Research Methods and Analysis

Research Methods and Analysis for
Business Studies: a holistic approach
G.Egziabher H.Selasie Gebru

Assistant Professor
MBA, MNE, MEP, EMIN
College of Business and Economics

Mekelle University, Mekelle
Ethiopia
To
My wife TTye
Preface
In the professions of Accounting and Finance, research is becoming such a demanding

task. Accounting and Finance researches are published in the reputed Journals such as;
Journal of Finance, Journal of Accounting and Finance, Journal of Financial Research,
Journal of Accounting Research, International Journal of Auditing and Journal of
Abacus. Current textbooks on business research methods such as Zikmund (2000) and
Babbie (1986) present the basic principles of doing research. But in most cases students
and junior researchers demand hands on research experiences in addition to the general
principles of research. This teaching material includes the basic principles of research and
it uses research papers published in peer reviewed Journals as illustrative examples.
The primary target audiences of this teaching material are undergraduate students of
Accounting and Finance. It helps them to be able to read and understand Journal article
and other academic publications within their profession. This teaching material is also of
help to postgraduate students and junior researchers in Accounting and Finance. Further,
this teaching material is also helpful to undergraduate students of Cooperatives,
Economics, Management, and Public Administration and Development Management.
This teaching material has seven chapters. The first chapter introduces the definition of
science and the concept of research. The second chapter presents the research design, that
is, the art of visualizing a research project before it is started. The third chapter explains
the research proposal, that is, the art of convincing others. Chapter four presents on the
practical issues of data collection and sampling. Chapter five is devoted to survey
research method and analysis. This chapter introduces the learner with survey data
analysis using STATA software routine. Chapter six is devoted to case study research
method and analysis. Chapter seven presents about scientific research report writing and
presentation. This chapter acquaints students with techniques regarding how to write
publishable research papers.
Gebregziabher Hailesilasie
Assistant Professor
Table of Contents
Content Page No.
Chapter 1: Introduction to Research
1.1 Introduction 1
1.2 What is Science? 1
1.3 What is Research? 2
1.4 Purposes of Research 3
1.5 Types of Research 5
1.6 Research Ethics 6
Chapter 2: The Research Design
2.1 Introduction 7
2.2 What is Research Design? 7
2.3 Types of Research Designs 9
2.4 Designing the Research Problem 12
2.5 Designing the Research Hypothesis 14
2.6 Designing the Data and Method 18
Chapter 3: The Research Proposal
3.1 Introduction 20
3.2 Types of Research Proposal 20
3.3 Sections of a Research Proposal 21
3.4 Sources of Research Funding 40
3.5 Evaluating the Research Proposal 40
Chapter 4: Data Collection and Sampling
4.1 Introduction 44
4.2 Unit of Analysis versus Unit of Observation 44
4.3 Sources of Primary Data 45
i
4.4 Overview of Primary Data Collection Techniques 46
4.5 Sources of Secondary Data 50
4.6 Overview of Secondary Data Collection Techniques 50
4.7 Biases in Data Collection 52
4.8 Sampling Terminology 53
4.9 Types of Probability Sampling Designs 55
4.10 Types of Non-Probability Sampling Designs 58
4.11 Sample Size Determination 60
Chapter 5: Survey Method and Analysis
5.1 Introduction 64
5.2 Characteristics of Surveys 64
5.3 Questionnaire Design 66
5.4 Testing Reliability and Validity of Questionnaire 74
5.5 Quantitative Analysis of Survey Data 76
5.6 Data Analysis Uisng STATA software routine 83
Chapter 6: Case Study Method and Analysis
6.1 Introduction 94
6.2 Scientific Paradigms 94
6.3 When to use case Studies? 97
6.4 Case Study Research Design 98
6.5 Data Analysis in Case Studies 105
6.6 Case Study Research Method versus Operations research Method 116
Chapter 7: Scientific Research Report Writing and Presentation
7.1 Introduction 118

7.2 Types of Scientific Research Reports 119
7.3 Deciding to Publish your Work 120
7.4 Writing a Scientific research report 121
7.5 Presenting a Scientific Research Report 132
ii
Introduction to Research Mekelle University, CBE
CHAPTER 1
INTRODUCTION TO RESEARCH
1.1 Introduction
Science is an enterprise dedicated to "finding out" no matter what you want to find out.
There are likely many ways of finding out. Suppose, for example, that you want to find
out whether Abay automobile is a good car for you. You could buy one and find out that
way. You could talk to Abay automobile owners, or talk to people who considered buying
one but didn't buy any. You could read newspapers or consumer magazines evaluating
Abay automobile. The same situation occurs in scientific inquiry. That is, scientific inquiry
comes down into making observations and interpreting what you have observed. Before
you can observe and analyze, however, you need a plan. You need to determine what you
are going to observe and analyze, why and how.
Key Words: Science, Research; Exploratory, Descriptive, and Explanatory; Pure, Applied, and
Dispensing Solutions, and Research Ethics.
Chapter Objectives:
At the end of this chapter, students are expected to be able to:

- Know the definitions of science and research
- Define and discuss the purposes of research
- Define and discuss the types of research
- Examine the characteristics of research ethics
1.2 What is science?
The word science came from the Latin word “Scientia” that means knowledge. Hence,
science is an enterprise of systematic knowledge. It is sometimes characterized as logic-
empirical that is, rationality and observation of empirical facts. In other words, scientific
understanding of a concept must “make sense” and correspond with what you observe.
By: G.Egziabher H.Selasie Gebru, Assistant Professor

1
Roughly speaking, scientific theory deals with the logical aspect of science and research
deals with the observational aspect. To put it differently, a scientific theory describes the
logical relationships that could exist among parts of the world and research offers means
for seeing whether those relationships actually exist in the real world. One point of caution
is that the relationship between scientific theory and research is more complex than what is
stated here.
Practically all human beings and many lower animals exhibit a desire to predict their future
circumstances. Moreover, we seem quite willing to undertake the task of using causal and
probabilistic reasoning. We usually assume that the present conditions the future. Hence,
we seem to plan for the future based on what we presently have. In fact science makes use
of the causality and probabilistic ways of reasoning in an objective manner. What makes
science different from human casual inquiry is that it makes the reasons more explicit and
provides techniques so as to deal with them more rigorously.
1.3 What is research?
Literally, research means to “search again”. The term research has been used in so many
contexts and with such a variety of meanings. For instance, if your project analysis and
management lecturer asks you to conduct a research on “project risk analysis” you may go
to the library and select some books on project management and you may take notes
selectively on the topic assigned. Is this research? To be frank, you have taken notes but
you did not do research! Similarly, assume that the department of Accounting and Finance
requires you to do a research paper as a pre-requisite for your graduation. If you select a
senior essay topic “internal auditing in Mekelle University” and you simply describe the
auditing system by taking notes from auditing manual of Mekelle University. Is this
research? This is not research rather it is information gathering. In short, research is not
just information gathering neither it is rearranging facts. However, some writes such as
Emory and Cooper (1991: p14) consider information gathering as research.

2
What is research then? Research is a quest driven by a specific question that needs an
answer. Research can be defined in terms of the component parts that can be decomposed
into nine parts as follows:
1. Establishing a research question,

2. Finding background information,
3. Formulating a research design,
4. Data collection,
5. Analyzing the data,
6. Establishing conclusion,
7. Looking for areas of further research,
8. Stating the values associated with the research knowledge, and
9. Publishing the research work for others to view.
1.4 Purposes of research
In general, research has three purposes namely; exploration, description, and explanation.
Although a given study can have a mix of two or more purposes, it will be useful to
separately examine each purpose of research. Research has exploratory purpose if you are
examining a relatively new and unstudied research topic. Exploratory studies are most
typically done for three purposes:
1. To satisfy your curiosity and a desire for better understanding

2. To test the feasibility of undertaking a more careful study.
3. To develop the methods to be employed in a more careful study
An example of exploratory research is a research paper on “Home accountants: exploring

their practices” by Northcott and Doolin (2000).
Example 1: Northcott and Doolin (2000)
In this paper, the writers explored accounting practice at home. The researchers undertook
ten cases based on interviews with both accountants and lay people accordingly, four broad

3
areas of accounting practice are identified. These are budgeting, recordkeeping, decision
making and long-term financial planning. The findings suggest that simple accounting
practices are home practiced in order to serve multiple purposes. Domestic cash budgets,
financial records and decision-making rules of thumb offer a valued sense of financial
control and security. Home accounts are also drawn on in rationalizing financial choices,
and accounting practices reinforce a personal sense of identity. This study revealed little
difference between the practices of accountants and those of lay people, suggesting that the
“mind-set” and practices of home accountants may be attributed to factors other than
command over technique. Any “theories” of home accounting and prescriptions for
practice may have limited relevance, therefore, if they fail to account for the particular
characteristics of the home as an accounting domain.
Descriptive research aims only at describing a situation or a set of circumstances. It tries to

answer questions such as how many, how, and what is happening. The researcher observes
and then describes what was observed. The researcher’s observation is not casual. Since
scientific observation is careful and deliberate, scientific descriptions are typically more
accurate and precise than casual descriptions.
Explanatory research goes beyond description. It examines different aspects of a described

situation or an event simultaneously. An example of explanatory research is a research
paper by the author of this teaching material on “determinants of public employees’
performance: evidence from Ethiopian public organizations”.
Example 2: Gebregziabher (2009a)
The writer explained the determinants of employees’ performance given Business Process
Re-engineering (BPR) program. Utilizing primary data collected from 100 staff
departments’ non-managerial work units from seven public organizations that have
adopted BPR program, employees’ performance was regressed on motivation, ability, role
perception and certain control variables. The results of the ordered logit model estimated
using STATA routine indicate that ability has a statistically zero effect on employees’
performance. On the other hand, motivation has positive non-linear relationship with the

4
probability of having good performance. Similarly, Role perception is found to be strongly

positively related to the likelihood of good performance. These findings were then
compared to related results in the literature on employees’ performance.
1.5 Types of research
Generally, there are three types of research: pure, applied and dispensing solutions. Pure
research is concerned with developing valid, complete, and coherent descriptions and
explanations. It is interested in organizing data into the most general and parsimonious
laws or qualified statements of uniformity. The emphasis is on comprehension or
understanding. It is motivated by curiosity and inquisitiveness about natural phenomena.
Typical example of pure research in accounting is “triple entry accounting” by Ijiri, Y.
(1988). This research borrowed the concepts of momentum, impulse, and force from
Newtonian physics and tried to up grade the traditional time-invariant financial statements
into dynamic financial statements that incorporate “time” as additional accounting
dimension.
Applied research is conducted to solve socio-economic or industrial problems; and thus

have policy implication. When applied research is followed by actions of decision or
policy makers it may be referred to as action research. There is also modification of
applied research that is evaluation research – focuses on evaluating efficiency and
effectiveness of specific decision units such as business entities, departments, work units
etc. Applied research is concerned with the discovery of solutions to practical problems
and places its emphasis upon those factual data, which have more immediate utility or
application. It is like learning phrases in a foreign language without really understanding
the whole language.
The third type of research is called dispensing solutions. Practitioners are concerned with
the direct application of principles and theories from one or more fields of business for the
purpose of dispensing solutions to individual human problems rather than being concerned
with the discovery and organization of knowledge. Strictly speaking, a practitioner such as
a bank teller is not a scientist, but that is not to say they are necessarily unscientific.

5
Practitioning is like memorizing sounds of a song in a foreign language without necessarily

knowing the language. While a practitioner may uncover a phenomenon of great
importance to the understanding of nature, that is not their primary focus.
1.6 Research Ethics
Most assuredly, there is a philosophical basis to research ethics, but instead, we shall focus
on the practical considerations. In fact, all researches are inherently unethical to some
degree. There is also no such thing as totally harmless research. At least research subjects
are going to be harmed either psychologically, socially, physically, or economically. This
is because, the researcher has to invade her subjects’ privacy in order to get any useful
information and this is psychological harm. Subjects are also socially and physically
harmed because the researcher has to take their time for data collection purposes. Of
course, it is not uncommon for research subjects to be at an economic disadvantage owing
to the fact that researchers usually exploit them by not remunerating their contribution. The
researcher becomes famous by writing a book about them, but they will always remain
research subjects.
Research ethics is particularly important for accounting and finance research because
financial data are mostly competitive. Therefore, you need to exercise utmost due care in
unanimously using financial data only for your research.
Chapter Summary:
- Science is an enterprise dedicated to finding out something. It is Logic-Empirical.

- Re-search is to search something again
- Exploratory researches are preliminary researches in relatively new research areas
- Descriptive researches answer questions such as how many, how, and what
- Explanatory research draws conclusion based on premises from descriptive research
- Pure researches deal with theoretical understanding of a research topic
- Applied researches deal with empirical understanding of a research topic
- No research is inherently ethical because researchers can not be value-free

6
The Research Design Mekelle University, CBE
CHAPTER 2
THE RESEARCH DESIGN
2.1 Introduction
The preceding chapter should have given you a broader picture of what research is all
about. This chapter presents the “blue print” of research, that is, research design. In
designing the research, the researcher acts like an architect. An architect prepares blue
print of a building. The blue print is given to a civil engineer in order to construct an
actual building. Likewise, the researcher prepares a research design before the actual
research is done. This chapter presents the components of a research design.
Key Words: Research Design, Research Process, Research Design Dimensions, Research
Problem, Hypothesis, Data and Method.
Chapter Objectives:
At the end of the chapter, students are expected to be able to:

- Create analogy between a researcher and an architect in the process of designing
- Know and differentiate the different types of research designs
- Develop art of visualizing the research process before the actual research starts
- Develop the art of mentally designing research problem for any research topic
- Produce research hypothesis for any research problem except exploratory ones
- Produce research question for any exploratory research problem
- Design the data and method in order to test a hypothesis
2.2 What is research design?
Research design is the plan and structure of investigation so conceived as to obtain

answers to research questions. The plan is an overall scheme or program of the research
process. It includes an outline of what the researcher will do starting from statement of
the problem and hypotheses formulation to the final analysis of data.

7
Figure 1: Schematic Presentation of the research process
Own or Others’
Scientific Literature Theory Experience
Population and
Conceptualization Operationalization Choice of Method
Sampling
Observation or Data
Collection
Data Analysis
Drawing Conclusion
Source: Adapted from Babbie (1986)
Figure 1 presents holistic view of the research process. Research design starts from
asking research questions. Not all questions are research questions though. For instance, a
question whose answer is known beforehand is not research question. Once a research
question is clearly defined in light to literature, the researcher has to plan for how the
actual research will be done. The how aspect of a research design has to clearly answer
questions pertaining to: 1) conceptualization and operationalization of key concepts in the
research, 2) the research method to be used and 3) the data source and research subjects
to whom conclusion is to be inferred. Besides, it is not uncommon for research designs to

8
include the expected outcomes and possible applications or possible value addition of the
research.
2.3 Types of research designs
Types of research designs are analogous to types of “blue prints”. An architect can
choose among different types of possible building designs depending on factors such as
purpose of the building, method of construction, time of construction and the like.
Similarly, researchers choose among different types of possible research designs
depending on such dimensions as purpose of the research, method of data collection,
control of variables, the time dimension, the research environment, depth of the research,
and the like.
Purpose of research is the first dimension of research design. As explained in the first
chapter of this teaching material, research has purpose of exploration, description or
explanation or a combination of them. Therefore, based on the “purpose of research”
dimension, there are at least three classes of research designs: exploratory, descriptive
and explanatory. Change of purpose, say from exploration to explanation, changes all
components of the research process in Figure 1.
Method of data collection is the second dimension resulting in new classes of research
designs. In chapter 4 of this teaching material, you are going to learn about the different
methods of data collection. Based on the level of involvement of the data collector we
may orderly classify data collection methods into participant observation, unstructured
interview, focus group discussion, non-participant observation, and mail questionnaire.
The level of data collector involvement decreases as you go on from participant
observation to mail questionnaire or survey. Hence, method of data collection generally
ranges between observation and survey. It is important to note that a combination of
purposes of research and method of data collection could result in new breeds of research
designs.
Control of variables is a third dimension that could result in additional classes of research
designs. As you are going to see later in this chapter, variables are operationalization of

9
concepts in your research. For example, wear and tear of a fixed asset can be considered
as a variable that operationalizes the concept of depreciation. Similarly, net income is a
variable that operationalizes the concept of wealth. You may want to do research on “the
relationship between depreciation and wealth” in order to come up with a generalized
statement or proposition regarding depreciation and wealth of a company. But you know
that depreciation is not the only factor that determines wealth. Can you put a boundary on
the part of wealth that is totally determined by depreciation? If you can then you have
control over your variable. Therefore, you may conduct an experiment. Particularly in
social sciences, it is very difficult to put clear boundary on a variable hence it is difficult
to conduct experiments. The only option you have is to observe past data about
depreciation of different companies and relate it with their past data about wealth.
Research design that plans to use past data, instead of experimental data, is known as ex
post design. In short, based on control over variables we differentiate research designs
between experimental and ex post designs.
The time dimension of the research is the fourth dimension of research design. You may
plan to collect data across geographical space, that is, from respondents located in
different geographical locations or you may plan to collect data across time. Data that
involve respondents from different geographical locations as of a single point of time are
known as cross-sectional data. For example, a research on “measuring relative efficiency
of Ethiopian banks” is cross-sectional study because it does not require observation or
survey across time periods. Data that involve a respondent in one geographical location
across different time periods are known as longitudinal data. For example, a research on
“On the dynamics of share holders’ wealth in Awash International Bank” is longitudinal
study because it requires observation or survey across time. In short, we classify research
design based on time dimension as cross-sectional and longitudinal designs.
Research environment is fifth dimension of research design. There are at least two
possible ways of getting data for your study. You may collect actual or real data from the
field, say, by interviewing respondents or you may also generate artificial data using
simulation. Simulation is a statistical method of generating artificial data based up on
statistical parameters such as mean and standard deviation and the type of distribution.

10
Based on the research environment dimension, we classify research designs into field and
simulation.
Last but not least, depth of research is the sixth dimension of research design. The depth
and breadth of research are on a constant tradeoff. If a research has to increase its depth it
usually gives up some part of its breadth. Research projects that emphasize on depth of
analysis are usually aimed at particularization. Whereas, research projects that emphasize
on breadth of analysis are usually aimed at generalization. Qualitative research designs
are usually meant for researches that require depth instead of breadth and quantitative
research designs are commonly used in research projects aiming at generalization.
Table 1: Dimensions of research designs
Purpose of the Method of Control of Time Research Depth of

Dimensions of the Research Data Variables Dimension Environ Research
Research Collection
Exy Des Exp Obs Sur Exl Pos Cro Lon Fie Sim Qnt Qul
Exploratory (Exy)
Descriptive (Des)
Explanatory (Exp)
Observation (Obs)
Survey (Sur)
Experimental (Exl)
Ex post (Pos)
Cross-sectional (Cro)
Longitudinal (Lon)
Field (Fie)
Simulation (Sim)
Quantitative (Qnt)
Qualitative (Qul)
Table 1 presents different types of research designs. New research designs can be formed
by combining the 13 different types of research designs in Table 1. For instance,
exploratory research design could be formed as a combination of observation method of
data collection, ex post design, cross-sectional time dimension, field research design and
qualitative analysis. Likewise, varieties of research designs can be identified by
combining the different research designs in Table 1. Given 13 alternatives among which
we have to select 6 (i.e. one for each dimension like the shaded boxes in Table 1), there
are 1,716 (i.e. 13 C 6 = 13!/(7!6!)) possible number of research designs. You need to

11
choose one of the 1,716 possible research designs and state your justification why you
did not select the rest possible designs.
Due to wide applicability in accounting and finance research, survey and case study
research designs are presented in chapter five and chapter six of this teaching material.
2.4 Designing the research problem
Research problem formulation is one of the most critical steps in scientific research. A
research problem is part of a research process that reflects a gap in understanding.
Research problem statement is usually written as an introduction part of a scientific
research paper. In Ethiopia, most student research papers present statement of the
problem in a separate section. In general, research problems can have three sources: 1)
own experience, 2) scientific literature, and 3) theories. First, researchers identify
research problems from personal observation or discussion with experts and stakeholders.
In order to state the identified research problem, it is common practice for researchers to
undertake literature review related to the personal observation. Secondly, research
problems can be identified from scientific literature. Researchers read findings of
published research papers and notice that some issues are not covered. This could end up
identifying a research problem. Thirdly, limitations of standard theories could lead to
research problem identification.
Example 1 is an introduction part of the paper titled “Facilitating total quality through
effective project management” by Hides et al. (2000). This illustration does reflect the
gap in utilizing potential benefits of Total Quality Management (TQM). Research
objective and organization of the research paper are presented towards end of the
introduction part.
Example 1: Hides et al. (2000)
The move towards total quality management (TQM) is usually undertaken as a company-
wide initiative with a timescale of many years, and as such is undertaken with great
caution. New products are an essential element of a company's strategy (Kagioglou,

12
1997) and consequently, there are few organizations that do not have new product
development projects in progress. Such timescales are relatively short, sometimes only
months, as companies must respond to customers' needs (Wheelwright and Clark, 1992).
Research (Hides, 1998) has been conducted that examines new product development and
uses the information from a project to improve successive work. The research approach
has been to introduce tools as and when they are required rather than try to implement a
company-wide total quality/continuous improvement initiative. Working with small
cross-functional teams, each new product project has advanced both new products and
the ability of the organization (albeit in a small way).
A survey of small and medium enterprises (SME) in the UK suggests that SMEs fail to
realize the “full” potential of TQM, despite their commitment to this (Parkin and Parkin,
1996). They continue by suggesting that this may be due to SMEs' inability to translate
management strategies into employee actions. Pitsika (1995) suggests that although small
businesses are economically disadvantaged against bigger organizations, they still have
some very important advantages over their bigger competitors. Jennings and Beaver
(1997) conclude that for a small enterprise to grow and develop, greater control and more
“formal management process” are needed. Many researchers, however, point out that the
only way for an SME to become a quality organization and indeed shift to TQM is
through a transition from ISO 9000 (DTI, 1996) or any other quality assurance standards
system (Williams, 1997; Mo and Chan, 1997).
Taylor and Meegan (1997) offer a framework for senior executives who wish to
undertake the transition of a company from ISO9000 to TQM. Bryde (1997) discusses the
journey towards quality project management and examines the application of TQM
principles to the project management process.
The aim of the paper is to examine the use of projects as a means for adopting total
quality principles. The first part of the paper introduces a framework for senior
executives undertaking the transition from ISO 9000 to TQM, which has previously been
proposed by Taylor and Meegan (1997). The second part of the paper compares the
Taylor and Meegan (1997) framework with a case study company to determine if the

13
theory can be translated to projects that are more general (than the specific quality
management transition). The final part of the paper offers recommendations for further
work.
2.5 Designing the research hypothesis
Hypothesis is a tentative explanation of observable phenomena. Hypothesis states the

researcher's expectations concerning relationship among variables in the research
problem. It can also be asserted that hypothesis states what the researcher thinks the
outcome of the study would be. In short, hypothesis is a refinement of the research
problem.
2.5.1 Hypothesis and Logic
Hypotheses are results of logical reasoning. In logic there are three types of reasoning:
deduction, induction, and abduction. Deduction is determining the conclusion given the
rules and preconditions. For instance, if corporate tax increases then weighted average
cost of capital (WACC) declines; corporate tax increases, therefore WACC declined.
Mathematicians are accustomed to deductive reasoning. Inductive reasoning is about
determining the rule given the conclusion and precondition. For instance, WACC
declined every time corporate tax has increased, therefore WACC declines when
corporate tax increases. Scientists are used to inductive reasoning. Abductive reasoning is
about determining the precondition given the rules and the conclusion. For instance,
when corporate tax increases WACC decreases; WACC decreased therefore corporate tax
might have increased. Detectives are accustomed to abductive reasoning.
2.5.2 Hypothesis, Constructs and Variables
One has to understand the nature of concepts, constructs, and variables in order to grasp
the logic behind research hypothesis design. Emory and Cooper (1991) defined concept
as “a bundle of meanings or characteristics associated with certain events, objects,
conditions, situations, and the like.” For instance, “depreciation” is an accounting concept

14
that can be objectively visualized as a wear and tear of fixed assets whereas “profit” is
beyond concept since it can not be objectively visualized. Concepts that can not be
objectively visualized are called constructs. For example, knowledge of accounting is
made up of basic constructs as business entity, going concern, periodicity, debits, credits
and the like. Accounting and finance theories are made up of complex relationships
among constructs. The relationship is at a higher level of abstraction as it is in
mathematical methods. Any research in accounting and finance dealing with structural
relationship among constructs at higher level of abstraction is deemed to be theoretical
research.
Variables are operational representation of constructs. That is, constructs need to be

further conceptualized to get equivalent concepts and further operationalized to get
equivalent variables. Hypothesis is the relationship between two variables. This
relationship is at a lower level of abstraction. Any research in accounting and finance
dealing with substantive relationship among variables at lower level of abstraction is
deemed to be empirical research.
2.5.3 Types of hypothesis
In general there are two types of hypotheses: descriptive and relational. Descriptive
hypothesis deals with tentative conclusion about one case. For example, “In Ethiopia,
power generation investment is increasing.” In this example, Ethiopia is the case and
investment is the variable. Descriptive hypothesis could be formulated to test
“truthfulness” of rules, conclusion, or preconditions in deductive, inductive or abductive
reasoning. Researchers prefer to use research question instead of descriptive hypothesis.
On the other hand, relational hypothesis deals with correlational or causal relationship
between two variables with respect to some case. For example, “In Ethiopia, increased
investment in power sector leads to increased economic growth”. In this example,
Ethiopia is the case and investment and economic growth are the two variables. Usage of
the term case is not the same as it is in case study. Instead, it refers to the unit on which
you are interested to conduct research or shortly the unit of analysis.

15
2.5.4 Hypothesis and Research Problem
In practical sense, research hypothesis formulation starts with a well stated research
problem. Further, researchers usually review related literature so as to better understand
the research topic of interest. Prior research works usually explain the important variables
in the research topic and different forms of relationships among the variables. For
instance, example 2 presents an excerpt from a research paper titled “Impact of warrant
introductions on the behavior of underlying stocks: Australian Evidence” by Aitken and
Segara (2005). It illustrates hypothesis formulation. The writers identified stock return
volatility and stock warrant introduction as key variables and hypothesized on the
relationship between the variables in question. From example 2, it is very important to
note that the hypothesis and the research topic are very much related.
Example 2: Aitken and Segara (2005)
The introduction of a derivative can enable a more efficient risk allocation across
securities and investors, resulting in an increased demand for the underlying stock with a
consequent decline in the spot price volatility. This idea is established by Detemple and
Selden (1991) in a detailed model of the interaction between the spot and derivative asset
markets. In sharp contrast, Ma and Rao (1988) demonstrate that stable stocks become
more volatile after listing because of the increased speculation in the derivatives market
from informed traders. A more recent view expressed in the published option
introduction literature is that as the option listing decision is endogenous, exchanges and
regulators will more likely list options on large and risky stocks (see Mayhew and Mihov,
2000; Bollen, 1998). The endogeneity of the option listing decision is also applicable to
warrant listings. Given the profit motives of warrant issuers and their ability to time when
to issue derivative warrants, a more favorable warrant premium will be secured during a
period when the underlying stock is anticipated to be more volatile and declining in
value. This view gives rise to the final hypothesis.
H3: Underlying stock return volatility should be elevated following warrant introduction.

16
Not all researches have research hypothesis though. For instance, exploratory researches
use research question instead of hypothesis. This is because, exploratory researches lack
sufficient prior researches. Similarly, descriptive researches commonly use research
question instead of hypothesis. In effect, hypothesis design is most common in
explanatory researches. Similarly, the use of research hypothesis depends on the
philosophy behind the research. If the research is designed from positivism (i.e. reality is
objective and independent) point of view then research hypothesis is required.
2.5.5 Null versus alternative hypothesis
Accounting and finance researches are accustomed to relational hypotheses that are
subject to statistical tests. In relational hypothesis two competing hypotheses are
formulated namely null hypothesis and alternative hypothesis. The null hypothesis is a
default position. It hypothesizes that there is no relationship between two variables. It is
usually labeled as H0. If the collected data could not contradict the null hypothesis for
more than 95% of the cases then the researcher do not have sufficient reason to conclude
that the null hypothesis is “false”. To put it differently, if the collected data contradicts
the null hypothesis for only less than 5% of the cases then the contradiction is considered
insignificant hence the null hypothesis is accepted as “true”. Alternative hypothesis, as
the name implies, is alternative to the null hypothesis. If the actually collected data
contradict the null hypothesis for more than 5% of the cases then the null hypothesis is
considered as “false” and hence rejected. If the null hypothesis is rejected then the
alternative hypothesis is accepted.
2.5.6 Criteria of Good Hypothesis
A hypothesis is considered as good if it is testable, simple, fruitful, literature driven, and

of wider scope of application. If it is difficult to test whether the hypothesis can be
falsified or supported by empirical data then the hypothesis is not good. Simplicity refers
to using fewest numbers of assumptions and postulates in formulating a hypothesis. A
hypothesis is considered fruitful if it can explain phenomena beyond statement of the
problem. It is considered as literature driven hypothesis if it follows prior research works

17
in the area. Last but not least, a hypothesis is good if it can be applied in multiple cases of
phenomena.
2.6 Designing the data and method
The data and method part of the research design is meant to plan for a hypothesis test.
Design of data and method in effect requires choice of research design in each of the
design dimensions in table 1. For instance, a researcher could decide to undertake a
research on “dynamics of foreign direct investment in Ethiopia” with explanatory
purpose, survey method of data collection, ex post control over variables, longitudinal
time dimension, field research environment, and wider breadth of research design. In fact,
if the researcher is time and budget constrained, as it is for graduating students, a bit
different research design could be adopted combining explanatory purpose, data
collection from secondary sources, ex post control over variables, longitudinal time
dimension, simulation research environment, and wider breadth of research design.
Table 2: Alternative designs for the research “dynamics of foreign direct investment in Ethiopia”
Dimension Research Design 1 Research Design 2 Research Design 3

Purpose of research Explanatory Explanatory Explanatory
Method of data collection Survey Secondary Source Secondary Source
Control over variables Ex post Ex post Ex post
Time dimension Longitudinal Longitudinal Longitudinal
Research environment Field Simulation Field
Depth of the research Quantitative Quantitative Quantitative
In table 2, if the researcher decides to adopt research design 1 then detailed survey
questionnaire has to be designed so as to collect data from the investors themselves. If the
researcher decides research design 2 then she has to be accustomed to simulation where
as research design 3 requires the researcher to look for data source such as the Ethiopian
Central Statistical Authority (CSA) or Ethiopian ministry of Finance and Economic
Development. Whenever possible the researcher has to enumerate the possible alternative
research designs such as in table 2 and justify why one is selected while the rest options
are not. If the research design selected is new to the research topic of interest, the

18
researcher has to convince readers that the selected research design will be more
appropriate.
The data and method part of a research design is very important because it tells to the
research committee or to the thesis supervisor how the researcher plans to tackle the
research problem. The guiding principle for writing the data source and method section is
that it should contain sufficient information for the reader to determine whether the
methodology is sound.
Chapter Summary:
- A research design looks the future research as architecture looks the future building
- A research design helps you to vividly see the whole research process ex ante
- Research design starts with topic selection and statement of the problem
- Hypothesis is a refined form of the problem statement
- Logical reasoning can be classified into deductive, inductive, and abductive
- Construct is type of concept at a higher level of abstraction (e.g. Poverty)
- Concepts are those in an intermediate level of abstraction (e.g. depreciation)
- Variables are operational representations of constructs and concepts (e.g. experience)
- Relationships among any two constructs is known as proposition
- Relationships among any two variables is known as hypothesis
- Data and method design is set of procedures to be followed to test hypothesis

19
The Research Proposal Mekelle University, CBE
CHAPTER 3
THE RESEARCH PROPOSAL
3.1 Introduction
Research proposal is a written document of research plan meant to convince specific

readers whereas research design deals with choosing a specific holistic research design
from among all possible research designs to address the research topic. Hence, it can be
claimed that a research proposal presumes availability of the research design. A research
proposal is usually required when the research project is to be commissioned and the
researcher is expected to compete with other researchers to get research fund or else
when the research project is a requirement for partial fulfillment of an academic degree
such as BSc, MSc or PhD. For example, a senior essay research proposal is intended to
convince your supervisor that your senior essay is a worthwhile research project and that
you have the competence and the work plan to complete it.
Key Words: Research Proposal, Theoretical Framework, Problem Statement, Conceptual

Framework, Literature Review, Referencing Style, Funding Source, Proposal Evaluation.
Chapter Objectives:

- Know the different parts of a research proposal
- Develop theoretical and conceptual frameworks and hypotheses thereof
- Review related theoretical and empirical literature using APA Style of referencing
- Know different websites of funding sources
- Prepare a full scale research proposal for funding purposes
- Evaluate and criticize research proposals
3.2 Types of research proposal
Research proposals can be generally classified into student, internal and external. Student
research proposals range between term paper and senior essay that are least complex to

20
MSc thesis and PhD dissertation that are most complex. The form and content of student
research proposal differs according to complexity of the student research. For instance,
term papers are meant for teaching students as to how to do research where as senior
essay, thesis or dissertation are supposed to develop students’ knowledge on a specialized
field. Research proposals are internal if they target an entity or organization and are
prepared by individual or a team of researchers who are part of the entity or organization.
For instance, academic and administrative staff research proposals of Mekelle University
addressing efficiency and effectiveness of the university are internal proposals. The
organization or entity usually provide budget for large scale internal research projects
even though it is also not uncommon for the researcher to self-sponsor small scale
internal research projects.
Research proposals that are not either in the student or internal proposal categories are
external research proposals. External research proposals are either solicited or
unsolicited. Solicited research proposals are when the customer asks for a proposal. For
instance, students that look for masters scholarships abroad are expected to sell
themselves by writing best research proposal according to the specific guidelines in the
call for scholarship. In fact research proposal is usually a mandatory requirement to get
PhD scholarship. Likewise, large scale research projects such as the NORAD of Norway,
the IUC of Belgium, and the NUFU of the Netherlands call for research proposals for
funding. Hence research proposals that are addressed to NORAD, IUC, NUFU etc funds
are all solicited external proposals.
3.3 Sections of a research proposal
Section of a research proposal varies from one type of research proposal to the other. In
addition, for practical reasons many research funding agencies prefer their own research
proposal format. For instance, Mekelle University has its own research proposal
guideline based on which staff research proposals of the university are assessed.
The most common sections of a large scale research proposal are: the title, abstract,
introduction, literature review, importance/benefits of the study, data and method, method

21
of data analysis, qualification of the researcher, facilities and other resources, budget,
time schedule, project management, bibliography, and appendix. For a small scale
research proposal, the most common sections could be: the title, introduction, literature
review, data and method, importance/benefits of the study, cost and time budget, and
bibliography.
3.3.1 Title
Title or topic of a research proposal tells what the research proposal is all about. It should
be catchy, concise and descriptive. Title selection presumes extensive literature review
related to the area of research interest. For example, the research topic on “Determinants
of Private Investment in Tigray Regional State” is good because it is concise and at the
same time contains the three basic elements of a topic: the thing that is going to be
explained (or explanandum), the thing that explains the explanandum (or explanan) and a
geographical scope. The explanandum of the aforementioned topic is private investment
because your research is expected to draw conclusions pertaining to private investment.
The explanans are captured by the word determinants. The actual factors that determine
private investment are not stated in the topic because it has to be very concise. Hence, the
key word “determinants” is used. And finally, the phrase “Tigray Regional State” puts a
geographical delimitation of the proposed research. Therefore, whenever you design your
topic it is better to think in terms of the most important explanan and explanandum and
delimitation of the proposed research.
3.3.2 Abstract
Abstract is a brief summary of approximately 300 words. It should include the research
question, the rationale for the study, the hypothesis (if any), the method, and expected
value addition. Abstract is usually done after you finish your research proposal. The basic
purpose of an abstract is to market the research idea especially when you require funding
for the research project. Funding organizations accept hundreds of research proposals
from different corners of the world. Similarly, senior essay supervisors at your university
usually supervise a number of students. Hence do not expect that the busy proposal

22
reviewer will be reading the whole contents of your research proposal instead she reads
the abstract part.
Therefore, if you have done excellent research proposal but you carelessly have written
the abstract then your research proposal is going to be rejected. This implies that you
need to carefully and elegantly write the abstract of your research proposal. Example 3 is
the abstract section of a research proposal titled “Small and Micro Enterprises financing
preference in Tigray region: Does POH hold?” prepared by the author of this teaching
material.
Example 3: Gebregziabher (2009b)
This paper aims to investigate the determinants of financing preferences of micro and
small enterprises (MSEs) owners in Tigray regional state of Ethiopia in light of the
Pecking Order Hypothesis (POH). Structured questionnaires will be used to collect data
from 120 MSEs in six zonal towns of Tigray regional state. Logistic regression model
will be used to empirically test the literature-driven hypotheses.
The abstract in example 3 presents that purpose of the research will be explanation,
method of data collection will be survey, the researcher will have ex post control of
variables, time dimension of the research will be cross-sectional, the research will be
carried on a field environment, and the research will have breadth than depth (i.e. it will
be quantitative research). It is worthwhile to note that good abstract of a research
proposal contains all dimensions of the research design.
3.3.3 Introduction
The introduction part summarizes the whole content of your research proposal in a
logical way that chains theoretical framework, problem statement, conceptual framework,
hypothesis or research question, and objective of the study. Theoretical framework is
derived from the scientific paradigm1 your research topic is preferably in. That is, a
1
Scientific Paradigm is more general than Theoretical Framework, because, various Theoretical Frameworks can be
developed with in one Scientific Paradigm.

23
scientific paradigm helps you design an “eye-glasses” for your thought based on which
you would see the natural and social worlds. Theoretical framework is the summarized
form of your theoretical literature review. It draws a boundary within which the research
problem is stated. In general, it serves as a background or context for your research
problem and in particular, it serves the following purposes; 1) it helps you to clearly see
the most important concepts of your research, 2) it can provide you with a general
framework for data analysis, and 3) it is essential in preparing research proposals that
have descriptive or explanatory purpose. In exploratory researches little is known about
the research topic hence you may not have theoretical framework.
Statement of the problem part of your research proposal reflects the gap and justifies that
the issue is worth researching. The problem statement is usually general, encompassing
all of the most important concepts in the topic. Hence, statement of the problem has to be
more focused making use of conceptual framework - an operationalization of the
theoretical framework. Conceptual framework is the summarized form of your empirical
literature review. It reflects your own position on the problem and it guides you how to
analyze your research problem.
The difference between theoretical and conceptual frameworks lies on the degree of
abstraction of the concepts among which interrelationships are to be created. The
theoretical framework creates interrelationship among higher level concepts or constructs
thereby forming a set of propositions and the conceptual framework creates
interrelationship among lower level concepts or variables thereby forming a set of
hypotheses. The conceptual framework needs to be summarized using hypothesis or
research question. The relationship among theoretical framework, statement of the
problem, conceptual framework, and hypothesis or research question is depicted in
Figure 1.
Once you optimally focus your research problem until a hypothesis or a research question
it is right time to write about objective of the study. A research problem may not be fully
captured by one hypothesis or one research question. In this case, your research proposal
may have one general objective and a set of specific objectives. The last paragraph of

24
your introduction has to summarize the problem you are addressing in your planned
research, that is, sell your point to the reader.
Figure 1: Structure of the introduction part

Operationalization
Operationalization
The Introduction part of a research proposal usually contains 300 to 500 words but the
number of words could be higher if your literature review is merged with the introduction
part. Example 4 is the introduction part of the research proposal on “Small and Micro
Enterprises financing preference in Tigray region: Does POH hold?” It illustrates as to
how to write the introduction part (consisting of background of the problem, statement of
the problem, and objective of the study) of a research proposal. On the way, it also
illustrates how to review the literature.
Currently, micro and small enterprises (MSEs) are recognized as socio-economic and
political development catalysts in both developed and developing economies (Hussain et
al., 2006; Gregory et al., 2005; Fielden et al., 2006). Literature in the area show that,
regardless of size, access to finance is one of the key impediments of MSEs’
development (Gregory et al., 2005; Van Auken, 2005). Compared to large firms, MSEs
face a relative disadvantage to raise finance from formal institutions such as banks.
Tucker and Lean (2003) articulated the extant knowledge on this matter as follows:

25
Smallest firms generally have smaller financial reserves geared compared to larger firms due to
the difficulty and expense of attracting new equity finance. Thus, such firms do not only bear
higher business risk but also higher financial distress risk. Banks tend to respond to this risk by
adopting a capital-gearing rather than an income-gearing approach to lending. Thus, rather than
focusing their attention on evaluating income streams flowing from an investment project they
may focus more on the value of collateral available in the event of financial distress. This creates
a problem for small firms in that they often do not have significant fixed assets to secure on in
their early years of establishment. The stage of development then may be an important
determinant of, and constraint on, the type and amount of external fund.
Particularly in developing countries like Ethiopia, where the financial market is weakly
efficient, there is high information asymmetry between financial institutions and MSEs.
As a result, financial institutions face severe problems of adverse selection. That is why
access to formal finance is usually difficult for MSEs.
Johnsen and McMahon (2005) summarized five competent theories of MSEs financing;
namely, Static Trade-off Theory, Agency Theory, Growth Cycle Theory, Alternative
Resource (or Bootstrapping) explanations and Pecking Order Theory. According to Static
Trade-off Theory, firms prefer external financing scheme to the extent that the marginal
benefit due to tax shield advantage of debt is exactly equal to the marginal cost of
financial distress (Ross et al., 2000, p. 504). This theory assumes that all factors are
assumed to be fixed but the debt-equity ratio. In this theory, value of the firm and total
debt are non-linearly related (Ross et al., 2000). Johnsen and McMahon (2005) stated that
other factors held constant firms with more intangible assets need to borrow less,
compared with firms with more tangible assets, because of the collateral factor.
The Static Trade-off Theory has been more important and sensitive area of research for
academics in that practicing managers could not follow as much (Silveira and Slack,
2001). Particularly for MSEs, the importance of static trade-off is expected to be low
because the theory requires practicing MSE owners to have financial sophistication, but
customarily MSE owners are rather product specialists. On the other hand, an application
of firm value optimization technique under static trade-off theory demands substantial

26
reliable data. However, MSEs usually have weakly organized accounting system adding
complexity to data generation problems.
Agency Theory (Jenson and Meckling, 1976) explains the principal-agent relationship
between equity holders and debt holders. In a principal-agent framework, the MSE is the
agent and the finance provider is the principal. This theory asserts that principals have
higher agency costs because equity-controlled firms have a tendency to invest sub
optimally to expropriate wealth from debt holders (Jordan et al., 1998) that in turn results
in incremental risk for the principal. That is, MSEs owners invest on high risk projects
that the cost of debt should have been higher to share on the marginal return on
investment. Johnsen and McMahon (2005) also pointed that firms experience higher
agency cost due to the fact that equity-controlled firms have a tendency to sub optimal
investment.
Tucker and Lean (2003) showed that MSE owners might fail to perform to their full
capabilities giving rise to the problem of moral hazard. This problem prevails in MSEs
due to higher degree of information asymmetry between financial institutions and the
owners of MSEs; that is why the former “regulate” performance of their loans by
designing “collateral incentives”. This incentive mechanism shapes the financing
behavior of the MSEs owners particularly for MSEs organized as sole proprietorship and
partnership forms of business. The issue of “collateral incentive” necessitates marginal
risk that is applicable on the personal assets of the MSE owners. Hence, MSE owners
sub-optimally finance their business through debt, other things being constant. They
forgo additional value that could be generated through financial leverage that can be
treated as a measure of “liability risk”. Therefore, a given financing theory needs to
embrace the “games” played between the agent, the firm, and the principal, the finance
provider.
Berger and Udell (1998) initiated the Growth Cycle Theory of small business financing.
This theory illustrates the dynamic financial needs, as the small business becomes more
experienced and enhanced informational transparency. In this theory, the firm gets better
access to venture capital as a source of equity and midterm-loans as a source of debt

27
(Gregory et al., 2005); yet, as the firm gets older and information-wise transparent it
tends to have better access to public equity and long-term financing. Gregory et al. (2005)
found that only firm size, as measured by total employees, could significantly determine
the decision of whether to use insider financing instead of going for public equity or
long-term financing.
Van Auken (2005) defined Bootstrap Financing as a method of acquiring the use
resources without resorting to debt or equity financing. He identified 28 possible sources
of Bootstrap Financing that in turn are categorized onto five factor dimensions, namely:
delaying payments, minimize accounts receivable, minimize investment, private owner-
financing and sharing resources. Bootstrap financing claims that MSEs exhaust the
possibilities of self-financing before going for external financing sources.
Myers (1984) proposed the Pecking Order Hypothesis (POH) asserting, small firms strive
for external sources of finance only if the internal sources are found inadequate. Besides,
firms prefer to exhaust their borrowing capability before issuance of new equity due to
reasons explained by Agency Theory (Jenson and Meckling, 1976), Information
Asymmetry (Myers and Majluf, 1984) and Signaling Theory (Ross, 1997). Empirical
evidence on the applicability of the POH in MSEs shows mixed results (Paul et al.,
2007). MSE owners are generally concerned with level of intrusion instead of optimizing
their firm value choosing the “right” debt-equity ratio (Hamilton and Fox, 1998). They
try to meet their finance needs with a pecking order of personal savings and retained
earnings, short-term borrowing, long-term debt and issuance of new equity (Hussain and
Matlay, 2007).
According to Paul et al. (2007), the POH has a relatively dominant explanation for small
business finance. Hence, in this paper, the author set out to investigate the determinant
factors of MSE owners’ financing preference in Tigray regional state of Ethiopia in light
of the POH.

28
3.3.4 Literature Review
Literature review is a body of text that aims to review the critical points of current
knowledge about your research topic. A well-structured literature review is characterized
by a logical flow of ideas, current and relevant references with consistent and appropriate
referencing style; proper use of terminology, and an unbiased and comprehensive view of
the previous research regarding your research topic.
Theoretical and empirical literature reviews
Literature review can be broadly classified into theoretical and empirical literature
review. The theoretical literature review builds the detailed theoretical framework for
your research that is an elaborated version of the one in the introduction part. This
detailed theoretical framework can be structured as: 1) present critical review of
alternative theories that can possibly help you as possible points of view to look at your
research problem, 2) make appropriate citation, 3) consult experts and add your informed
reflections on the theories, 4) illustrate your theoretical framework (i.e. interrelationships
among constructs) using a diagram, and 5) state the most important propositions that
characterize your research problem. On the other hand, you can build the detailed
conceptual framework for your research proposal in the literature review part the same
way as that of detailed theoretical framework.
The literature review serves several important functions some of which are: 1) Ensures
that you are not “reinventing the wheel”, 2) Gives credits to those who have laid the
groundwork for your research, 3) Demonstrates your knowledge of the research problem,
4) Demonstrates your understanding of the theoretical and research issues related to your
research question, 5) Shows your ability to critically evaluate relevant literature
information, 6) Convinces your reader that your proposed research will make a
significant and substantial contribution to the literature (i.e., resolving an important
theoretical issue or filling a major gap in the literature), 7) Indicates your ability to
integrate and synthesize the existing literature, and 8) Provides new theoretical insights or
develops a new model as the conceptual framework for your research.

29
Referencing Styles
The term ‘reference’ refers to those consulted materials that are actually cited in-text, that
is, when your research includes paraphrased empirical and theoretical quotations from
published papers of other writers. On the other hand, bibliography refers to all materials
consulted regardless of whether they are in-text referenced or not. There are two types of
references: To write reference bibliography, there are a number of referencing styles such
as Harvard Style, Chicago Style, American Psychological Association (APA) Style and
etcetera. In this teaching material we use American Psychological Association (APA)
style of referencing because of its predominant use in social sciences. However, you need
to remember the fact that sometimes the sponsor requires specific style of referencing.
In-Text Citation using APA style
In-text citation in the APA style requires the author and the year method. Page numbers
are required for direct quotes. In-text citations rely on signal phrases that indicate the
reader where you got your information using such verbs as: stated, noted, found etc. You
can also have in-text citation without signal phrases using phrases such as: some authors
explored, one study concluded, one writer asserted etc. If you are summarizing or
paraphrasing others’ work you need to put it within quotation. For example, you may cite
as follows: According to Gebregziabher (2009b), self-initiated startup businesses follow
the Pecking Order Hypothesis (p.333). Alternatively, the same quotation can be re-
written as; self-initiated startup businesses follow the Pecking Order Hypothesis
(Gebregziabher, 2009b, p.333).
There are two types of quotations these are short and long quotation. A direct quotation
of less than 40 words is called short quotation. For example, you may cite as follows: In
his 2009 study, Gebregziabher noted, “MSE owners that started their business based on
self-initiation exhibit a strong tendency to use internal sources before looking for external
source of finance” (p.333). Long quotation is direct quotation of more than 40 words.
Long quotation can be cited as in the second paragraph of Example 4 but page number
has to be included at the end of the indented paragraph.

30
In-text citation from a research paper of two authors should always include last names of
both authors. For example, you may cite as: Gemechu and Amare (2008, p.15)
emphasized …. If the in-text citation is from a research paper of three to five authors
simply list name of the first author and add the phrase “et.al.”. For example, you may cite
as: Alemtsehay, et.al. (2004), concluded ….
If the author is unknown, use the title of the work in the signal phrase or give the first
word or two of the title in the parenthetical citation. For example, you may cite as: one
2008 guide, “APA In-Text Citation Guidelines” stated that an in-text citation should
include author and year of publication. Alternatively, this can be re-written without signal
as follows: “In an APA style of In-Text Citation author name and year of publication are
written” (“APA In-Text Citation Guidelines”, 2008).
If the author is an organization or government agency, use the name of the organization
in your signal phrase or parenthetical citation. For example, you may cite as: according to
Mekelle University (2009), “BPR implementation resulted in improved research
performance”. Alternatively, this can be re-written without signal as follows: One study
has shown that BPR implementation resulted in improved research performance (Mekelle
University, 2009). You can also use an abbreviation in your in-text citation as follows: In
its study, MU (2009) concluded that BPR implementation resulted in improved research
performance.
If your in-text citation involves citing memos, letters, emails, interviews etc you need to
label the type of communication and the exact date on which it occurred but you need not
write it in your reference list. For example, you may cite as follows: Meseret (Personal
Interview, November 05, 2010) indicated that her study on class participation of students
verified the importance of active participation.
Example 5 is the literature review part of the research proposal on “Small and Micro
Enterprises financing preference in Tigray region: Does POH hold?” It illustrates the way
hypothesis can be formulated based upon prior researches. H1, H2, and H3 are hypothesis
number 1, 2, and 3 respectively.

31
Education versus financing preference
Since the seminal paper of Modigliani and Miller (1958) on the irrelevance of capital
structure, scholars widely argue about the relationship between financing decision and
firm value, though the stock of literature on relevance of financing decision to firm value
greatly dominated the other. Firm value optimization with “right” debt-equity mix
requires financial sophistication but substantial MSEs owners usually lack adequate
knowledge on financial sophistication, as they are often product or service specialists, not
specialists in the area of finance (Tucker and Lean, 2003). Thus the information
asymmetry problem could be partly attributed to difficulties in communication and
credibility. This problem is even more evident for newly established MSEs since they are
unable to provide evidence of a good financial performance track record (Tucker and
Lean, 2003; Groth and Anderson, 1997). Rather, MSE owners base their financing
decisions on matters of intrusion (Hamilton and Fox, 1998). They hardly prefer financing
sources, such as equity, for fear of higher interference level:
H1: The less educated MSE owners are the higher conformity to the POH
Ownership type versus financing preference
Business organizations are commonly classified into Joint venture, ordinary partnership,
general partnership, limited partnership, private limited company and share company
(Commercial Code of Ethiopia, 1960, Art 212). Based on degree of intrusion and
unlimited liability, business organizations could be reclassified as sole proprietorship,
partnership and company. Degree of intrusion increases and the degree of unlimited
liability decreases as MSEs change their organizational form from sole proprietorship to
partnership and to a company.
MSE owners that are established as either sole proprietorships or partnerships prefer to
exhaust internal sources of finance before going for debt or equity because of the
“collateral incentives” of financing institutions that have bearings on personal assets of

32
the MSE owners thereby following the pecking order. On the other hand, for private
limited companies, since investment risk is limited to the amount of capital contributed,
owners prefer to choose financing options that could maximize firm value even if internal
sources are not yet exhausted:
H2: The lower level of MSE intrusion, operationalized by forms of business organization,
the more conformances to POH
Entrepreneurability versus financing preference
MSE market entry behavior significantly determines financing preferences. MSE owners
who initiate their own businesses have implicit demand for further growth and better
entrepreneurial character in raising finance from nontraditional sources of capital (Van
Auken, 2005). There is some form of entrepreneurial continuum between voluntary and
forced entrance conditions. Voluntary entrants have better entrepreneurial orientation
than their forced entrant counterparts. Therefore, the degree of voluntary entrance can
serve as proxy for the degree of entrepreneurability. The MSE owner who voluntarily
started business has a relatively good deal of entrepreneurship behavior and relatively
higher preference for external financing schemes:
H3: The lower the degree of entrepreneurability, as measured by the degree of voluntary
entrance, the higher conformance to POH
What is plagiarism?
Plagiarism is wrongful appropriation of others’ idea, thoughts, expression, close imitation

etc as if it were your own original work. However, this does not mean that you can not
use research papers of others’ in your research proposal. You can use others’ work but
you must acknowledge the source using an appropriate reference. Plagiarism is not only
copying and pasting of a text without reference but also presentation of others’ idea as
once own. If you use your own previous research work in your current research proposal
without citation it is called self plagiarism.

33
3.3.5 Data and Method
The introduction, literature review and importance of the study sections of your research
proposal communicate the “what” questions namely: what objective the research has and
what value will your research add to the sponsor. The data and method section of your
research proposal answers the “how” questions since it provides your work plan and
describes the activities necessary for the completion of your project. The data and method
part of your research proposal has to include such subsections as: 1) study design, 2) data
type and source, 3) population and sample size, 4) model specification (if any), 5) data
source and method of data collection, 6) data collection instrument and procedure, 7)
instrument validity and reliability, and 8) data analysis techniques and any computer
program to be used.
The guiding principle for writing the method section is that it should contain sufficient
information for the reader to determine whether your methodology is sound. Some even
argue that a good proposal should contain sufficient details such that another qualified
could be able to replicate the research whenever necessary. Besides, you need to
demonstrate your knowledge of alternative methods and make the case that your
approach is the most appropriate and most valid way to address your research question.
It is worthwhile to note that your research question may be best answered by qualitative
research. However, since many reviewers are still biased against qualitative research, you
may need to justify your qualitative method. Furthermore, since there are no well
established and widely accepted canons in qualitative analysis, your method section
needs to be more elaborate than what is required for traditional quantitative research.
More importantly, the data collection process in qualitative research has a far greater
impact on the results as compared to quantitative research.
3.3.6 Method of data analysis
In small scale research proposals the method of data analysis is one part of the data and
method section of the research proposal. For large scale research proposals and PhD
dissertation however, it may be preferable to present the method of data analysis in a

34
separate section. Method of data analysis encompasses sub-sections of the research

proposal starting from data collection up to the nature and form of results. If a research
proposal uses an operations research or econometric model then the method of data
analysis is relatively easy to write since standard operations research and econometric
models are available on software routines such as SPSS, STATA, SAS, MATLAB and
the like. Sometimes, the method of data analysis part of large scale research proposals
and PhD dissertations might have a plan to develop new software program for data
analysis purposes or extend an existing software routine by writing programming codes.
If a research proposal adopts qualitative research, as it is in most case study researches,

the method of data analysis part has to be explicitly stated in light to the available
literature. Some of the qualitative methods of data analysis are: 1) analytic induction, 2)
grounded theory, 3) domain analysis, 4) logical analysis, 5) quasi statistics, 6) event
analysis, and 7) content analysis. Qualitative method of data analysis is broadly presented
in chapter 6 of this teaching material.
3.3.7 Importance/Benefits of the study
This section allows you to write about why the research has to be done. Especially for
large scale commissioned research proposals the importance/benefits of the study section
is considered as justification to get fund. Hence, this section helps you to understand the
research outcomes your sponsor expects. Moreover, this section indicates how clearly
you understood the research problem. If you fail to correctly understand the research
problem then you may face difficulties in writing the importance/benefits of the study.
The converse is also true. If you face difficulties in writing the importance/benefits of the
study then you failed to correctly understand the research problem.
3.3.8 Qualification of the researcher
For internal and external research proposals and that of PhD proposal, the researcher’s
qualification section of a research proposal informs the sponsor about research capability
of the researcher. Large scale research proposals are usually prepared by a team of
researchers in which one of them is principal investigator. In student research proposals,

35
the student is the principal investigator. Content of the qualification of the researcher
section has to refer to the relevant areas of expertise and experience only. However,
Curriculum Vitae (CV) of the researchers can be attached as an appendix if the Request
for Proposal (RFP) requires it.
3.3.9 Facilities and other resources
Types of facilities and special resources required to undertake the research project
depends on the research design. For instance, if a research design uses a quantitative
research design with survey method of data collection then the researcher needs at least
transportation facility to reach the research subjects. Whereas, if the research design uses
simulation instead of field research then the researcher needs computer and simulation
software programs. If the research design uses observation method of data collection
instead of survey, then facilities such as conference halls and special equipments such as
recording tapes might be required.
Internet facility is a very important resource for researchers because millions of published
research papers are available on different journal databases such as J-store
(http://www.jstor.org), Science Direct (http://www.sciencedirect.com), Emerald
(http://www.emeraldinsight.com), and the like. Further, researchers use the internet to look
for research funds.
Sometimes, research projects are also supported by stakeholders other than the sponsor.
For example, if a research project targets auditing practice in Tigray regional state then
audit bureau of the regional state could provide assistance in terms of facilities and
human resource. These types of institutional collaboration have to be taken into account
in the research proposal.
3.3.10 Budget
For large scale research proposals, the budget section of a proposal should be presented
in accordance to the budget format provided by the sponsor. For example, the first grant
application form of International Foundation for Science (IFS) that can be downloaded

36
from http://www.ifs.se includes a budget format composed of: 1) Equipment, 2)

Expendable supplies, 3) literature, documentation, and information, 4) local travel, 5)
extra manpower, and 6) other costs. Detailed presentation of the budget section of a
research proposal could help you to ex ante manage possible budget constraint.
3.3.11 Time Schedule
A research project has a beginning and an ending time. During this project period, the
research requires both material and human resource deployment. Preferably a research
project has to be modularized into major milestones. For example, a pilot study can be
the first milestone that could help you to complete your research proposal. Completion of
a research proposal can be taken as the second milestone. You might want to test your
questionnaire so as to revise it before it is administered. Hence, questionnaire testing is
the third milestone. In general, a research process may be classified into component
milestones and each milestone may have specific activities. Time duration has to be
assigned to each of the activities.
3.3.12 Project management
Large scale research projects such as PhD dissertation take 3 to 4 years. Hence, the
project demands better organization of material and human resources. You can use the
logical framework tool of project management. The logical framework (logframe)
approach to project management top-down integrates the project goal or impact with
project purpose, intermediate objectives, outputs, and activities of the project.
Goal or impact of the research project along with its indicators and means of verification
has to be defined in measurable terms. Purpose or outcomes of the research project have
to be formulated along with: their indicators, means of verification, and important
assumptions. The researcher has to formulate intermediate objectives that are necessary
to achieve the purpose or outcomes. Each intermediate objective shall have its own
indicators, means of verification, and important assumptions.

37
Each intermediate objective has a number of outputs that in turn are made up of many
activities. If each activity is assigned budget variable and fixed costs then each output can
have total budget figure. Estimated budget of each intermediate objective will be the sum
of output budgets under it. Similarly, the total project budget will be the sum of estimated
budgets of each project purpose or outcome. The activity or task scheduling, calculation
of critical path, and generation of different types of project reports can be managed using
project management software such as MS Project 2007.
3.3.13 Bibliography
Your bibliography might include Books, Journal Articles, Magazines and Newspapers,
and sometimes other media such as Television and Radio depending on the nature of data
your research demands. Bear in mind that Books and Text-Books are different. Usually,
Text-Books are not preferred sources of literature for your research because they are
“tertiary” sources of data. In other words, Text-Books are neither primary nor secondary
sources of data collection.
APA Style of referencing Books
A book is referenced by writing name of the authors, year of publication in bracket, title
of the book (in italics), edition, publisher, and place of publication respectively. For
example: Gitman, L. (2003). Managerial Finance. Dryden Press, Hinsdale Illinois. If the
book has no author then you need information regarding title of the book, city where the
book was published, and publisher. Oxford Dictionary, (2nd ed.), (1991). Oxford
University Press, USA. Similarly, citation of an online book, journal, or any other online
material for that matter has to include the date it was viewed. For instance: Trochim,
W.M. (2004). The research methods knowledge base, 2nd ed. Retrieved November 14,
2009, from http://www.socialresearchmethods.met/kb/index.htm.
APA Style of referencing Journal Articles
A journal article can be referenced by writing name of the authors, year of publication,
‘title of article’ (in single quotation marks), title of the journal (in italics), volume

38
number, issue number, and page numbers. For example: Gebregziabher (2009b).
‘Financing preferences of micro and small enterprise owners in Tigray: does POH hold?’
Journal of Small Business and Enterprise Development, 16 (2), 322-334. If a journal
article has no author then it can be cited as follows: ‘Building human resources instead of
landfills’ 2000. Biocycle, 41 (12), 28–9.
APA Style of referencing Magazines and Newspapers
Magazine and newspaper articles can be used to support an empirical fact. Magazine
articles are cited similar to that of journal articles except that the date of publication
should be written. For example: Kluger, J. (2008, January 28). Why we love. Time, 171
(4), 54-60. Newspaper article can be cited as: Tesfaye, K. (2010, September 10).
Unchanged Trade Flows May Nullify Impact of Devaluation. Addis Fortune, 11 (541).
Retrieved from http://www.addisfortune.com/economic_commentary.htm. Similarly,
articles from web 2.0 services such as wikipedia can also be cited as: Research Funding.
(2010, August 27). In Wikipedia, the free encyclopedia. Retrieved, September 16, 2010,
from http://en.wikipedia.org/wiki/Research_funding.
APA Style of referencing Audio-Visual Media
Sometimes, audio-visual media can also be referenced. Audio-visual references shall

include the following: name and function of the primary contributors (e.g., producer,
director), date, title, the medium in brackets, location or place of production, and name of
the distributor. For example: Anderson, R., & Morgan, C. (producers). (2008, June 20).
60 Minutes [Television broadcast]. Washington, DC: CBS News.
3.3.14 Appendix
The data collection apparatus such as the questionnaire and curriculum vitae of the
researchers and other very important documents are attached as an appendix to the
research proposal. For a material to be attached as an appendix it should be very
important to the research proposal.

39
3.4 Sources of research funding
Sources of research funding can be broadly classified into two as: government and
corporate. Government research funds are common in developed countries. Most of these
countries spend 1.5 % to 3% of their GDP on research (“Research Funding”, 2010).
Governments give research grants mainly to universities and research institutes. Most
university researches aim at publication whereas industry researches focus on
commercialization. The universities and research institutes usually invite researchers and
prospective Masters and PhD students in the form of call for research grants and
scholarships.
Some of the government sponsored student research grants targeting African students are:
1) For Undergraduate Thesis Award: http://globetrotter.berkeley.edu/RADW/index.html
2) For Masters Thesis Award: http://www.ascleiden.nl/Research/Award/
3) For different types of research grants: http://www.ifs.se/
In order to prepare a research proposal for funding it is advisable to refer web sites such
as: http://www.learnerassociates.net/proposal/.
3.5 Evaluating the research proposal
To recall your memory, we defined a research proposal as a written document of research

plan that is meant to convince specific readers. Hence, every research proposal is subject
to evaluation. On top of this, most research grants are very competitive. That is, you
research proposal has to compete with other research proposals in order to get fund. In
this teaching material, we shall give emphasis to the evaluation of undergraduate senior
essay research proposal.
The research proposal evaluation format in Table 1, despite some modifications, has been
prepared and used to evaluate senior essay research proposals in the department of
Management of Mekelle University. This research proposal evaluation format presumes
the scientific paradigm of positivism. As you will learn in chapter 6, if you resort to the
scientific paradigm of constructivism, Table 1 is not the proper evaluation format.

40
Table 1: Research Proposal Evaluation Format
Points Out
Sections of the Research Proposal
of 10
Preliminaries
1 Are Preliminaries (cover page, typographic accuracy, 1.15 line-spacing,
Times New Roman of 12 font size, Titles in bold …) properly formatted
and organized?
2 If it is hand written, does it approximate the preliminaries in number 1?
Introduction
Theoretical Framework
3 Does the summarized theoretical framework present a critical evaluation of
up-to-date theoretical knowledge about the topic of interest?
4 Does the theoretical framework provide sufficient justification for the
choice of constructs?
5 Does the theoretical framework clearly show the propositional relationship
between constructs?
Statement of the problem
6 Does the statement of the problem show a clear gap in knowledge/practice
in the literature/own experience that needs to be filled in?
Conceptual Framework
7 Does the conceptual framework present a critical evaluation of up-to-date
empirical knowledge about the topic of interest?
8 Does the conceptual framework provide sufficient justification for the
choice of variables and the hypothetical relationships thereof?
Hypothesis or Research question
9 Does the hypothesis or research question fruitfully put the research problem
on focus? Is there any part of the problem outside the focus?
Research Objective
10 Does the general objective clearly state the core intention of the study? Is
the general objective focused, feasible and measurable?
11 If any, are all the specific objectives derived from the well established
conceptual framework? Are they directly linked to the research problem?
Selling the Point
12 Does the research proposal include a point for sale in a summary form?
Review of related literature
13 Does the theoretical literature review elaborate the theoretical framework in
the introduction part? Does the empirical literature review elaborate the
conceptual framework in the introduction part?
14 Are variety of issues and ideas gathered from peer-reviewed papers such as
Journals, abstracts, books and cited properly using APA style?

41
Data and Methods

If Survey Method is used
15 Is the population and working population clearly defined?
16 Is sampling procedure and the sample(s) size clearly stated and justified?
17 Are the procedures for collecting data clearly stated step-by-step (general
plan for collecting the data)?
18 Are the data collection method(s) proper to achieve the objectives of the
research?
19 Is the data collection apparatus (e.g. questionnaire) tested for validity and
reliability?
20 Are the mathematical and statistical tools (e.g. operations research models,
econometric models, etc) justified and clearly formulated?
21 Are the data analysis procedures (e.g. data cleaning, data coding, data entry,
triangulation, etc) specified and labeled accurately?
If Case Study Method is used
15 Does the proposal justify the use of case study method to address the
problem? Does it explicitly state and reason out its scientific paradigm?
16 Does the proposal clearly define the units of analysis (e.g. a program, an
entity, an event, a person, or group of people)?
17 Are the research questions carefully formulated? If descriptive or
explanatory, are the research questions translated onto propositions?
18 Does the proposal justify the selection of case(s)? If multiple cases are used,
does the proposal put clear boundaries around the cases?
19 Are the procedures for collecting case study data clearly stated step-by-step
(general plan for collecting the data)?
20 Is the data collection apparatus (e.g. interview questionnaire) tested for
validity and reliability?
21 Are the analysis procedures (e.g. data cleaning, data coding, data entry,
triangulation, etc) specified and labeled accurately?
General
22 Does the research proposal make a review of the authoritative research
papers in the research topic of interest?
23 Is there consistency on the citation and referencing style (i.e. APA Style)
and has it been properly used throughout the research proposal?
24 Does the research proposal exhibit good organizational skills: flow of
thought, effective transitions from section to section, & good writing skills?
25 Is the research proposal convincing?
Source: Adapted from the Research proposal evaluation format of the department of
Management, College of Business and Economics, Mekelle University.

42
The evaluation uses a ten-point rating scale in which an evaluation point of 1 denotes that
with respect to the evaluation criterion the research proposal is extremely poor and an
evaluation point of 10 denotes that with respect to the evaluation criterion the research
proposal is excellent. According to the research proposal evaluation format in question, a
research proposal can get a maximum of 250 points.
Chapter Summary:
- The three types of proposals are student, internal, and external proposals
- Research topic usually includes explanandum, explanan, and the case
- Theoretical framework establishes interrelationships among constructs
- Conceptual framework establishes interrelationships among variables
- Two types of literature reviews are theoretical review and empirical review
- The term ‘reference’ refers to consulted materials that are actually cited in-text
- APA referencing style is common in social sciences
- Plagiarism is wrongful appropriation of others’ idea, thoughts, expression, etc
- The data and method section answers the question “how to do the research?”
- Qualitative research method requires much more justification than quantitative
- The benefits of the study section present answers to “why do we research on it?”
- Bibliography refers to all materials consulted even those that are not cited in-text
- Most of the developed countries spend 1.5% up to 3% of their GDP on research

43
Data Collection and Sampling Mekelle University, CBE
CHAPTER 4
DATA COLLECTION AND SAMPLING
4.1 Introduction
If your research is not theoretical then you need to plan for empirical data collection.
Identification of sources of data and sampling design are pre-requisites of an actual data
collection. Data sources can be generally classified into: primary and secondary sources.
Ideally data are collected from all members of a given population (or people, objects,
work units, departments, organizations). Nevertheless, because of time and cost
constraints you may need to take representative sample such that conclusions based upon
the sample data could be generalized onto the defined population if certain statistical
conditions are met.
Key Words: Unit of Analysis, Unit of Observation, Primary Data Source, Secondary Data
Source, Sampling, Statistical Inference.
Chapter Objectives:

- Identify unit of analysis and unit of observation of any research topic
- Develop skill of collecting primary data
- Develop the skill of identification and collection of secondary data
- Design theoretical and working population for any survey research
- Design sampling procedures and random selection of the actual sample respondents
- Know the art and science of sample size determination
4.2 Unit of Analysis versus Unit of Observation
Unit of analysis is the subject of a research about which conclusion is to be drawn. In

survey research the random sample are the units of analysis because the generalized

44
conclusion in effect explores, describes, or explains the population from which the
sample is drawn.
Unit of observation is the person, event, or object from which the data, regarding the
units of analysis, is collected. The unit of analysis and unit of observation can be
identical. That is, data about the sample can be collected from the sample it self. On the
other hand, the unit of observation of a research can be different from the unit of analysis.
For example, the sample of your research could be a set of small and micro enterprises in
Ethiopia but your unit of observation could be owners of the small and micro enterprises
in question.
4.3 Sources of primary data
Primary data are original or un-interpreted accounts pertaining to people, objects, or

events that are collected for the task at hand. For example, you might be interested to
study “the impact of board composition on company performance”. Information
regarding each members of board of directors of a company can be collected either by 1)
directly asking a set of questions to the board members themselves, 2) tracking audio and
video speech recordings of the board members, 3) examining minutes of board meetings,
4) letters, memos or emails exchanged, 5) newspaper articles at the time of event, 6)
asking a set of questions to respondents with firsthand information about the board
members and etcetera.
Primary data can also be generated on a controlled environment that is simulation. For
instance, to study “the impact of board composition on company performance” you may
assign board member roles to selected students thereby establishing control and
experimental groups. Structured questionnaire or similar data collection apparatus could
be prepared to collect the experimental data. Likewise, primary data can also be
generated using simulation software if the statistical properties of the explanandum are
known. Specifically, simulation requires information regarding statistical distribution, a
measure of central tendency or mean, a measure of dispersion or standard deviation, a
measure of peakedness or kurtosis, and a measure of tail length or skewness.

45
4.4 Overview of primary data collection techniques
There are a number of data collection techniques some of which are: observation, content
analysis, questionnaire, interview, key informants, focus group discussion, nominal group
technique, and Delphi method.
4.4.1 Observation
Observation is a technique that involves systematically selecting, watching and recording

behavior and non-behavior characteristics of living beings, objects or phenomena.
Observation of human behavior is a much-used data collection technique. For instance, if
you want to study on the topic “Impact of energy efficiency on energy expenditure at
Adihaqi Campus” one method of data collection could be observation of student and staff
behavior with regard to energy efficiency.
Observation method of data collection can be designed as either participant or non-

participant. In participant observation, the researcher actually participates in the event or
phenomenon being observed. Good example of participant observation is the Ethiopian
Television drama titled “Gemena” televised since August 2009. In participant
observation the observer may be known by all of the observed, known only by some of
the observed, or may be totally concealed from the observed. On the other hand, in non-
participant observation the observer is outside the social setting of the observed.
Observation could be directly observed an observer that is physically present and

personally monitors what takes place. As an alternative, observation could also be
indirectly observed with the help of mechanical recordings, cameras, or other electronic
equipments.
Observational studies can be classified based upon the degree structure and researcher
involvement in the natural environment. Based on degree of structure, an observational
study can be either structured or unstructured. Structured observation can be conducted
either within natural environment or within a laboratory. Similarly, unstructured
observation can be made either within natural environment or within a laboratory. An

46
unstructured study within natural environment requires you to be part of the social
setting. For example, you may want to undertake an exploratory study on “the degree of
corrupt behavior in the Audit Bureau of Tigray regional state”. One way of doing this
type of research is to be an employee of the bureau in order to observe the natural
behavior employees and managers in the bureau. On the other hand, in a structured study
within a laboratory environment, questionnaire-like checklists are used to collect the
actual data. The observation may be at a factual level such as age, gender, height, etc or it
may involve inference such as the degree of motivation of the observed, attitude of the
observed towards a stimulus and the like.
4.4.2 Content Analysis
Content analysis systematically describes the form or content of written and/or spoken
material. For example, you may want to study “Role of forensic accounting in fighting
corruption: A case study of Mekelle University”. Content analysis could help you collect
secondhand information related to speeches of the university staff members under study,
written letters and email conversations, and the like.
Content analysis is a tedious process due to the requirement that each data source be
analyzed along a number of pre-determined dimensions. It may be inductive (identifies
themes and patterns) or deductive (quantifies frequencies of data) or it may also use
abductive inference.
4.4.3 Questionnaire
A written questionnaire, also referred to as self-administered questionnaire, is a data

collection tool in which written questions are presented and given to respondents.
Structured questionnaire is common data collection tool in survey researches. It can be
administered to respondents in different ways such as 1) Sending questionnaires by mail
with clear instructions on how to answer the questions and asking for mailed responses,
2) Gathering all or part of the respondents in one place at one time, giving oral or written
instructions, and letting the respondents fill out the questionnaires, or 3) Hand-delivering
questionnaires to respondents and collecting them later. Chapter five of this teaching

47
material presents detailed account of structured questionnaire design and methods of

testing questionnaire validity and reliability.
4.4.4 Interview
Interview is a data-collection technique that involves oral questioning of respondents,

either individually or as a group. Answers to the questions posed during an interview can
be recorded by writing them down (either during the interview itself or immediately after
the interview) or by tape-recording the responses, or by a combination of both. Interviews
can be either unstructured or structured. Unstructured interview is very important to study
sensitive issues such as assessment on the prevalence of corruption in an organization. If
you are using unstructured interview for your research then you may use a list of topics
rather than fixed questions. Sequence of the topics has to be determined by the flow of
the discussion. That is, unstructured interview is more flexible. On the other hand,
structured interview presumes that you are aware of the subject matter that is why
interview questionnaire is used in structured interviews. In general, unstructured
interview method of data collection is particularly useful in exploratory researches
whereas structured interview is good for descriptive and explanatory researches.
4.4.5 Key informants
The use of key informants is another important technique to gain access to potentially
available information. Key informants could be knowledgeable community leaders or
administrative staff at various levels and one or two informative members of the target
group of your research. For instance, if you want to study “time series analysis of energy
cost efficiency in Mekelle University” you may collect primary data using the method of
key informants. Method of key informants is good when the types of data you need are
relatively objective - like energy expense of Mekelle University.
4.4.6 Focus group discussions (FGD)
The focus group is a special type of group in terms of purpose, size, composition, and
procedures. A focus group is typically composed of seven to twelve participants who are

48
unfamiliar with each other and conducted by a trained interviewer. You may be interested
to study “the determinants of tax evasion in Tigray regional state”. You may create a
permissive environment in the focus group that nurtures different perceptions and points
of view, without pressuring participants to vote, plan, or reach consensus about tax
evasion in the region. In FGD the facilitator sets the agenda and lets the participants
brainstorm on the agenda thereby specifying the points of discussion. The group
discussion is conducted several times with similar types of participants to identify trends
and patterns in perceptions. FGD can be conducted in a meeting hall or using FM radio or
television as a medium. Moreover, online services such as Facebook and Twitter can also
be used to run a focus group discussion.
4.4.7 Nominal group technique (NGT)
The nominal group technique is a method to identify issues of concern to special interest
groups or the public at large. The NGT is used to obtain insight into group issues,
behaviors and future research needs. In NGT discussants are given the chance to list
down and prioritize their own idea about the topic in question. Then each NGT member
presents the prioritized points and accordingly group discussion starts. As a result of the
group discussion, a consolidated list of issues is developed. Finally, NGT members
prioritize the consolidated list of issues by voting on it. The NGT could be helpful in
developing structured questionnaire.
4.4.8 Delphi method
The Delphi method was developed to structure discussions and summarizes options from
a selected group so as to avoid meetings, collect information/expertise from individuals
spread out over a large geographic area, and save time through the elimination of direct
contact. The technique requires a series of questionnaires and feedback reports to a group
of individuals. Each series is analyzed and the instrument/statements are revised to reflect
the responses of the group. A new questionnaire is prepared that includes the new
material, and the process is repeated until a consensus is reached. Although the data may
prove to be valuable, the collection process is very time consuming. When time is

49
available and respondents are willing to be queried over a period of time, the technique
can be very powerful in identifying trends and predicting future events.
Delphi method of data collection can be exercised using FM radio or television as a

medium. Moreover, online services such as personal web blogs, Facebook and Twitter
can also be used to apply the Delphi method.
4.5 Sources of secondary data
Secondary sources are the analyzed, evaluated, interpreted or criticized form of primary
data. Secondary data are originally collected for a task other than the task at hand. For
example, if you refer to an article that presents a scientific experiment along with inferred
conclusion, as most scientific papers do, then your data source is secondary. To this end,
most scientific papers are secondary data sources. Researchers keep database of their
primary data for future use. If you get access to these databases then you are using
secondary data source because the data were not collected for your research’s purpose.
Collection of primary data is not only very costly but also difficult. Recognizing this fact,
governments have an agent organization in charge of primary data collection. For
example, the Central Statistical Agency (CSA) of Ethiopia is in charge of collecting
primary data from all over Ethiopia and making it available for researchers. The CSA
covers population, agriculture, industry, commerce, and other economic data. To get
more information about CSA, you can browse http://www.csa.gov.et. Similarly, the
United Nations has huge database of primary data collected from all over the world. The
UN data covers such diverse issues as finance, industry, energy, ICT, labor, population,
food and agriculture, gender, health, and the like. For more information you can browse
http://data.un.org. Published data are often preferred due to convenience, relatively low
cost, and higher reliability if it is collected by reputable organization.
4.6 Overview of secondary data collection techniques
Secondary data collection techniques are mainly related to data collection from available
database, data collection from documents, content analysis, and meta-analysis.

50
4.6.1 Data collection from available database
Usually there is a large amount of data that has already been collected by others, although
it may not necessarily have been analyzed or published. Locating these sources and
retrieving the information is a good starting point in any data collection effort. For
example, the author of this research had collected primary data for the paper on
“Financing preferences of micro and small enterprise owners in Tigray: does POH hold?”
You can use this data if your topic is similar to the paper in question. Another example is
data collected by governmental and non-governmental organizations for their own
purposes.
4.6.2 Data collection from documents
Data collection from documents or otherwise called data mining is commonly used in
accounting and finance researches. For example, you may be interested to conduct
research on “efficiency of land lease contracts in Tigray regional state”. In this research,
the best candidate for data collection method is data collection from documents; that is,
land lease contracts in Tigray regional state. Similarly, most researches in auditing
intensively use data collection from documents related to financial statements of an
entity.
Data collection from documents is particularly helpful method of secondary data

collection to identify research questions. It is also helpful to validate your research
findings. In other words, after you finish your research, you may compare your
conclusions with the documentary facts in the organization in question.
4.6.3 Data collection using Meta-Analysis
Meta-analysis combines the results of studies being reviewed. For instance, you may
want to conduct extensive literature review on “climate change financing”. The natural
starting point is to collect theoretical and empirical research papers in the area of climate
change financing. You may utilize statistical techniques to estimate the strength of a
given set of findings across many different studies. This allows the creation of a context

51
from which future research can emerge and determine the reliability of a finding by
examining results from many different studies. Meta-analysis findings form a basis for
establishing new theories, models and concepts.
4.7 Biases in data collection
Bias in data collection is a distortion in the collected data so that it does not represent
reality. Data collection could be distorted due to the use of a defective data collection
apparatus (e.g. defective questionnaire), intentional or unintentional biases from the
observer or interviewer, intentional or unintentional biases from the respondent, and/or
poor quality of the data.
4.7.1 Defective instruments
Commonly, biases emanate from data collection instruments. For example, your
questionnaire may fail to include questions that are relevant for your research. On the
other hand, your questionnaire might be with closed questions on topics about which little
is known for the respondent and the questionnaire might include open-ended questions
without guidelines on how to ask or how to answer them. In addition, the logical flow of
questions may influence respondents to respond in a predetermined way compared to a
randomized order of questions. Defects in questionnaires could also arise due to not-
standardized weighing scales or other measuring equipments.
These sources of bias can be prevented by carefully planning the data collection process
and by pre-testing the data collection tools. In addition, questionnaire validity and
reliability can be measured to test whether the defectiveness of the instrument goes
beyond a tolerable limit.
4.7.2 Observer or Interviewer bias
Observer bias can easily occur when conducting observations or utilizing loosely
structured group or individual interviews. The cognitive, affective, and behavioral setup
of the observer or interviewer creates bias in data collection. Training is one way of

52
minimizing observer bias. The observer or interviewer has to get sufficient training as to
how to collect the required data. Hence observation protocol or interview guidelines have
to be prepared in order to minimize observer bias. The observed data or results of an
interview have to be discussed in group and recorded as immediate as possible.
Electronic equipments such as camera and tape recorder can be used in order to better
control observer bias.
4.7.3 Respondent bias
Respondent bias occurs when research subjects give answers to your questions in a way
they expect you wanted them rather than in accordance their true feelings. Similarly, it
occurs when research subjects give misleading responses by the time they do not want to
reveal information about sensitive matters. For example, if your research deals with audit
inspection then respondent bias is inevitable.
Respondent bias can be reduced by adequately introducing the purpose of the study to
respondents, by phrasing questions on sensitive issues in a positive way, by taking
sufficient time for the interview, and by assuring respondents that the data collected will
be confidential.
4.7.4 Information bias
Sometimes the information itself has weaknesses. Records may have many blanks or be
unreadable. This tells something about the quality of the data and has to be recorded.
Another common information bias is due to gaps in people’s memory. Information bias
can be minimized by collecting triangulated data from two or more sources. If you do not
fully succeed to control information bias then it is important to report honestly in what
ways the data may be biased.
4.8 Sampling terminology
Any research work has conclusion inferred from set of premises. The conclusion could be
about groups of people, groups of organizations, employees in an organization, contracts

53
between organizations, and the like. The broad class of people, objects, or events that are
targeted by the conclusion is known as population or universe. A population or universe
has individual members called population element. Population has countable population
elements such as people whereas a universe has uncountable population elements such as
stock price index. The study of all individual members of a population is known as
census.
Population can be divided into theoretical and accessible. Theoretical population draws
the boundary of the scope of conclusion of the research. That is, conclusion of a research
is irrelevant for entities beyond the theoretical population’s boundary. Accessible
population (also known as working population) is subset of a population whose members
can possibly be accessed by the researcher. The enumerated list of all population
elements of an accessible population from which sample can be drawn is called sampling
frame. For instance, if you want to do research on “determinants of incentive
incompatibility in purchase bids: the case of Mekelle University”. Theoretical population
of your research will be the set of all purchase bids since the establishment of Mekelle
University. This is because, the conclusion of your research generally applies to purchase
bids in Mekelle University. However, not all purchase bids of Mekelle University are
accessible. Hence, the set of all accessible purchase bids of Mekelle University
constitutes accessible population and the enumerated list (e.g. in temporal order) of the
accessible purchase bids is the sampling frame.
Sampling unit is a member of a sample selected from a sampling frame. Sampling unit
can be either an individual member a group of members of the sampling frame. If the
sampling unit is a group of members of the sampling frame, you can also take sample
from the sampling unit itself. Random selection is how you draw the sample of people,
objects, or events for your study from a population. If the sampling units are randomly
selected from the sampling frame then the sample-based conclusion of your research can
be generalized onto the population, that is, your research has external validity. Random
assignment refers to categorization of sample units into treatment groups (or control and
treatment groups) based on certain matching criteria. Random assignment is related to
internal validity of the research design. Sample units are randomly assigned in order to

54
help assure that the treatment groups are similar to each other (i.e. equivalent) prior to the
treatment.
Starting from operational definition of the population until the final selection of the
sampling unit you may commit errors. Magnitude of these errors can be measured as a
difference between results using census and sample. These errors are called random
sampling error. On the other hand, non-sampling errors are those discussed in section 4.6
of this chapter.
4.9 Types of probability sampling designs
Probability sampling is a sampling technique in which each member of a population has a

known non-zero probability of being chosen. Simple random sampling, stratified random
sampling, systematic random sampling, cluster random sampling, and multi-stage
sampling are the common types of probability sampling.
4.9.1 Simple Random Sampling
Assume the working population of your research has N number of population elements
out of which you decided to select x number of sampling units. Simple random sampling
sets a criterion of “equal probability of selecting each x from N”. Simple random
sampling is most appropriate if the population elements are highly homogenous. For
example, you may want to study “attitude of Business and Economics students towards
quantitative courses”. Assume that the total number of students within Business and
Economics College is 1000 out of which you want to take a 10% sample (i.e. 100/1000 =
0.1 sampling fraction). There are billions and billions of ways of selecting 100 sampling
units from the population of 1000 elements.
In order to randomly select the sampling units you can use the lottery method, that is
write down names or codes of each element in the sampling frame on separate pieces of
paper, mix them up, close your eyes and pull out the first 100. As an alternative, you can
also use table of random numbers.

55
4.9.2 Stratified Random Sampling
Stratified random sampling (otherwise called quota sampling) involves dividing your
population into homogeneous subgroups and then taking a simple random sample from
each subgroup. Each of the homogenous sub-group is known as strata. If sizes of each
stratum are comparable then equal sampling fraction can be used. This is called
proportionate stratified random sampling. The use of unequal sampling fractions for each
stratum is known as disproportionate stratified random sampling. For each stratum simple
random sampling is applied.
Example 1 is the “sample data characteristics” part of the research paper titled
“predicting probability of default of Indian corporate bonds: logistic and Z-score model
approaches” by Bandyopadhyay (2006). It illustrates stratified random sampling.
Example 1: Bandyopadhyay (2006)
Sample data characteristics
The information on defaulted and solvent firms is collected from CRISIL’s annual ratings
of long-term bonds issued by 542 companies from 1998 until 2004. These companies are
then matched by their asset size, year, and industry affiliation. Following these criteria in
a random selection, we finally got a sample of 52 solvent firms and 52 defaulted firms.
The defaulted group is a class of manufacturers whose long-term bonds have been
defaulted between 1998 and 2003. The solvent firms are chosen on a stratified random
basis drawn from CRISIL rating database. The mean asset size of the firms in the solvent
group (Rs. 948.51 Crore) is slightly greater than that of the defaulted group (Rs. 818.62
Crore), but matching exact asset size of the two groups is not always necessary. The
financial information of these 104 companies over the period of 1998 to 2003 is obtained
from the CMIE Prowess database. In addition to the accounting data, information on ISO
certification is obtained from the Q-Prod’s directory. In order to test the predictive
accuracy of our estimated sample, a holdout sample of another 50 (25 solvent and 25
defaulted) companies is being created for the years 2003 and 2004. Our estimated sample
of 104 firms have been classified into 11 industry categories based on their major

56
economic activities after matching them with the National Industrial Classification (NIC)
codes (see Table I).
Industry dummy Industry type Number of firms
IND1 Food products/sugar

/tea/tobacco/beverages 4
IND2 Paper 3
IND3 Textile 9
IND4 Chemical 27
IND5 Machine/electrical/computers 14
IND6 Metal/non-metal 23
IND7 Auto/parts 6
IND8 Power 2
IND9 Diversified 5
IND10 Service 9
IND11 Other manufacturing 2
Total 104
4.9.3 Systematic Random Sampling
Systematic random sampling is a technique of selecting sampling units using a fixed

interval on a randomly ordered list of the sampling frame. It involves selection of every
kth element in the sampling frame where k is the ratio between number of elements in the
population and the sample size (i.e. N/x). As homogeneity is basic characteristic of
simple random sampling random order is basic characteristic of systematic random
sampling. This type of probability sampling is particularly useful when the number of
elements in the population is too cumbersome to be handled by either simple random
sampling or stratified random sampling. In using systematic random sampling you need
to take care of the sampling interval size because it might hide important patterns.
For example, you may want to study on the topic “on the willingness to pay tax in
Mekelle”. The tax payers regularly visit the customs and duties office of Mekelle city.
Hence, you can take the total number of tax payers in the city from the customs office
and divide it by the sample size from which you want to collect data thereby determining
the sampling interval. Then you can collect data using systematic random sampling from
tax payers visiting the customs and duties office.

57
4.9.4 Cluster or Area Random Sampling
Cluster or area random sampling is a probability sampling technique whose sampling

units are homogeneous sub-groups (usually geographic areas) called clusters. The
elements within each cluster are heterogeneous. For example, you may want to study on
the topic “role of religion on household saving in Ethiopia”. In each Wereda of Ethiopia,
there are heterogeneous religions. So, you may consider a list of more than 600 Weredas
as your sampling frame and accordingly select cluster sampling units. In effect you are
aggregating set of households into Wereda clusters in order to minimize cost of data
collection. In cluster random sampling, data analysis will be at a cluster level this is
called single stage cluster random sampling. In contrast, stratified random sampling
analyzes at individual household level not at Wereda level. In addition, stratified random
sampling aims at precision whereas cluster random sampling aims at cost reduction.
4.9.5 Multi-Stage Sampling
Multi-stage sampling is a two or more stage cluster random sampling. For instance, if you
want to conduct an experiment on the topic “Ethiopian small landholding farmers’
willingness to sell their land: an experiment on private ownership of land”, you need to
collect data from Ethiopian farmers but Ethiopian farmers are more than 68 million.
Hence you can use multi-stage sampling. You can randomly select cluster of Weredas as
sampling units. From each sampling unit of Wereda you can also randomly select clusters
of Kebeles as second stage sampling units. Further, you may randomly select clusters of
sub-Kebeles in a third stage cluster sampling. This way you can reach small landholding
farmers in Ethiopia. In effect, multi-stage sampling creates clusters within clusters.
4.10 Types of non-probability sampling designs
Unlike probability sampling, non-probability sampling designs do not involve random

selection of sampling units. In general, researchers prefer probability sampling over non-
probability sampling. But, there may be cases where random sampling is not practical or
it does not make theoretical sense. Non-probability sampling techniques are classified
into convenience and purposive sampling.

58
4.10.1 Convenience Sampling
Convenience sampling is when you take sample simply by asking for volunteers. Clearly,
the problem with convenience sampling is that we have no evidence that they are
representative of the population in question.
4.10.2 Purposive Sampling
Purposive sampling is a non-probability sampling technique that takes sample with a

purpose in mind. Some of the common types of purposive sampling are Modal instance
sampling, expert sampling, quota sampling, heterogeneity sampling, and snowball
sampling.
Expert Sampling involves the assembling of a sample of persons with known or

demonstrable experience and expertise in the research area of interest. Expert sampling is
the best way to elicit the views of persons who have specific expertise. In addition, it
could provide evidence for the validity of another sampling approach you have chosen.
Quota Sampling involves non-randomly selection of sampling units according to some

fixed quota. For instance, if you know the population has 40% women and 60% men, and
that you want a total sample size of 100, you will continue sampling until you get those
percentages and then you will stop.
Modal Instance Sampling is a non-probability sampling technique that takes a modal

instance or the most frequent case as sample. On the other hand, Heterogeneity Sampling
is when you want to include all opinions or views, and you are not concerned about
representing these views proportionately. Heterogeneity sampling is, in this sense, almost
the opposite of modal instance sampling.
Snowball Sampling is a non-probability sampling technique that begins by identifying

someone who meets the criteria for inclusion in your study. You then ask them to
recommend others who they may know who also meet the criteria. Although this method
would hardly lead to representative samples, there are times when it may be the best

59
method available. For example, you may want to study the topic on “role of family
income on female dropouts in Mekelle University”. Using snowball sampling you can
ask one female dropout and ask her to recommend you another dropout she knew and so
on.
4.11 Sample size determination
Once the actual samples are selected and accordingly data is collected, you perform data
analysis. The statistical results of the sample data are called sample statistics. Whereas,
had you had collected census data from the population instead of sample, the statistical
results of the census would have explained the population variables called population
parameters. Sample statistics are used to make inferences about the population
parameters with the help of inferential statistics.
4.11.1 Ratio or Interval Scaled Explanandum
In general, the larger the sample size the more power sample statistics has to infer about
the population parameter. Where S is the sample standard deviation and n is the sample
size, the central limit theorem implies that one standard error of the sample mean is given
by equation 4.11.1. That is, as n tends to infinity, the standard error of mean approaches
zero. On the other hand, as n tends to zero the standard error of mean gets larger.
S2 S
SX = or S X = (4.10.1)
n n
Confidence intervals work not only for one standard deviation (i.e. Z-value of 1) that
covers around 68% of the normal curve but also other confidence intervals. For example,
a Z-value of 1.96 standard deviations away from the mean covers 95% of the normal
curve and a Z-value of 2.57 standard deviations away from the mean covers 99% of the
normal curve. Therefore, the standard error of mean in equation 4.11.1 can be generalized
to an error factor (E) as in equation 4.10.2.
S
E=Z (4.10.2)
n

60
Using some algebra and solving equation 4.11.2 for n, we get sample size determination
formulae as in equation (4.10.3).
2
Z S
n=   (4.10.2)
 E 
Sample size depends on three variables: heterogeneity of the population or population
standard deviation (approximated by S), magnitude of acceptable error (E), and
confidence level (Z). Heterogeneity of the population varies from most homogeneous to
most heterogeneous. The more heterogeneous the population the higher expected
population standard deviation and accordingly the larger sample size is required. The
standard deviation of a population parameter can be estimated using sample standard
deviation from a pilot study or it can also be estimated using rule of thumb.
The range of possible random error or confidence interval is the second important factor
in sample size determination. Confidence intervals of some population parameters are
known ex ante. For instance, height of Ethiopian men could be assumed to range between
1.60 meters up to 1.90 meters though there are Ethiopian men whose height goes beyond
1.90 meters or below 1.60 meters. Hence, the confidence interval (i.e. a range with in
which the population parameter lies) of height of Ethiopian men is [1.60, 1.90] with a
standard error of 0.15 meters. Magnitude of error is estimated as a product of the Z-value
and standard error of the sample mean or using personal judgment.
Confidence level is a long run percentage value that tells you to what extent you can be
confident as to the inclusion of the population parameter within the confidence interval.
The higher confidence level you require the larger sample size you need to have. The
most commonly used confidence levels are 95% and 99%.
For example, if you want to study the topic “Does salary increment bolster research
performance of the academic staff of Mekelle University?” how many sample units
would you select out of the nearly 900 academic staff members of the University? Keep
in mind that the factor explained or the explanandum is research performance. Assume
that research performance can be measured in percentage terms such that the possible
values lie in between 0 and 100. Given regulatory requirements such as the legislation, if

61
you reasonably believe that research performance of staff members of Mekelle University
is normally distributed between 70% and 100% and if you decide a 95% confidence level
then your sample size would be 123 (i.e. [(1.96*85)/15]2).
4.11.2 Ordinal or Nominal Scaled Explanandum
In section 4.11.1 we discussed about sample size determination for explanandum that is
measured in terms of either ratio or interval scales. But the explanandum can also be
measured in terms of either nominal or ordinal scales. For example, research on the topic
“Ethiopian small landholding farmers’ willingness to sell their land: an experiment on
private ownership of land” has a nominally scaled explanandum, if the respondent
farmers are expected to answer either “willing to sell” or “not willing to sell” by the time
they are subjected to the experimental treatment, that is, land privatization.
Standard error of the proportion is similar to the standard error of the mean in Equation
4.11.1 except that the former uses multiple of p and q instead of the sample standard
deviation squared. The magnitude of acceptable margin of error of proportion is also
determined as the product of Z-value and the standard error of proportion. Accordingly,
sample size of an ordinal or nominally scaled explanandum is given by Equation 4.11.3.
Z 2 pq
n= (4.10.3)
E2
For example, in the research titled “Ethiopian small landholding farmers’ willingness to
sell their land: an experiment on private ownership of land” how many random sample
units would you select? Assume that you reasonably expect 60% of Ethiopian farmers
would irrationally sell their land if Ethiopian land were privately owned (if you are not
sure about the proportion you may use 50%). The first think you need to decide about is
the maximum allowance for sampling error. For example, if you allow your expectation
of 60% to vary between 55% and 65% then your maximum allowance for sampling error
is 5%. Accordingly, the expected sample size at confidence level of 95% is:
1.96 2 0.6 * 0.4
n= = 369 Farmers
0.05 2

62
Similarly, the sample size for different allowances for sampling errors and different
confidence intervals are given in Table 1 below.
Table 1: Sample size at different confidence interval and error margins

Confidence + 1% Error + 3% Error + 5% Error + 7% Error
Interval
95% 9,220 1,024 369 188
99% 15,852 1,761 634 323
It is important to note that sample size basically depends on the degree of external
validity you need to have, the time and cost budget you want to allocate. The more
external validity you require and the more time and cost budget you allocate for the
research the larger sample size you need to take.
Chapter Summary:
- Unit of analysis is the subject of a research about which conclusion is to be drawn.

- Unit of observation is the person, event, or object from which the data is collected.
- Primary data are original or un-interpreted accounts collected for the task at hand.
- The unit of analysis and unit of observation can be identical.
- Primary data can also be generated on a controlled environment that is simulation.
- Observation is watching and recording behavior and non-behavior characteristics
- Content analysis describes the form or content of written and/or spoken material.
- Questionnaire is where questions are written in a document & given to respondents.
- Interview is a data-collection technique that involves oral questioning of respondents
- Key informants are experts such as community leaders or administrative staff
- Focus group is typically composed of seven to twelve participants
- In NGT discussants are given the chance to list down and prioritize their own idea
- Delphi method the instrument/statements are revised to reflect responses of the group

63
The Survey Method and Analysis Mekelle University, CBE
CHAPTER 5
THE SURVEY METHOD AND ANALYSIS
5.1 Introduction
In the last four chapters you dealt with the basics of research design, research proposal
writing, data collection and sampling. Survey method refers to primary data collection
from respondents using questionnaires. Quality of questionnaire determines quality of
your research hence your questionnaire has to be tested for reliability and validity via
pilot study. Further, survey data analysis is usually quantitative. It is too cumbersome to
manually analyze survey data. Therefore, the use of specific statistics or econometrics
software package is inevitable.
Key Words: Survey, Generalization, Questionnaire Design, Reliability, Validity, Data Matrix,
STATA.
Chapter Objective:

- Know the detailed characteristics of survey research method
- Differentiate survey research from case study research
- To design a written questionnaire for any topic that has survey research design
- To test the reliability and validity of a questionnaire
- To clean, edit, and create data matrix for any survey data
- To analyze survey data using STATA software routine
5.2 Characteristics of Surveys
Zikmund (1997) defined survey as “a research technique in which information is gathered

from a sample of people by use of a questionnaire”. The questionnaire of survey research
could be either structured or semi-structured so long as data are primarily collected from
representative sample of respondents. Sometimes, survey research method might use
interview technique of primary data collection from representative sample of
By: G.Egziabher H.Selasie Gebru, Assistatnt Professor

64
respondents. Most survey researches are descriptive but survey research can be designed
for explanatory or exploratory purposes.
Survey research is different from case study research because it considers many units of
analysis based on which it opts to generalize. As can be seen form Table 1, the units of
observation could vary between one, few, and many. So long as intention of the survey
research is generalization, the units of analysis can also be few. On the other hand, case
study research considers one unit of analysis but the units of observation can be one, few,
or many in number. Case study research aims at particularization instead of
generalization. If a case study research project considers few units of analysis then it is
known as multiple case study research.
Table 1: Differences between survey and case study researches
Units of Analysis
One Few Many
Observation
Units of
One Case Study Survey

Few Case Study Survey
Many Case Study Survey
Particularization Generalization
(Case Study) (Survey)
Surveys can be classified according to methods of communicating with the respondent,

structured and disguised questions, and based on temporal basis. As you have learned in
the last chapter, there are different ways of collecting data from respondents. Accordingly
survey researches can be classified as mail survey, telephone interview survey, personal
interview survey and etcetera. The nature of questions with in a questionnaire also
determines classification of surveys. Accordingly, survey researches can be classified
into structured and undisguised, structured and disguised, unstructured and undisguised,
and unstructured disguised surveys. Survey researches can also be classified based on
temporal dimension as cross-sectional survey, longitudinal survey, and panel survey.

65
Cross-section surveys consider sample respondents from different categories of the

working population as a single point in time. Longitudinal surveys consider sample
respondents from different categories of the working population across time.
Longitudinal surveys are preferably analyzed using time series models. Panel surveys are
special type of longitudinal survey that considers the same sample of respondents across
time.
Some of the advantages of survey method are 1) it is relatively cheap especially if mail
questionnaires are used, 2) a population of very large number of members can only be
reached via survey, 3) if carefully designed, higher instrument reliability can be achieved,
and 4) because of standardization usually they are free from various errors. Some of the
disadvantages of survey researches are 1) quality of survey data depends on motivation,
memory, and ability of respondents, and 2) non-response and self selection biases.
5.3 Questionnaire Design
Questionnaire is the most important data collection apparatus for survey research. In this
section you are going to learn about measurement of variables, question formulation,
measurement scale selection, and characteristics of good questionnaire.
5.3.1 Measurement of variables
In terms of research, a variable is the operationalized form of a concept. For example,

household-head salary is a variable that operationalizes the concept of income. In order to
get numeric or quasi-numeric value of a variable, respondents are asked a question or set
of questions. Responses of the respondents are coded using numerical values to make it
ready for analysis.
A level of measurement is the precision by which a variable is measured. Measurement

has four levels of precision namely; nominal, ordinal, interval, and ratio scales. Nominal
scale of measurement has the lowest precision whereas ratio scale has the highest
precision.

66
Nominal and Ordinal Scales
Nominal level of measurement describes variables that are categorical in nature. In other
words, nominal level of measurement deals with classifying a variable into different
distinct groups. If there are a limited number of distinct categories then you are dealing
with discrete variables. For instance, gender is a variable that can be nominally measured
classifying the respondents either female or male. Virtually any operationalized concept
can at least be nominally measured.
The ordinal level of measurement describes variables that can be ordered or ranked in
some order of importance. Ordinal variables include all the properties of nominal
variables. It describes most judgments about things, such as big or little, strong or weak.
For instance, employee satisfaction variable has ordinal level of measurement because
you need to ask respondents whether they are highly dissatisfied, somehow dissatisfied,
indifferent, somehow satisfied, or highly satisfied. This type of scale of measurement is
commonly referred to as the “Likert scale”. Advanced statistical analysis requires more
precise level of measurement than nominal and ordinal scales.
Interval and Ratio Scales
Interval level of measurement describes variables that have more or less equal intervals,
or simply meaningful distances between their ranks. For example, if you were to ask
respondents if they were first, second, or third batch graduates of Mekelle University, the
assumption is that the distance or number of years, between each batch is the same. In
this example, distance between two consecutive batches has algebraic value but the
transition point between two batches does not have meaningful algebraic value. For
example, the “numbers” in Degree Celsius type of temperature measurement are not pure
numbers. That is, a temperature of zero Degree Celsius is not the “true zero” therefore 20
Degree Celsius temperature is not twice as hot as 10 Degree Celsius. In interval scale of
measurement what matters is the presence of meaningful distance between scale items.
These distances may not necessarily be equal.

67
Ratio scale of measurement describes variables that have equal intervals and a fixed
reference point. For instance, the variable income has ratio scale of measurement because
it has a reference point of zero. That is, a zero income is meaningful in real life.
Similarly, the variable stock price has ratio scale of measurement because distance
between two consecutive stock prices has algebraic value and the transition point
between two stock prices have meaningful algebraic value. Hence, a stock price of $20 is
twice as higher as a stock price of $10. It is helpful to note that measurement scales
involving “not at all”, “often”, and “twice as often” might qualify as ratio level
measurement.
5.3.2 Influence of Respondents’ Attitude
Attitude is an enduring disposition to consistently respond in a given manner to various

aspects of the world. Respondents of a survey research are mostly human beings.
Therefore respondents’ attitude creates significant bias in survey data collection if it is
not controlled ex ante. Attitude is composed of three elements: general feeling or
emotions toward an object, that is affective, one’s awareness about an object, that is
cognitive, and intentions and behavioral expectations that is behavior.
In survey researches respondents are usually asked to respond on questions 1) regarding

themselves, their family, or their property ownership or possession 2) regarding their
organization: their work department, their colleagues, their subordinates, their bosses, the
work they perform, the work others perform, and the like, 3) regarding phenomena that
happened: events that the respondents were involved in, events that the respondents were
not involved in, 4) regarding their attitude towards people, organization, events etcetera.
Respondents’ attitude has strong influence on the data collected depending on the object
about which respondents are asked to respond and depending on sensitivity of the
question asked. For example, if your unit of analysis is concerned with managers of a
company and your unit of observation involves employees of the company in question
the respondents may not dare to reveal sensitive information about their bosses. Whereas,
if your unit of analysis targets managers of a company but your unit of observation

68
considers customers of the company in question then respondents may not be afraid of
revealing sensitive information.
5.3.3 Question Formulation
Questions of a survey questionnaire need to be formulated in a way that maximizes the

incentive to the unit of observation to reveal private information regarding the unit of
analysis. Some of the common types of question formulation are: Simple Attitude
Scaling, Category Scaling, Likert Scaling, Semantic Differential, Numerical Scaling,
Simple Attitude, Category, and Likert Scaling
In its most basic form attitude scaling requires that an individual agrees or disagrees with
a statement or a single question. Simple attitude scaling limits the type of statistical
analysis that may be applied. You can use simple descriptive statistics such as tabulation,
percentages, measures of central tendency, factor analysis, and the like.
Category scaling expands the response categories of simple attitude scaling thereby
providing the respondent more flexibility in the rating task. For instance, quality can be
measured on a four-item rating scale as excellent, good, fair and poor. Alternatively, it
can also be measured on a five-item rating scale as very good, fairly good, neither good
nor bad, not very good and not good at all.
A Likert Scale is a measure of attitudes ranging from very positive to very negative. The
total score is the summation of the weights assigned to an individual’s total response. For
example, employee motivation can be measured using a total score on a set of questions
formulated using Liket scale items. More advanced statistical analysis such as measures
of dispersion, measures of association and correlation, and regression can be applied on
Likert scale measured variables.
Semantic Differential and Stapel Scale
Semantic differential is a type of rating scale intended to measure connotative meanings

of events, objects, or concepts. Set of bipolar adjectives such as fair – unfair, high - low,

69
valuable – worthless, and the like. The responses are connected via line to form a graphic
picture. For example, you may want to research on the topic “Leadership Style and
Effectiveness of Department Chairpersons in Mekelle University” you can use semantic
differential way of question formulation.
The staple scale is a uni-polar measure of attitude that places a single adjective in the
center of an even-number range of numerical values. It is used to measure the direction
and intensity of an attitude simultaneously. For example, you may want to research on the
topic “Support staff attitude towards independent auditors: the case of Mekelle
University”. In measuring the support staff’s attitude towards the independent auditor
one of your questions may look like:
The Independent Auditor

+3
+2
+1
Integrity
-1
-2
-3
Respondents are expected to select a plus number for the level of integrity the
independent auditor has or a minus number for the degree of lack of integrity. The plus
signs may be higher than +3 if lack of integrity is rarely expected. The converse is also
true.
Constant Sum Scale
Constant Sum Scale is a measure of attitude in which respondents are asked to divide a
constant sum among attributes to indicate the relative importance of the attributes. For
example, you may want to do research on the topic “dynamics of expenditure behavior of
households in Ethiopia”. In your questionnaire, one of the questions could be: “If you had
$3,000 in benefits per month; how much would you like to allocate to food, energy,
medical insurance, and retirement plan?”

70
5.3.4 Selecting a Measurement Scale
Selection of a measurement scale can has strong influence on validity and reliability of
the data collected. A measurement scale is selected for every question based on certain
purpose in accordance to the research design.
Level of Precision
Choice of level of precision (i.e. nominal, ordinal, interval, or ratio scale) largely depends
on the type of statistical analysis that is desired. If the data analysis part of your research
design selects advanced descriptive and inferential statistics such as econometric
modeling then the questions in your questionnaire have to be measured using interval or
ratio scale.
Monadic versus Comparative Questions
Questions in a questionnaire could be monadic or comparative. A monadic question asks

the respondent to evaluate only one object at a time and the resulting data set is either
interval or ratio scale. For example, if you ask respondents the degree they favored Abay
automobile then the question is monadic or it has metric scale. On the other hand,
comparative rating scale is any measure of attitudes that asks respondents to rate a
concept, an object, or an event in comparison with a benchmark explicitly used as a
frame of reference. You need to choose either monadic or comparative questions
depending on your research design.
Balanced and Unbalanced Rating Scale
Balanced rating scale is a fixed-alternative rating scale that has an equal number of
positive and negative categories in which a neutral or indifference point is at the center of
the scale. Balanced rating scale assumes normal distribution. Unbalanced rating scale
refers to a rating scale that has unequal positive and negative categories. That is,
responses in unbalanced rating scale are expected to be skewed to the positive or negative
side. For each question in your questionnaire, you need to choose between balanced and

71
unbalanced rating scale in accordance to empirical literature. For example, if empirical

researches state that Mekelle University students have more than average academic
performance then you need to frame the question such that the scale items could range
between average performance and the highest performance.
Single Measure or Index Measure
A variable that can be measured using one question is called single measure variable. For
example, net income is a single measure variable since you can get it by asking only one
question. A variable that can be measured using more than one question is called an index
measure variable. For example, the variable accountability is too complex to be measured
using only one question. Hence, you may prefer to ask set of questions that indicate
accountability thereby aggregating the responses into an index.
Maturity and Education of Respondents
Choice on the type and number of category level depends on the maturity and educational
level of respondents. That is, you need to consider the number of meaningful distinctions
that respondents could make. Otherwise, respondents may not understand the question
hence they may not respond to it or they may give misleading answers.
5.3.5 Characteristics of Good Questionnaire
It is widely said that a research survey is as good as the questions it asks. This is because,
if the data input is wrong, no doubt the output will also be wrong. Best questionnaires
have some common characteristics some of which are: 1) relevant and accurate questions,
2) correctly phrased questions, 3) non-forced choice of scale, 4) avoids leading and
loaded questions, 5) avoids order bias, 6) best layout, and 7) pre-tested and revised.
A questionnaire is relevant if it does not include unnecessary questions. It is said to be

accurate if reliable and valid data can be collected in response to the questions.
Reliability and validity are the most important characteristics of a good survey
questionnaire.

72
Question phrasing is related to the choice between open-ended and close-ended

questions. The cost of open-ended questions is substantially greater than that of fixed-
alternative questions, because the job of coding, editing, and analyzing the data is quite
extensive. One problem of close-ended questions is that in many situations a respondent
may not form an attitude toward a concept and simply may not provide an answer. A
good questionnaire includes scale categories such as “Do not know”, “Can not say”, “No
opinion” etc.
Good questionnaire avoids loaded and leading questions. A leading question is a question
that suggests or implies certain answers. A loaded question is a question that suggests
socially desirable answers or is emotionally charged answers. Good questionnaire avoids
questions that may induce bias because they cover two issues at once these questions are
termed as double-barreled items. Respondents usually do not want to trouble themselves
in giving you responses hence; it is better to choose operationalization, which is easy
enough hence a good questionnaire avoids burdensome questions that intensively use the
respondent’s memory. Further, a good questionnaire avoids implicit assumptions instead
it asks respondents about the assumptions themselves.
Order bias is a bias caused by the influence of earlier questions in a questionnaire or by

an answer’s position in a set of answers. Good questionnaire makes sure that any pattern
would not create biasedness in the responses. It uses funnel technique - a procedure
whereby general questions are asked before specific questions in order to obtain unbiased
response.
Best questionnaires have well designed layout. Questionnaire layout is of great

importance especially for mailed questionnaires. Respondents are usually motivated to
give responses to questionnaires of good layout. A good questionnaire is brief, small, and
not overcrowded. Last but not least, a good questionnaire undertakes pilot study before
the actual data are gathered thereby testing whether it is understandable by respondents.
If the questionnaire is not understandable to the respondents in the pilot study then it has
to be revised.

73
5.4 Testing Reliability and Validity of Questionnaire
Reliability refers to consistency or replicability of findings. Validity refers to the

truthfulness of findings. Questionnaire validity asks whether the questionnaire measures
what it is intended to measure. A study can be reliable but not valid – a consistently
wrong finding. But a study cannot be valid without first being reliable. You cannot
assume validity no matter how reliable your measurements are.
5.4.1 Testing Questionnaire Reliability
There are four good methods of measuring questionnaire reliability namely; test-retest,
multiple forms, inter-rater, and split-half methods. The test-retest technique of measuring
questionnaire reliability involves administering the test, instrument, survey, or measure to
the same group of people at different points in time. The questionnaire is considered
reliable if the correlation coefficient between two scores on the same group is
significantly higher. The question “how higher is ‘higher’?” is an empirical question
demanding empirical answer.
The multiple forms technique, otherwise called parallel forms or disguised test-retest, is
simply the scrambling or mixing up of the questions on the survey and administering it
twice to the same group. The questionnaire is considered reliable if the correlation
between the original questionnaire and the questionnaire with mixed up questions is
significantly higher. The question “how higher is ‘higher’?” is an empirical question
demanding empirical answer.
Inter-rater reliability is questionnaire reliability measurement technique that reports the

percentage of agreement on the same subject between two raters or data collectors. Inter-
rater reliability is most appropriate when you use assistants to do interviewing or content
analysis for you. The questionnaire is considered reliable if the raters or data collectors
collect similar responses on similar questions.
The split half method of reliability measurement classifies items of the questionnaire into
two. Correlation between the split up parts measures internal validity or consistency of

74
the questionnaire. Higher the correlation between the split up parts the more internal
validity the questionnaire has. Standard Statistics and Econometrics software routines
such as SPSS and STATA have features that calculate internal validity of a questionnaire
using split half method. The Cronbach’s Alpha and the Kuder-Richardson Coefficient are
cases in point.
5.4.2 Testing Questionnaire Validity
There are four methods of measuring validity or truthfulness of a questionnaire. These

are; face validity, content validity, criterion validity, and construct validity. Face validity
is an assertion on the researcher's part claiming that the researcher has reasonably
measured what she intended to measure. Usually, colleagues or experts in the field are
asked to vouch for the questionnaire items for face validity. Good questionnaires are
discussed on a panel of experts in the field.
Content validity looks into the correctness of conceptualization and further

operationalization of constructs and concepts into variables. Content validity can be
examined looking into the literature review part of the research proposal. A questionnaire
has content validity if all dimensions of a given construct or concept are included in the
form of operationalized questions. In other words, a questionnaire lacks content validity
if important indicators are overlooked.
Criterion validity uses some standard or benchmark that is known to be a good indicator.
Criterion validity can be classified into literature-based, concurrent and predictive
criterion validity. Validity of a questionnaire can be measured by comparing it with a
benchmark questionnaire in the literature. Concurrent validity refers to how well the
questionnaire estimates actual day-by-day behavior. For example, if the questionnaire is
related to “integrity of the independent auditor” then it has to include the data-by-day
behavior of a typical independent auditor. For that matter you can check whether your
questionnaire has concurrent criterion validity by observing the day-by-day behavior of
an independent auditor in your pilot study. Predictive validity refers to how well the
questionnaire estimates some future event or manifestation that has not happened yet. For

75
example, if your questionnaire deals with “determinants of financial embezzlement” you

can select an individual that participated in financial embezzlement and retrospectively
consider his life history before participation in financial embezzlement and then
investigate if the predictive factors on your scale include all important indicators that
determine financial embezzlement.
Construct validity refers to the extent in which the questions in your questionnaire are
derived from an underlying theory or model of behavior. Construct validity is the most
difficult validity to achieve.
5.5 Quantitative Analysis of Survey Data
Data analysis is the process of cleaning, transforming, and modeling data with a goal of
extracting useful information in the form of conclusion. Survey data are collected from
sample respondents using questionnaire. One way of analyzing survey data is the use of
descriptive statistics such as percentage, measures of central tendency, and measures of
dispersion. Unlike inferential statistics, descriptive statistics do not give results beyond
description.
5.5.1 Data Cleaning
Data cleaning is the process of identifying erroneous data and whenever possible correct
it in accordance to the dataset design. Dataset is a data matrix consisting of variables as
column entries and observations as row entries. Some of the filled questionnaires may
contain incomplete data. Other filled questionnaires may have conflicting responses.
These types of problems should be corrected before data entry. In effect, survey data are
cleaned before data entry.
5.5.2 The Data Matrix
Figure 1 shows an example of data matrix in MS Excel. The data matrix has 15
observations and 10 variables. That is, each observation has 10 variables. Data are

76
entered from the filled and cleaned survey questionnaire to the data matrix such as in
Figure 1.
Figure 1: The data matrix
5.5.3 Data Entry
The process of entering data from the filled and cleaned questionnaire to the data matrix
is called data entry. Each survey questionnaire is considered as an observation consisting
of a number of variables. Some variables are measured in the form of index. For example,
variables such as net income and depreciation are monadic variables. They can be
directly entered into the data matrix. But, variables such as accountability and motivation
are usually index variables. Therefore you need to calculate the index for accountability
and motivation for each sample observation. Accordingly, index value of accountability
and motivation can be entered into the data matrix.

77
5.5.4 Descriptive Statistics
Descriptive statistics are used to describe the basic features of a survey data. Summary
descriptive statistics are usually represented using simple graphs such as bar graph, pie
chart and line graph. Descriptive statistics can help you visualize distribution of the
sample. Descriptive statistics can be easily calculated and graphs can be generated used
MS Excel, STATA routine or other statistical packages. At the end of this chapter you are
going to be introduced to STATA econometric package.
5.5.5 Correlation
Descriptive statistics describes a single variable in the dataset whereas correlation

measures the degree of linear relationship between two variables in a dataset. Correlation
coefficient varies between +1 and -1. Two variables are perfectly negatively correlated if
all observations lie on any downward sloping linear line (i.e. correlation does not make
distinction between different slope values) and they are perfectly positively correlated if
all observations lie on any upward sloping linear line. A correlation coefficient of 0
designates that the two variables are not related to each other, that is, when all
observations are away from any possible linear line.
The most common methods of correlation coefficient calculation are Pearson correlation
and Rank correlation. Pearson correlation is obtained by dividing covariance between
variables by the product of their standard deviations as in Equation 5.5.1.
ρ=
∑ i
( xi − x )( yi − y )
5.5.1
∑ i
( xi − x ) 2 ∑ i
( yi − y ) 2
Rank correlation measure the extent to which relationship between two variables can be
described using monotonic function (not necessarily linear). If tied ranks do not exist,
Spearman’s rank correlation is given by Equation 5.5.2.
6 ∑ d i2
ρ =1− 5.5.2
n (n 2 − 1)

78
Where, d i = xi − y i and n refers to the number of raw data subject to ranking. If there
exist tied rank then rank correlation uses Equation 5.5.1.
Your dataset may contain more than two variables. In this case, partial correlation is used
to calculate the degree of relationship between any two variables holding the effect of all
other variables constant.
Given the dataset, you can manually calculate correlation and covariance between any
two variables. But, for a larger sample size manual calculation is tedious. Therefore, at
the end of this chapter you are going to learn how to compute correlation and covariance
using STATA routine.
5.5.6 Linear Regression
Correlation or partial correlation coefficient is a measure of goodness of fit. Correlation

values close to 1 indicate highest goodness of fit and correlation values close to 0 indicate
lowest goodness of fit. If the correlation coefficient is squared it measures the proportion
of variability in one variable that is accounted for with the knowledge of the other
variables.
Regression is a modeling technique that establishes a functional relationship between the

dependent and one or more independent variables. Linear regression analysis establishes
a theoretical linear line called “best fit line” that minimizes the squared distance between
an observation and the theoretical linear line. The best fit linear line can be
mathematically described using slope and intercept. Regression analysis explains the
change in the dependent variable as one of the independent variables change by one unit
holding all other independent variables constant.
Where yi is the dependent variable for observation i, x pi is independent variable p and
observation i, β p is slope of the independent variable p holding all other independent
variables constant, and ε i is the error term for observation i, the general multiple

79
regression model can be formulated as in Equation 5.5.3. The dependent variable yi in

Equation 5.3.3 is required to be continuous variable.
y i = β 0 + β 1 x1i + β 2 x 2i + . . . + β P x Pi + ε i 5.5.3
The regression model specification in Equation 5.5.3 puts several assumptions the most
important of which are: 1) representative sample, 2) regression parameters are linear but
the independent variables could be non-linear, 3) number of observations exceed number
of parameters, 4) error is a random variable with mean value of zero and constant
variance (i.e. no hetroscedasticity), 5) independent variables are linearly independent (i.e.
no multicolinearity), 6) errors are uncorrelated (i.e. no autocorrelation), and 7) value of
the dependent variable is postulated to depend on value of the independent variables (no
specification error).
The beta coefficients of Equation 5.5.3 are to be estimated using STATA routine or other
statistical software packages.
5.5.7 Analysis of Variance
Analysis of variance is a technique of evaluating the overall goodness of fit of the

regression model. The difference between the observed value yi and the mean value y is
known as total error. If the total error of each observation is squared and summed up it
results into total sum of squares (SS). If it is adjusted for degrees of freedom, the total
sum of squares (SS) measures variance or spread of observations away from their mean.
That is, variance is equal to total sum of squares divided by n – 1 where n stands for the
number of observations. Normalizing the total sum of squares for the degrees of freedom
is important especially when you need to compare variances of different sample sizes.
The total sum of squares can be divided into error explained by the regression model (i.e.
difference between predicted and mean value) and error not explained by the regression
model (i.e. difference between observed and predicted value). Part of the total sum of
squares that is not explained by the regression model is known as residual. The

80
coefficient of determination, also known as r – squared, can be computed as the ratio

between explained sum of squares and total sum of squares. But, the F-ratio is calculated
as a proportion of mean sum of squares from the regression model and mean sum of
squares from the residual. The F-ratio is used to test the null hypothesis that all of the
independent variables do not significantly determine the independent variable. If the
calculated F-ratio is greater than the table value at k level of significance and d degrees of
freedom then the joint null hypothesis is rejected. But, if the calculated F-ratio is lower
than the corresponding table value then the joint null hypothesis is accepted.
5.5.8 Qualitative Response Regression Models
The multiple linear regression model in Equation 5.5.3 requires the dependent variable to
be a continuous variable. However, it is not uncommon for Accounting and Finance
researches to use discretely valued dependent variable. These types of econometric
models are known as qualitative response regression models. Though the dependent
variable is discretely valued, numbers have to be assigned in order to use standard
qualitative response regression models. Equation 5.5.4 summarizes the total contribution
of all independent variables which is going to be used as an input to the logistic function
f(z), a function the converts z (that ranges from negative infinity to positive infinity) into
defined discrete values such as 0 and 1 for binary logistic function in Equation 5.5.5.
z = β 0 + β1 x1i + β 2 x2i + . . . + β P x Pi 5.5.4
ez 1
f ( x) = = 5.5.5
e + 1 1 + e −z
z
For example, if the dependent variable for your research had dichotomous values of
whether to invest (Y = 1) or not to invest (Y = 0) then the binary logistic function assigns
probability values (i.e. the probability of investing) between 0 and 1 depending on the
positive or negative overall contribution of the independent variables. Hence, the
probability of not-investing is 1 – f(x).

81
The beta coefficient estimates of logistic regression model measure the probability that
Y = 1 as the independent variable increases by one unit holding all other independent
variables constant. For interpretation purposes, the beta coefficient estimates of a logistic
regression model can be converted into odds ratio as in Equation 5.5.6.
βp
Odds Ratio = e 5.5.6
An odds ratio of 2 indicates, the probability that Y = 1 is twice as likely as the value of an
independent variable increases by one unit holding all other independent variables
constant; an odds ratio of 1 indicates that there is no relationship between the dependent
variable and the independent variable in question; and an odds ratio of 0.5 indicates, the
probability that Y = 1 is half as likely as the value of an independent variable increases
by one unit.
5.5.9 Regression versus Causal Relationships
Causality is the functional relationship between the cause and the effect. In logic, causes
are often classified into necessary cause and sufficient cause. Event A is a necessary
cause of event B if the presence of event B necessarily implies that event A is present but
the presence of A may not necessarily imply that event B is present. On the other hand,
event A is a sufficient cause of event B if the presence of A necessarily implies that event
B is present but the presence of event B may not necessarily imply that event A is
present. Panel 1 of Figure 2 illustrates causation in that, event A is both necessary and
sufficient cause of event B. In panel 2, event A is correlated to event B not because event
A is the cause but there exist an extraneous event C that is correlated to both events A
and B. In panel 3, event A is sufficient cause of event B and the extraneous event C is
also sufficient cause of B. But neither event A nor event C is necessary cause of B.
Regression analysis deals with prediction of values for a dependent variable using known
values of an independent variable. The influences of common variables and confounding
variables, as in Figure 2, prevent regression analysis from implying causation.
Nevertheless, if regression analysis is conducted on an experimental setting, that captures

82
the influence of common and extraneous variables, regression results may imply
causation.
Figure 2: Regression versus Causation
A variable or an event is considered as a cause of another variable or event if 1) it is

necessary and sufficient cause of the effect 2) the association between the cause and the
effect is strong, 3) different studies show that the association between the cause and the
effect is consistent, and 4) the cause precedes the effect in time.
5.6 Data Analysis Using STATA Routine
In this section you are going to learn the basic features of STATA version 11 and
interpretation of STATA results. Before all, STATA 11 has to be installed in your
computer before starting your analysis. You can start the program by selecting the
STATA 11 icon from the start menu. The results window opens automatically. STATA
has two windows: the results window and the data editor window. The results window
has a smaller window called command. In STATA 11 you can use either the user friendly
windows or commands. The data entry is performed in the data editor window and results
of analysis are displayed in the results window. The data editor window looks like an
excel sheet.
Let us assume that you want to do research on the impact of advertising expenditure on
viewers’ impression. You randomly selected 21 firms and surveyed their annual
advertising expenditure and the corresponding number of viewers. In effect, you have

83
two variables and 21 observations. Your dependent variable is impression (i.e. measured
in millions of viewers) and your independent variable is advertising expenditure.
Before you start working on STATA 11, always you need to log into it following file >>
log >> begin. STATA 11 asks you to save your work on a file with “.smcl” extension. By
the time you finish your job you need to close STATA 11 following file >> log >> close.
The data editor and STATA 11 graphs are saved on separate files with “.dta” and “.gph”
file extensions respectively. To save data editor and graphs follow file >> save as >>
save.
5.6.1 The Data Editor of STATA 11
To open the data editor window of Figure 3, you can go to the menu bar of the results
window and follow data >> data editor >> data editor (Edit) or you can simply write the
word edit in the command window then you get an Excel-like window – the data editor.
When you enter your data, observations are given roll number but you can also write
either name or code for your observation in our example names of observations are
written. STATA 11 gives red color to letter characters and black color to numeric
characters. STATA assigns default names to variables as Var1, Var2, Var3 and the like.
These default names have to be changed. In Figure 3, Var1 and Var2 are changed into the
variable names “Firms” and “Impression” but Var3 is not yet changed. To change
variable names you can follow data >> variable manager or you can simply double click
on the Var3 to get the dialog box containing fields for variable information. You can not
use space or reserved characters such as %, &, and # in the “name” field. If need be, you
can write any further information about the variable the “label” field. In the “label” field,
you can use space and the reserved characters in question.
There are three ways of data entry into the data editor of STATA 11. First, if you have
dataset in Excel format you can follow file >> import >> XML or if you have dataset in
ASCII format you can follow file >> import >> ASCII data in fixed format. Secondly,
you can simply copy data from Excel and past it into the data editor window. Thirdly,
you can also manually enter data directly into the data editor.

84
Figure 3: Data entry into STATA 11
5.6.2 The Statistics Menu of STATA 11
Given your dataset entered onto the data editor of STATA as shown in Figure 3 (or you
open an existing dataset of “.dta” file extension) you are ready for analysis. The statistics
menu of the menu bar in the results window contains varieties of descriptive statistics and
econometric models including linear regression, qualitative response models, time series
models, and panel data models.
Descriptive Statistics in STATA 11
To describe your dataset using summary statistics such as measures of central tendency,
measures of dispersion, measures of association, etc using the windows version of

85
STATA 11 you can follow statistics >> summaries, tables, and tests >> summary and
descriptive statistics >> summary statistics. Accordingly, the summary statistics dialog
box pops up.
Figure 4: STATA 11 output for Summary Statistics
Click on the drop down scroll in the summary statistics dialog box and click on the
variable name you want to calculate summary statistics. Automatically the results
window in Figure 4 appears. Results in Figure 4 show that mean value of Impression and
AdvertExp variables are 40.4667 million viewers and 50.37143 million dollars
respectively. Likewise, standard deviation, minimum and maximum values are also
displayed. Bear in mind that you can also include more statistics in the table.
You can also use STATA 11 commands to calculate summary statistics instead of using
windows version. If you write the following command summarize Impressions

86
AdvertExp and press enter, you will get exactly the same result as in Figure 4. STATA
understands short forms of its commands. For example, instead of writing summarize
you can shorten it to sum still STATA understands it as summarize!
Figure 5: STATA 11 output for linear regression model
Linear Regression in STATA 11
To run linear regression model you can use either windows or command. If you use
windows version of STATA 11 then follow statistics >> linear models and related >>
linear regression as a result a linear regression dialog box appears. The dialog box has a
heading that reads “regress – Linear regression”. The first word, “regress” in this case is
always the STATA command. In the dialog box select the dependent variable and
independent variables using the drop down scrolls. Accordingly, you get STATA 11
output in the results window as shown in Figure 5.

87
Bear in mind that when you use the windows version of STATA 11 (e.g. if you do not
know the command) STATA 11 returns the command along with the output. For
example, in Figure 5 the 10th row that reads regress Impression AdvertExp is the
appropriate command to get the results. Note that STATA 11 understands reg as the short
form of regress. Hence the command reg Impression AdvertExp can give you the result
in Figure 5.
Interpretation of Linear Regression Results
The linear regression model result in Figure 5 has two important tables: a table for
Analysis of Variance (ANOVA) and another table for regression estimates. The ANOVA
table is summarized using the F-ratio and the r-squared. The F-ratio is 14.02 (i.e. Mean
Squares of the Model divided by Mean Square of Residuals) and the probability greater
than F-ratio is 0.0014. That is, up to 99.86% of the cases calculated F-ratio of 14.02 is
greater than the table value. Therefore at 5% level of significance or even at 1% level of
significance (because the p value is only 0.14%) the joint null hypothesis is rejected.
Therefore, you can conclude that advertising expenditure has statistically significant
impact on impression of viewers.
The r-squared of 0.4246 (i.e. Sum of Squares from the Model divided by Total Sum of
Squares) it measures the goodness of fit of the regression model. In our example, the
linear regression model can explain only 42.46% of the cases the rest 57.54% of the
variation is accounted by residual factors. The adjusted r-square is similar to r-square
except that it includes an adjustment for the number of explanatory variables. That is,
adjusted r-square accounts for the loss of degrees of freedom as a result of including new
independent variables into the model.
The root MSE of 23.488 (i.e. the square root of the Mean Squares from residuals)
measures the standard error of the estimate. The root MSE is whose value is minimized
during parameter estimation process. Hence, root MSE can be used to make decisions on
whether to include additional explanatory variable into the model.

88
The coefficient column (shown as Coef.) of the linear regression result table in Figure 5
shows the sample estimates of the true population parameters and the standard error
column measures how well the sample represent the population. The bigger the standard
error the less representative is the sample. The t-statistic column is computed as the ratio
between the coefficient estimate and the respective standard error. If this calculated t-
statistic is greater than the table value for the t-statistic at a given confidence interval,
then the sample estimate can be used to approximate the population parameter. In our
example, assuming 95% confidence interval or 5% significance level, the calculated t-
statistic values of AdvertExp and the constant are 3.74 and 3.33 respectively are both
greater than the two tailed t-test table value of 2.093. Therefore, the null hypothesis is
rejected.
Similarly, the p values and the confidence interval columns can be used to test the null
hypothesis. Accordingly if the p value of a coefficient is lower than 5% (i.e. at 95%
confidence interval) then we reject the null hypothesis thereby accepting the alternative
hypothesis that claims the independent variable in question has statistically significant
influence on the dependent variable. Likewise, if the confidence interval does not contain
0 in between the lower and upper bound then the null hypothesis is rejected. In other
words, if the confidence interval includes 0 in between the lower and upper bound then
the null hypothesis is accepted, that is, the coefficient estimate can not be used to
represent the population parameter.
Bear in mind that the linear regression model for our example on “the impact of
advertising budget on viewers’ impression” was specified as in Equation 5.6.1. Since the
population parameters are estimated using sample statistics in Figure 5, the fitted
regression equation is given by Equation 5.6.2.
Impressioni = β 0 + β 1 AdvertExpi + ε i 5.6.1
Impression = 22.16 + 0.36 AdvertExp 5.6.2

89
The slope coefficient of Equation 5.6.2 can be interpreted as follows: as a firm increases
advertisement expenditure by 1 million dollars, the number of impressed viewers
increases by 360,000 viewers.
Figure 6: STATA 11 output for binary logistic regression model
Binary Logistic Regression Model in STATA 11
Binary logistic regression model has a nominally measured dependent variable. In order
to run binary Logistic regression model let us modify our dependent variable into a
binary variable, say, if the number of viewers is below 30 million then the firm did not
create impression on viewers but if the number of viewers is greater than 30 million then
the firm created impression on viewers. To put it differently, the “did not create
impression” category is coded as 0 and the “created impression” category is coded as 1.

90
Figure 6 shows STATA 11 output for binary logistic regression model. At 95%
confidence, the coefficient estimates of the model are significant because the p – values
of both the independent variable and the constant are below 5% and the confidence
intervals do not contain 0 with in the range. Therefore, the binary logistic model can be
written as in Equation 5.6.1 where the value of z is given by Equation 5.6.2.
1
Chance of Impressing = 5.6.1
1 + e −z
z = − 2.59 + 0.08 AdvertExp 5.6.2
In the binary logistic regression fit of Equation 5.6.2, as advertising expenditure increases
by 1 million dollar, the firm has an 8% probability of impressing viewers. If you want to
predict the probability of impressing viewers as a firm spends 10 million dollars on
advertisement, first you need to calculate z as follows: z = - 2.59 + 0.08 *10 = - 1.79.
Then you need to substitute z = - 1.79 into Equation 5.6.1 you will get 14.3%. Hence, a
firm that spends 10 million dollar on advertisement budget has 0.143 probability of
impressing viewers.
5.6.3 The Graphics Menu of STATA 11
To illustrate varieties of graphs let us re-consider the original Example on the impact of
advertisement expenditure on viewers’ impression. To create a scatter plot graph using
the windows version of STATA 11 follow graphics >> two way graph (scatter, line, etc)
>> create. As an alternative you can directly write the STATA command twoway
(scatter Impressions AdvertExp) in the command window to get Figure 7.
STATA 11 has a special feature of graph editor. You can edit your graph in the graph
window following file >> start graph editor. You can double click on the graph it self to
edit different aspects of the graph.
To create a histogram along with normal density plot then on the results window follow
graphics >> histogram >> add normal density plot. You can get the “add normal density

91
plot” option in the density plot tab of the dialog box of histogram. Alternatively, you can
use the STATA command histogram Impressions, bin(20) normal density to plot
histogram for Impressions variable using 20 bars as shown in Figure 8.
Figure 7: STATA 11 Scatter plot output
5.6.4 How to Use Help in STATA 11
STATA 11 has a well documented help system. If you want to read STATA manuals, in
the results window, you can follow help >> PDF documentation. The documentation has
15 book volumes. If you want to specifically study STATA commands then you can
follow help >> advise. Further, the best way of learning STATA 11 is to Google your
question! Many STATA funs explain a lot of tricks.

92
Figure 8: STATA 11 Histogram Plot
Chapter Summary:
- Survey deals with generalization but case study deals with particularization
- The four levels of measurement are: nominal, ordinal, interval and ratio scales
- Questionnaire is constructed using different types of questions
- Internal validity of a questionnaire measures the degree of questionnaire consistency
- Quantitative survey data analysis usually uses computer software routine
- Regression analysis dependent variable with its independent variables
- Regression analysis can have a nominally or ordinal measured dependent variable
- Regression analysis does not necessarily imply causal relationship
- STATA 11 has easy to use help feature

93
Case Study Method and Analysis Mekelle University, CBE
CHAPTER 6
CASE STUDY METHOD AND ANALYSIS
6.1 Introduction
A case study is an empirical inquiry that investigates contemporary phenomena within its
real life context, especially when the boundaries between phenomenon and context are
not clearly evident. In addition, phenomena within a given context have overlaps or
interaction. Case study method is helpful to study a phenomenon while it interacts with
other phenomena. Most organizational researches employ case study approach because
organizational problems or gaps can not be researched out of their intra and inter
organizational context. In case studies, data collection and analysis method has to be
thoroughly planned ex ante. A mix of qualitative and quantitative approaches can be
used. The varieties of models in operations research method are typical instances of a
quantitative approach to case study methods.
Key Words: Case study, Particularization, Scientific Paradigm, Positivism, Constructivism,

Factor Analysis, Analytic Induction, Life History Method.
Chapter Objectives:

- Know the detailed characteristics of case study research method
- Differentiate positivism paradigm from constructivism paradigm
- Acquire the skill of analyzing case study data
- Understand Operations Research method as one type of research method
6.2 Scientific Paradigms
Scientific paradigm is the conceptual framework that helps you look into the social
world. Scientific paradigms are analogous to “eyeglasses”. If you wear black eyeglasses
you might observe your environment as black and if you wear 3D eyeglasses you might

94
strangely observe your surrounding environment. Similarly, social reality could depend
on the scientific paradigm you are in. Most disciplines have their own scientific
paradigm. For example, the dominant scientific paradigms of Accounting and Finance are
different from the dominant scientific paradigms of Sociology. However, with the advent
of inter-disciplinary approach to research scientific paradigms of many disciplines are
getting “cross-fertilized”. Especially, if you want to do qualitative research, it is very
important to know the scientific paradigm you are in as it is important to know the type
of eyeglasses you worn in order to visually describe the environment.
A scientific paradigm is established on the top of philosophical assumptions that are the
basic beliefs of human beings about the natural and social world. The philosophical
assumptions can be categorized into four dimensions namely, ontology, epistemology,
axiology, and methodology.
Ontology is branch of philosophy that deals with the nature of reality, classifications of
reality, and relation among the classes of reality. The two poles of ontology are realism
and nominalism. The realist views the social world as an aggregate of tangible or
objective abstract reality that exists independent from human beings’ thought faculty. But
for the nominalist the social world is an aggregate of manmade subjective concepts that
do not really exist outside human beings’ thought faculty. The nominalist believes in
realities that differ from person to person, that is, existence of multiple realities. For
example, assume that you are interested to do research on the topic “role of institutions in
growth and development: a case study of EDIRS1 in Mekelle city”. Does the concept of
EDIR have an objective existence independent of your mind? It depends on whether you
are realist or nominalist. If you believe that it has an objective and independent reality,
then you are claiming that the concept can be explained using mathematical entities and
accordingly general laws can be established thereto.
Epistemology is second dimension of the philosophical assumption of a scientific

paradigm. Epistemology is concerned with the nature and scope or limitation of
knowledge. The polar epistemological stances are positivism and phenomenalism.
1
EDIR is a traditional form of institution in Ethiopia.
95
Positivism refers to a perspective in which the scientific method as the best method to
study both physical and social worlds. A positivist believes that the social world can be
objectively studied and looks into component parts in search of general patterns that
could explain the whole. The positivist formulates hypotheses and tests them against the
objective facts on the ground. The contemporary version of positivism is logical
positivism or neopositivism. Logical positivism attempts to reduce statements and
propositions into pure logic and it adheres to the concept of verifiability. As opposed to
verification, Karl Popper amends logical positivism by introducing the concept of
falsification. He claimed that a theory can not be verified but it can be falsified. He called
his theory post-positivism. In general, behind quantitative researches there exists
positivism, logical positivism, or post-positivism as a guiding scientific paradigm.
On the other hand, phenomenalism is the study of phenomena which literally means
appearances as opposed to reality. It claims that the social world is socially constructed
and the observer is part of it and hence the unit of analysis can be studied by fixing the
frame of reference of the observed in action. Phenomenalism tries to understand what is
happening looking at the unit of analysis in its totality and within its context and develops
ideas from the data through induction. Phenomenological paradigm can be divided into
critical theory and constructivism among others. Critical theory is a social theory oriented
toward critiquing and changing society as a whole in contrast to the traditional theory that
is oriented only to understanding or explaining it. Constructivism is an epistemological
philosophy that asserts human beings construct their own understanding of the natural
and social worlds from within themselves. In effect, it claims that reality is subjective.
Axiology is the third dimension of philosophical assumption in scientific paradigms. It is

concerned with the notion of value, that is, issues related to right or wrong, good or bad,
beautiful or ugly, smart or dull, and the like. Axiology is particularly linked to the social
world.
Methodology is the fourth dimension of philosophical assumption in scientific

paradigms. Methodology is the study of research methods. It deals with the “how” aspect
of research. Methodology can be divided into two types of paradigms: quantitative and

96
qualitative. Quantitative method intensively employs the scientific method as implied by

logical positivism. The method asserts that mathematical or statistical approach can
explain both the natural and the social worlds. Qualitative method looks into the
subjective context of the unit of analysis with in the social world as implied by
phenomenologism.
6.3 When to use case studies?
A case study is a method of learning about a complex instance, based on a comprehensive

understanding of that instance obtained through extensive description and analysis of that
instance taken as a whole and in its context. The most important characteristics of a case
study are: 1) presence of a complex instance, 2) comprehensive understanding of the
context, 3) holistically studied instance within its context, and 4) focuses on
particularization instead of generalization.
A case study research method is appropriate if you are interested on the “why” and
“how” questions pertaining to people, objects, phenomena, etcetera. On the other hand, if
you are interested on the “what”, “who”, “where”, and “how many”, the appropriate
method is survey or document analysis. Hence, case study method is more appropriate in
descriptive and explanatory researches. Accordingly, there are three types of case studies
namely descriptive, explanatory, and evaluative case study.
Descriptive Case Studies
Descriptive case study describes an intervention within the context it operates.

Descriptive case studies could be illustrative, pilot, or critical instance. Illustrative
descriptive case studies provide in-depth examples of a program, project, or policy.
Illustrative case studies are used to make the unfamiliar familiar by using description
method. Pilot descriptive case studies are case studies performed before a large scale
research is done. Pilot descriptive studies are particularly important when the large scale
research project is on a topic of interest where little is known about the intervention or the
potential outcomes. Critical instance descriptive case studies focus on examination of a
single instance that serves as a critical test of an assertion about a program, policy,
97
problem, or strategy. Critical instance descriptive case studies are not interested in
generalization.
Explanatory Case Studies
Explanatory case study is concerned with contextual explanation of causal links between
interventions and program, project, or policy outcomes. Explanatory case studies could
be process oriented or means-end oriented. Process oriented explanatory case studies
normatively investigate operations in different departments or divisions of an
organization or different organizations. It deals with program or project implementation.
Means-end oriented explanatory case studies establish cause and effect relationship
between means and ends. For example, case studies dealing with contextual impacts of an
intervention, such as a program, a project, or a policy, are means-end oriented
explanatory case studies.
Evaluative Case Studies
Evaluative case studies are systematic assessment of worth of a program, policy,

technology, person, need, activity, and so on. Evaluative case studies end up on
qualitative judgments such as good or bad, right or wrong, worthy or worthless,
successful or not successful, and the like. Evaluative case studies can be broadly
classified as formative and summative. Formative evaluative case studies are used to
improve the thing that is evaluated. Some examples of formative evaluative case studies
are: needs assessment, evaluability assessment (i.e. whether a given evaluation system is
feasible), implementation evaluation, and process evaluation. Summative evaluative case
studies deal with the outcome of an intervention. Some examples of summative
evaluative case studies are: outcome evaluation, impact evaluation, and cost benefit
analysis.
6.4 Case study research design
Case study research design is the plan and structure of investigation so conceived as to
obtain answers to case study research questions. A case study research design within

98
positivist paradigm is different from a case study research design within constructivist
paradigm. The positivist paradigm takes a deductive approach to case study design in
which the definition of research questions and propositions come in advance of data
collection and quantitative analysis is preferred. The constructivist paradigm takes an
inductive approach to case study design in which contextualized data collection precedes
definition of research questions, propositions, and insights.
6.4.1 Case Study Design using Positivist Paradigm
Positivist case study research design has at least seven components namely: 1) Research
question formulation, 2) The study’s propositions, 3) The study’s units of analysis, 4)
Methods of case study data collection, 5) Generalization, validity, and reliability, 6)
Methods of case study data analysis, and 7) The criteria for interpreting findings.
Research question formulation
Since positivist paradigm presumes existence of objective reality in both the natural and
social worlds, the positivist researcher tries to look for a research problem in light to the
established literature in the area of interest. A conceptual or theoretical framework is
developed based upon existing literature in order to equip yourself with an outlook such
that you could be able to vividly comprehend the research problem. A synthesis of both
theoretical and empirical literature in the area of interest is helpful in order for you to
reflect the gap your research opts to fill in. If you are able to successfully reflect the gap
then you defined your research problem. Your research problem can be even clearer if it
is summarized in the form of research question. The research question is a center of
gravity for your research project.
However, availability of theoretical and empirical literature is not the same for
exploratory, descriptive, and explanatory case study research topics. If your case study
research topic has exploratory purpose then you may try to focus on the conceptual
framework and try to reflect the gap in understanding. This is going to be your research
problem.

99
The Study’s propositions
The main difference between positivism and constructivism is that the former puts study
propositions ex ante whereas the latter extracts those propositions ex post. In other words,
the former follows deduction logic whereas the latter follows inductive logic. Hence, in a
positivist paradigm, descriptive and explanatory case studies need propositions. Research
questions are translated into propositions such that you would be able to evaluate them as
either “true” or “false” based on the empirical case study data you are going to collect.
Study propositions are like hypotheses of survey research, they are accepted if empirical
evidences are in support and they are rejected if empirical evidences are against.
Case studies that have exploratory purpose do not have study’s propositions since the
topic of interest involves a gap in understanding. Hence, for exploratory case studies,
study propositions are the conclusions based up on which descriptive and explanatory
case studies might follow.
Study’s unit of analysis
The unit of analysis is the basis for any case study. The unit of analysis of a case study
may be an individual person, group of persons, an event (such as a decision, a program,
an implementation process, or organizational change), a department within an
organization, an organization, or group of organizations. Sometimes it is difficult to
identify boundaries of the unit of analysis. Yet it is very important to delimit the unit of
analysis. Selection and delimitation of the units of analysis is determined by the research
purpose, questions, propositions and theoretical context.
Method of case study data collection
In a positivist paradigm of a case study research, the study’s propositions are supposed to
be tested using empirical evidence. Method of data collection section of a case study
research design links the study’s propositions with empirical data. You may use any one
or a combination of the primary and secondary data collection tools and techniques
discussed in chapter 4 of this teaching material. In any case, your method of data

100
collection should result in sufficient empirical data that can be used as a set of evidence
to evaluate the study’s propositions as “true” or “false”. In other words, your case study
data collection apparatus (e.g. observation checklist, interview questionnaire, focus group
discussion agenda etcetera) are expected to be valid and reliable.
The actual data collection of your case study research should be guided by a case study
protocol. This protocol needs to include sections such as: 1) an overview of the case
study project, 2) field procedures or observation checklist and access arrangements to
these sources, 3) case study questions that you need to keep in mind when collecting data.
The case study protocol is particularly important if you intend to use multiple case studies
in your case study research project. Multiple case studies require a team of researchers
gathering case study evidence. Case study protocol is one way of standardizing the case
study data collection process.
Regardless of the source of data you are planning to use in your case study research, you
need to include triangulation and case study database as part of your case study research
design. Triangulation is a method of assuring case study data reliability by comparing
two or more sources of case study data collection on the same fact. If these different case
study data sources result in convergent findings then the case study data in question is
said to be triangulated. In addition, you can also use several investigators from one case
study data source. The resulting degree of convergence measures the strength of
triangulation. Moreover, you need to design a database in a way that could help to index
the triangulated case study data. Your database may index case notes made by the
investigators, case study documents that are collected during a case study, interview notes
or transcripts, and analysis of the evidence. Case study databases are kept for further
reference.
Generalization, Validity and Reliability
A case study research within positivist paradigm can be included into knowledge base of
the field of study if it is characterized by generalizability, validity, and reliability. In case
studies within positivist paradigm, the term “generalization” does not refer to the

101
statistical generalization such as in survey research method. It refers to the analytic

generalization in which the case study research design is deduced from the theoretical or
conceptual framework. Generalization would be more meaningful and value adding to the
existing literature if research results of two or more case studies support the same theory.
You need to state the analytic generalization quality of your case study research and you
need to explain how your case study research adds value to the existing knowledge base.
Method of case study data analysis
Your case study research design has to include a section on method of case study data
analysis. This section presumes the presence of a case study research database that is
tested for validity and reliability. Based upon the nature of your case study’s propositions
and your method of case study data collection you need to code, tabulate, and cross-
tabulate if you ex ante know the distribution of the responses. For example, if your
observation checklist requires you to observe customer’s satisfaction level, you ex ante
know the expected distribution of your responses. Therefore, you can design the coding
and tabulation of some case study questions. On the other hand, there are case study
questions whose distribution of the responses are known but ex post. These questions are
not included in the case study research design but they are going to be coded and
tabulated after the actual case study data is collected.
In method of case study data analysis design what matters is the logical link between
analysis and the crafted propositions. Can the method of analysis satisfactorily evaluate
the study’s propositions? This question has to be answered. If the proposed method of
case study data analysis can not help you generate sufficient evidence to evaluate your
study’s propositions then you need to look for another method or combination of methods
of case study data analysis. You can use a combination of quantitative and qualitative
methods of case study data analysis.
The criteria for interpreting findings
You may evaluate your study’s propositions as “true” or “false” using sufficient
evidence. Moreover, while analyzing the case study data you may find important patterns
102
that require attention. These are all case study research findings. They need to be
synthesized in order for the case study research to imply important conclusions in
response to the research questions asked. In other words, evaluative meanings are
attached onto the case study research findings in order to give answers to the research
questions asked. You need to define the evaluative meanings, that is, your case study
research design has to include the criteria for interpretation of findings.
Exploratory case study is an exception
Exploratory case study research designs are a bit different from descriptive and
explanatory case study research designs. This is because, study’s propositions of an
exploratory case study research are not known ex ante. But you need to provide a
conceptual or theoretical framework to guide your case study research and in light of
which the case study research questions are crafted. The study’s propositions are known
ex post; in fact the main purpose of most exploratory case studies is to create better
understanding about the unit of analysis. At best, an exploratory case study comes up
with propositions.
Therefore, research design of an exploratory case study has at least the following
component parts 1) Research question formulation, 2) The study’s units of analysis, 3)
Methods of case study data collection, 4) Generalization, validity, and reliability, 5)
Methods of case study data analysis, and 6) The criteria for interpreting findings.
6.4.2 Case Study Design using Constructivist Paradigm
Unlike positivism, constructivism views reality as subjective construction of human

beings. That is, people (including the researcher) construct the realities in which they
participate. Constructivist case study design follows an inductive logic approach of
proposition development in that grounded propositions are formulated from the empirical
data. In contrast, the positivist case study design follows a deductive logic approach in
which literature driven propositions are formulated ex ante and empirical data are used as
evidences to test whether each proposition is “true” or “false”. Therefore, case study
design using constructivist paradigm starts with the experience and asks how members of
103
the unit of observation construct a given reality. A constructivist case study research
design may have at least three component parts namely: 1) The case study’s units of
observation, 2) The case study’s context, and 3) Methods of case study data collection.
The case study’s units of observation
The constructivist case study research design has to include a clear description of the case
study’s units of observation. Identification of units of observation deals with definition of
a system within which the units of observation operate. For example, if you want to
research on “impact of ‘staff empire building’ on quality of education in Mekelle
University: A constructivist approach” then your analytical inquiry may focus on how
staff members of Mekelle University build empires and how they perceive educational
quality. Hence, Mekelle University is your case study’s organizational delimitation and
the staff members of Mekelle University that are involved on ‘empire building’ are your
units of observation. A case study, be it positivist or constructivist, can have few or many
units of observation but it usually has few units of analysis.
The case study’s context
Context can be defined as the attitude, surroundings, circumstances, environment,

background or settings which determine, specify, or clarify the meaning of an event,
object, behavior, or any part of the social world. Case entity refers to the entity for which
the case study research is going to be done. For example, in the case study research
“impact of ‘staff empire building’ on quality of education in Mekelle University: A
constructivist approach” the case entity is Mekelle University. The boundary of the case
entity in question (e.g. Mekelle University’s boundary) defines scope of the case study
research. The boundary can have political, economic, social, technological,
environmental, legal, and cognitive dimensions. Case entity elements refer to the human
resources, objects, actions, phenomena, and etcetera that are possessed and or owned by
the case entity. Case entity’s boundary is used to differentiate case entity elements that
are internal to the entity (e.g. resources and actions that belong to Mekelle University)
from elements that are external to the entity.

104
A context can be broadly classified as internal and external to the case entity. Internal
context refers to the social reality that binds all case entity elements. In constructivist
paradigm the social reality is presumed to be constructed by human beings since
inception of the entity. Therefore, internal context mainly captures the case entity’s
culture. The case entity’s culture is the effect of horizontal and vertical interactions
among cognitive, affective, and behavioral dimensions of the human resources within the
case entity. Horizontal interaction refers to the cognitive-to-cognitive, affective-to-
affective, or behavior-to-behavior influences among human resources within the case
entity. Vertical interaction refers to the bi-directional influences between cognitive of an
individual and affective or behavior of another individual within the case entity. In fact,
culture is not the only factor that establishes internal context for your case study research.
The internal context looks into the effect of political, economic, social, technological,
environmental, and legal shocks on the case entity’s elements and/or on their network of
interactions.
External context refers to the political, economic, social, technological, environmental,

and legal system of, usually, a sovereign country within which the case study research is
going to be done. Positioning a case study within its internal and external context helps
you to holistically and vividly see your case study (and of course including yourself since
you are part of the reality!) and the multitudes of dynamic interactions among the case
entity elements.
Methods of case study data collection
A constructivist case study research design could benefit from the statement of methods
of case study data collection. You may use one or more of the primary and secondary
data collection methods explained in chapter four of this teaching material.
6.5 Data Analysis in Case Studies
The philosophical assumptions and type of case study determine the type data analysis to
be used in case studies. For example, for a positivist descriptive case study that aims at
exploring a topic of interest, factor analysis can be more appropriate. On the other hand,
105
for a constructivist descriptive case study life history method of data analysis is more
appropriate.
Table 1 presents some of the appropriate methods of data analysis for different numbers
of units of observation even though the list is not exhaustive. According to Table 1, the
degree of qualitativeness of a case study analysis increases as the number of units of
observation decreases. This section presents Factor analysis, Analytic Induction, and Life
History methods of case study data analysis.
Table 1: Methods of data analysis in case study and survey researches
Units of Analysis
One Few Many
Observation
Units of
One Life History Statistical

Few Analytic Induction Statistical
Many Factor Analysis Statistical
Particularization Generalization
(Case Study) (Survey)
Unlike quantitative data, that is collected using mail questionnaire, qualitative data are
usually not ready for analysis. Therefore, you need to perform series of editing and
coding so as to make the qualitative data ready for analysis. Editing and coding are very
limited if the underlying paradigm of the case study is constructivism. But this does not
mean that editing and coding are not practiced in constructivist case studies.
Editing case study data
To ensure the completeness, consistency, and meaningfulness of the data, you need to
editing procedures have to be planned ahead so that the data will be checked for
consistency, completeness, and meaningfulness. Case study data can be edited in the field
or in-house after the data is actually collected. Field editing is usually done by the data
collectors and their supervisors, if any. For example, in an interview, the data collector
106
may clarify a question if the respondent tends to give inconsistent answers. On the other
hand, after the data collectors finish collecting data, the supervisor may compare the data
from different data collectors and check the overall consistency. In-house editing is
detailed investigation on the overall completeness, consistency, and meaningfulness of
the data. One problem of data editing is that the editor might create bias by introducing
subjectivity.
Coding case study data
Coding is the process of systematically classifying information and assigning codes such
as numbers. For example, if some of the responses are “I don’t know” or “not applicable
for the respondent, then you can assign a number that you can not forget for instance you
may use the number 99 to designate “I don’t know” and you may use the number 77 to
code the “not applicable” response.
The codes you assign need to be recorded. It is recommended to use a code book in that
all the codes are recorded along with the content they denote. Coding makes data analysis
a lot easier. Nevertheless, if the case study involves a single unit of analysis, a single unit
of observation, and if it follows a constructivist paradigm then you may not dwell much
on coding. Because, you are expected to show how the unit of analysis construct her or
his own world within a defined context.
6.5.1 Factor analysis
Factor analysis is a statistical technique used to describe variability of observed variables

using small number of unobserved variables called factors. Factor analysis can be used to
analyze case study data that have a higher number of variables and units of observations,
that is, large datasets. Factor analysis presumes a large dataset in order for the correlation
coefficients between variables to be reliable. Factor analysis is used for different
purposes such as: 1) data reduction, that is, to reduce a relatively dataset into small
number of common factors, 2) empirical classification or description, 3) development of
a scale, 4) reduce correlated independent variables into uncorrelated factors in order to

107
apply multiple regression technique, 5) exploration of unknown domains, and 6) mapping

the social phenomena using factors and sources of variations.
The two types of factor analysis are exploratory and confirmatory. Exploratory factor
analysis tries to extract the underlying dimensions or factors given a set of variables.
Exploratory factor analysis does not assume presence of a theory or propositions ex ante.
Hence, it can be useful for data analysis in constructivist case studies. Confirmatory
factor analysis starts with a theory or set of propositions involving complex relationships
among concepts. It tries to test whether a data reduction technique on a multiple measure
could result in the same concepts as in the theory. Confirmatory factor analysis is
particularly useful for data analysis in positivist case studies.
The easiest way to understand factor analysis is to look into the intercorrelation patterns
of the variables in your dataset. Here, it is assumed that variables that have high
correlation are due to the fact that they measure the same object, phenomena, event, and
the like. Hence, highly correlated variables can be clustered to form fewer factors.
To illustrate factor analysis let us use data from the paper by Gebregziabher (2009a) titled
“the determinants of public employees’ performance: evidence from Ethiopian public
organizations”. In this paper, employees’ performance was modeled as a function of
ability, motivation, and role perception. The writer used four questions to measure role
perception to collect data from 120 respondents. Would these four questions be reduced
to smaller number of factors? The questions are listed below.
1. How complex is your job now compared to your job before BPR Program?
Very simple 1 2 3 4 5 Very complex
2. Do you clearly know the core, major and minor roles that the result oriented
system require you to perform? No 1 2 3 4 5 Yes
3. Do you feel that you are performing according to the objectives of Result
Oriented System of Performance Assessment? No 1 2 3 4 5 Yes
4. How important, do you think is your job to achieve the overall objectives of your
organization? Less important 1 2 3 4 5 Very important

108
Dataset in Figure 1 below shows a portion of the numeric responses to the four questions
mentioned above entered onto data editor window of STATA.
Figure 1: Dataset ready for factor analysis
To get the results on Figure 2, on the results window of STATA follow statistics >>
multivariate analysis >> factor and principal component analysis >> factor analysis and
then select the variables you want factors to be extracted from. Before interpreting Figure
2, it is better to define the important jargons of factor analysis.
Factor loading is the first jargon in factor analysis. It measures the degree of correlation
between a variable (i.e. questions in our example) and a factor that has been extracted
from the data. For example, Figure 2 shows that factor loadings of question1 are 0.6061
and - 0.0966 respectively. That is, the question1 is highly correlated to factor1.

109
Figure 2: Factor analysis result (unrotated)
The second jargon is communality, that is, the total amount of variance in the variables
that is explained by the extracted factors. Communality is the sum of squares of factor
loadings for a given variable. For example, from Figure 2, communality of question1 is
equal to 0.3767 (or 0.60612 + 0.09662 = 0.3767). This means, 37.67% of the variability in
question1 can be explained by the two factors extracted. Similarly, communality of
question4 is equal to 0.2001 (or 0.40502 + 0.18992 = 0.2001). That is, 20.01% of the
variability in question4 can be explained by the two factors extracted. In Figure 2, the
uniqueness column measures the unexplained total variance of each variable (i.e. 100
minus communality of a variable in percentage).
The third jargon is Eigenvalue. The Eigenvalues of a factor analysis are calculated as the
square summation of factor loadings of all variables with a given factor. For example,
Eigenvalue of factor1 is 1.02842 (i.e. 0.60612 + 0.64992 + 0.27322 + 0.40502 = 1.02839).
110
The difference is due to rounding error. According to the principal components method,
each variable is standardized to have 0 mean and + 1 standard deviation. That is, each
variable has 0 mean and 1 variance. In our example, the total variance to be explained is
4 (i.e. the sum of variances of each variable). Hence, in our example, a factor is
considered as important if the associated Eigenvalue is greater than 1 and it is considered
as perfect explanatory of the total variance if Eigenvalue is 4. For example, in Figure 2,
Eigenvalue of factor1 is 1.02842. The factor explains 25.71% (i.e. 1.02842/4 = 0.2571) of
the total variance the principal component analysis presumes. Therefore, in our example,
only factor1 is important.
In the table for Eigenvalues of Figure 2, the difference column measures the difference
between Eigenvalues of consecutive factors. For example, the difference value of
0.89474 for factor1 is given by 1.02842 – 0.13368 = 0.89474. Similarly, the difference
value of 0.27355 for factor2 is given by the difference between 0.13368 and – 0.13987.
The proportion column of the same table in Figure 2, measures the relative proportion of
each Eigenvalue relative to the total sum of the Eigenvalues that is 0.76979. Hence, in
Figure 2, the proportion value of 0.1737 for factor2 is given by 0.13368/0.76979 =
0.1737.
You may use the principal components method of analysis as an alternative to factor
analysis in Figure 2. In STATA, the principal components method of analysis can be
done in the results window following statistics >> multivariate analysis >> factor and
principal component analysis >> principal component analysis as to get results as in
Figure 3. You may compare and contrast the factors in Figure 2 with the components in
Figure 3.
Graphic approach is the second way to extract factors. The graph is known as Scree plot.
It plots the Eigenvalues against the number of variances. In STATA, the Scree plot can
be plotted after the initial solution is determined using either factor analysis or principal
component analysis or other method. To get the Scree plot in Figure 4, in the result
window of STATA follow statistics >> multivariate analysis >> factor and principal
component analysis >> post estimation >> Scree plot of Eigenvalues. In the Scree plot,

111
you can easily observe the factors with Eigenvalue greater than 1. Factor plots that are
above Eigenvalue of 1 are considered important. For example, in Figure 4, only factor1
has Eigenvalue above 1.
Figure 3: Factor analysis using principal component analysis method
Factor rotation
The initial solution of factor analysis as in Figures 2, 3, and 4 is not the final solution.
The solution may be improved by rotating the axes such that the variables would better fit
to the nearest axes. A factor analysis is said to have an improved solution if the total
variance the factors explain increases. In order to perform factor rotation, in the results
window of STATA follow statistics >> multivariate analysis >> factor and principal
component analysis >> post estimation >> rotate loadings. Keep in mind that factor
rotation is performed after the initial solution is calculated using either factor analysis or
112
principal component analysis. Factor rotation result of our example, based upon the initial
solution in Figure 2, is shown in Figure 5. If you want to run factor rotation for a
principal component analysis then you need to perform post estimation based upon
results in Figure 3.
Figure 4: Scree plot of Eigenvalues
Scree plot of eigenvalues after factor

1
.5
Eigenvalues
0
-.5
1 2 3 4
Number
Naming the factors
Once factors or components are extracted, you need to assign meaningful labels to them.
For example, in Figure 2, the most important factor is factor1 and the highest factor
loadings are that of question1 and question2. The questions were “How complex is your
job now compared to your job before BPR Program?” and “Do you clearly know the
core, major and minor roles that the result oriented system require you to perform?” The
first option is to select a term or phrase that can represent both questions that are
clustered within factor1. For example, the label “relative job complexity” may be used to
label factor1. However, the label may have different connotations to different people. The
113
second option of factor labeling is using symbols such as the A, B, and C classification of
inventories.
Figure 5: Factor analysis result (rotated)
6.5.3 Analytic induction method
Analytic induction is method of case study data analysis that seeks to accurately represent
the reality of an event, situation, or object using the induction method of logical
reasoning. Based on the empirical data on the ground, analytic induction makes tentative
“universal” statements or propositions that are going to be modified if exceptional cases
are discovered. The modification could be on the explanandum or on the explanans or on
the definition of the phenomena. Analytic induction could result in better causal
relationships among variables if you are considering multiple case studies. If more cases
are considered, analytic induction could result in “universal” explanations usually known
114
as grounded theory. Analytic induction method sufficiently deals with the necessary
causes but not with the sufficient causes of a given explanandum. For example, the data
analysis method does not help you predict “who will be involved in corruption” but it can
help you investigate “who was involved in corruption” if the phenomenon of corruption
has already happened.
Johnson (1998) explained the analytic induction method of case study data analysis using
four steps namely 1) gaining access to the phenomenon, 2) operational definition of the
phenomenon and identifying variations, 3) case features and causal analysis, and 4)
theoretical explanation of variance in the phenomenon. Gaining access to the
phenomenon includes description of the techniques used, the challenges during data
collection and the way you manage it, and the mechanisms you used to maximize validity
and reliability of your case study data. If you have the right access to the phenomenon
then you can pursue the rest steps in the analytic induction method of data analysis.
As part of your case study data analysis, the underlying phenomenon has to be defined
clearly and in detail. Clear definition of the phenomenon increases your ability to figure
out major variables and mapping them based on their magnitude and frequency of
variation. Further, you may classify the major variables in the defined phenomenon based
on degree of importance. You can apply the tools of tabulation and cross-tabulation to
come up with better way of categorization. Operational concepts have to be assigned to
new dimensions and the whole classification system has to be explained in greater detail.
Identification and categorization of main variables within the defined phenomenon and
tracking their pattern of variation is merely the “anatomy” of the case study. This does
not show the causal relationship among the variables. Hence, in your case study data
analysis you need to account for the specific features (i.e. features that make the case
unique) of each case in light to the operational definition of your phenomenon.
Accordingly, establish cause and effect relationship among the main variables within the
phenomena.

115
If your case study involves multiple cases and your analytic induction process reaches a
point of theoretical saturation (i.e. a point whereby the existing explanation remains
stable for all possible new cases) then you may proceed to the theoretical explanation of
the variation within the phenomenon.
6.5.4 Life history method
The life history method of case study data analysis presumes constructivist paradigm. It
gives the sense of reality people have in their own world. The life history method views
an organization as a network of relationships among individuals and groups of people
who have their own way of interpreting the world. Similarly, the method presumes that
the researcher has her/his own way of interpreting the interpretations of the research
subjects. In life history method of case study data analysis, data collection and analysis
are not separate.
According to Musson (1998), life history method is important to study such research
questions as how does organizational culture emerge? How are accounting information
understood? How do some managerial styles come to be accepted as ‘natural’? What is
the role of particular individuals and groups in creating and maintaining organizational
culture? In his case study that uses life history method of analysis, Musson (1998) tried to
figure out set of oppositions within an individual, oppositions between individuals as
members of a specific group, and opposition between individuals and the researcher’s
construction of events.
The norms, values, and practices are deduced from life histories of individuals. These can
help you explain out why oppositions, contradictions, and ambiguities exist among
individuals. The method is time consuming for both the researcher and the subject.
6.6 Case Study Method versus Operations Research Method
Positivist case studies commonly use quantitative method of data analysis. The use of
advanced analytical techniques to improve decision making is known as operations
research. For example, if you want to do research on “Optimizing transportation system

116
of Mekelle University”, you may use a case study research design from positivist
paradigm point of view. Pertaining to the method of case study data analysis, you may
apply the standard transportation model of operations research and come up with optimal
operational solution. In using operations research method in your case study research, the
most important skill required is the ability to convert word problems into equivalent
mathematical and statistical expressions. Most operations research models have two
parts: the objective function and a set of constraint equations. If both the objective
function and its constraints are linear functions then the operations research model is
known as linear programming model or simply linear model. Operations research models
try to attain the stated objective taking the constraints into account. Hence, you can use
operations research model for your positivist oriented case study research so long as you
are able to translate your case study propositions into mathematical expressions involving
an objective function and a set of important constraints.
Chapter Summary:
- Scientific paradigm is helpful in case studies so as to understand the general context

- Positivism assumes objective reality whereas constructivism assumes subjective reality
- There is no hypothesis in constructivist case study because reality is not objective
- Case study is most appropriate method to answer the “why” and “how” questions
- The three types of case studies are descriptive, explanatory, and evaluative case study
- Descriptive case studies are classified into illustrative, pilot, and critical instance
- Factor analyses reduces many variables about a concept into few and basic dimensions
- Factor loading measures the degree of correlation between a variable and a factor
- Communality is the sum of squares of factor loadings for a given variable
- Eigenvalue is square summation of factor loadings of all variables with a given factor

117
Scientific Research Report Writing and Presentation Mekelle University, CBE
CHAPTER 7
SCIENTIFIC RESEARCH REPORT WRITING AND PRESENTATION
7.1 Introduction
The ultimate purpose of any research project is to communicate solutions of a research

problem. Communication system has two inseparable parties: the sender and the receiver.
In the case of scientific report writing, the researcher is the sender and the target audience
is the receiver. Like any other communication system, scientific reports have to be
prepared in such a way that the target audience with a reasonable understanding of the
research area would sufficiently understand it. Hence, in order to write a scientific report,
your critical task is definition of the audience you want to put on target. For example, if
your research output deals with role of microfinance on rural livelihoods in Ethiopia then
the target beneficiaries are farmers. At least two different types of scientific reports can
be generated, that is, one for the scientific community in the area and another report for
farmers. Scientific research reports are usually presented to different types of audiences
in such forums as seminars, workshops, and conferences. Like that of scientific reports,
presentation involves communication of research results to target audience.
Key Words: Journal Article, Conference Paper, Senior Essay, Thesis, Dissertation, Books,
Publication, Peer Review, Power Point Presentation.
Chapter Objectives:

- Write scientific research report for publication and conference presentation purposes
- Know how to publish a research work in peer reviewed journals
- Develop the art of using appropriate words in their report using Bloom’s taxonomy
- Point out common mistakes from any research paper such as senior essay
- Develop the skill of research paper presentation by managing stage fright
- Prepare good power point presentation slides

118
7.2 Types of scientific research reports
In general, scientific research papers can be classified into four: Journal Articles;
Conference Papers; Senior Essay, Thesis and Dissertations; and Books. Journal Articles
are in turn classified based on the reviewing process or the source of data the research
uses. Based on the review process, Journal Articles are categorized into Peer Reviewed
Articles and Double Blind Reviewed Articles. A Peer Reviewed Article is a research
paper that is examined by peers whether the research work maintains standards
accordingly increasing research paper quality and credibility. The process of reviewing a
research paper is known as referring. An article is Double Blind Reviewed Article if
identity of both the researcher and the referee are not known. That is, when referees
review the research article without knowing personal information about the writer and the
writer does not know who is reviewing her paper.
Conference Papers are papers submitted and/or presented at conferences, workshops,

seminars and other forums. Conference Papers may be classified as Paper with the
Respondent, Panel Presentation, Roundtable, and Poster. In the Paper with Respondent
type of Conference Paper a speaker submits a thirty-minute paper and a respondent
responds to the paper for about fifteen minutes and the speaker gives response to the
respondent for about fifteen minutes. Panel Presentation paper is prepared such that it
would be presented in a setting involving panel sessions led by 3-4 panel speakers each
of who talks about fifteen minutes. The panel may involve individual or group
discussants that give responses in reaction to the panel speakers. Roundtable Papers are
submitted so as to be presented in a setting that involves five or more speakers who speak
about ten minutes each. Poster Papers are visual presentations of research findings such
as posting a hypothesis and an outline of the findings.
Senior Essay, Thesis, and Dissertation are research works that are done in partial
fulfillment of Bachelor, Masters, and PhD degrees respectively. Senior Essay, Thesis, and
Dissertation are structurally the same but differ in the degree of complexity of the
research problem on the one hand and degree of reliability and validity of the research
methodology adopted on the other.

119
Books, as different from textbook (i.e. a manual of instruction in any field of study), are
compilation of different research works about one specific field of study. Alternatively, a
Book can also be one research work divided into several chapters. Books are usually
known as monographs.
7.3 Deciding to publish your research work
To publish means to make any content available to the public and the act of publishing is
called publication. If you decide to publish your research paper the first question you
need to ask is what to publish? You can publish either an abstract or full report from your
research paper. The next thing you need to consider is choice of publication outlet.
Depending on the quality of research you have you may look for Journals starting from
the least reputed up to the most reputed ones. Alternatively, you may publish your
research paper in a book form or else you may extract Conference Paper.
Journals are usually classified based on thematic research areas in a discipline. For
example, Journal of Finance is the most reputable Journal in the field of Accounting and
Finance. This Journal is most reputable because, many of the writers that publish their
research papers in Journal of Finance are known professors some of which are Nobel
Prize winners. Likewise, Journal of Accounting Research, Journal of Financial Research,
International Journal of Accounting, Auditing, and Accountability, Abacus, and etcetera
are some of the well cited Journals of Accounting and Finance. As an alternative, you can
also look for Accounting and Finance Journals that are published by Ethiopian
Universities and research institutes.
You may give a look at the publication guidelines for authors in the website of each
Journal in the area of Accounting and Finance and select one that matches your research
and prepare your article following every requirement in the author guideline and online
submit your article to the Journal editor. An editor is assigned for your paper and if she or
he decides to further scrutinize your article then the editor sends your article to two
anonymous reviewers for a double-blind reviewing process. Based on comments from
anonymous reviewers, the editor decides one of the following decisions: 1) accept with

120
minor revision, 2) accept with major revision, and 3) reject. The publication process may
take between 6 to 12 months.
7.4 Writing a scientific research report
Main purpose of a scientific research report is to communicate research results to specific

set of audience. Though there are no hard and fast rules guiding scientific research report
writing, ideas need to flow to the audience as concisely and effectively as possible.
Therefore, before writing your scientific research report you need to know your audience.
In general, you may categorize your potential readers or audience into three as: readers
with deep understanding about your research topic, readers with reasonable
understanding about your research topic, and readers that do not reasonable
understanding about your research topic.
If you intend to write a scientific research report that is to be published in a specific

Journal, in most cases your intended audiences are those readers with deep understanding
about your research topic. Hence, you need to exclude unnecessary details on concepts,
variables, or phenomena etc that is already known by your audience. In fact, you need to
exercise due care to manage the risk of excluding some details that are very important to
your research topic.
Conference papers usually have mixed audiences unless the conference or workshop is
organized by a specific discipline and its participants are selected based on their depth of
understanding of the conference or workshop topic.
The audiences of senior essay, thesis, and dissertation are students, university lecturers or
professors, researchers, and stakeholders of the research project. In this case, usually, you
may need to write two types of scientific research reports assuming; 1) readers with deep
understanding about your research topic (i.e. university lecturers or professors and
researchers), and 2) readers that do not reasonable understanding about your research
topic (e.g. beneficiaries of your research). Nevertheless, for the purpose of earning your
academic degree it is enough if you produce a scientific research report assuming your
audience has deep understanding about your research topic.

121
A book or monograph is usually written once in a lifetime. It comprises scholarly

treatises provide evidence that the academic is carrying out research in the field. A
monograph usually brings new light to the subject. For example, the monograph on
“Wealth of Nations” was a monograph written by Adam Smith that established the field
of Economics as a field. A monograph refines the academic specialty of the author and
establishes the author as an authority on the topic.
7.4.1 Word Choice
Your scientific research report not only has to be free of any grammatical and
typographic error but also words have to be carefully chosen. A grammatical or
typographical error in your scientific research report might signal to the reader that you
failed to exercise maximum care in doing your research and therefore the reader may
doubt your research results.
Bloom (1956) classified cognitive understanding into six levels of taxonomy as

knowledge, comprehension, application, analysis, synthesis, and evaluation. The
knowledge level refers to basic understanding of the subject matter. At knowledge levels,
among the most common verbs are: define, label, state, underline, review, list, relate,
show, distinguish, and describe. Comprehension goes one step higher than knowledge. It
deals with the ability to state the problem or interpret the fact. Some of the most common
words used at comprehension level are: restate, identify, give examples of, discuss,
paraphrase, describe, report, recognize, review, observe, outline, account for, interpret,
explain, summarize, and ask. Application level of Bloom’s taxonomy refers to the third
level of understanding that focuses on applying the information you have, in form of
comprehended knowledge, to solve new problems. Some of the common verbs at
application level of the taxonomy are: manipulate, exhibit, illustrate, prepare, experiment,
practice, apply, relate, operate, record, compute, employ, show, solve, demonstrate,
dramatize, construct, and use. Analysis focuses on explanation of patterns or meanings.
Common verbs that refer to the analysis level of understanding are: distinguish,
experiment, inspect, examine, inquire, arrange, investigate, criticize, compare, scrutinize,
discover, detect, point out, analyze, contrast, classify, organize, differentiate, and deduce.

122
Table 1: Bloom’s Taxonomy (1956)
Taxonomy Process Verbs Product Verbs

Level
Judge, rate, validate, assess, score, revise, infer, Investigation, opinion, debate,
criteria, determine, prioritize, tell why, evaluate, verdict, conclusion,
Evaluation compare, defend, select, measure, choose, conclude, recommendation, panel,
deduce, debate, justify, recommend, discriminate, judgment, report, editorial, scale,
appraise, value, probe, argue, decide, estimate, evaluation
criticize, rank, award
Compose, assemble, manage, pretend, arrange, Film, story, project, blueprint,
organize, invent, generalize, systematize, show, plan, solution, new game, song,
Synthesis compile, forecast, modify, devise, derive, propose, video, newspaper, painting,
construct, plan, revise, collect, prepare, develop, media product, advertisement,
originate, imagine, generate, predict, combine, poem, formula, machine, goal,
organize, write, suppose, formulate, set up, design, play, cartoon, invention,
blend, create, produce, hypothesize, predict, concoct, product, event, design
infer, act, compile, reorganize, role-play, improve
Distinguish, contract, question, appraise, experiment, Diagram, investigation, graph,
inspect, examine, probe, separate, inquire, arrange, conclusion, category,
Analysis investigate, sift, research, calculate, criticize, solve, questionnaire, illustration,
interpret, compare, inventory, scrutinize, discover, inventory, spreadsheet, checklist,
survey, detect, group, order, point out, sequence, test, chart, outline, list, plan, summary,
debate, analyze, diagram, contrast, relate, dissect, survey, database, abstract, report
categorize, point out, classify, organize, differentiate,
deduce, discriminate
Translate, manipulate, exhibit, illustrate, calculate, Prediction, photograph, illustration,
sketch, interpret, prepare, make, experiment, list, simulation, sculpture, experiment,
Application practice, apply, relate, operate, interview, paint, interview, performance,
change, record, produce, compute, sequence, employ, presentation, demonstration,
show, solve, schedule, collection, demonstrate, puzzle, relate, diary, report, poster,
dramatize, construct, use, teach, adapt, draw diagram, lesson, model, journal,
map
Restate, identify, discuss, locate, retell, research, Recitation, summary, reproduction,
convert, annotate, translate, give examples of, collection, explanation,
Comprehension paraphrase, describe, report, recognize, review, dramatization, show & tell, story
observe, locate, outline, account for, interpret, give problems, example, definition,
main idea, explain, tell, express, summarize, ask, quiz, list, test, label, debate, outline
calculate, expand upon
Define, name, record, match, select, underline, cite, Quiz, definition, fact, worksheet,
sort, know, repeat, label, recall, listen, group, recite, reproduction, label, list, test,
Knowledge choose, review, quote, list, memorize, relate, show, workbook
locate, distinguish, give example, reproduce, describe,
state
Source: http://nerds.unl.edu/pages/preser/sec/articles/blooms.html
Synthesis level of understanding has a lot to do with making predictions and discussing
possibilities. Some of the common verbs at synthesis level are: Compose, organize,
invent, generalize, systematize, compile, forecast, devise, derive, propose, construct,

123
develop, imagine, generate, combine, formulate, design, produce, hypothesize, predict,

concoct, and infer. According to Bloom (1956), evaluation is the highest level of
understanding. Evaluation refers to drawing conclusions, attaching value sentiment to the
conclusion, and making recommendation. Some of the common verbs of evaluation are:
Judge, validate, assess, criteria, determine, prioritize, tell why, evaluate, compare, defend,
select, choose, conclude, deduce, debate, justify, recommend, value, argue, decide,
criticize, rank, and award.
The most important lesson you may draw from Bloom’s Taxonomy is that you need to
choose appropriate words that designate the level of understanding you are dealing with.
7.4.2 Sections of a scientific research report
As discussed in chapter 6 of this reaching material, scientific researches may be classified

based on their purpose (i.e. generalization versus particularization) into survey and case
study researches. Chapter 6 thoroughly went through scientific research report for both
positivist and constructivist case study researches. Therefore, this section provides
scientific research report of survey research.
A typical survey research report aimed at publication in a Journal has the following parts:
Abstract, Introduction, Literature Review, Results and Discussion, Conclusion,
Acknowledgement, References, and Appendix (if any). Many of these have been
explained and illustrated in chapter 3 of this teaching material. Hence, emphasis shall be
given to the common mistakes in each section of the scientific research report.
A. Abstract
As explained in chapter 3, the abstract section of a scientific research report is done after
completion of the actual research. Abstract of a research proposal differs from abstract of
a scientific research report only because the former is written using future tense since the
actual research is yet to be done. Besides, abstract section of a research proposal includes
summary of the expected research output instead of including summary of the research
results. On the other hand, abstract section of a scientific research report uses past tense

124
and includes summary of the actual research results but it does not include the expected
research output because, at this level, the research is already done.
Therefore, abstract section of a scientific research report includes: the problem

investigated, purpose of the research, methods used in the research, the main findings of
the research, and conclusion of the research. A typical abstract contains 200 up to 300
words only.
Some of the common mistakes in abstract writing are: 1) too much information, 2)
figures and images, 3) using an in-text reference, and 4) using acronyms and
abbreviations. If your abstract contains too much information then some part of the
potential readers, for example very busy professors, will throw your paper away. Since
the very purpose of an abstract is to provide concise summary to a very busy reader, you
need to exercise due care in balancing the risks of too much information and too little
information. An abstract should not include figures and images but textual words. Figures
and images provide detail information about your topic but the main purpose of an
abstract is to provide summarized information. Since an abstract is meant to summarize
your research it is preferable if your abstract does not include in-text references and
figures and images of other writers. Last but not least, refrain from using acronyms and
abbreviations because the very busy readers may not understand them and accordingly
may throw your research paper away.
B. Introduction
Chapter 3 describes and illustrates how to write introduction for your research proposal.
Introduction part of a research proposal and introduction part of a scientific research
report are essentially the same. In other words, you can directly use the introduction part
of your research proposal as an introduction part for the actual research report. A typical
introduction has a 300 to 500 words length, that is, 1 up to 1.5 pages.
As you have learnt in chapter 3, the introduction part includes broad information on the
topic taken from the available literature related to your topic, narrower background
information that could reflect the problem or gap, statement of focus of the paper using

125
hypothesis or research question, and summary of the problem so as to sell your point to
the reader. In other words an introduction part of a scientific research report contains
context, problem, hypothesis or research question, and a point for sale.
Some of the most common errors in writing the introduction part of a scientific research
report are: 1) too much or not enough information, 2) irrelevant background information,
unclear research objective, and 3) confusing structure. As mentioned in the abstract
section, an introduction containing too much or not enough information does not serve
the purpose. Hence, both risks of too much and too little information need to be balanced
for the sub sections (i.e. context, problem, hypothesis, and the point for sale) of an
introduction part of a paper. It is not unusual for students to write irrelevant background
information. For example, most undergraduate students in Ethiopian universities like to
write detailed notes about mission, vision, and objectives of the case organization their
research considers. But, detailed note pertaining to their case organization can not state
the context, problem, hypothesis or research question, and a point for sale.
Confusing structure is another common mistake in writing the introduction part of a

scientific research report. The logical flow of introduction part of your scientific research
report determines readability of your research and it is the most important factor that let
the reader decide either to continue or stop reading your research report. The logical flow
has to put the following in order: the research context, research problem statement,
hypothesis formulation or research question formulation, and the point for sale.
Especially for Journal articles, it is common to add one paragraph regarding section-wise
organization of the remaining parts of the research paper.
C. Literature Review
Literature review has been discussed in chapter 3 of this teaching material in light to the
APA style of referencing. Literature review for your research proposal is essentially the
same as that of literature review for your scientific research report except that literature
review of a research report is more extensive than the respective literature review of a
research proposal. The literature review part of your scientific research report provides

126
detailed account of the theoretical and empirical frameworks of your research. That is,
literature review section is an elaboration to the introduction part. However, bear in mind
that not all research proposals have literature review. For example, a research proposal
aimed at exploration, by its very nature, can not have separate literature review section.
The most common error in the literature review part of a scientific research report are: 1)
failure to collect relevant literature, 2) verbosity, 3) lack of logical flow and failure to
attract readers’ attention, 4) too many fillers, and 5) do not focus on building theoretical
and conceptual frameworks. The failure to collect relevant information for your research
proposal is a critical mistake. Undergraduate students of Ethiopian universities usually
select their research topic and start their senior essay research without reading the
relevant literature and without keeping an appropriate research design in their mind.
Accordingly it is not uncommon for undergraduate students to write a “lecture note” type
of bulky literature review by copying from “tertiary” data sources such as textbooks and
modules. This common mistake has three dimensions: first relevant literature is not
collected before topic selection, second academic crime of plagiarism is committed due
to lack of appropriate reference, and thirdly the copies are not from primary or secondary
sources of data.
Verbosity is another common mistake in writing the literature review. An idea that can be
stated in one sentence becomes a paragraph or more creating hard time for the reader’s
attention. Moreover, many student scientific research reports fail to catch the maximum
attention of readers because of problems in the logical flow. Likewise, it is common
mistake for student scientific research reports to use so many fillers. Fillers are set of
paragraphs or sections within a scientific research report that do not have direct
relationship to the topic. Some of the reasons why students prefer to use fillers is in order
for their research report to reach the minimum page limit required that in turn is due to
the implicit value on “the more number of pages the more effort exerted the more
valuable the research work is”. Last but not least common mistake in writing literature
review is the failure to integrate the theoretical literature review with the theoretical
framework and accordingly with the statement of the problem and the failure to integrate

127
the empirical literature review with the conceptual framework and accordingly with the
hypothesis formulation.
D. Data and Methods
As explained elsewhere in this teaching material the data and method section tells the
reader how the research was done and provides instruction on exactly how to repeat or
replicate the research if need arises. The only difference between data and methods of
your research proposal and the data and method section of your actual research report is
the use of tenses. Research proposal uses future tense whereas the actual research report
uses past tense. Some of the most common mistakes in writing the data and method
section of a scientific research proposal are: 1) too little information, 2) information from
introduction and literature review, 3) verbosity and too many filers, and 4) results are
reported.
In most of the cases, undergraduate students in Ethiopian universities fail to provide

sufficient information in the data and method section of their scientific research report. It
is common to write “… this research will use both qualitative and quantitative methods
…” without providing further details. However, the data and method section has to
provide sufficient instruction for a possible replication by an interested independent party
in your absence. Some research reports wrongly repeat information that are already stated
in the introduction and literature review parts thereby missing the purpose of data and
method section.
Verbosity is another common mistake in writing the data and method section of a
scientific research report. It is common for students to unnecessarily provide a kind of
“lecture note” on the different types of research methods for purposes other than choice
of method and try to explain an idea using bunch of words. Similarly, student research
reports present an unnecessarily elaborated description of the study area. Last but not
least, some research reports provide preliminary results of the research in the data and
methods part rather than presenting it in the results sub section of the results and
discussion section of the scientific research report.

128
E. Results and Discussion
The results and discussion section of a scientific research report presents summary of the
data, either statistical or analytical, and your interpretation of it by comparing it with
results of prior researches in the literature. There are two modes of presenting the results
and discussion part of your scientific research report. The first mode is to separately
present the results from the interpretations thereof or discussion. This mode is most
common in survey researches. Research results are first presented in the form of
descriptive statistics and model estimation results.
Hypotheses are tested in the discussion section. Software routines such as STATA by
default assume null hypothesis and generate statistical tests of significance that can help
you either accept or reject the null hypothesis. It is up to you to discuss on how to relate
the results of your research with your hypotheses and accordingly with your research
problem. Especially some results may need further interpretation and, if possible, use
triangulation technique by comparing your quantitative results with qualitative data
collected using key informant interview and focus group discussion and/or by comparing
your results with the results in the literature. Further, in the discussion section you can
discuss on the implications of the study and you can also pinpoint the potentials for
improvement and further research.
The second mode is to concurrently present results and discussion. This is most common
in case study researches. That is, first present results in relation to one of your case
study’s propositions (if you have no research proposition then in relation to one of your
research questions) in the form of factual evidence then test your case study research
proposition or research question. If the factual evidence could not significantly contradict
with your case study proposition then accept your proposition otherwise reject it.
Interpretation or discussion on the first result follows before you go on presenting the
next result.
Some of the common mistakes in presenting the results of a scientific research report are:
1) raw data, 2) redundancy, 3) incorrect presentation of figures and tables, and 4) data

129
and methods reported. Presenting raw data is the most frequently occurring common
mistake. Students in Ethiopian universities prefer to summarize the data they collect
(even survey data) using percentages. Most student research papers in Ethiopian
universities neglect the fact that numbers need to be interpreted. Another common
mistake is redundancy. Depicting a result in Table and graph is one form of redundancy.
Common mistakes in writing discussion part of a scientific research report are: 1) broad
statements, 2) failing to interpret the figures, 3) incorrectly discussing inconclusive
results, 4) ambiguous data source, and 5) missing information. Student research papers in
Ethiopian universities commonly use general statements, such as statements that go
outside scope of the result under discussion, while interpreting the results. It is also
common for student research papers not to interpret the figures. For example, a research
report might present that “60% of respondents are in favor of micro-financing services”
or another research report might present that “female students of Mekelle University
account for 40%” but if meaning could not be deliberately attached to these results it will
be very difficult to test the hypothesis or to answer the research question and ultimately
to solve the research problem.
F. Summary and conclusion
In the results and discussion section of your scientific research report you presented all
necessary facts and accordingly hypotheses are tested. As shown in Figure 1, the tested
hypothesis need be linked to the problem statement part of your research report. You can
link the tested hypothesis to the statement of the problem by showing how your research
solves the stated problem. Accordingly, the summary and conclusion part of your
scientific research report has to present solution to the research problem. It is good for the
summary part of your scientific research report if it could present a translation of the
tested hypothesis to factual premises (i.e. premises based upon evidence presented in the
results and discussion) and accordingly onto propositions – statements that can be
evaluated as true or false. That is, the factual premises and hence propositions would
setup your logical argument. For example, you may state your factual premises in
propositional form as “X is Y” and “Y is Z” and the like.

130
Figure 1: The link between Problem Statement and Tested Hypothesis
The nature of conclusion of your scientific research report depends on whether your
survey research has exploratory, descriptive, or explanatory purpose. If your survey
research has explanatory purpose then clearly your conclusion will follow from the set of
propositions that summarize your results and discussion. For example, if the summarized
propositions were “X is Y” and “Y is Z” then using the logical axiom of transitivity you
can conclude that “X is Z”. Of course, the conclusion will be true if both premises are
evaluated as true in the results and discussion part. The most important thing about the
conclusion of an explanatory research is that it is not a factual premise or proposition
thereof which is directly observed but it is a result of logical reasoning. Exploratory and
descriptive researches do not go beyond the factual premises. For example, the summary
and conclusion part of a descriptive research simply contains summary of the factual
premises thereby providing holistic picture about a complex problem.
G. Recommendation
A research has target beneficiaries recommendation is addressed to either directly to the

target beneficiaries or indirectly through policy makers or decision makers at any level of

131
the bureaucracy. In fact, recommendation should not be shallow. Otherwise, no body will
make use of it. For example, assume that you have done a research on “Factors affecting
VAT collection in Tigray region” and assume that you described the major factors and
their degree of influence on VAT collection. Further assume that one of your
recommendations goes “training must be given to the VAT collectors”. Of course your
recommendation seems to be good but it will be shallow. The reason is that you need to
conduct intensive cost-benefit analysis before you recommend training to the VAT
collectors. However, cost-benefit analyses by it self is one form of research. Hence, if it
is going to be shallow it is better not to write any recommendation. Instead, you can
recommend areas for further research. For example, you may recommend that research
has to be done with respect to “cost-benefit analysis on training VAT collectors”.
The most common mistake in writing recommendation part of a scientific research report
are: 1) too many recommendations, and 2) recommendations beyond scope of the
research paper. Student research papers in Ethiopian universities usually present too
many recommendations at a time regardless of whether they can be sold or not. Secondly,
the research papers tend to forward recommendations that are not analyzed in the paper.
In other words, the recommendations go beyond scope of the paper.
H. References
In Chapter 3, the details of referencing have been presented using APA style.
Consistency is the most important things you need to remember while writing reference.
For example, if the Journal to which you want to submit your research article prefers
Chicago Style of referencing then you need to consistently follow Chicago Style. Peer-
reviewed Journal articles, abstracts, and books are sources of secondary data that should
be referenced. Whereas, Non-peer-reviewed works, personal communications, and
textbooks should not be referenced.
7.5 Presenting a scientific research paper
Once you finish writing a scientific research report, you have many opportunities to
present your paper. How can you make an effective presentation? It depends on the

132
understanding level of your audience. For example, if you want to present your senior
essay to the evaluation committees you need to consider the fact that your audience has
higher level of understanding with regard to your research topic therefore try not to
present unnecessary details. But, if you want to present your paper to the beneficiaries,
such as farmers, you need to consider the fact that your audience has lower level of
understanding about your research topic and hence try to prepare your presentation in a
way your audience can understand it and try not to use jargons.
7.5.1 Types of presentations
Presentation can be classified into informative, instructional, arousing, and persuasive

types. Informative presentation has the purpose of informing the audience about an idea,
object, phenomena, etc using concise and to the point presentation. Informative
presentation has to exclude complications because it is aimed at informing the basic facts.
An informative presenter has to be concerned as to how to make the presentation brief
and to the point. Informative presentations are common in events such as seminars,
workshops, and conferences where the presenter and the audience do not have similar
level of understanding on the subject matter.
Instructional presentation has the purpose of acquainting the audience with new
knowledge and skill. Instructional presentation puts a detailed account of a topic.
Instructional presentation is very common in a class room teaching, public lectures;
seminars, workshops, and conferences on technical issues of a discipline. In instructional
presentation, the presenter has more knowledge and skill on the topic than the audience.
Arousing presentation has the purpose of making the audience think about a problem, an
event, an object, an idea, etc and thereby attaching value onto the event, the object, the
idea, etc. Arousing presentation is common when the audience is composed of decision
makers such as the members of state council of Tigray regional state of Ethiopia. You
may want to first help the policy makers attach value onto your research problem then try
to sell your point. You may start your arousing presentation by using local proverbs that
are very common to the audience, or by giving to the audience a thought experiment that

133
is directly related to your topic, or by using catchy anecdotes. The proverb, thought
experiment, or anecdote make your audience starts to think about your topic. After you
raise emotion of the audience you can sell your point.
Persuasive presentation has the purpose of convincing the audience. The best example of
persuasive presentation is defending your research proposal. In a persuasive presentation
you are required to present sufficient logic in order to convince the audience to take your
view. For example, research proposal presentations by graduating students of Mekelle
University are persuasive presentations. In general, a good persuasive presentation may
have three parts, namely; gap reflection or problem statement, need for intervention, and
your proposed intervention.
7.5.2 How to prepare good power point presentation
Power point presentation is the preferred method of preparing your presentation. If you
do not like power point presentation, you can prepare presentation cards the resemble
power point slides. The following are some of the most common tips in preparing good
power point slides.
a. Think about your presentation
You need to think about the content and logical flow of your presentation before you start
writing about it either on your power point slides or on the presentation card. You can
start writing on the power point slides one you have clear idea as to what to present.
b. Map out what you are going to say
Depending on the type of presentation you are preparing for (i.e. informative,
instructional, arousing, and persuasive) write an outline of presentation for each power
point slide. The number of power point slides depends on the presentation time allocated.
For example, if time allocated for your presentation is 15 minutes then the maximum
number of slides should be 15 – that is one slide for one minute presentation. Keep in

134
mind that one slide has to contain no more than 6 bullet points and one bullet point has to
contain no more than 6 words.
c. Slide layout
Make sure that your power point slides have maximum contrast between color of the text
and the slide background. For example, if you choose a white slide background obviously
color of your text should be black; if you choose a black slide background then color of
your text should be white and so on. It is advisable to use minimum font sizes of 24 and
34 for the content and title respectively. With respect to font styles, use a font style that
does not have stroke on letter ends. For example, Sans Serif, Tahoma, and Arial are good
font style for power point presentation.
d. Slide sequence
In the first slide you should write title of your presentation, your name and department or
position (if any), the date and venue of presentation. Title of your presentation should be
catchy. If your presentation is either arousing or persuasive the second slide should try to
seize attention of your audience using proverbs or anecdotes. However, if your
presentation is either informative or instructional your second slide should present overall
summary of the facts in your presentation. Your third slide has to present the structure of
your presentation. It is better if one sub-topic of your presentation could not exceed 3
slides otherwise it would be boring for the audience to attend your presentation. In
addition, if you try to use pictures that are relevant for your presentation you can get extra
attention from the audience because pictures are more powerful than words.
e. End of presentation
The last but one slide of your presentation has to summarize the point you want to sell
and articulate the conclusion of your presentation. A good last slide gives response to the
“so what?” question of your audience. Finally, you need to have one slide that gives a
thank you message to the audience thereby inviting the audience to make comments and
ask questions.

135
7.5.3 Stage fright and its management
Fear has long been the research agenda of psychologists. Stage fright otherwise known as
performance anxiety is one type of fear that starts to develop as an individual is required
to perform in front of an audience. The familiar physical symptoms of stage fright are
increased heart bit, breathing problem, sweating, weakening knees, expanding pupil,
trembling hands and legs, and the like. These physical responses are said to be caused by
an excessive generation of adrenaline hormone.
The hormone is excessively generated when the presenter perceives the audience as
judges and he or she has fear of failing to perform. If the presenter does not perceive the
audience as judges but he or she has fear of failing to perform then stage fright can not
occur. For example, if you were to make a presentation to kindergarten school students
you may not develop stage fright regardless of your level of fear of failing to perform.
Likewise, if the presenter does not have fear of failing to perform but he or she perceives
the audience as judges then stage fright can not occur. That is, if the presenter has
sufficient confidence on the presentation topic he or she can not develop stage fright
regardless of his or her attitude towards to audience. In effect, stage fright can be
managed by controlling either your attitude towards the audience or your fear of failing to
perform.
Some of the stage fright management techniques are presented as follows:
1. Change “fright” energy onto “fun” energy
Most of the physiological responses of fear such as increased heart bit, breathing
problem, sweating, weakening knees, expanding pupil, trembling hands and legs, and the
like are also physiological responses to happiness. You need not try to get rid off the
physiological feelings you feel while standing in front of the audience but try to switch
the “fright” onto “fun”. What psycho-physiological response do you immediately give to
the happiness-driven physiological responses in question? Perhaps you may out cry, hag
friends around you, jumping etc. You can have similar psycho-physiological responses to
change “fright” onto “fun”. For example, you may tell the audience a joke or act in a

136
funny way so that both the audience and you would burst into laughter immediately the
“fright” will be changed onto “fun”. Another example is to walk around the stage and try
to deep breath in order to convert “fright” energy onto “fun” energy.
2. Take deep breath from the diaphragm
One of the symptoms of stage fright is rapid breathes that affects tone of the speech. If
you happen to deep breathe from your diaphragm on the event of stage fright, you are
reducing the frequency of inflow of oxygen to your lung and outflow of carbon dioxide
from your lung. Deep breathing has the capability of restoring your breathing system
back to normal.
3. Prepare, Prepare, and Prepare
Preparation is the best way of managing stage fright ex ante. While preparing for
presentation at home, you may assume presence of the audience in front of you and try to
loudly present for your self. It is preferable if you could recite the main points of your
presentation.
Chapter Summary:
- Research works are Published as Journal Articles; Conference Papers, and Books
- Journal Articles are peer reviewed and double blind reviewed articles
- Journals are usually classified based on thematic research areas in a discipline
- A book or monograph is usually written once in a lifetime
- Presentation can be classified into informative, instructional, arousing, and persuasive
- Stage fright develops as an individual is required to perform in front of an audience

137
Bibliography
Ackert, L. F., Church, B. K. and Gillette, A. B. (2004), ‘Immediate Disclosure or

Secrecy? The Release of Information in Experimental Asset Markets’. Financial
Markets, Institutions & Instruments, 13 (5), 219-243.
Aitken, M. and Segara, R. (2005). ‘Impact of Warrant Introductions on the Behavior of

Underlying Stocks: Australian evidence’. Journal of Accounting and Finance, 45, 127–
144.
Amaratunga, D. and Baldry, D. (2001). ‘Case study methodology as a means of theory building:
performance measurement in facilities management organizations’. Journal of Work
Study, 50 (3), 95-104.
Babbie, E. (1973). Survey Research Methods. Belmont Wadsworth Publishing Company, Inc.
Babbie, E. (1986). The Practice of Social Research. 4th ed. Belmont: Wadsworth Publishing Co.
Bandyopadhyay, A. (2006). ‘Predicting Probability of Default of Indian Corporate

Bonds: Logistic and Z-score model approaches’. The Journal of Risk Finance, 7 (3),
255-272.
Bauera, R., Ottenb, R. and Radc, A. T. (2006). ‘New Zealand Mutual Funds: measuring
performance and persistence in performance’. Accounting and Finance, 46, 347–363.
Beatti, V. and Jones, M. J. (2002), ‘Measurement distortion of graphs in corporate

reports: an experimental study’, Accounting, Auditing and Accountability Journal, 15 (4),
546 – 564.
Bloom, B. S. (1956). Taxonomy of Educational Objectives, the classification of educational

goals – Handbook I: Cognitive Domain, New York: McKay.
Ferguson, A. C., Francis, J. R. and Stokesd, D. J. (2006), ‘What Matters in Audit Pricing:
industry specialization or overall market leadership?’ Journal of Accounting and
Finance, 46, 97–106.
Gebregziabher, H. (2009a), ‘Determinants of Public Non-Managerial Employees’ Performance:

Empirical Evidence from Ethiopia’. International Journal of Productivity and
Performance Management, 53 (3), 238 – 253.
Gebregziabher (2009b), ‘Financing preferences of micro and small enterprise owners in Tigray:
does POH hold?’ Journal of Small Business and Enterprise Development, 16 (2), 322-
334.
138
Guedhami, O. and Pittman, J. A. (2006). ‘Ownership Concentration in Privatized Firms: The
Role of Disclosure Standards, Auditor Choice, and Auditing Infrastructure’. Journal of
Accounting Research, 44 (5), 889-929.
Hemsley-Brown, J. (2004). ‘Facilitating Research Utilization: A cross-sector review of research

evidence’. The International Journal of Public Sector Management, 17 (6), 534-552.
Hides, M.T., Irani, Z., Polychronakis, I. and Sharp J.M. (2000). ‘Facilitating Total Quality
Through Effective Project Management’. International Journal of Quality &
Reliability Management, 17 (4/5), 407-422.
Johnsen, P. C. and McMahon, R.G.P. (2005). ‘Cross-industry Differences in SME Financing

Behavior an Australian Perspective’. Journal of Small Business and Enterprise
Development, 12 (2), 160-177.
Llewellyn, S. (1999). ‘Methodological Themes: Narratives in Accounting and Management

Research’. Accounting, Auditing & Accountability Journal, 12 (2), 220-236.
Northcott, D. and Doolin, B. (2000). ‘Home Accountants: exploring their practices’.

Accounting, Auditing & Accountability Journal, 13 (4), 475-501.
Perry, C. (1998). ‘Processes of a case study methodology for postgraduate research in

marketing’. European Journal of Marketing, 32 (9/10), 785-802.
Quattrone, P. (2000). ‘Constructivism and Accounting Research: Towards a trans-

disciplinary perspective’. Accounting, Auditing & Accountability Journal, 13 (2), 130-155.
Rowley, J. (2002). ‘Using Case Studies in Research’. Management Research News, 25 (1), 16-
27.
Stathopoulou, S., Psaltopoulos, D. and Skuras, D. (2004). ‘Rural entrepreneurship in Europe A

research framework and agenda’. International Journal of Entrepreneurial Behavior &
Research, 10 (6), 404-425.
Voss, C., Tsikriktsis, N. and Frohlich, M. (2002). ‘Case Research in Operations Management’.
International Journal of Operations and Production Management, 22 (2), 195-219.
Whybark, D. C. (1997). ‘GMRG survey research in operations management’.

International Journal of Operations & Production Management, 17 (7), 686-696.
Zikmund, W. G., (2000). Business Research Methods. 6th ed., The Dryden press Harcourt Brace
College Publishers, Toronto, Canada.
139

Research Methods and Analysis

Uploaded by

Document Information

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 and Analysis

Uploaded by

Copyright:

Available Formats

Research Methods and Analysis for

Business Studies: a holistic approach

G.Egziabher H.Selasie Gebru

College of Business and Economics

In the professions of Accounting and Finance, research is becoming such a demanding

Content Page No.

Chapter 1: Introduction to Research

Chapter 2: The Research Design

Chapter 3: The Research Proposal

Chapter 4: Data Collection and Sampling

Chapter 5: Survey Method and Analysis

Chapter 6: Case Study Method and Analysis

Chapter 7: Scientific Research Report Writing and Presentation

7.1 Introduction 118

At the end of this chapter, students are expected to be able to:

1.2 What is science?

By: G.Egziabher H.Selasie Gebru, Assistant Professor

1.3 What is research?

By: G.Egziabher H.Selasie Gebru, Assistant Professor

1. Establishing a research question,

1.4 Purposes of research

1. To satisfy your curiosity and a desire for better understanding

An example of exploratory research is a research paper on “Home accountants: exploring

Example 1: Northcott and Doolin (2000)

By: G.Egziabher H.Selasie Gebru, Assistant Professor

Descriptive research aims only at describing a situation or a set of circumstances. It tries to

Explanatory research goes beyond description. It examines different aspects of a described

Example 2: Gebregziabher (2009a)

By: G.Egziabher H.Selasie Gebru, Assistant Professor

probability of having good performance. Similarly, Role perception is found to be strongly

1.5 Types of research

Applied research is conducted to solve socio-economic or industrial problems; and thus

By: G.Egziabher H.Selasie Gebru, Assistant Professor

Practitioning is like memorizing sounds of a song in a foreign language without necessarily

1.6 Research Ethics

- Science is an enterprise dedicated to finding out something. It is Logic-Empirical.

By: G.Egziabher H.Selasie Gebru, Assistant Professor

THE RESEARCH DESIGN

At the end of the chapter, students are expected to be able to:

2.2 What is research design?

Research design is the plan and structure of investigation so conceived as to obtain

By: G.Egziabher H.Selasie Gebru, Assistant Professor

Figure 1: Schematic Presentation of the research process

Source: Adapted from Babbie (1986)

By: G.Egziabher H.Selasie Gebru, Assistant Professor

2.3 Types of research designs

By: G.Egziabher H.Selasie Gebru, Assistant Professor

By: G.Egziabher H.Selasie Gebru, Assistant Professor

Table 1: Dimensions of research designs

Purpose of the Method of Control of Time Research Depth of

By: G.Egziabher H.Selasie Gebru, Assistant Professor

2.4 Designing the research problem

Example 1: Hides et al. (2000)

By: G.Egziabher H.Selasie Gebru, Assistant Professor

By: G.Egziabher H.Selasie Gebru, Assistant Professor

2.5 Designing the research hypothesis

Hypothesis is a tentative explanation of observable phenomena. Hypothesis states the

2.5.1 Hypothesis and Logic

2.5.2 Hypothesis, Constructs and Variables

By: G.Egziabher H.Selasie Gebru, Assistant Professor

Variables are operational representation of constructs. That is, constructs need to be

2.5.3 Types of hypothesis

By: G.Egziabher H.Selasie Gebru, Assistant Professor

2.5.4 Hypothesis and Research Problem