., 298 Coefficient of correlation, 208
Austin, J. L., 300, 304 Bryman, A., 41, 42, 49 Cohort, 60–61
Budge, I., 268–270, 278 Cole, G.D.H., 88
Bachrach, P. and Baratz, M., 5, 26, Burke, K., 299 Cole, John, 81–82, 83
36 n.9 Burnham, P., Killand, K., Collett, Peter, 172 n.2
Barthes, Roland, 281, 298 Grant, W., and Layton- Comparative method, 51, 55–60
Bashir, Martin, interview of Henry, Z., 4, 8, 21–22, 137, Comte, Auguste, 23
Princess Diana: case study: 149–150, 196, 278, 283 Constructed memory, 60
narrative analysis, 302–303 Bush, George W.: Content analysis, 263–264
Baton signals, 164–165 State of the Union Address Hamlet II, 271–276, 278 n.4
Baudrillard, Jean, 282 (2005), 271–276 qualitative, 264–266
BBC News, 161 State of the Union Address software for, 270–271
Behaviouralism, 25–27 (2007): case study: structural, 267–268
Behaviourism, 25 Fairclough’s critical discourse substantive, 268–270
Bell. J., 309, 316, 317 analysis, 291–297 Counterfactual, 27, 36 n.12
Benn, Tony, 83–84 Covariation, 31
Beresford, P. and Hoban, M., 63, CACI, 148 Critical social theory, 62
66 n.16 CAQDAS (Computer-Assisted Crossman, Richard, 81–82, 83
Berg, B.L., 129 (box) Qualitative Data Analysis Crouch, C., 44, 48
Berger, A.A., 298 System) software, 41–42 Crystal, D., 281, 282
336 Index
Curt, Beryl C., 279 Discovered actions, 163 Facts, 85

Czarniawska, B., 299 Disraeli, Benjamin, 183 Fairclough, Norman, 280, 304
Domhoff, D.W., 27 Fairclough’s critical discourse
Dahl, Robert, 25–27, 35 Downs, A., 74 analysis:
Data: Durkheim, Emile, 23, 24–25 case study: George W. Bush’s
accuracy of, 85 State of the Union Address
dichotomous nominal data, 234 EEC Directive 96/96, 13 (2007), 291–297
reliability of, 83 EFTA (European Free Trade Area), case study: interview of
validation, 178–179 21 n.4 Margaret Thatcher, 290
Data analysing, 175–176 Electoral Commission, 54 case study: New Labour,
assembly of data, 178 Elite dialect, 286 290–291
hierarchy of, 177–178 Elite interviews: description, 286–289
process, 176–177 approaching elites, 121–123 explanation, 289
reduction, 179 defining elite, 119 interpretation, 289
validation, 178–179 introductory letter, 122 (box) Fairy tales, 298
Data Protection Act (1988), 13 post interview, 127 Falsification/falsibility, 24, 43
Data Protection Co-ordinator, 14 preparation and procedure, Feminism, 28–29
Data protection legislation, 13–15 123–125 Field, A., 237
Databases, 102 reasons for interviews, 119–120 Fields of experience, 128
De Vaus, D.A., 21, 35, 98, 159, 204 rules, 120–121 Filstead, W.J., 41
Deductive research, 33 structure, 125–127 Focus groups, 134–136
Denscombe, M., 313, 318 Empiricism, 24–25 Foddy, W., 132, 138
Derrida, Jacques, 282, 284 Engels, F., 55–56 Foot, Michael, 84
Descriptive statistics, 183–187 Enlightenment, 23 Footnotes, 106 (box), 311
calculation, 190–193 Enumeration district (ED), 103 Foucault, Michel, 81, 279,
cardinal numbers, 193–194 Epistemology, 22 282–283, 284
categorical numbers, 193 ESRC, 12, 21n.2 Freedom of Information Act
grouped frequency distribution, Ethics. See harm; Nuremberg Code (2000), 103–104
188–190 Friedrich, C.J., 119
and n, 188 Eugenics, 25, 208–209
and N, 188 Eurobarometer, 61 Galbraith, J.K., 50 n.8
standard derivation, 188 European Convention of Human Gallie, W.B., 279
variance, 187–188 Rights & Fundamental Galton, Francis, 208–209
Diachrony, 281 Freedoms, 13, 21 n.3 Garfunkel, Harold, 283
Dialects, 286 Evaluating information case study, Gaze behaviour, 165
Dichotomous nominal data, 234 86–88 George V, 89
Discourse analysis, 242, 279–281 Excel, 190–193, 200, 223 Gestures, 164
critical discourse analysis calculating multiple linear Gibbs, A., 135
(CDA), 284 regression equations, 214–216 Giddens, Anthony, 27, 285, 301
constituents, 301 calculating regression statistics, Gilbert, N., 65, 159
linguistics, 281–282 210–212 Gini coefficient, 166
Marxist school, 284 Poisson distribution, 236 Glaser, Barney, 33–34, 259
philosophy, 282–283 time series analysis, 230–232 Glaser, B.G. and Strauss, A.L., 35
positivist school, 283 Experimental effect, 17 Glottal stop, 305 n.27
post structuralists, 284 Experimental method, 51, 54–55 Goffman, Erving, 283, 50 n.6
realist school, 283–284 Eye-pupils, 166 Gramsci, Antonio, 285
semiotics, 282 Greenfield, S. et al, 4
See also Fairclough’s critical Factor analysis, 220–224 Grigg, J., 84
discourse analysis using SPSS, 224–230 Grounded research, 33–34
Index 337
Grouped frequency distribution, with non-elites, 127–129, 129 Mace, C.A., 4

188–190 (box) Mach, Ernst, 23
Gubrium, J.F. and Holstein, J.A., pariah topics, 130–132 Manifesto Research Group, 268
35, 261 projective technique, 132–133 Manning, P.K. and
Guttman scale, 149–151, 151 (table) questions, 117–118 Cullum-Swan, B., 298
recording, 129–130, 138 n.1 Marlow, A., 177
Hajer, M., 280 vignettes, 133–134 Marr, Andrew, 84–85
Halo effect, 17 Marsh, D. and Furlong, D., 35
Hamlet II, 271–276, 278 n.4 Jary, D. and Jary, J., 74, 83, 281 Marsh, D. and Stoker, G., 5, 8
Harm, types of, 10 Jenkins, Roy, 82 Mayer, L.C., 64
Harris, K., 244–251 John Henry effect, 16 Mayo, Elton, 15–16, 124
Harrison, L., 49, 98, 138, 159, 277 Journals of political science, 70 McEvoy, J., 138
Hart, C., 110 McLuhan, Marshall, 283
Harvard system, 106 (box) Kanji, G.K., 205, 219 McRobbie, A., 299
Harvey, L., 65 Kinnear, P.R. and Gray, C.D., 237 Measures of central tendency, 185
Hawthorne effect, 15–16 Kohler Reissman, C., 299 Median, 185
Healey, Dennis, 84 Krippendorff, K., 277 Meek, James, 175
Heath, A., and Taylor, B., 44 Members’ resource, 285
Heath, Ted and Harold Wilson: Lacan, Jacques, 281 Memoing, 259–260
qualitative information analysis Laclau, E. and Mouffe, C., 284 Method of agreement, 57
case study, 243–259 Landman, T., 59–60, 64 Method of concomitant
Heidegger, Martin, 299 Language, 281 variations, 57
Held, D. and Leftwich, A., 8 Langue, 281 Method of difference, 57
Heywood, A., 29 Large-n samples, 58, 59 Miles, M.B. and Huberman, A.M.,
Hill, Anita, 300–301 Last Time, The, 309 177, 243, 259, 261, 262
Hobbes, Thomas, 23 Law of large numbers, 53, 65 n.1 Mill, John Stuart, 23, 55, 56, 57
Honderich, T., 32 Laws, 22 Mixed actions, 164
Hopkin, J., 64 Lee, R.L., 65 Mixed methods, 47–49, 48, 49
Hopp, Bruno, 271 Leftwich, A., 8, 64 Mochmann, Ekkehard, 271
Howarth, D., 280, 283–285, 304 Levine, D. and Stephan, D., Modernity, 185
Hume, David, 23, 30–31, 32 204, 219 Moore, B., 59
Hypothesis, 22, 23 Levi-Strauss, Claude, 281 Morris, Desmond, 163–168
Lewis-Beck, M.S., 219 Most different systems design
Ideology, 286 Likert, Rensis. See Likert scale (MDSD), 58
Inductive research, 32 Likert scale, 149, 153–154, 187 Most similar systems design
Inferential statistics, 197–198 Literature reviews: (MSSD), 58
central limit theorem, 199 examples of reviews, 107–108, Multiple linear regression
national opinion polls, 201–202 110–116 (box) equations, 213–217
normal distribution, 198–199 official records, 103–105 Mumby, D.K., 298
null hypothesis, 202–204 plagiarism, 108–109
standard error of the purpose of, 100–102 N and n, 188
proportion, 201 reading literature, 105–106 Narrative analysis, 297–298
standard error of the sample referencing, 106–107 case study: Martin Bashir’s
mean, 199–201 searching for literature, interview of Princess Diana,
Intermediaries, 92 102–103 302–303
Interquartile range, 186 writing the review, 106 definitions, 299–301
Interviews, 85–86 Longitudinal method, 51, 60–62 origins, 298–299
with elites. See elite interviews Lovenduski, J., 29 Narrative-of-the-self, 302
focus groups, 134–136 Lukes, Stephen, 5, 27 National opinion polls, 201–202
338 Index
National Readership Survey, Prejudices proforma, 19–20 Questionnaire surveys:

146–147 Pre-knowledge, 298–299 access, 141–143
Naturalism, 27–28 Probability, 198 administering, 143
Natural science, 23, 36 Probability samples, 93–95 coding, 153–154
Neuman, W.L., 29, 36, 98, 138, Property ownership, 109 (table) definition, 140–141
177, 178, 197, 37 n.17 Propositions, 22–23 demonstration questionnaire,
New Labour: Propp, V., 298–299 154–158
case study: Fairclough’s critical Przeworski, A., Alvarez, M., design, 144–145
discourse analysis, 290–291 Cheibub, J.A. and Guttman scale, 149–151, 151
Nicholson, Harold, 86–89 Limongi, F., 64, 66 n.8 (table)
Nietzsche, Friedrich, 47 Pygmalion effect, 17 Likert scale, 149, 153–154, 187
Nomination, 91 motivating respondents, 143–144
Non-probability samples, 91–93 QAA, 5 n-th address technique, 142
Non-verbal leakage, 165 Qualitative analysis compared personal safety, 152–153
Normal distribution, 198–199 with quantitative analysis, postcodes, 148
Norris, P., 29 178 (table) proxy-indicators, 145–147
Null hypothesis, 202–204 Qualitative information, analysing, random walk technique, 142
Nuremberg Code (1947), 10–12 235 ranking, 151–152
case study: Harold Wilson and scale rating, 150–151
Oakley, A., 28 Ted Heath, 243–259 semantic differential, 148
Official records, 103–105 case study meetings, 260 show cards, 147
Ontology, 22 coding, 242–243 Questions. See Interviews;
Orwell, George, 13 content analysis, 241 Questionnaire surveys
Osgood, Charles E., 148 discourse analysis, 242 Quota samples, 95–96
O’Sullivan, T., 283 interim case study, 260
Outlier effect, 185 memoing, 259–260 Race, K.E., 136
second-level pattern coding, RAE (Research Assessment
Panel, 61 254–257 Exercise), 78 n.3
Parker, I., 284 tagging and coding, 243 Range, 185–186
Park, R.E., 221 typology, 257 RDN (Research Discovery
Peacock effect, 17 Qualitative methods: Network), 103
Pennings, P., Keman, H. and characteristics of, 43 (table) Reed, M. and March, D., 50
Kleinnijenhuis, J., 64, 65, criticisms and weaknesses, 46 Referencing, 106–107
196, 219 defence and counter-claims on Regression analysis, 209
Personal life histories, 60 behalf of, 46–47 Regression analysis, calculating,
Pilot project, 54–55 strengths, 45–46 210–213
Placebo effect, 16 See also Mixed methods Reinharrz, 9
Plagiarism, 108–109 Quantitative analysis compared Reliability of data, 83
Plummer, K., 299 with qualitative analysis, 178 Research:
Poisson distribution, 235–236 (table) action research method, 52,
Poisson, Siméon Denis, 235 Quantitative methods: 62–63
Politics as an academic subject, 4–6 characteristics of Table, 43 choice of, 52–53
Popper, Karl, 23, 24, 43, 202 defence and counter-claims on comparative method, 51, 55–60
Population, 91, 185 behalf of, 44–45 deductive research, 33
Positivism, 23–24 strengths, 42–44 definition, 3–4
See also Feminism; Naturalism weaknesses and criticisms, 44 experimental method, 51, 54–55
Post-modernism, 29 See also Mixed methods grounded research, 33–34
Postural echo, 165 Quarterly deviations, method of, inductive research, 32
Powell, Enoch, 120 232–234 longitudinal method, 51, 60–62
Index 339
RAE (Research Assessment Smith, M.J., 36 Thomas, Mae, 301

Exercise), 78 n.3 Snowball samples, 92 Thompson, Walter, 82
research component analysis Social facts, 24 Time series analysis, 230–232
(RCA), 70–73 Social gatekeeper, 91–92 Trained actions, 163–164
Rose’s ABCDE validity analysis Social Sciences Citation Index, 102 Treaty of Rome (Article 108), 13
deciphering model, 73–76 Socio-economic classification Triangulation, 89–90, 120, 138 n.4
single case study, 53–54 scheme, 146 (box) Truth, 85
research component analysis Sources 80–83 Twain, Mark, 183
(RCA), 70–73 Spearman, Charles, 25
Research effects, 15–17 Spencer, L., Ritchie, J. and United Nations Declaration of
Researchers and subjects, 9–10 O’Connor, W., 180 Human Rights, 13
Richards, D., 138 SPSS, 224
Ritchie, J. and Lewis, J., 177, 262 binomial test, 235 Validity, 74–76, 83
Robson, C., 65 data view, 226–227 Variable-orientated analysis, 58
Rose, G. See Rose’s ABCDE factor analysis using, 224–230 Variance, 187–188
validity analysis deciphering rotated component matrix, 229 verstehen, 45
model text books on, 237 Vienna Circle, 23, 24
Rose’s ABCDE validity analysis total variance explained, 228 Voice-transcription software, 130
deciphering model, 73–76 variable view, 224–226 Volunteer samples, 92–93
Rotated component matrix, 229 Squire, C., 280
Rumsfeld, Donald, 101 Standard error of the
Webb, B. and Webb, S., 18, 21 n.9
Russell, Bertrand, 79 proportion, 202
Webb, Beatrice, 18
Standard error of the sample mean,
Saint-Simon, Comte de, 23 Weber, Max, 29, 32, 263
199–201
Sampling: Williams, Raymond, 286, 304
Statistics, 206
errors, 94, 199 Wilson, Harold, 298
official, 84
large-n samples, 58, 59 and Ted Heath: qualitative
regression statistics and
non-probability samples, 91–93 information analysis case study
association, 210–213
probability samples, 93–95 case, 243–259
See also descriptive statistics;
quota samples, 95–96 WinStat, 223
inferential statistics
small-n samples, 58, 59–60 Wirth, L., 221
Strauss, Anselm, 33–34
snowball samples, 92 Wittgenstein, Ludwig, 23, 282
Stringer, E., 65
volunteer samples, 92–93 Woodruff, William, 105
Structuralism, 281
Sanders, D., 27 Writing-up, 309–310
Stubbs, M., 280
Saussure, Ferdinand de, 281 final draft, review of, 316
Subjects and researchers, 9–10
Scepticism, 4 literature reviews, 106
Suicide, 24–25
Schattschneider, E.E., 32 publication rules, 311
Synchrony, 281
Schramm, W., 128 readership, 310
Second-level pattern coding, Tagging and coding, 243 re-writing, 316
254–257 Taylor, W.F., 21 n.7 structure, 311
Serendipity, 3–4 Terms, 185 structure and contents, 313–316
Series, 185 Thatcher, Margaret, 266 when to write up, 312–313
Shakespeare, William, 302 case study: Fairclough’s critical word distribution, 316–317
Silverman, David, 176, 262 discourse analysis, 290
Single case study, 53–54 Theoretical saturation, 34 Yates, S.J., 160
Skinner, B. F., 25 Theory, 22
Small-n samples, 58, 59–60 Thick/thin approach, 54 Zipf, George, 266
Smith, M., 129 (box) Thomas, Clarence, 300–301 Zipf’s Law, 266–267

Research Methods in Politics - A Practical Guide

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Research Methods in Politics - A Practical Guide

Uploaded by

Copyright:

Available Formats

Los Angeles • London • New Delhi • Singapore

© Roger Pierce 2008

First Published 2008

SAGE Publications Ltd

SAGE Publications Inc.

SAGE Publications India Pvt Ltd

SAGE Publications Asia-Pacific Pte Ltd

Library of Congress Control Number: 2007934984

British Library Cataloguing in Publication data

Typeset by CEPHA Imaging Pvt. Ltd., Bangalore, India

About This Book v

2 Power in Research, Ethics, Data Protection and Bias 9

3 The Philosophy and Principles of Research 22

4 Qualitative Versus Quantitative Methods: A Relevant

5 Collecting Data Sets: Case Studies, Experimental,

PART III COLLECTING INFORMATION 67

6 Critically Evaluating Published Research 69

7 Evaluating Information: Validity, Reliability, Accuracy,

8 Completing a Literature Review: Accessing Published (β)

9 Asking Questions: Effective Elite Interviews, Other Interviews,

10 Questionnaire Surveys 140

PART IV DATA ANALYSIS 173

12 Analysing Research Data: The Process 175

Part IVA Quantitative Analysis 181

13 Calculating and Interpreting Descriptive Statistics 183

14 Using and Understanding Inferential Statistics 197

15 Testing for Association 206

16 Applying Factor Analysis and Other Advanced Techniques 220

PART IVB Qualitative Analysis 239

17 Analysing Qualitative Information: Classifying, Coding

18 Using Content Analysis 263

19 Understanding and Adopting Discourse and Narrative

PART V COMMUNICATING RESEARCH 307

Glossary of Terms 319

Key Formulae and Symbols 329

Hobbes, T. (1651/1996) Leviathan. Oxford, Oxford University Press.

Teaching and learning strategy

1. text (supplemented by diagrams and other illustrations) communicating the underlying

How to use the book

Where best to begin?

A three-stage approach is recommended using the following proforma. You can

1. Assess your starting research skills.

Training needs self-evaluation proforma

Chapter Chapter Title Existing Knowledge and Understanding

Chapter Chapter Title Existing Knowledge and Understanding

Chapter Chapter Title Existing Knowledge and Understanding

There is nothing as useful as a good theory.10

and, to Patrick Brindle of Sage who encouraged me to submit a proposal. However,

1 Panegyric, public speech.

Teaching and learning objectives:

1. To answer the question: What is research?

Research is essentially a process of systematic inquiry. Its core activities are:

• goal orientated and purposeful

Academic research essentially involves a systematic process. It begins with a

• testing the applicability and relevance of theory to new contexts

society (associated with feminism); interpretative theory (anti-foundationalism); and,

[Politics is concerned with] … developing a knowledge and understanding of government

Illustration 1 Ask yourself, Father, what do we really mean by ‘power’?

Textbook structure and contents

Part IV introduces methods for analysing and interpreting the information

Questions for discussion or assignments

1. What is Politics? Is ‘political science’ more appropriate, a misnomer or an