(Theories of Institutional Design) Keith Dowding, Peter John - Exits, Voices and Social Investment - Citizens' Reaction To Public Services-Cambridge University Press (2012)

E x i t s , Voic e s a n d S o c i a l I n ve s t m e n t
Over fifty years ago, Albert Hirschman argued that dissatisfied consumers could either
voice complaint or exit when they were dissatisfied with goods or services. Loyal con-
sumers would voice rather than exit. Hirschman argued that making exit easier from
publicly provided services, such as health or education, would reduce voice, taking
the richest and most articulate away and this would lead to the deterioration of public
services. This book provides the first thorough empirical study of these ideas. Using
a modified version of Hirschman’s account, examining private and collective voice,
and viewing loyalty as a form of social investment, it is grounded on a dedicated five-
year panel study of British citizens. Given government policies over the past decade or
more which make exit easier from public providers, this is a timely publication for all
those who care about the quality of government services.
k e i t h d ow di ng is Professor of Political Science in the School of Politics and

International Relations, Research School of Social Sciences, and Research Director
of the College of Arts and Social Sciences at the Australian National University,
Canberra.
pe t e r joh n is Professor of Political Science and Public Policy, School of Public

Policy at University College London.
t h eor i e s of i nst it u t iona l de sign
Series Editor
Robert E. Goodin
Research School of Social Sciences
Australian National University
Advisory Editors
Russell Hardin, Carole Pateman, Barry Weingast, Claus Offe,
Susan Rose-Ackerman, Keith Dowding, Jeremy Waldron
Social scientists have rediscovered institutions. They have been increasingly concerned with
the myriad ways in which social and political institutions shape the patterns of individual inter-
actions which produce social phenomena. They are equally concerned with the ways in which
those institutions emerge from such interactions.
This series is devoted to the exploration of the more normative aspects of these issues. What
makes one set of institutions better than another? How, if at all, might we move from the less
desirable set of institutions to a more desirable set? Alongside the questions of what institutions
we would design, if we were designing them afresh, are pragmatic questions of how we can best
get from here to there: from our present institutions to new revitalized ones.
Theories of institutional design is insistently multidisciplinary and interdisciplinary, both in
the institutions on which it focuses, and in the methodologies used to study them. There are
interesting sociological questions to be asked about legal institutions, interesting legal questions
to be asked about economic institutions, and interesting social, economic and legal questions
to be asked about political institutions. By juxtaposing these approaches in print, this series
aims to enrich normative discourse surrounding important issues of designing and redesigning,
shaping and reshaping the social, political and economic institutions of contemporary society.
Other books in the series

Brent Fisse and John Braithwaite, Corporations, Crime and Accountability
Robert E. Goodin (editor), The Theory of Institutional Design
Itai Sened, The Political Institution of Private Property
Mark Bovens, The Quest for Responsibility: Accountability and Citizenship in Complex
Organisations
Bo Rothstein, Just Institutions Matter: The Moral and Political Logic of the Universal Welfare
State
Jon Elster, Claus Offe and Ulrich K. Preuss, Institutional Design in Post-Communist Societies:
Rebuilding the Ship at Sea
Adrienne Héritier, Policy-Making and Diversity in Europe: Escape from Deadlock
Geoffrey Brennan and Alan Hamlin, Democratic Devices and Desires
Eric M. Patashnik, Putting Trust in the US Budget: Federal Trust Funds and the Politics of
Commitment
Benjamin Reilly, Democracy in Divided Societies: Electoral Engineering for Conflict
Management
John S. Dryzek and Leslie Templeman Holmes, Post-Communist Democratization: Political
Discourses across Thirteen Countries
Huib Pellikaan and Robert J. van der Veen, Environmental Dilemmas and Policy Design
Maarten A. Hajer and Hendrik Wagenaar (editors), Deliberative Policy Analysis: Understanding
Governance in the Network Society
Jürg Steiner, André Bächtiger, Markus Spörndli and Marco R. Steenbergen, Deliberative
Politics in Action: Analyzing Parliamentary Discourse
Bo Rothstein, Social Traps and the Problem of Trust
Jonathan G. S. Koppell, The Politics of Quasi-Government: Hybrid Organizations and the
Dynamics of Bureaucratic Control
Mark E. Warren and Hilary Pearse (editors), Designing Deliberative Democracy: The British
Columbia Citizens Assembly
Graham Smith, Democratic Innovations: Designing Institutions for Citizen Participation
Exits, Voices and
Social Investment
Citizens’ Reaction to Public Services
Keith Dowding
and
Peter John
c a m br i d g e u n i v e r s i t y p r e s s
Cambridge, New York, Melbourne, Madrid, Cape Town,
Singapore, São Paulo, Delhi, Mexico City
Cambridge University Press
The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9781107022423
© Keith Dowding and Peter John 2012
This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2012
Printed in the United Kingdom at the University Press, Cambridge
A catalogue record for this publication is available from the British Library
Library of Congress Cataloguing in Publication data

Dowding, Keith M.
Exits, voices and social investment : citizens’ reaction to public services / Keith Dowding,
Peter John.
p. cm. – (Theories of institutional design)
Includes bibliographical references and index.
ISBN 978-1-107-02242-3 (hardback)
1. Human services. 2. Consumer satisfaction. 3. Public welfare
administration. 4. Total quality management in human services.
I. John, Peter, 1960– II. Title.
HV40.D69 2012
361.0068–dc23
2011052106
ISBN 978-1-107-02242-3 Hardback
Cambridge University Press has no responsibility for the persistence or

accuracy of URLs for external or third-party internet websites referred to in
this publication, and does not guarantee that any content on such websites is,
or will remain, accurate or appropriate.
Contents
List of figures page vi

List of tables vii
Preface and acknowledgements ix
1 Hirschman’s original idea 1

2 Exits, voices and the object of loyalty 25
3 Exit, voice, loyalty and neglect 51
4 The structure of UK public services and some
simple relationships 75
5 Evidence of the major EVL relationships 102
6 Exit, voice and welfare 130
Appendix A: Summary of the empirical literature testing EVL 142
Appendix B: Note on statistical methods 150
Appendix C: The survey instrument 152
Bibliography 175
Index 188
v
Figures
1.1 The voice–exit decision tree page 11

1.2 Dynamic relationship between voice and exit 12
3.1 EVLN responses to dissatisfaction 60
3.2 The Lyons et al. EVLN framework 66
5.1 Intentions to exit geographically following
previous individual voice 108
5.2 Intentions to exit privately following previous
individual voice, NHS 109
5.3 Private school if previously complained and did not like
state secondary school 110
vi
Tables
1.1 Responses to decline in quality page 11

1.2 Decline primarily arouses 15
2.1 Sub-game perfect Nash equilibriums from the EVL game 31
2.2 Sub-game perfect Nash equilibriums from the
extended EVL game 32
2.3 Types of goods provision 35
2.4 Types of exit in public provision 38
2.5 Types of voice in public provision 44
3.1 Hypothesized relationships between propensity
to invoke responses to dissatisfaction and the determinants
of responses 67
4.1 Numbers and response rates during the panel 78
4.2 Characteristics of respondents in the panel
and in the UK population 78
4.3 The incidence of exit and voice 85
4.4 The dynamics of exit and voice 86
4.5 Rotated factor loadings for voice and exit using
principal components (N=9944) 90
4.6 Voice and exit compared 91
4.7 Factor analysis: voice and exit compared 92
4.8 Change in satisfaction by change in
individual voice (complaints) 93
4.9 Change in dissatisfaction by change in intention
to exit geographically 95
4.10 Change in dissatisfaction by change in intention
to exit jurisdiction 96
vii
viii L ist of ta bl e s
4.11 Cross-tabulation of change in expectation of timely

treatment for an illness or injury by change in intention
to private exit 97
4.12 Change in expectation of timely treatment for an
illness or injury by change in locked into the NHS 97
4.13 Dissatisfaction, voice and exit 98
5.1 Change in intention to exit geographically by change
in individual voice to schools and satisfaction 105
5.2 Satisfaction by geographical exit in current year and income 107
5.3 Change in intention to exit geographically by change
in individual voice sorted out or not 111
5.4 Multilevel mixed-effects linear regression on
intention to exit geographically (voicers only) 112
5.5 Multilevel mixed-effects linear regression on intention
to exit and actual exit in health care 116
5.6 Multilevel mixed-effects linear regression on use
of private education 117
5.7 Multilevel mixed-effects logistic regression: local voting 122
5.8 Multilevel mixed-effects Poisson regression: participation 123
5.9 Multilevel mixed-effects Poisson regression: individual voice 125
5.10 Multilevel mixed-effects logistic regression: individual
voice to the NHS 126
5.11 Multilevel mixed-effects logistic regression: individual
voice in education 127
5.12 Summary of main EVL relationships 128
A.1 Main empirical relationships found in the literature 146
B.1 Summary of BIC/AIC values for the variance–covariance
for models of intentions to vote 151
Preface and acknowledgements
We started looking at the sorts of issues considered in this book as far back
as 1989 when we drafted a paper, ‘Rational Choice Approaches to Local
Government’ (delivered at the Political Studies Association conference in
Durham, 1990) on the functional distribution of goods and services in
urban communities. Our aim was to consider the efficient nature of pub-
lic service provision in terms of the characteristics of different goods and
services, and the changing demands of the public in differing economic
conditions. One side of that efficiency question is the possibility of mirror-
ing market efficiencies through choice of providers, either through quasi-
markets or through competition at the local level through ‘Tiebout exit’
that we briefly discuss in this book. Armed with an Economic and Social
Research Council (ESRC) grant (R000 23 3815 ‘Population Movements
in Response to Taxes and Services’) in 1993, with a follow-up grant four
years later (‘Citizen Choice and Population Movements: The Impacts of
Local Taxes and Services on Location Decisions’, R000 9000 236658), we
examined the ‘exit’ side with studies of geographical mobility in response
to local tax and services. A number of articles resulted from those grants
(Dowding et al. 1994; John et al. 1995; Dowding 1996; Dowding and John
1996; 1997; John 1997; Dowding and Mergoupis 2003; Dowding 2008;
2008) including a piece that examined Hirschman’s exit, voice and loyalty
relationship (Dowding et al. 2000) as it became increasingly clear that the
politics or ‘voice’ side of citizen satisfaction was crucial in considering effi-
ciency in public services.
The current research was conducted under the ESRC public services
initiative directed by Professor Christopher Hood, Gladstone Professor of
Comparative Politics at Oxford University. We received two grants: ‘Public
ix
x Pr e face a n d ack now l e d ge m e n ts
Services: Exit and Voice as a Means of Enhancing Service Delivery’ RES-

153-25-0056, and ‘Public Services: Exit and Voice as a Means of Enhancing
Service Delivery: Phase II’ RES-166-25-0012. This enabled us to collect
the five-year panel data on attitudes to public services that our empirical
results in this book are based upon. We thank the Economic and Social
Research Council for its support of our work over the years, especially the
latter awards under the Hood programme. This is the first of two planned
books based upon these research awards.
The survey reported on in Chapters 5 and 6 was conducted by YouGov
over a five-year period from 2005 to 2009. We would like to thank all the
people at YouGov, especially Briony Gunstone, for their advice and help in
facilitating the reasonably smooth process of conducting the survey and
providing the data in the form we required. The data are lodged at the
ESRC data archive for replication purposes and further research (www.
data-archive.ac.uk/).
A number of people have been helpful over the years. We would first
like to thank Thanos Mergoupis, who worked with us on our second ESRC
award, ‘Citizen Choice and Population Movements: the Impacts of Local
Taxes and Services on Location Decisions’, and from whom we learned
much. Together with Mark Van Vugt he helped us draft ‘Exit, Voice and
Loyalty: Analytic and Empirical Developments’ (European Journal of
Political Research vol. 37, no. 4, 2000, pp. 469–95). Parts of that review
appear in a few paragraphs especially in Chapter 3 and Appendix A, which
precises some of the review. We thank Thanos and Mark for allowing us to
use parts of the original article.
The bulk of this book was written at the Australian National University
in Canberra, when Peter John was a Visiting Fellow of the Politics Program
at the Research School of Social Sciences for eight weeks in July and August
2009 and he would like to thank the ANU for supporting that visit. We
profited from the unique opportunity it gave us to complete the first draft
of the book, prising Peter John away from his many projects in the UK. We
thank Emlyn Williams (ANU Statistical Consulting Unit), Tony Bertelli
and Ian Plewis for their advice on the modelling.
We also thank Stephen Biggs, George Boyne, Patrick Dunleavy, Dave
Lowery, Ken Newton, Lin Ostrom, Mark Schneider, Bob Stein and Paul
Teske, for discussions of some of the issues concerning efficient public ser-
vice provision over the years. We would also like to thank Perri 6, André
Alves, Alan Fenna, Anne Gelling, Andrew Hindmoor, Oliver James and
three anonymous reviewers for their comments on the manuscript of this
book. We also thank Bob Goodin and John Haslam for advice on restruc-
turing the original draft to its present form.
1
Hirschman’s original idea
New-formula Coca-Cola
In 1985 Coca-Cola introduced a new version of its trademark brand of soft
drink. Unofficially called ‘New Coke’ until it was rebranded officially as
Coca-Cola II in 1992, the new brand proved controversial. On its intro-
duction the company stopped production of the original formula. Three
months later, however, following protests and falling sales, Coca-Cola
reintroduced the original brand under a new name, Coca-Cola Classic.
What had gone wrong?
The company had seemingly done its homework. Blind tastings amongst
focus groups had shown most people preferred the new formula to the old
and also to its rival Pepsi. (It was that rivalry that had led to Coca-Cola’s
innovation, in an effort to regain customers from Pepsi.) Why did New
Coke fail when it was preferred by a majority of people? The early focus
group research might give some indication, for those groups often con-
tained a vocal minority who preferred the old formula. And that vocal
minority could sometimes turn focus groups away from the new drink.
That is what happened nationally. Whilst sales of New Coke at first held
up, there were vociferous groups of people who campaigned against it.
Gary Mullins formed a pressure group ‘Coca-Cola Drinkers of America’
to lobby the company to reintroduce the old formula or to sell it to a new
company. In the Deep South there were some street protests; bottling com-
panies became worried and also lobbied Coca-Cola. Finally Coca-Cola
announced that it would reintroduce the old formula, calling it Coca-Cola
Classic, alongside the New Coke. What seemed to have happened is that
some people were so upset about the new version of their favourite drink
1
2 H i r s ch m a n ’s or igi na l i de a
that they campaigned against it, managing to turn others against it too.
This behaviour is an example of what Hirschman calls ‘voice’. What really
mattered to Coca-Cola of course were not the voices against their new soft
drink but falling sales. Consumers stopped buying Coca-Cola and shifted
to Pepsi or other soft drinks. Moving away from a brand is what Hirschman
calls ‘exit’. Furthermore, the story might also reveal something about the
role of what Hirschman calls ‘loyalty’. At the press conference announcing
the reintroduction of Classic Coke, Coca-Cola stated ‘the simple fact is
that all the time and money and skill poured into market research on the
new Coca-Cola could not measure the deep and abiding emotional attach-
ment to original Coca-Cola felt by so many people’. It seems loyalty to the
old product was an important determinant of the failure of the new one. So
in one example there are all the key Hirschman concepts: exit, voice and
loyalty. This book examines how they might play out in different contexts.
Exit, voice and loyalty

In 1970 the development economist Albert O. Hirschman published a
short book entitled Exit, Voice and Loyalty: Responses to Decline in Firms,
Organizations and States. The book had an immediate impact upon aca-
demia in various disciplines including political science, management and
social psychology. It reached the status of a classic almost on publication
(Barry 1974) and has since then been massively cited and utilized in the
academy.1
The argument of the book is simple. When there is a decline in the
quality of the production of a firm – or any organization including the
state – there are two general ways in which consumers or citizens might
react. They might exit from the product or they might complain about its
decline. For example, imagine one likes to pop into a local bakery a few
mornings a week to buy a particular cake they sell. Then one day one’s
regular order has changed in some way. Perhaps the owner has started
baking it in a larger size, which is more than one wants with one’s morn-
ing coffee. Or perhaps the price has suddenly doubled. Or maybe the cake
looks the same at the regular price, but the baker has changed the mix
slightly and it just doesn’t taste quite the same anymore. As a consumer
one might remark to the baker that you are not so keen on the new cake.
You might inquire about the new mix and say you preferred the old version
and ask if they are going to go back to the original recipe. Or complain that
the cake is a little too much for morning coffee or that whilst you still love
1
Inputting ‘exit, voice and loyalty’ to Google Scholar gives ‘about 13,600’ hits (May 11,
2011).
E x i t, voice a n d l oya lt y 3
it the price is beyond your budget now. In other words, by some friendly
remark or complaint, you have signalled that in your opinion the quality
of the cake has declined (in taste, size or in terms of value-for-money).
The point of such remarks is to suggest to the baker that she returns to
the original recipe, size or price. Making the complaint might be difficult,
especially if the baker has become a friend over the years, but is certainly
possible. And given that one is a regular customer, especially in a small
bakery, one’s comment might well have an effect. The baker will want to
keep customers happy.
Alternatively one might simply stop buying cakes from that bakery. If
one’s favourite cake is no longer on sale, or not so much to one’s taste,
one might try different cakes from a different bakery or shop nearby. One
might choose a different type of product for the morning snack – a bar
of chocolate or a jam tart – or save money by skipping sweet food with
morning coffee. These are different exit options. The exit option is to with-
draw one’s custom from the bakery. And indeed, perhaps the voice option
could only really be effective if the exit option is available as an implied
threat. If the bakery is the only place that sells cakes locally, then short of
stopping eating cakes altogether, one might be forced to carry on buying
the less favoured cake. The baker might well prefer to keep her customers
happy, but if they are willing to go on buying the new cakes she has less of
an incentive to go back to the old recipe or price – especially if she makes
more money than before. So voice might not be an effective option without
the threat of exit, even if that threat is only an implied one.
Of course, the two responses are not direct competitors. One might try
complaining first and only if the baker does not respond would one then
take one’s business elsewhere. Indeed if the customer has a friendly long-
term relationship with a local bakery that is the most likely response. If the
customer purchases his cakes in a busy supermarket where the checkout
operators change frequently and so there is no personal relationship then
he might be much less likely to signal his dissatisfaction through voice. If
one does not have a personal relationship, then making a complaint might
be more difficult, or more pertinently, making a complaint might be much
less effective. Complaining at a checkout that one’s favourite cakes are no
longer available might receive a sympathetic response, but the checkout
operator can do nothing about it. Unlike the baker, checkout operators
have no control over the products on sale at the supermarket. To com-
plain about the changed buying policy of a given supermarket the cus-
tomer would have to speak to the supermarket manager; even then the
local manager might have little control over the products on the shelves.
Multinational supermarkets often make buying decisions at the regional
level or even, for big policy decisions, at the headquarters in another
country. To take one’s suggestion to such a high level requires a great deal
more effort than merely mentioning one’s disquiet over the counter as one
might do in a local bakery. And if one does make that effort with the big
company one might expect a less helpful response. A local baker is highly
likely to be interested in the views of one of their customers; the cake-
buyer supreme of a multinational supermarket corporation is less likely
to be influenced by one complaint. As the cost of voice goes up and the
expected benefits go down, the exit response to quality decline becomes
more likely. In these cases voice just does not seem worthwhile.
We might note also that the implied threat of exit is much greater for
the local bakery than for the supermarket. A local bakery might notice the
impact of one of their regular customers taking their business elsewhere; a
large supermarket chain would not. Though both businesses must take the
general threat of exit equally seriously – since in either case too many cus-
tomers exiting could seriously jeopardize their business – the size of the
customer base affects the efficacy of each individual voice or exit to change
policy. Exit in the general sense, however, is the universal motor of com-
petitive markets. The importance of this element in the Coca-Cola story
is that Gary Mullins organized a campaign and the effect of his campaign
snowballing into a broad movement caused the multinational company to
respond.
Hirschman complicates his story by suggesting that a third variable, ‘loy-
alty’, might mediate in some manner between the two possible responses
to the decline he identifies. Essentially if one is loyal to a firm, organiza-
tion or state then one might be less prepared to exit than if one felt no such
loyalty. So if the customer felt some loyalty to the baker, say the baker and
customer had built up a friendship over the years, he might be less inclined
to exit. To suddenly stop going into the baker’s might be noticed, and that
would affect the friendship. The customer might find himself continuing
to buy cakes there even though he does not like them as much, or has dis-
covered nicer or better-value ones in a new delicatessen nearby. Loyalty
might lock one into a particular organization, even though one can see
that one might be better off if one took one’s custom elsewhere.
The fact that loyalty might make exit more difficult also might entail
that it makes voice more likely. One might simply remain loyal to the local
bakery and continue to buy cakes there (especially if it is run, say, by one’s
sister-in-law). Indeed, making a complaint about a business matter to a
friend can be more difficult than making that same complaint to a stran-
ger. Out of loyalty one might simply suffer in silence. Silence and non-exit,
of course, are also possible responses to quality decline, and loyalty might
also make these more likely than either exit or voice. It might depend
on the nature of the loyalty and the nature of the product decline. The
important issue for Hirschman, however, is that if the exit option seems
too costly or impossible, and if the decline in quality is marked, one might
get up the nerve to say something. However gently one might explore the
issue, the customer notes the cakes are not the same as they once were
and whilst nice, one prefers the old recipe or size. Or that one cannot buy
the cakes as often as the price is a bit beyond one’s means. The customer
might find some other excuse for shopping elsewhere, or might simply
exit. Hirschman hypothesized, reasonably enough, that loyalty makes exit
less likely, and given that voice is the only other positive response possible,
loyalty, on average, tends to make voice more likely. In the supermarket
case the only reasonable response (given the high cost of voice) is either to
suffer the change in silence or to exit. In the bakery case, exit remains an
option, but voicing complaint is a real option that might work. For Gary
Mullins, exiting from new-formula Coca-Cola to Pepsi did not satisfy him.
Remaining loyal to old-formula Coca-Cola, he had to voice in order to try
to get it reinstated.
Hirschman argues that exiting is the standard economic response to
quality decline. Firms tend to respond to signals given to them by their
customers. If custom drops off because their clientele chooses to purchase
products from rival firms, the company must respond to those signals.
They might have to reduce price if custom drops rapidly after a price hike,
or return to an original size or recipe if custom falls just after an alter-
ation in the product. In pointing out the potential signals of voice activ-
ity Hirschman was directing economists’ attention to consumer response
outside of the general market equilibrium models, albeit a response well
known to manufacturers themselves who often pilot new brands and
product changes carefully taking account of consumer feedback, and
who conduct regular market research into the opinions of their customer
base. Though – as the Coca-Cola example demonstrates – not always
successfully.
Whilst exit and voice are two possible responses to a decline in quality
in some regard, they can be used in different ways. The important elem-
ent in Hirschman’s book, however, is that the two responses – exit and
voice – can be the result of a decline in quality, but they can be mediated or
affected by the third psychological variable, loyalty, which makes exit less
likely. Loyalty can be seen simply in cost terms. In this sense one’s loyalty
is simply a measure of how likely it is that one will exit, given the rela-
tive objective costs and benefits of voicing and exiting. We might explain
the amount of loyalty we find in such a calculation of the relative costs
and benefits of voice and exit by a variety of means, as we see later in this
book. However, the amount of loyalty a person has might be from rela-
tive cost–benefit calculations over exit or voice behaviour. It can also be
conceptualized theoretically to appear in some circumstances rather than

others.
Hirschman drew a number of implications from these simple reflections
on exit and voice that we will briefly consider in this opening chapter. We
then turn to various applications of the argument and some problems that
have been noted in the literature. First however, we consider Hirschman’s
own original application of his exit, voice and loyalty framework.2
Hirschman’s puzzle
Ideas, concepts, models and frameworks in the social sciences are usually
developed in order to help solve a puzzle noted by the originator. In that
sense they are problem-solving devices. It is often useful or important to
remember the original puzzle that motivated the creator since ideas get
taken up by others and put into contexts that do not always quite fit the
original puzzle. Modifications occur to ideas and concepts and they end
up being used quite differently from how the creator envisaged them. As
we shall see, this has happened to Hirschman’s exit, voice and loyalty idea.
So what was his original puzzle?
Hirschman uses the exit, voice and loyalty framework to consider the
issue of introducing competition into a system of public services. In that
sense, though the book was published over forty years ago its resonance is
very contemporary. He wrote the book in order to defend an observation
he had made about the efficiency of the Nigerian railway system he had
discussed in his previous book (Hirschman 1967), which he felt could be
extended to a whole realm of public services. What we refer to as his ‘exit,
voice and loyalty framework’ was devoted to explaining a puzzle, a puz-
zle that is particularly pertinent to economics and the idea that competi-
tion improves the efficiency of services through the standard exit response.
Hirschman noted that the Nigerian public railway system deteriorated at
just the time that competition with other forms of transport was increas-
ing. He argues that it was the availability of ready alternatives to rail trans-
port that facilitated that decline in quality. He suggests that where there is
a state monopoly in some service, such as transport or education, then, in
a free and democratic society, declining quality will lead their customers
2
We describe Hirschman’s account as a framework. We take it that hypotheses can be
logically drawn (and subsequently tested) from formally specified models. Hirschman’s
framework is not a model in that formal sense (and we examine attempts to formally
model the framework in Chapter 2). However, he does draw some important hypotheses
from his argument that we attempt to examine and test in this book. We term such a non-
formal model a framework (Morton 1999).
H i r s ch m a n ’s pu z z l e 7
to complain and these complaints will put pressure on government to

ensure that the service improves. In that sense voice is the standard polit-
ical response to decline in quality. However, if there are alternatives that
people can choose to use instead then they are less likely to put pressure
on government. So when the services of the Nigerian railway declined,
their passengers did not pressurize government to do something about it;
rather, many of them chose the cheaper and more immediately effective
response of departing from the rail system altogether by using buses or
private cars.
Without the signal of complaints the government did nothing to improve
the state monopoly rail service, and as the rail service was not a private com-
pany it did not have strong incentives to respond to the exit signals. Rather,
as its customer base eroded the quality of its services declined further. The
general idea that underlies this account of the Nigerian railway system in
the 1960s is that providing an exit option, or more generally making exit
easier, might have the effect of driving out voice. If voice is a relatively
ineffective process for arresting quality decline, increasing the propensity
to exit will not necessarily be a bad thing; indeed it might increase effi-
ciency gains in a given product market. However, if voice responses do
have qualities missing from the exit signals then driving out voice might
have an overall deleterious effect on quality by increasing exit opportun-
ities. Hirschman notes this particularly with regard to the state-run rail-
way systems and to state education. He suggests that those who can afford
private cars are more likely to move from the railway system than those
who cannot afford them. Those able to shift to cars are likely to be richer
and more educated and so also those who could lobby government most
effectively. Similarly, if the state education system declines, then those who
are most likely to remove their children from state-funded schools are the
middle and upper classes. Again, on average, it is these people who are
most able to effectively vocalize their complaints to teachers and adminis-
trators within the school system. Their voice and votes are also likely to be
important within the democratic system. Once those voices are lost within
the state school system there are fewer incentives for teachers, adminis-
trators and politicians to arrest the decline in quality. Exit, hypothesizes
Hirschman, drives out voice.
One of Hirschman’s major theses in his book is that making exit eas-
ier causes voice to decline, possibly leaving behind a class of people who
have no practical exit option, and are unable to voice effectively. Thus if
state welfare systems – whether in transport or education, or health or
whatever – start to decline causing exit they will continue to do so. This is
an important implication to which we will return both theoretically and
empirically in this book. Its pertinence to modern debates about ways of
improving public services should be obvious, as new public management

(Dunleavy and Hood 1994; Ferlie 1996; Barzeley 2001; McLaughlin et al.
2002), entrepreneurial government (Osborne and Gaebler 1992; Wood and
Waterman 1994), public value management (Moore 1995; Stoker 2006)
and choice in public services (Le Grand 2007) are pushed throughout both
the developed and the developing world. He also noted various conceptual
features of voice, exit and loyalty, and we will take each of these in turn,
keeping in mind his primary hypothesis about exit driving out voice with
regard to state-monopoly public services.
Elements of the three elements

E x it
Exit only performs the function of signalling deteriorating quality if those
consumers who exit are not replaced with incoming ones. For example,
the bakery’s new cake might not be to the taste of the regular customer
who takes his business elsewhere. But if it is preferred by other customers
the bakery might not notice the departure, or if it does notice it, the loss is
worthwhile for the extra gains in business. Of course, if it is truly the case
that the problem is a decline in quality, then gaining some replacement
consumers for those exiting might seem unlikely. However, our bakery
example does not necessarily entail a decline. Rather the baker is offer-
ing a slightly different product, and whilst some regular customers might
not like the new variant, others might prefer it. The issue of decline versus
change is an important one to which we return when considering the rela-
tionships of voice and exit in empirical examples.
In order for exit to perform the signalling function that the firm can
respond to, the speed of exit is important. If all consumers exited imme-
diately and completely from a company’s products then the firm might
go out of business before it could respond. So the customer base needs to
erode more slowly if the company is to take note, find out what the problem
is and respond. Hirschman writes about the different types of consumers,
whom he calls the alert and inert, and suggests both are needed for effi-
cient signalling. An organization needs some consumers who respond fast
to quality decline, some who respond more slowly and others that perhaps
do not respond at all. He suggests that the exit mechanism works most effi-
ciently with the right mix of alert and inert consumers.
The exit signal can also be difficult to interpret. For complex products
like computers or cars, for example, quite why consumers stop buying one
model and choose another is not obvious. Similarly, as we see in the public
sector, if the exit option is used geographically by households who move
E l e m e n ts of t h e t h r e e e l e m e n ts 9
from one local authority jurisdiction to another, how far these geographic
exits are related to specific services on offer, or ones on offer in other areas,
or whether they are the result of other considerations entirely might not be
obvious to the authorities. In other words the exit signal can be very indis-
tinct and unclear. In fact so crude is the exit signal, that when companies
realize they are losing customers, often their first response is to find out the
cause of the dissatisfaction. They commission voice through customer or
public surveys to try to find out what the problem is. Thus exit is a dramatic
signal to which competitive firms must respond, but it is not necessarily
a clear one. The nuances of the problem need voice. The indistinctness of
exit signals in the public sector might be even greater and whilst public
authorities also commission surveys to gain greater information through
voice, we argue below that these voice signals can be even more unclear in
the public context.
Voice
Exit might seem to be a fairly straightforward response (though in
Chapter 2 we distinguish between different types of exit). Voice is less
straightforward, however. Hirschman (1970, p. 3) defines voice as ‘any
attempt to change, rather than escape from, an objectionable state of
affairs, whether through individual or collective petition to the man-
agement directly in charge, through appeal to higher authority with the
intentions of forcing a change in management, or through various types
of action and protests, including those that are meant to mobilize pub-
lic opinions’. We see therefore that voice is a multiform and complex
response to quality decline. Again in Chapter 2 we unpack voice some-
what to distinguish different elements and forms in which it might be
realized. However, we can see immediately that there could be a form
of private voice – our customer complaining directly to the baker – or
a more collective voice where consumers might organize a complaint to
the regulatory authorities over, say, the amount of a chemical used in the
baking of cakes. Hirschman somewhat privileges the latter, as he makes
a direct comparison between exit as an economic response and voice as
a political one. He seems to see voice very much as an interest articula-
tion rather than as a simple complaint. However, both are important and
might interact in different ways to potential ways of exiting. We explore
these issues later in this book.
Exit tends to be a binary response. If one is dissatisfied one can either
leave or not. Voice is more nuanced and one can not only demonstrate
dissatisfaction (or indeed satisfaction) but also the degree of dissatisfac-
tion and state what the organization can change to increase satisfaction.
Hirschman (1970, p. 31) argues that there might be negative returns to

voice particularly in the political sphere. If people harass government, the
state might not only not comply but become more determined not to give
in to protest. Hirschman again argues that the right mix of alert and inert
citizens is needed at the political level to ensure that government receives
signals and that apathy does not suddenly burst forth with too much vig-
our. Like exit, to be effective voice must provide the signal and then allow
the organization time to respond before escalating.
Voice can be a residual of exit; that is, voice becomes the only option
when exit is simply impossible for, say, a monopoly supply of necessities.
The degree of voice might depend upon inelastic demand or the lack of
opportunity for exit. So the amount of voice would increase as the oppor-
tunity for exit decreases up to the point where exit is impossible. What
should we expect to happen where exit is impossible? On the one hand,
voice is the only option for change. On the other hand, if exit is impos-
sible firms might have no incentive to respond to voice and if firms have
no incentive to respond to voice then consumers have little incentive to
use it in the first place. This might be the situation in non-democratic
authoritarian societies. Citizens are dissatisfied, exit is almost impossible
with closed borders, but government has little incentive to respond to voice
demands. So the citizens suffer in silence. Of course, protest is possible
even under the most authoritarian regimes, but the costs of such protest
can be immense (Clark et al. 2007; Wright 2007). When considering the
feasibility of exit and voice strategies we must always bear in mind the
likely costs and benefits. We might see an example of the changing costs
of voice with the opening up of Eastern Europe. Voice had been costly
and so sparingly used despite great dissatisfaction with the communist
regimes. With the break-up of the Soviet Union and perceived relaxation
of control in communist Europe, voice flowered with many protests and
eventual regime change (Hirschman 1995; Blanz et al. 1998; Pfaff and Kim
2003; Pfaff 2006).
Hirschman generally views voice as a complement to exit rather than
as an alternative. As the implied threat of exit generally needs to underlie
voice for the latter to be effective the two are not strict alternatives. Voice
rather can act as an early signal for organizations to respond to before exit
kicks in. But voice might be an alternative in the sense that once someone
has exited then they do not have an effective voice. Voice needs to be the
alternative used first, with exit as the final option (Hirschman 1970, p. 37).
Table 1.1 shows a two-by-two table of the possibilities.
Category 1 occurs when the product has declined in quality and shows
no sign of improving despite the consumers’ voicing complaint. Exiting
silently is the response of the individual who does not discern any point in
Table 1.1 Responses to decline in quality
Exit Non-exit
Voice 1 2
Silence 3 4
VOICE
Y N
EXIT
Y N Y N
1 2 3 4
Figure 1.1 The voice–exit decision tree
complaining and simply shifts to another product. Finally there is a cat-

egory of people who suffer in silence. In fact it may be better to represent
this as a stylized decision tree, where at different points in time consumers
decide whether or not to exit or voice in response to the changing quality
of a product.
In Figure 1.1 the decision tree begins with the decision on whether to
voice following dissatisfaction. If the answer is no then the consumer must
decide whether or not to exit, leading to points 4 and 3 corresponding
to those places in the matrix. If the individual uses voice, she must then
decide whether or not to exit as well, leading to points 1 and 2. But certainly
in this first fork there are two distinct ways in which we could reach points
1 and 2 in the matrix, which are illustrated in Figure 1.2. The individual
could exit and voice simultaneously – a ‘strident exit’ in which voice is
used to advertise one’s exit decision. Or the decision to exit could be taken
at a later date, following the failure of voice to raise standards. Hence the
decision tree begins with the individual’s product satisfaction level, and
the satisfaction level needs to be judged before each decision on whether to
voice and exit. We ignore the path where a consumer is satisfied, assuming
(for the moment, but we examine that course later in this book) that she
t1
Y N SATISFACTION
Y N VOICE
Y N Y N EXIT
t2
SATISFACTION Y N
VOICE Y N
EXIT Y N Y N
Figure 1.2 Dynamic relationship between voice and exit
will neither voice nor exit when satisfied, and track the decision whether
to voice or exit, or do both. After that decision, the consumer must give
the firm or organization or group time to react. How long a time she might
give is relative to the situation she faces. A citizen complaining to a local
government about a streetlight not working may expect almost immedi-
ate action and could be satisfied soon after. But if one voiced concern over
the general course of the government then one may allow several years for
a response – if any – and only subsequently would satisfaction perhaps
follow. Either way, whenever we judge the pertinent point in time, satisfac-
tion is gauged and a decision over whether to voice again or to exit (or do
both) is made. A decision not to bother to voice, given its earlier failure,
but rather to exit after t2 might be perfectly rational; so might a decision
just to voice again, if the costs of exit are high, or it might be rational to do
nothing given the high costs of exit and the failure of voice the first time.
Hence the more complex nature of the decision tree in Figure 1.2 where
we also label satisfaction as a decision point in time (labelled t1…t.n). This
might be modelled as a game against nature, where the reaction function
of organizations affects the probabilities assigned to different outcomes for
the player.
Voice can be delivered in a variety of nuanced ways. Hirschman sug-

gests it is an art that can be developed. As well as the problem of ‘hearing’
voice if the most articulate consumers choose to exit, he also suggests that
if exit is too easy, then people might not learn the art of voice: ‘the pres-
ence of the exit alternative can therefore tend to atrophy the development
of the art of voice’ (Hirschman 1970, p. 43, emphasis removed). He sees
this argument as the central point of his book and it is clearly related to his
view that voice is a political activity in a democracy, which should not be
threatened by commercializing or introducing markets into aspects of the
political process. His major example of this process outside of his starting
case study of the Nigerian railway system is state education.
Hirschman argues that a state monopoly can be superior to a com-
petitive system if exit is ineffective as an improving mechanism but voice
can be made into an effective mechanism. Exit will be ineffective for state
monopolies if exiters are also those who would put pressure on both the
service providers and elected politicians if there were no exit option. But
given they leave, they do not provide that pressure. The exiters might get
more satisfaction from their new private service providers, but their exit
will not help arrest decline in the education service they have left. This
would not matter if all could exit. However, if some are locked in to the
state system because they do not have the means to purchase the goods
privately then the quality of their services will decline. This is Hirschman’s
central insight, and the one upon which we concentrate a great deal of
attention in this book.
Loyalty
Loyalty is the third element in Hirschman’s framework that affects the
exit–voice balance. For Hirschman, consumers must balance the trade-off
between the certainties of exit (the certainty of leaving behind what they
had for something else) against the uncertainty that their voice might lead
to improvements. He suggests that the willingness to trade is affected by
loyalty described as ‘a special attachment to an organization’ (Hirschman
1970, p. 77). He suggests
as a rule … loyalty holds exit at bay and activates voice. It is true that, in the
face of discontent with the way things are going in an organization, an individ-
ual can remain loyal without being influential himself, but hardly without the
expectation that someone will act, or something will happen to improve mat-
ters. That paradigm of loyalty ‘our country, right or wrong,’ surely makes no
sense whatever if it were expected that ‘our’ country were to continue forever
to do nothing but wrong. (Hirschman, 1970, p. 78)
Hirschman defends loyalty as rational in the sense that exit can be a

destructive behaviour, and too fast an exit means that an organization
might not be able to respond. Showing some loyalty, hoping things will
turn round and encouraging change through voice, is thus a perfectly rea-
sonable course of action for an unhappy individual. Hirschman suggests
that loyalty is likely to operate to a greater extent between products that are
close substitutes because he thinks loyalty mitigates against fast change.
Where there is no close substitute Hirschman thinks consumers might
stay with the merchandise until the quality has declined massively and
then move on to a substitute product. So Hirschman suggests that there is
a paradox in that loyalty is most functional when it appears to be needed
the least – that is, where there are close substitutes. There is no extant evi-
dence that we have found that supports this claim and it seems odd to
suggest that consumers would stick with Coca-Cola longer if it declined
in quality if there was a close substitute such as Pepsi than if they could
only switch to water. We do not attempt to examine this claim ourselves
in this book
Loyalty mediates the relationship between exit and voice in two ways.
First, since it makes exit less likely, it seems to increase the likelihood of
voice by default. Secondly, loyalty might increase the effectiveness of voice
thereby making voice more likely (Hirschman 1970, p. 83). To go back to
our bakery example, a loyal customer, that is a regular one, who complains
about the change of ingredients in her favourite cake might be listened to
more readily than someone who shops there irregularly.
As well as the positive effects of loyalty, Hirschman points out that
it might have inefficient results. Loyalty might lead people to be too
unwilling to exit, thereby slowing down improvements in products; and
organizations might promote loyalty not only to reduce exit, but also to
repress voice. Whilst feedback in the form of exit and voice might be in
the long-term interests of an organization, the short-term interests of
management might be to repress both in order to give the impression
that all is well. If the price of exit is high but the cost of entry is zero, then
loyalty might have deleterious effects. Hirschman gives the examples of
family, gangs and country to illustrate this possibility. He suggests that
exit from these sorts of organization brings high costs as it is seen as
betrayal, but depending on the nature of the organization, voice might
also be discouraged and difficult. Criticizing one’s country, the gang or
family values might seem as big a betrayal as exiting them. Indeed in one
sense, in such cases full exit might seem almost impossible. Even if one
leaves one’s country, exits a gang or has nothing further to do with one’s
family, one might still care deeply about it or them and what is happen-
ing to those left behind.
Cr i t icisms a n d e x t e nsions of t h e EV L f r a m ewor k 15
Table 1.2 Decline primarily arouses
Exit Voice
Exit organization is Competitive business Organization where
sensitive dissent is allowed but
institutionalized
Primarily to voice Public enterprise subject Democratically
to competition responsible
organizations
commanding loyalty
Modified from Hirschman 1970, p. 122
The EVL mix

Hirschman tries to consider the optimal exit–voice mixture that can miti-
gate decline in quality. He recognizes that the mix will not be the same for
all products or organizations, which he illustrates in a table we reproduce
(Table 1.2). Efficient processes occur in the top left-hand quadrant and in
the bottom right-hand one. In the former, exit provides firms offering rela-
tively simple products in competitive markets with the signals they need to
see how their products fare in relationship to their rivals. They can readily
see how well they do relative to other firms and change their productive pro
cesses accordingly. At the bottom right, voice leads to responsiveness from
organizations composed of loyal members. The problematic cases occur
where voice is used primarily but exit is what managers would respond
to, or where exit occurs too quickly for response and voice mechanisms
might have created more responsiveness. Hirschman concludes his study
by suggesting that optimal mixes of voice and exit are difficult to achieve,
and that organizations that rely primarily upon one mode of response can
be improved with a healthy dose of the other at decent intervals.
Criticisms and extensions of the EVL framework

We introduce considerable extensions to the EVL framework in Chapter 2.
Prior to that, however, we consider some early criticisms of the framework
and presage some of our extensions here.
Collective action problems

Hirschman’s definition of voice equivocates across two distinct ways in
which it may operate. One may individually petition a firm or one may
take part in some collective protest. One might not choose to petition the
firm because one feels that the effort is not worthwhile. Writing to the
multinational supermarket chain is troublesome and unlikely to change
the policies of the company. In the latter case one might feel that a col-
lective response is what is required. Coca-Cola might have reintroduced
the original formula coke simply because too many people were exiting
from their new version, but Gary Mullins’ pressure group, ‘Coca-Cola
Drinkers of America’, undoubtedly publicized the issue and was almost
certainly highly influential in the company’s decision. Collective protest
can be directed at a company itself. A petition complaining about the
changed product might lead the executives to reconsider the issues if there
are enough names on the petition. Collective protest can also be directed
at the wider public in order to change their attitudes. Collective voice can
encourage collective exit. Guillermo O’Donnell (1986) makes a distinction
between ‘vertical voice’ and ‘horizontal voice’. The former can be seen as
voice directed at the organization or leaders themselves, such as a peti-
tion to a company; horizontal voice as directed at other citizens or con-
sumers. Complaining about the change to one’s friends and neighbours
counts as horizontal voice, and so does directing pressure activity towards
the public.
Any form of collective voice is subject to the well-known collective
action problem (Olson 1971; Sandler 1992). The collective action problem
arises in this context: though it is in everyone’s interest for there to be a
protest or set of complaints about the decline in some public service, given
that voice is costly, it is also in each person’s interest to leave the voice
activity to someone else. If everyone leaves it to others then there will be
no voice activity at all; there will be no protest about declining quality
and hence no signal to providers. Whilst Hirschman ignores the collective
action problem in his 1970 book, he does reflect on the problem in later
essays (Hirschman 1986) and then dismisses Olson’s idea of the collective
action problem as ‘nonsensical’ and ‘obviously absurd’ (Hirschman 1990,
p. 159). In fact Hirschman sees people engaging in collective action for
expressive reasons – that is for fun and because they care about public
issues despite potential costs. (Hirschman might take this line because as
a young man he was very active in protests and later worked with Varian
Fry to help smuggle people out of Vichy France to escape the Nazis.)
Hirschman (1986) suggests that only vertical voice is costly and that hori-
zontal voice brings expressive benefits. The former is subject to the col-
lective action problem and the latter not. However, horizontal voice may
be costly in certain circumstances – complaining about the government
under repressive regimes, for example – and the larger the group the less
beneficial each horizontally voiced concern will be in overcoming the col-

lective action problem of vertical voice. We can see that the relative costs
and relative benefits of exit and voice will vary massively across contexts.
Thus the cost–benefit calculations involved in voice activity might differ
greatly. This will affect the nature of exit–voice trade-offs.
Whilst there is a great deal of sense in Hirschman’s arguments about
collective action they are not anti-Olsonian as he seems to think. Olson
recognizes that collective action occurs because of selective incentives, and
amongst those selective incentives are expressive reasons. Nevertheless,
expressive reasons for engaging in collective action are more likely in
some contexts than they are in others. It is all very well to take part in
some public rally in the summer where there are opportunities to picnic
and meet friends; or in riots where there might be some fun in destroying
others’ property; there is not so much fun in writing a letter to the coun-
cil about the poor refuse-collection service; nor indeed in voting, espe-
cially if one is alone and it is raining. Indeed, the truth in the Olsonian
collective action problem is that where goods are collective there will be
under-mobilization – that is, fewer people will voice complaint than there
are potential complainers. We see under-mobilization in all areas where
there are collective action problems, though it will be greater the lower
the potential benefits, the lower the selective incentives and the higher the
costs. Where interaction between people is more likely then horizontal
voice might lead to greater vertical voice. In Chapter 2 these considerations
lead us to distinguish between individual voice and two sorts of collective
voice, but here we will simplify and merely note the distinction between
individual and collective voice. One might individually complain about a
particular problem that one has had with a service – say, a complaint to the
head teacher about the attitude of a teacher to one’s child; while collective
voice is directed at a service as a whole that everyone using that service
might appreciate – say, voting for a political party that has promised to
increase expenditure on education. Therefore we make an analytic distinc-
tion between two forms of voice. One we call ‘individual voice’ and the
other ‘collective voice’. We define them:
Individual voice = actions where the intention of the individual in act-
ing is to bring about the desired effect solely through that action.
Collective voice = actions where the intention of the individual in act-
ing is to contribute to the desired effect through that action.
The distinction depends upon the motivation of the actor, and mirrors the
familiar distinction in economics between private and public goods. A
public good is often defined by two conditions:
1. A good is ‘jointly supplied’ (or ‘non-rival’) if one person’s consump-

tion does not reduce the amount available to others. For example,
the nation’s defence force protects my neighbour equally with myself
from other nations’ aggression, and his consumption of that protec-
tion does not reduce my protection.
2. A good is ‘non-excludable’ if the consumption of one person makes
it available to other group members. If I sweep the snow and ice from
the front of my house so I do not slip, I will also enable others not to
slip when they walk by my house.
So a purely private good is rival and excludable whilst a purely public good
is non-rival and non-excludable. Most goods are neither purely private nor
purely public and similarly our distinction between individual and col-
lective voice will have many behaviours falling between the two. However,
what is important is that for voice activities that share some collective fea-
tures there is a collective action problem of a greater or lesser extent. That
is crucial when we consider the exit–voice trade-off, as we see in Chapters
4 and 5. It is only those empirical issues that really matter for making this
analytic distinction.
Citizens must decide whether individual or collective voice is likely to
be effective. Citizens’ use of individual voice depends upon their belief in
its expected efficacy, which similarly depends upon their legal rights, and
the degree of competition, both political and economic, plus any expres-
sive benefits of the act itself (what Hirschman 1982, chapter 5, has called
the ‘rebound effect’). We should not expect to see the same relationship
between individual and collective voice, and exit. Nor should we expect
the same relationship between dissatisfaction and both sorts of voice.
Product ‘quality’
Hirschman tends to assume that there is an objective function ‘quality’ of a
product and if this objective function falls then consumers have an incentive
to act through exit or voice to try to halt the decline. However, as we have
seen in our bakery example, consumers may have different attitudes towards
the quality of a product. This is most easily seen in political terms, where
some people may want, say, a local government to provide a programme
of ‘left-wing’ policies, where others may want a menu of ‘right-wing’ ones.
Within each set of policies there may be some agreement – over, say, refuse
collection – and over which both groups agree an objective quality func-
tion. But one set of consumers or citizens may think the overall quality of
the local government services is declining as those services take on a differ-
ent form, for example, they become more ‘right wing’. Contracting out and
privatizing service provision is controversial. Some people believe contract-

ing out and privatizing will overall bring quality gains. Others might think
they bring no gain and more problems, or might be concerned about the
employment conditions of contracted-out companies, or want the govern-
ment to provide more redistributive services and so on. In other words, dif-
ferent people might have different views about the nature of some collective
goods they all receive which are not really about the ‘quality’ of the good but
about the type of goods and services to be provided (Barry 1974). Such con-
flicts might concern the supply of private goods, disbursed at point of sale,
or privately or collectively consumed goods disbursed through the public
purse (such as local government services). Later in this book we concentrate
upon the latter, as this appears to be the most obvious and important set of
conflicts and the one to which most empirical literature in political science
has been devoted.
This conflict over the nature of the goods and services changes both the
exit and the voice functions, though in different ways. Consider the example
of political conflict over local goods and services. First, as the nature of the
goods and services changes within a local authority jurisdiction the incen-
tive to exit from that area might grow for one group of people, yet provide
an incentive to enter for another set of people. Secondly, different groups
might have an incentive to ‘voice’ but to voice in opposite directions. Those
who are becoming more dissatisfied have an incentive to work politically
to change the colour of their local authority, whilst those who applaud the
political direction have an incentive to voice their satisfaction and protect
the changes taking place. The decision tree of Figure 1.2 needs to be exam-
ined from the Y line from time t1 since it may be rational to collectively
voice (at least) even if one is satisfied with goods and services provided.
This is a major modification to Hirschman’s original conception and we
discuss this further below as it is an important theoretical and empirical
complication of what appears at first to be a simple model.3
Loyalty as an ‘equation filler’

Loyalty does seem to be a real phenomenon that perhaps was not insig-
nificant in the failure of the new brand of Coca-Cola. Nevertheless, early
commentators on Hirschman criticized the idea, suggesting it had no role
to perform (Barry 1974; Laver 1976). Brian Barry suggested that it simply
acted as an ‘equation filler’ to be waved around to try to explain why people
3
In Hirschman’s (1970, chapter 6) discussion of voice where there is conflict over ‘quality’
he becomes embroiled in the reaction of different parties to the conflict, taking issue with
the ‘median voter theorem’ (Black 1958; Downs 1957) but we will not take this up here.
are irrationally voicing rather than exiting in terms of more obvious cost
benefit calculations. In other words he was suggesting that ‘loyalty’ is not
an empirically measurable phenomenon but an unobservable one that is
theorized to ensure that otherwise empirically puzzling results do not
serve to falsify the framework.
As we saw, Hirschman introduces the concept of loyalty to try to under-
stand why exit is virtually ruled out in certain contexts, such as family,
tribe, church and state. He suggests that loyalty clearly makes exit less
likely, but queries whether it also gives more scope to voice. He suggests
that the principal determinants of individuals’ readiness to resort to voice
rather than exit are (1) willingness to trade the certainty of exit for the
uncertainties of improvement following voice, and (2) the estimate of their
abilities to influence the organization. Despite his interest in general social
categories he introduces loyalty in terms of ‘brand loyalty’ whose connec-
tion to identity issues such as tribe, country or family is not immediately
apparent. He states ‘a member with a considerable attachment to a prod-
uct or organization will often search for ways to make himself influential,
especially when the organization moves in what he believes is the wrong
direction’ (Hirschman 1970, pp. 77–8). This is true but there seem to be two
completely different concepts of loyalty caught in this sentence. Loyalty to
a ‘product’ is surely very different from loyalty to an ‘organization’ such
as the firm one works for, or Hirschman’s other examples such as fam-
ily or tribe, which are more often considered in terms of a ‘group’ with
which one identifies. The first is ‘brand loyalty’. Individuals who display
brand loyalty are displaying a psychological resistance to change or a con-
servative attachment to their accustomed product. Either that or differ-
entiation between rival products (such as cars) is so great that individuals
may feel they have nowhere to exit and there is no need to bring loyalty
into the equation (Laver 1976, pp. 477–81). However, group loyalty is some-
thing quite different. Group loyalty depends upon one’s identification with
the group itself. A Londoner may feel loyalty to his country because he
is ‘British’, and that is one of the ways in which he identifies himself. He
may retain that loyalty to his country no matter how much he disagrees
with the policies of the government representing the country. To exit from
such a group just because he disagrees with the policies of the government
is an exceedingly costly and painful process. Similarly a person may feel
great loyalty to the firm for which she works, despite considering its cur-
rent senior management to be fools and buffoons ruining the organization.
In both cases the object of loyalty is different from that of the people held
responsible for its decline, though loyalty to those people may still be dis-
played in contexts where they represent the object of loyalty. For example a
middle manager may criticize his boss to fellow workers but defend her at
a conference when talking to rival firms. Or a US citizen might be highly

critical of his president when at home in New York, but defend that same
president when on vacation in Paris.
‘Loyalty’ as used by Hirschman seems to equate more with ‘brand loy-
alty’, since it operates simply as a tax upon competing products or as an
‘exit tax’, as for example when Hirschman (1970, p. 136) suggests we can
subtract the ‘cost of disloyalty’ from the value of the competing product
(Barry 1974; Laver 1976, p. 478). But loyalty in the broader sense surely
means much more. Barry (1974, p. 98) suggests ‘loyalty does not normally
mean a mere reluctance to leave a group but rather a positive commit-
ment to further its welfare by working for it, fighting for it and – where
one thinks it has gone astray – seeking to change it. Thus, voice (as well as
other forms of activity) is already built into the concept of loyalty.’4 Barry
goes on to criticize Hirschman’s attempts to analyse the effects of loyalty in
terms of one’s efficacy in using voice. We will leave that aside here because
we feel that it is possible to bring in an account of loyalty which allows us
to capture the type of loyalty Barry refers to without completely collapsing
voice into the concept.
The move partly depends upon recognizing the distinction between
the object of loyalty and the precise product which one receives from that
object of loyalty. The degree of loyalty that one has depends upon, first,
one’s identification with the object of loyalty, and secondly, the amount
one has invested in that object. People often identify with something to the
degree that it is tied to their personal history. We identify with objects to
the extent that they form part of us. Thus a woman sees herself as female,
and through this may recognize a shared interest with other women to
the extent of shared gender. When asked to describe ourselves we often
make reference to our job since that takes up a large part of our time and
is important to our self-reference. Both gender and work are important
to our personal history. That, of course, is why we identify ourselves with
our nation or our nation-state, because it forms an important part of our
personal history, especially if we have been taught about the history of our
nation, our subjection to other nations, our heroes and heroines. This can
also be true of a local community or a geographical region, particularly
if we were born and grew up happily there.5 Investment in one’s nation:
fighting for it; or one’s community: buying a house, working to improve
4
However, as Anthony Birch (1975) suggests, ‘loyalty’ in this sense might also lead one
eventually to exit silently when one chooses to exit. We discuss ‘Birch’ loyalty in various
contexts later in the book.
5
We can note that this sense of identity is involved both in Bentley’s old anti-Marxist
‘group theory’ (Bentley 1908) and in Sen’s (2006) account of identity, which is intimately
connected to his notion of commitment and rationality (Sen 2002; 2005).
the local schools, developing a large number of friends and so on; or one’s
firm: working to improve sales, being proud of the product and the quality
one has instilled in it; all these may increase one’s loyalty to that object –
the nation, the local community, the company. Now at least some of these
activities constitute, in the broadest sense, voice. But the fact that past
voice, together with other aspects of one’s personal history, may increase
one’s present loyalty does not mean that we cannot bring loyalty into an
equation predicting future voice, or predicting relative exit–voice tenden-
cies in individuals who share satisfaction levels but differ in amounts of
loyalty. By tying loyalty to investment in the object of loyalty we suggest
an empirically verifiable way in which Hirschman’s concept of loyalty may
be seen to affect relative exit–voice tendencies. Social investment here is
closely related to recent writing on social capital (Coleman 1990; Putnam
1993; 1995; 2000). Without developing the links here, we might say that the
social capital of some group or community is some function of the sum of
individual investment in that group or community. The rate of return of
this investment can be hypothesized to be determined by the efficiency of
the social structures as measured by the nature of the networks and other
factors in any group or community. We return to these issues later in the
book. The important point we wish to make at this juncture is that loyalty
is a complex phenomenon that needs a great deal of unpacking for it to
prove useful in empirical contexts with Hirschman’s framework. Loyalty
is not simply the ‘equation filler’ suggested by Barry, but provides a valid
theoretical category that may be used in empirical contexts and modelled
in an equation.
Conclusions
This chapter has outlined the key ideas in Hirschman’s scheme to explain
how they are supposed to interact through a series of examples and in dif-
ferent contexts. Already some elements of our general argument have been
seen to work. We have sought to be precise about what Hirschman meant
about his key terms to avoid confusion. We have sought to show carefully
how the elements of his scheme relate together, and how his ideas relate
to many common situations human beings face in their everyday lives as
well as in problems social scientists seek to explain. We have sought to
defend Hirschman against some of his critics. We have also elaborated
some refinements to his scheme, in particular his concept of loyalty. In
this way, we have put down the foundations for further theoretical elab-
oration in the rest of the book, where we seek to move away from using
exit and voice as a metaphor and toward a valid theoretical framework to
analyse a variety of political choices and situations. We aim to take the best
C onc lusions 23
from Hirschman’s argument and stay true to his aims, but at the same time
toughen up the approach ready for a wider and more sustained use within
political science and other allied disciplines.
The next chapter takes forward theoretical argument by setting out the
nature of Hirschman’s framework, and some recent attempts to formally
model it. Then the chapter sets out the current policy context within which
to examine the main choices open to citizens, which then generates a more
modern conceptualization of three different kinds of exit. The chapter re-
examines voice distinguishing three types. By the time the chapter is com-
plete, we have elaborated our framework and set out the relationships we
expect to see.
Chapter 3 reviews an important and complementary approach to that
of our own: the exit, voice, loyalty and neglect (EVLN) framework, which
offers a model of psychological responses to different sources of dissat-
isfaction. We review this important approach, and set our framework
alongside it, which allows us to sharpen our argument as well as learn
from previous work of political scientists. This approach also allows us to
introduce the context for our own study, as the EVLN model was tested in
urban communities in the United States. The urban or local context offers
many choices for citizens, in particular generating different forms of exit
and clear routes to exercise voice when consuming public services. At the
end of the book, we set out how our framework applies to a wider range of
contexts where citizens make choices and can choose to voice. In this way,
the urban context serves as a laboratory for our approach, which can then
be used more generally in cases where citizens and decision-makers face a
choice between voice and exit.
Chapter 4 is the first of two empirical chapters. It sets out the back-
ground to the study, first giving details about its design. The study is based
on a five-wave panel dataset of a representative sample of 4,026 citizens
specially commissioned by the authors. In our survey we asked respond-
ents about satisfaction with basic public services in the urban context, with
the National Health Service (NHS) and with schools. We asked about their
past individual and collective voice activities, their satisfaction levels with
basic services, and past exit and intention to exit activities. These dedicated
panel survey data allow us to examine the interactive effects of past voice
on satisfaction, and upon exit; and the interactive effects on the intentions
to exit upon voice and satisfaction and so to truly test not only Hirschman’s
original hypotheses about exit and voice, but also the more complex three
exit, three voice and loyalty relationships we identify in Chapter 2. These
data enable the first thorough test of Hirschman’s hypotheses derived from
his exit, voice and loyalty framework as developed by us since his book was
published over forty years ago.
Chapter 4 reports some basic descriptive statistics and examines a few

of the foundational relationships we identify. Chapter 5 then tests our
three exit, three voice and loyalty relationships more extensively, seeking
to explore the implications of our typology, through tabular and regres-
sion analysis using the panel data. The concluding chapter reviews what we
find in terms of these relationships, interprets the findings in the growing
literature examining Hirschman and suggests future lines of research in a
dynamic context.
Running through the book is the intersection of citizens and policy-
makers. Citizens make the choices and react to the signals, but policy-
makers help create the signals and react to the choices citizens make.
Various parts of the book pick up on this. First, we review the extent to
which governments have changed how far citizens can exercise choices
through reforms of public services. We briefly examine the extent to which
bureaucrats can react to signals about services in Chapter 2. In the con-
cluding chapter we again review these arguments and suggest what, if
anything, policy-makers should do in the light of our findings. The basic
message is that the choices citizens make over exit and voice form part of a
wider system of institutions and policies that are affected by exit and voice,
but also structure them. These are wider issues about how best to design
institutions to make the best of the positive aspects of exit and choice and
to minimize the loss of voice.
In our short Appendix A we very briefly consider the literature in other
subjects and disciplines and make some suggestions as to how our frame-
work can be applied in these other fields. Table A.1 in the appendix sum-
marizes the empirical results from other fields which might usefully be
compared to our own findings.
2
Exits, voices and the object of loyalty
Introduction
It is probably fair to conclude that the use and application of Hirschman
has not been as great as might have been expected from the flurry of inter-
est in his model in the 1970s. We know that Hirschman has been cited
extensively, but EVL has not been used much as a framework to guide
research questions. In fact, this judgement is not entirely fair as there has
been a steady stream of studies using Hirschman’s model, in particular
testing whether there is in fact a negative trade-off between exit and voice,
which we briefly review below and summarize in Appendix A of this book.
However, in spite of these advances in knowledge, there is still the lurk-
ing problem that Hirschman did not set out his model with enough ana-
lytical precision to generate testable hypotheses. Hirschman seems to be
applying a set of labels in a loose way, which is inspiring to researchers
working in a wider range of contexts, but also encourages researchers to
indulge in backward induction whereby any action counts as a relationship
between voice and exit and where any outcome may be interpreted within
its terms so as not to contradict the model, but in fact to embody it. One
example of this kind of approach is the phenomenon of ‘noisy exit’. This
is the idea that people will voice before they exit. This would appear to
flatly contradict Hirschman’s prediction that there is a negative trade-off
between voice and exit. In fact, a careful reading of Hirschman shows how
this is compatible: people will negatively trade exit and voice, but once they
have worked out there is nothing to lose they may voice their concerns just
before leaving, just like the discontented employee who keeps quiet until
she or he has found a new job, and then says what they think at the final
25
26 E x i ts , voice s a n d t h e obj e c t of l oya lt y
farewell celebration. And there is an interesting line of research that takes

Hirschman’s insight, following on from the noisy exit of people from the
former German Democratic Republic (Pfaff and Kim 2003, Pfaff 2006).
As this example, and our argument in the introduction, makes clear,
some of the problems of too loose a model can be overcome through care-
ful definitions of the concepts and a meticulous reading of Hirschman
himself. Nevertheless, there are some conceptual problems inherent in
Hirschman’s work and so in this chapter we try to go beyond Hirschman.
We seek to clarify and extend his concepts but remain within his con-
ceptual framework. We seek to build up the framework in part to make
it more sensitive to the variety of contexts that citizens find themselves
in, but most of all to present it in such a way as to generate some testable
implications.
It is with this aim that the chapter investigates what kind of framework
Hirschman advocated and how it contrasts with other attempts at theory
building in the social sciences. Then it considers a line of formal mod-
els, which have tried to model some of the relationships Hirschman notes
between the three elements he identifies. After suggesting that these mod-
els do not get us far enough, the chapter reviews some recent changes in
the context of citizen relationships to public authorities as a preview to our
reformulation of Hirschman’s framework. We then set out our framework
and discuss its implications for testing Hirschman.
Hirschman’s framework
Hirschman’s EVL approach is what we term a framework. It is not a for-
mal model with precise predictions. A social-scientific model is a set of
statements related formally or analytically to generate testable hypotheses
or predictions. They are deductive and have a set of precise assumptions
which together logically generate the predictions. Models are simplified
representations of something else in the world and are used to help us
understand the world. Models are designed to abstract (some) import-
ant features so that we may examine some of their causal effects in theory
and try to examine them more closely in reality. We begin to model when
we try to put together data (by which we mean any reasonably systematic
information about the world) into a set of propositions by which we relate
the data together. We start by collecting information about the world and
conjecturing about how the data fit together in terms of identity and causal
relationships. In other words we start to make descriptive and causal
inferences. The former is about using observations of the world to draw
conclusions about non-observed features of the world. The latter is about
explaining the causes of features of the world. We begin to model when
H i r s ch m a n ’s f r a m ewor k 27
we fit these together more precisely. Much of social science does not use
formal models in this way. Rather arguments are produced which follow
this general form but are not so precisely analytical. We draw a distinction
between models, which produce clear testable hypotheses, and ‘frame-
works’, which do not. We can draw hypotheses out of frameworks but their
implications are not always so clear. Sometimes a framework is called a
‘non-formal model’. We see Hirschman’s exit, voice and loyalty account as
a framework in this sense. It is more analytical than some frameworks, but
does not have the precise characteristics of a formal model. Nevertheless,
from his verbal framework he draws out some implications about the rela-
tionships between his concepts of exit, voice and loyalty. We can consider
the nature of the mechanisms involved in his hypothesized relationships
between his concepts and his assumptions about human motivation.
Hirschman assumes that individuals prefer better-quality products and
are prepared to act in order to bring these about. Their potential actions
within his framework are to exit to a better-quality product where one
is available or to complain about the product where there are no better-
quality alternatives. His loyalty motivation is that people want to maintain
their relationship with a product (we use this strange ‘managerial’ phrase
to cover many sins, see below) and will not switch to potentially better
alternatives at least until they can see whether their favoured product
improves. We suggest below that this surely must mean that there is some-
thing else about the product to which they maintain loyalty, other than
the satisfaction they gain from consuming it. If not, then loyalty would be
entirely mysterious.
As was discussed in Chapter 1, one aspect of Hirschman’s motivational
account which might prove problematic is his wilful refusal to acknow-
ledge the collective action problem. Where voice is used collectively indi-
viduals might not be so motivated to use their voice hoping others will
do so on their behalf. We discuss this further below. Another aspect of
Hirschman’s account is the response of the producers to the voice and exit
behaviour of their consumers. In a competitive context we can expect loss
of customers to motivate managers to improve quality otherwise their firm
will have lower profits and might eventually go out of business. In that
context we might expect voice only to lead to a response from a firm if it
contains the threat of exit. For a monopolist whose customer base is truly
captured for they literally have nowhere to exit, voice activity might have
no effect. Indeed this is the expectation engendered by faceless bureaucrats
under dictatorships. Here bureaucrats have no incentives to help clients,
but rather merely to process complaints. Such systems generate corruption
as the only way people can motivate bureaucrats is to pay them to help
them out. A market is created where one does not already exist.
Managers in some public services might well not fear exit but positively
welcome it as it might reduce their burdens. Oversubscribed doctors’ sur-
geries or welfare offices might welcome some clients exiting to other doc-
tors or welfare agencies. And without the threat of exit being a threat (as
opposed to an offer!) why should they respond to voice?
There are several motivations through which public servants might
respond to voice even if exit cannot be a threat. The first is that we should
not assume that all public servants are entirely self-interested, or in Julian
Le Grand’s (2003) terms, ‘knaves’. Many public servants have entered
their chosen profession as doctors or as specialists in welfare departments
because they want to help the public. They wish to provide a public ser-
vice and take pride in their profession and professionalism and thus if
poor standards or treatment are brought to their attention through voice
activity they want to act on it. In Le Grand’s terms these public servants
are ‘knights’. As Le Grand argues, we should not be too blinded by the
(often useful) ‘self-interest’ assumption of much of modern social science,
to ignore the real effects of knightly behaviour, especially in the public sec-
tor. Thus voice, especially if articulated well, can provide the appropriate
response in the producers.
Second, there are top-down control mechanisms that try to ensure that
the public’s dissatisfaction is responded to by the public service. Especially
in democracies, our main concern in this book, politicians have a strong
interest in the public being satisfied with the goods and services provided
by the state and hence an interest in the appropriate response from state
bureaucracies. In the UK, where our empirical evidence comes from, gov-
ernments over the past twenty years or more have provided legislative and
executive incentives for local governments, the education and health ser-
vices amongst others to respond properly to complaints. During the time
of our survey the Audit Commission, a public corporation independent of
the government designed to improve efficiency and effectiveness in local
government, housing and the NHS, collected indicators of service qual-
ity including response times to complaints. Local councils, education and
health authorities also conduct surveys of their clients to see if they are sat-
isfied with services, and satisfied with the procedures in place to respond
to complaints. As we have just suggested, we might consider that a type of
exit mechanism underlies this second incentive to respond to voice activ-
ity. Politicians are in a competitive market for votes, and we view voters’
shifting support from one politician or party to another in terms of the exit
metaphor. Here the competition to win votes underpins the incentives to
ensure non-elected public servants serve the public well. (We should note
here that below we use the idea of votes as a form of collective voice rather
than exit because we see it as a gauge of satisfaction with services, and not
For m a l i z i ng H i r s ch m a n 29
as a response to the politicians themselves. We see it as a signal to politi-

cians about services, rather than a signal to politicians about politicians.
In other contexts seeing votes as part of the exit mechanism is equally
valid. What matters is the role it performs within the mechanism, model
or framework.)
We should not think that exit will always be viewed positively by public
servants. Whilst there might be examples where doctors, welfare workers,
schools, might welcome a shrinking clientele base, they would not want to
lose too many clients. Under-subscribed schools, health centres, welfare
offices and so on will face closure when the auditing mechanisms turn
their attention to them. Similarly a dramatic fall in clients will be noticed
by managers who will fear jeopardy if the decline cannot be arrested. Of
course, if the bureaucracy is truly monolithic then there might be no pos-
sibility of exit: however, even then the reasons given above about why voice
might be effective will still operate.
Thus in the public sector we can expect the exit–voice mechanism to
work from the supply side, though not necessarily in the same manner as
it does in the private sector. We can see, however, that there are complica-
tions from Hirschman’s original story, which involve different types of both
exit and voice; and the mechanisms which bind these together to produce
a response from public sector producers. We develop these in this chap-
ter. First, however, we examine some attempts to formalize Hirschman’s
account, which bring out the importance of the threat of exit to provide
motivations for producers to respond to voice activity.
Formalizing Hirschman
Few people have attempted to model mathematically the relationships
between exit, voice and loyalty to draw out deductive hypotheses more
formally. There are several reasons why this might be so. The basic rela-
tionships between exit and voice as specified by Hirschman are not very
interesting mathematically, and once one moves beyond trivial relation-
ships to include complications such as collective goods, multiple solutions
emerge. The conditions of these solutions might be interesting, but then
only testable with strong data. Some recent studies have attempted to spe-
cify Hirschman’s framework more formally in game-theoretic terms how-
ever, and these models demonstrate some of the aspects of the relationship
between voice and exit in the public sector we have just discussed.
Those that have attempted to mathematically model EVL relationships
tend to concentrate upon the exit and voice relationship to the virtual
exclusion of loyalty. Gehlbach (2006) presents a simple general complete
information game-theoretic model analysing the relationship between exit
and voice. In the model there is a game between the leadership of some
organization (a club or a nation, for example). It assumes that exit is costly
in the sense that the joint payoff to the leadership and membership is
greater without exit. Payoffs are between 0 and 1 with payoff x to member-
ship and (1-x) to the leadership. The leaders have the freedom to set policy
to the extent members can exit. He then assumes that the members can
organize collectively to bargain with leaders whose result is a Nash equi-
librium such that the gains of this trade are split equally. Without collect-
ive organization, leaders offer a take-it-or-leave-it offer which citizens can
either accept (in silence) or exit. There are costs of organizing.
The results are interesting for they show that payoffs for members once
organized must be higher to ensure they do not exit; without organizing
leaders can simply offer the utility exactly in line with membership exit.
This then specifies the Hirschman voice argument that members gain
more when they engage in voice activity. Also in Gehlbach’s model exit
never occurs because the leadership always offers enough to ensure exit is
avoided, and thus ‘silence’ emerges endogenously much as it seems to in a
great deal of organizational life. (Most people neither exit nor voice most
of the time.) Members choose to organize if their payoff from sharing the
surplus is greater than the cost of organizing plus cost of exit (the payoff
of a take-it-or-leave-it offer). So here exit will reduce voice when the exit
payoff is large (that is, the costs of exit are low). However, it is still the case
that voice is most powerful when exit is easiest. As Gehlbach (2006, p. 402)
puts it ‘voice and exit are complements once voice has been established,
but are substitutes when seen from the perspective of the decision to exer-
cise voice to begin with’. This nicely demonstrates the importance of the
possibility of exit for voice to be effective, but that if voice is effective then
exit is unnecessary. For example, if the democratic exit mechanism of the
threat to stop supporting one candidate or party if public services did not
improve was perfect, then politicians and their agents within the public
service would respond to all voice activity, and no one would ever switch
their vote. But of course, it is not. For one thing there are many services
and it is not clear what the switching of votes signals to politicians about
any particular service. Second, voters do not only vote on the quality of
services but over other issues below, thus again the signal is noisy.
On loyalty, Gehlbach (2006) shows that it leaves members worse off
when it is defined as a tax on exit – it makes members more reluctant to
leave – but increases welfare when it is a voice subsidy – that is, when loy-
alty increases the propensity to voice. Both aspects of loyalty might be
present – the tendency to suffer in silence, and the desire to improve the
organization even though other alternatives are available. Hence Gehlbach
makes sense of the two sides of loyalty we discuss in Chapter 3.
For m a l i z i ng H i r s ch m a n 31
Table 2.1 Sub-game perfect Nash equilibriums from the EVL game
# Equilibrium Required conditions Outcome

E1 (Loyalty, Loyalty; Ignore) E≤0 Citizen remains loyal,
state keeps benefit
E2 (Exit, Exit; Ignore) L≤ 1, E > 0 Citizen exits, state
keeps benefit
E3 (Voice, Exit; Respond) L > 1, 0 < E ≤ 1 ¡ C Citizen uses voice,
state responds and
returns benefit
E4 (Exit, Exit; Respond) L > 1, E > 0, E > 1 ¡ C Citizen exits, state
keeps benefit
Notes: Equilibriums are written (citizen’s first action, citizen’s second action;
state’s action). All equilibriums assume C, L > 0. (Other assumptions are
required: see original article.)
Source: Clark et al. 2007, p. 7
Clark et al. (2007) model exit, voice and loyalty with each compo-
nent being a separate behavioural category; loyalty in fact is simply silent
non-exit in their model, again illustrating the theoretical difficulty found
when dealing with loyalty. They run a three-stage game with two actions
by citizens followed by the state’s response. The equilibriums are given in
Table 2.1, taken directly from their article.
The state’s response is defined as dependent where the state values citi-
zen loyalty more than the gains it extracts from them; or autonomous if
the state values their loyalty equally or less than the gains it extracts. The
state’s response defines whether citizens have a credible exit strategy. If
E ≤ 0 then the citizen has no credible exit threat, in which case the state
will extract all the benefits. If E > 1 – C then the citizen’s exit payoff is so
high she will never use voice even if it were to be effective; where E ≤ 1 – C
then the citizen will use voice. So states only respond to citizens in this
model where citizens have a credible threat of exit, and the state is depend-
ent on them, that is, the state does not want them to exit.
Contrary to Hirschman, where product decline is arbitrary or acciden-
tal, in the Clark et al. model the state sets policies and so decides when
its services will decline or improve. In situations where the state does not
ignore citizens, why would it ever introduce policies that cause voice? To
answer this Clark et al. (2007) run another game where the state moves
first and must decide whether to prey or not on the citizens (Table 2.2).
Table 2.2 Sub-game perfect Nash equilibriums from the extended EVL
game
Required
# Equilibrium conditions Outcome
E5 (Predate, Ignore; Loyalty, E≤0 State predates and
Loyalty, Loyalty) keeps benefit, citizen
demonstrates loyalty
E6 (Predate, Ignore; Exit, L ≤ 1, E > 0 State predates and
Exit, Loyalty) keeps benefit, citizen
exits
E7 (Don’t Predate, Respond; L > 1, 0 < E ≤ 1 State does not predate,
Voice, Exit, Loyalty) –C citizen demonstrates
loyalty and keeps
benefit
E8 (Don’t Predate, Respond; L>1, E>0, E>1≤ C State does not predate,
Exit, Exit, Loyalty) citizen demonstrates
loyalty and keeps
benefit
Notes: Equilibriums are: (state’s first action, state’s second action; citizen’s first
action, citizen’s second action, citizen’s action after state’s second action). All
equilibriums assume C; CS; L > 0 and 1 > E. (Other assumptions are made:
consult the original article.)
Source: Clark et al. 2007, p. 13
Given predation, citizens act twice and then the state acts and the citizens
act again.
Clarke et al. find the autonomous state always predates; the dependent
state only where citizens have no credible exit strategy. It follows that citi-
zens with credible exit options can wield considerable influence, the rest
cannot. This supports other findings that the rich do not need to act to get
what they want (Dowding 1991) and Marxist accounts of the state where
capitalists have denied credible exit strategies to workers (Block 1977;
Przeworski and Wallerstein 1988). In other words, to the extent that citi-
zens’ activity does not affect the welfare of the state the state can ignore
them; where citizens can affect state welfare then the state will respond
only where exit is a credible alternative.
This model is not a model of democracy as such, but we can see how
politicians might be dependent or autonomous depending on the degree
to which they are threatened by electoral processes. If citizens can threaten
to exit from supporting politicians if services are not improved then there
Choice a n d voice 33
are incentives to get them to improve. However, to the extent that signals
are noisy, again the politician has little incentive to take action. However,
voice is almost entirely absent from this model.
To explain why citizens ever voice or why states do not respond to voice
requires moving to incomplete information games, and Clark et al. (2007)
argue that incomplete information has an asymmetric effect on the game
as it can help citizens but it never helps the state. A pooling equilibrium
exists where the state is unsure whether or not citizens might exit, which
gives all citizens the incentive to voice.
These studies demonstrate that there are clear implications for the prob-
ability of voice activities given the ease of exit. Under certain conditions
an organization (the state) will respond to citizens only if there are credible
exit opportunities which affect the interests of the organization. Applied to
public service providers the extent to which losing clients will not harm the
provider gives them no incentive to respond to the threat of exit. However,
where losing clients becomes a threat the provider will respond to voice
because of the threat of exit. Making exit easier will reduce voice under
certain conditions. If there is no other possibility than voice, then even if
the costs of voice are high it can be worth the while of citizens to utilize
voice. In the next section we examine recent changes in the way citizens
interact with the government, which have increased exit opportunities.
Choice and voice

Criticisms of monolithic state bureaucracies from the 1960s onwards often
utilized the argument that bureaucrats had little to gain from responding
to criticisms and suggestions from citizens. It has been argued that the
lack of competitive pressures underpinned by exit possibilities, together
with the fact that elected politicians and the public service that supports
them respond more to lobbies and pressures through organized interests
than to electoral processes, leads to an expanding but unresponsive state
bureaucracy (Olson 1982; Mueller 1993, chapter 21; Niskanen 1994). The
response of many writers from the 1970s was to argue that competition
within the public service both through privatization and contracting out,
would create efficiencies (for example Savas 1987; Miranda and Lerner
1995) as would internal competition (for example Le Grand and Bartlett
1993), though many cautioned against these ideas (for example Miller and
Moe 1983). Governments throughout the world have in the past thirty
years introduced these competitive principles but in the UK they were seen
initially with the Thatcher and Major governments of the 1980s and early
1990s increasing competition, and then with the Blair government of the
late 1990s and 2000s concentrating upon better service delivery and choice
within public service delivery.
In many regards these initiatives have, according to objective measures

of service quality, improved services in local government, in education
and in health care (Burgess et al. 2005; Bevan and Hood 2006a; Le Grand
2007, chapters 3 and 4), though how far this improvement is derived from
top-down managerial initiative and how much from choice and competi-
tion is moot (Bevan and Hood 2006b; Le Grand 2007). Some of the argu-
ments for increasing choice have considered Hirschman’s EVL claims (Le
Grand 2007), though none has done so in great detail. Hirschman argues
that once publicly provided services start to decline and there are ready
rival sources of supply that customers might exit to, the decline in quality
of public services will continue. For Hirschman competition and choice
will not (necessarily) lead to quality improvements. His reason is that exit
(or choice) can reduce voice. For example, if the quality of a state-funded
school declines and enough middle-class parents are willing and able to
send their children to private schools these parents will no longer press the
school board, governors, head teacher and other staff members to arrest
the decline in quality of the state school. Moreover, they will have less of
an incentive to press the local or national government to improve state
education, and will be less prepared to pay taxes to improve the quality of
state education. With their voice taken away there is less of an incentive for
those running the schools, from teachers to politicians, to arrest quality
decline. Those parents who are unwilling to send their children to private
schools, or unable to afford to, will be left behind. They may be willing to
voice, but to the extent that education and wealth are correlated the poorer
cohort of parents left behind might be less able to express their discontent
and be less able to mobilize themselves in order to pressurize government
to improve state education. Furthermore with the exit of many of their
number their political clout declines. In fact, given the correlation between
electoral turnout and the wealth and education of eligible voters, that pol-
itical power declines more than proportionately with the numbers leaving
the state education system. Thus the movements across the world to make
exit from state-funded services easier for people through tax rebates on
private school fees and on private health insurance, and in some cases dir-
ect subsidies to the private sector, might make public services worse even
as competition gets fiercer. Of course the champions of these competitive
and choice-driven alternatives disagree with Hirschman and it is worth
considering why.
We have to remember in this context that many, though certainly not
all, publicly funded services are funded because of market failure. The rea-
son we have public provision for some goods is either (a) that there is mar-
ket failure and hence allocative inefficiency due to the non-excludable and
jointly supplied nature of public goods; or (b) because of the inequitable
Choice a n d voice 35
Table 2.3 Types of goods provision
Production Disbursement Consumption

private private private
private collective private
private private collective
private collective collective
collective private private
collective collective private
collective private collective
collective collective collective
distribution associated with the market provision of some goods such as

education or health care. With the former the market system does not
produce signals that reflect true demand and so it might be thought that
democratic processes of preference aggregation provide better signals.
With the second, the equality of rights for certain fundamental aspects of
human welfare underpins social rather than market demand. The fact that
markets fail to allocate the social goods that we as a society wish to have
promoted prompts government to get into the public goods supply business
in the first place. However, it is simplistic thereby to assume that there is
no place for competition or market production for public goods. The rela-
tionship between production and consumption can be viewed as a three-
part process of production, disbursement and consumption (Dowding and
Dunleavy 1996; see also Ostrom et al. 1961). Each of these three can be
private or collective, as illustrated in Table 2.3. A good might be produced
by a private company or by public officials or workers, or by some mix of
public and private provision. Disbursement might be private paid for at
point of delivery, or public (or indeed some mix). And consumption can
be private in the sense that the good is rival and excludable; or collective
being non-rival or excludable. Again most goods are some mix of public
and private.
It is the threefold category that enables the complex service delivery
processes that exist in the modern state. At one end of the spectrum we
have private goods produced in the private sector and supplied directly
(or through private agents such as shops) to the consumer and paid for
at the point of delivery. At the other end we have public sector producers
and providers giving goods directly to the public paid for through taxation
with no money transferred at the point of delivery. In between we have a
host of other possibilities. Public subsidy of private transactions; private

companies selling goods to public organizations that are passed on to the
public paid for publicly (such as contract nurses in state-run hospitals); pri-
vate companies providing goods directly to the public but paid for through
public disbursement (such as refuse-collection contractors) and so on. The
important aspect for our argument here is disbursement.
Where disbursement is partially or fully public then the relation-
ship between the producer and consumer is either direct or indirect. In
the first the producer is either the public organization paid for through
public taxation, or there is a subsidy paid to the consumer who then dir-
ectly contracts with a private provider. In the second, the public organ-
ization contracts with the private producer on behalf of consumers who
then must make their representations to the public body. We can see then
that competition and choice can thus enter through at least two stages.
Competition between producers can be generated as public bodies – such
as local authorities or health care trusts – open up production of some of
their services to competitive tendering. Such competition does not in itself
entail any form of consumer choice. A process that leads a local author-
ity to hire one particular company to collect household waste rather than
another does not imply consumers have a choice of producers. The public
provider chooses the producer. The public provider also decides on the
nature and quality of the good or service. Such provision of the good might
be decided by specific rules, perhaps governed by cost and equity. So the
local authority decides that each household will have its refuse collected
once a week, along with secondary rules that might govern how much, say,
garden waste can be left out for collection. The provider also decides the
amount to be spent on refuse collection. Consumer input into the decision
process occurs through the traditional democratic processes – one aspect
of what Hirschman calls voice.
Competition might also be promoted through the public providers
themselves. Individuals can be given a choice of which public provider they
go to for a given service. In health care, for example, a patient might have
a choice of which surgeon she wants to carry out her operation at a given
hospital, or which hospital she wishes to attend for her outpatient care. In
the former case, the surgeon ‘produces’ the operation. In the case of the
choice between hospitals consumers may not in fact be selecting between
different producers. Two different hospitals may use the same private con-
tractor for their nursing staff, cleaners and so on. For local authorities two
different authorities might use the same company to collect household
waste. In other words, choice of public provider does not necessarily imply
choice of producer; and choice of producer does not necessarily imply
choice of public provider.
T h r e e e x i ts 37
The idea in both of these circumstances is that competition provides the

motor for efficiency. The firms’ costs are driven down as they compete for
business either with the public authorities or directly to the public, and
the buyers, again either the public authorities or the public themselves,
will purchase the highest-quality products. Market competition will drive
efficiency and quality. The idea, of course, is that the private sector is more
efficient than the public sector because of the discipline of competition.
Firms that do not produce the products people want at the price they want
are driven out of business by those that do. Competition provides the incen-
tives to cut costs to drive down prices, and to produce the products people
want: so both productive and allocative efficiency are driven by market
competition. Consumer choice is the motor that drives competition which,
along with free entry into the industry, is supposed to ensure that multiple
firms compete. In recent years the emphasis has grown on the ‘choice’ side
of this process. Shifting the emphasis away from the competitive aspects
of the process to the choice features softens the driving edge and concen-
trates attention on to the consumer side. It emphasizes that people can be
in control of their public services rather than being serviced as clients by
a rule-driven bureaucracy. Choice also fits nicely with the idea of freedom
as control, which has greater modern resonance than the terms equality
or welfare.
So these market- or competition-driven incentives vie with the
Hirschman considerations over the danger to public provision through
increasing exit opportunities. Choice in public services might dampen
down voice and exit though such choice might leave the nuanced voice
possibilities bereft of their most vocal and articulate elements. We notice
here that there are two sets of choices that individuals might have: exit
from public service provision altogether into the private sector, and exit
from one public sector provider to another. These would not have the same
consequences on Hirschman’s account. Choice between public sector
authorities does not rob the public sector of voice in the same way as exit
to the private sector might. In fact we see that Hirschman’s simple exit,
voice and loyalty framework hides a variety of exits and voices that may
not bear the same relationships that he claims within the simple model. In
the rest of this chapter we take up these complications.
Three exits
There are four ways that a member of the public might exit from a public
service provider (see Table 2.4). The first, which like Hirschman, we will
hereafter largely ignore, is to exit from using the good altogether. We term
this ‘complete exit’. If one simply stopped going to the doctor or sending
Table 2.4 Types of exit in public provision
Type of exit Description Disbursement

Internal Exit (EI) Leaving one public provider for Public
another public provider
Private Exit (EP) Leaving public service provision for Public to private
private service provision
Geographical Exit Leaving the jurisdiction of one public Public to public
(EG) provider for another
Tiebout Exit (ET) Geographical exit where the reason Public to public
for moving is to move provider
Complete Exit (EC) Stopping the service altogether Public to public
(private to
nothing in
private sector)
one’s children to school (and not providing any educational replacement

at all) then one has exited entirely. Whilst this form of exit is important,
and to some extent the problem it poses is what motivated public services
in many countries – for example it was the British armed forces’ inabil-
ity to recruit healthy and educated soldiers to fight in the Boer War that
stimulated the government’s resolve to improve public health and educa-
tion programmes in late nineteenth-century Britain (Thane 1982) – we do
not consider it further here. It was not simply to provide fit soldiers for the
armed forces, but egalitarian reasons that motivated the setting up of the
British NHS in 1948 and many other welfare services around the world. In
that sense complete exit is a vital motivating force for many public services.
However, we are more interested in the effects of competition between sets
of service providers than the reasons for having public provision in the
first place.
Discounting exiting altogether from a given good or service, we iden-
tify three ways individuals might exit. Two we have mentioned already.
First, individuals might exit from one public service provider to another:
for example, from one state-run school to another, or from one public hos-
pital to another. We call this ‘internal exit’, as it is exit internal to the public
sector and is not expected to have the consequences for public services in
general that Hirschman addresses in his book. Internal exit has not always
been possible in British public service provision. For example, children
were once required to attend their local school and there were only cer-
tain exceptions to attending a school where one did not live in the relevant
T h r e e e x i ts 39
‘catchment area’. Similarly people were expected to register with their local
medical general practitioner (GP) and were not given the opportunity to
shop around to find a GP they preferred. However, greater citizen willing-
ness to exercise choice over such things as schools and GPs, and a series of
government reforms enabling greater choice in these areas, have made some
form of choice more viable than ever before. Parents can now apply for dif-
ferent schools (though criteria such as ‘catchment area’ still apply, affecting
the chance of getting your child in to your preferred school), and standard-
ized tests and school inspections have led to ‘league tables’ of school quality
providing more information to enable rational choice of schools. Similarly
patients have more choice over GP practices than was once the case.
Even so, choice over many services might be hard to achieve in prac-
tice because of the constraints over supply and lack of information citizens
have about choice. There is also a general belief that citizens expect public
services to be provided first over and above the preference for a diversity of
service provision (Curtice and Heath 2009).
Partly for this reason, and partly because we have little usable data on
such forms of internal exit we say less about this form of exit than others.1
Theoretically it is less interesting in terms of Hirschman’s hypotheses.
Hirschman was concerned specifically with the decline in public service
provision as people departed the public sector, taking their voice with
them. Shifting providers within the public sector will not cause that decline
in voice within the public sector. Of course, it might cause a decline of voice
within organizations within the public sector: specific schools might con-
tinue to decline or specific hospitals or doctor services (that is, general
practitioners). However, to the extent that choice and competition within
the public sector are justified, that is what is supposed to happen. Bad pro-
viders should go out of business just as in the private sector. As we argue in
the final chapter, however, government is rarely prepared to take its logic of
choice and competition to its inevitable conclusion in this regard.
The second form of exit is that the public can leave the public sector
for the private sector. We call this ‘private exit’. Consumers might exit
from public provision to private provision. Parents may take their children
out of state schools for private education. In the UK patients may remove
themselves from NHS provision to private health care. This is the form
of exit that drove Hirschman’s original conception of EVL and drives the
1
Our data are from the UK and whilst choice of public service providers has been intro-
duced during the time of our study notably in health care, and was available to the public
to some extent in education, the numbers of people who take up these options are small.
In part this is due to legal restrictions and lack of information. We discuss these issues in
our final chapter.
problem that his EVL argument was designed to solve. In modern welfare
states with services free at the point of delivery, this kind of exit is usually
regarded as the prerogative of the rich, who wish to have a more personal
service or are prepared to pay for higher quality of service, or in the case
of education, buy social exclusivity that might help their children gain fur-
ther access to elite networks and sources of employment. Most citizens
remain happy to consume the basic services of the welfare state and even
the rich will continue to use public health care because of its high qual-
ity and then supplement it with extra privately bought services. In recent
years, with rising demand for some services, higher incomes, greater sup-
port from employers and constrained public budgets, the private provision
of services has once again become more fashionable and has been increas-
ingly adopted by policy-makers (Brekke and Sørgard 2006).
The third form of exit is geographical relocation. We call this ‘geograph-
ical exit’. Hirschman considers this largely in terms of relocation across bor-
ders from one state to another both in his original book and in later work
(Hirschman 1975; 1995). It can also occur within states where citizens move
from one public sector jurisdiction to another. If that exit occurs in order to
access (or more correctly (see below), facilitate) a better public service, then
we term this ‘Tiebout exit’. The title comes from a famous article by Charles
Tiebout (1956) who argued that the public good problem for local public
goods could be solved by such geographical exit. It is so important in this
regard, that we will discuss this form of exit in more detail.
Tiebout exit
Tiebout exit is a subset of geographical exit whose effects are supposed
to generate a competitive solution to local collective goods problems.2
What are local collective goods problems? As we saw in Chapter 1, a pub-
lic or collective good is defined by the conditions of ‘jointness of supply’
(or ‘non-rivalness’) and non-excludability. Together these two conditions
entail that there is a problem for the public revelation of preferences for
such goods that does not exist for private goods. If a collective good is
supplied to some members of the public then it is supplied to every mem-
ber of that public and none can be feasibly excluded. This means that it is
difficult to collect payment at the point of delivery since people have an
incentive to claim that they do not value the good that is supplied. I might
claim that I gain no benefit from national defence and refuse to pay for it.
Or I might claim that even though I use the public park every day, I only
2
We use the terms ‘public goods’ and ‘collective goods’ interchangeably.
T i e bou t e x i t 41
value it to the tune of a few pennies annually, far too little for it to be eco-
nomically feasible to claim from me. Thus public valuation of collective
goods might be massively lower than their true value, where ‘true value’
means something like the total amount people would be prepared to pay
if these goods had the properties of private goods, that is if they were rival
in supply and excludable. Samuelson (1954) argued that it is the technical
features of such goods that lead to the justification of state intervention.
We forcibly collect taxes from the public to pay for the collective goods that
the state provides. Preference revelation must come through people telling
government through voice activity, such as talking to politicians, pressure
group activity and the ballot box.
Of course many goods that are provided by government do not have the
character of pure public goods, if indeed any of them do. Many welfare
services, including of course education and health, can have ready private
markets. We do not have to supply either publicly and doing so often has
more to do with ideas of equality and public welfare than with infeasibility
constraints strictly understood. There might, though, be negative exter-
nalities in not having state education and health services. For example,
employers might gain from a more educated and technically competent
workforce, and also from a healthier one with lower morbidity rates.
Similarly, a public service such as the provision of refuse collection could
be charged at the point of supply. However, there are severe negative exter-
nalities that could follow if some households do not bother to have their
refuse collected, affecting the welfare of their neighbours; and households
might choose to dump waste in public places rather than pay to have it col-
lected. Enforcing orders to ensure that people do not have to suffer from
the dirtiness of their neighbours or policing the illegal dumping of waste
might be more expensive than enforcing payment for regular household
refuse collection. In other words, public supply of household refuse collec-
tion paid for through taxation is efficient.
It is also the case that technical change and changing demand and sup-
ply conditions often change the nature of goods from public to private and
back again. For example, television signals once had the nature of pure
public goods as anyone with a receiver and TV set could watch a broadcast
service. However, encryption and cable digitization have made it possible
to charge at the point of supply.
The history of local government in the UK, for example, can be told in
these public good terms. Local government in the UK developed massively
in the nineteenth century through petitions to parliament for the right to
pass by-laws to stop negative externalities (such as of rendering carcasses
in the street) and the right to collect local taxation for local public goods
(Prest 1990). Some of the public goods came about through direct failure
in markets for items such as street lighting, water supply and sewerage,
though later technical and demand conditions (i.e. a richer public) enabled
these functions to be transferred to the private sector for profit in more
recent times.
Tiebout (1956) provides a potential answer to these preference revelation
problems for local collective goods that is supposed to mirror market mech-
anisms. Whereas the market has competition between rival firms, Tiebout
suggests that local governments or local suppliers of local collective goods
can compete for households that will move to the jurisdiction that provides
the services they want. He says ‘there is no way in which the consumer can
avoid revealing his preferences in a spatial economy. Spatial mobility pro-
vides the local-goods counterpart to the private market’s shopping trip.’ So
geographical mobility is the fundamental element (Tiebout, 1956, p. 420):
‘The act of moving or failing to move is crucial. Moving or failing to move
replaces the usual market test of willingness to buy a good and reveals the
consumer-voter’s demand for public goods. Thus each locality has a revenue
and expenditure pattern that reflects the desires of its residents.’ And in one
of the most famous paragraphs in urban political economy:
The consumer-voter may be viewed as picking that community which best
satisfies his preference pattern for public goods. At the central level the pref-
erences of the consumer-voter are given, and the government tries to adjust
to the pattern of those preferences, whereas at the local level various govern-
ments have their revenue and expenditure more or less fixed. Given these
revenue and expenditure patterns, the consumer-voter moves to that com-
munity whose local government best satisfies his set of preferences. (Tiebout
1956, p. 418)
The distinct property of a market mechanism from the viewpoint of effi-

ciency is preference revelation. Residential mobility is seen as expressing
preference revelation. By ‘voting with their feet’ households are supposed
to reveal their preferences for local public goods: just as firms in the
marketplace are supposed to respond to the preferences revealed by their
customers, so local authorities are supposed to respond to the preferences
revealed by mobile citizens. We return to the efficiency claims of Tiebout
exit in the final chapter of this book. All we note here is that Tiebout exit
provides one way through which citizens might respond to poor-quality
public services, and this constitutes an important aspect of Hirschman’s
EVL claims. With Samuelsonian failure in preference revelation for public
goods, voice activities provide the only means by which government can
learn what level and types of services the public wants. If Tiebout exit could
provide that revelation then such voice activities might no longer be neces-
sary. Or, to put the point in Hirschman’s manner, if bad local governments
T h r e e voice s 43
lost citizens to neighbouring jurisdictions and those who moved are the
more discerning, wealthier and better educated, then the voice of those left
behind to improve their services might be irreparably damaged, ensuring
that services will continue to decline as the local governors will have few
incentives to improve them.
There are different types of ways that such Tiebout exit can occur. First,
households may relocate themselves from one local government jurisdic-
tion to another; second they might physically move away from the catch-
ment area of one provider to that of another. Thus parents move to the
catchment area of a good school and away from that of one with a poor repu-
tation. Citizens may move from the locality of one health authority to that
of another. Such relocational decisions are known to take place. Physical
relocation from one school catchment area to another is well known and
has a large capitalization effect upon house prices (Jud and Watts 1981;
Teske et al. 1993; Bogart and Cromwell 1997; Ogawa and Dutton 1997;
Cheshire and Sheppard 1998). Relocation in order to take advantage of
the ‘postcode lottery’ in health provision has not been empirically demon-
strated though anecdote suggests that it happens occasionally.3 Household
relocation across jurisdiction boundaries also takes place (Aronson 1974;
Davies 1982; Percy and Hawkins 1992; Dowding et al. 1994; John et al.
1995; Percy et al. 1995; Dowding and John 1996; 1997; Dowding and
Mergoupis 2003).
Our theoretical justification for identifying these three forms of exit is
that internal exit might be thought not to have the deleterious consequences
for voice that Hirschman identified; whilst private exit and Tiebout exit
can be thought to potentially have those deleterious consequences, they
are obviously very different in character and provide very different sorts of
signals to providers of public goods and services.
Three voices
As we have seen in our review of the Hirschman literature, many different
forms of voice have been identified. Some of these conceptualizations and
operationalizations seem to bear little relationship to what we ordinarily
think of as voice; others bear a close connection. We want to mention
two here again. First O’Donnell (1986) makes a distinction between ‘ver-
tical’ and ‘horizontal’ voice. The former is voicing to the provider, such
as complaining to a firm or the government. The latter is discussing the
good or service with family, friends or neighbours, or perhaps through
3
Such as the man moving from Durham to Scotland to obtain cancer drugs, Independent,
15 January 2007.
Table 2.5 Types of voice in public provision
Type of voice Description

Individual voice Individual communication about service either
direct to provider or to some public body, e.g.
letter to hospital; conversation with elected
representative
Collective voice voting Voting in election
Collective voice participation Participating in some collective voice activity,
e.g. belonging to some lobby group, signing a
petition, going on a demonstration
the mass media. Hirschman himself noted that theoretically the distinc-
tion is interesting because the former might be thought to bear costs to
the voicer, whereas the latter might bring expressive benefits and thus not
be costly. This distinction between horizontal and vertical voice is inter-
esting and we will return to it later when we consider aspects of loyalty.
However, the distinction is not relevant to the three forms of voice we
identify. Second we want to mention Luchak (2003) who discusses direct
voice, which is a direct contact between an employee and his firm, and
representative voice, where the complaint is communicated indirectly to
the firm, perhaps through a trade union. This distinction is similar to the
individual and collective voice distinction we introduced in Chapter 1
and which we take further here.
In keeping with there being three types of exit we identify three forms of
voice activity. Again our theoretical justification for distinguishing these
three forms is that they might affect the nature of the exit–voice relation-
ship that Hirschman identifies in different ways. We have already men-
tioned the three forms of voice when we discussed failure in preference
revelation for public goods through market mechanisms. We said then:
‘preference revelation must come through people telling government
through voice activity such as talking to politicians, pressure group activ-
ity and the ballot box’. The three forms of voice are: direct comments or
complaints to public officials, which we call individual voice (IV); par-
ticipation such as pressure group activity, petitions, demonstrations and
so on, which we call collective voice participation (CVP) (see Table 2.5);
and through the ballot box, which we call collective voice vote (CVV). The
important theoretical distinction is between private voice and collective
voice; we distinguish the two forms of collective voice for two reasons, one
theoretical and one empirical, as we explain below.
T h r e e voice s 45
We introduced the individual and collective distinction between voice

activities as important conceptually in Chapter 1. We pointed out that
individual voice does not suffer from the collective action problem that
collective voice might. In that sense the individual and collective voice dis-
tinction mirrors the private and public good distinction.
We generally view individual voice as personally complaining about
a good or service that the consumer has received. In the context of our
empirical research we see individual voice exclusively in terms of com-
plaints. For example, someone might complain to a local housing officer
about some problem with their house; or they might complain to a council
official about a problem with street lighting, or a hole in the road, or about
the quality of refuse collection; or they might make such a private com-
plaint to an elected councillor or their local MP. We call these ‘individual
voice’. Of course, it is possible for people to voice individually in order to
pass on compliments to service providers, and this does sometimes happen.
In this way a company or public provider can learn that a specific service
is much appreciated, or learn that a particular operative is doing a good
job. Within firms for example, individual voice might well include compli-
mentary comments or suggestions. Firms introduced ‘suggestion boxes’ in
the 1970s and 1980s for such positive reinforcements and congratulations
as well as for suggestions on improvements. As Hirschman suggests, voice
is a more nuanced response than the binary exit or non-exit response. In
the main, however, people tend to voice individually only to complain, and
it is of these negative individual voice activities that Hirschman normally
writes.
If the individual complaint is about the service that the individual or
family has received, then there is no element of the collective action prob-
lem involved. For example, if the streetlight outside a house has gone out
and a resident nearby points this out to the local government, that person is
acting out of personal self-interest (even though there are positive external-
ities for neighbours and other street users). Similarly, if a person complains
about poor hospital service, part of the purpose is some sort of individ-
ual recompense, or at least to gain some personal psychological satisfaction
that the problems pointed out will be addressed. This will not only impact,
perhaps, on future service the person receives at the hospital but will also
affect others. Individual complaints can be considered on a par with private
goods problems, even though they have important positive externalities for
all hospital users. (Similarly of course if a person makes a complaint about
a private good, pointing out a defect in some product, the company might
realize that there is a design or production problem which they can address
to the benefit of future customers. We can see therefore that individual voice
in the context of private goods also has positive externalities for others.)
People might also engage in collective voice activity. There are two
broad forms of such collective voice activity. Voting is one form. Another
is a more obviously collective or joint action, such as joining and cam-
paigning through a pressure group, signing a petition, going on a march
and so on; the latter, though not necessarily the former, involves horizon-
tal voice. With collective voice activity we might expect to see a greater
collective action problem. There are disincentives too, since a single vote
might be thought not to make a significant difference to the outcome and
if there are costs of voting people might decide not to bother. Similarly
if there are costs of participating in collective acts such as lobbying or
demonstrating, then people are less likely to turn out to do so. A more
important difference from our point of view, however, is that in our
applications of EVL we would ordinarily only expect to see the individ-
ual voice if there is a problem that a consumer wishes to see corrected.
However, collective voice activity might occur even when consumers are
perfectly content with the services as they are currently provided. There
are strong incentives to defend service provision of a certain type if it is
thought to be threatened by change. Whilst everyone is thought to have
similar views about salient goods and services, people have very different
views about the nature of other goods. Some prefer lower taxes and lower
levels of provision of welfare goods; others are prepared to pay higher
taxes if local services can be maintained. Parents are willing to pay more
for better education, perhaps, than those without children; those who
consume hospital services might be willing to pay more; some want their
refuse collected weekly, others are happy with fortnightly collections and
so on. Many of the goods and services we consider in this book are sali-
ent ones; but we have to bear in mind that in many forms of collective
activity views differ.
Collective voice is used to defend the nature or level of services as well
as to try to improve them. Citizens might be motivated to vote to main-
tain the level of services if one party or set of politicians were standing
on a tax-cutting ticket promising to remove, scale down or reduce spend-
ing on some services. Campaigning is also often directed at keeping the
status quo rather than changing the current arrangements.4 Collective
voice activities might not be only about falling quality in the manner that
Hirschman assumes that voice will be directed. This will affect the nature
of the relationship between satisfaction and (collective) voice; and between
(collective) voice and forms of exit.
4
There are other suggestions to sub-divide exit, voice and loyalty extant in the literature.
We review these in Dowding et al. 2000.
Sat isfac t ion 47
Satisfaction
Hirschman assumes that the individual positive response to falling qual-
ity is either exit or voice. The key variable that triggers that response there-
fore is satisfaction with a service. If people are satisfied they feel no need
to exit or individually voice. Thus satisfaction is a key variable that affects
any potential exit–voice trade-off. As we have argued, individual voice
activity is only likely to occur if citizens are dissatisfied with the services
they receive. However, they may use collective voice even if they are per-
fectly satisfied with services but fear they may deteriorate if another party
gains control at an election. The relationship between satisfaction and the
two exit strategies is also complex. All things being equal, dissatisfaction
with public services should be correlated with higher intentions to exit,
and with higher exiting itself. People are more likely to shift to the private
sector if they are dissatisfied with the services provided in the public sec-
tor than if they are satisfied with them. However, they might still exit even
if they state they are satisfied with those services. For example they might
feel that the local school provides a good education and be satisfied with
the service it provides; but, if they can afford it, still send their children
to a private school if they believe that this would bring extra advantages
to their child. One might even believe that the educational standards at a
local state school are superior to a private school (as regulations govern-
ing, for example, teaching qualifications are more demanding in the state
sector) but believe there are social advantages to sending one’s child to a
private school. Or one might have certain religious views that mean one
exits from state provision without being dissatisfied with that provision
on its own terms. One might be satisfied with a public service but still
think the private sector is superior. Satisfaction, then, is a relative con-
cept. Nevertheless, we should still expect to see dissatisfaction correlated
with exit.
Households rarely exit a local government jurisdiction purely for
Tiebout exit reasons. Even if one is highly dissatisfied with the provision
of local government services finding out whether they are better nearby
is not straightforward. Even if one is convinced that a nearby jurisdiction
has superior services moving is costly – by repute one of the most stress-
ful activities that people undertake. We might expect low levels of Tiebout
exiting since it would rarely motivate moving home all on its own. Thus
we expect that households move for a variety of reasons with Tiebout rea-
sons sometimes being part of the package of reasons (John et al. 1995).
Standard location models use a push–pull model for relocation decisions.
That is, people decide to leave a given location for a set of reasons and
decide to move to another for a convergent but not identical set of reasons
(Rossi 1980; John et al. 1995). It has been shown that once a household has
taken the decision to move some compare the tax service packages on offer
from different local authorities within the same metropolitan area (John
et al 1995; Dowding and John 1996) and that people do consider moving
because of the nature of local services (Percy and Hawkins 1992; Percy
et al. 1995; Devereux and Weisbrod 2006). However, comparison with
the relative packages on offer shows that, strictly speaking, exit does not
depend upon dissatisfaction with public services. Dissatisfaction thus has a
complex relationship with both voice and exit, though the general expect-
ation is that dissatisfaction spurs both voice and exit.
From where do people’s satisfaction levels arise? This is not an easy ques-
tion to answer. People become dissatisfied with a service when it does not
meet their expectations. But from where do these expectations arise? What
sorts of comparisons do people make? Do people compare public sector
provision with the private sector? Do they compare services provided in
their local authority with those provided in neighbouring ones? Do people
compare the services they receive with some national standard, gleaned
perhaps from information gained from family, friends, government and
the media? Or perhaps they compare today’s provision with what they have
received in the past. In other words, satisfaction with any given service is
not necessarily correlated with any ‘objective’ indicators of the efficiency
of that service. Whether people are ‘rationally’ dissatisfied or not is irrele-
vant to the exit–voice trade-off. All that matters to our analysis is what the
nature of that trade-off is, given their level of satisfaction. However, even if
it does not matter in theory, it might matter for policy-makers, who wish
to address dissatisfaction, in the hope of reducing exit, as it is a chimera
that is hard to shift.
It should also be noted that it is known that levels of satisfaction with
services vary with social class, educational attainment and employment
status. Those in employment, with higher education and higher social class
tend to be less satisfied with services. This higher level of dissatisfaction is
probably due to higher expectations. It is also the case that, independent
of the level of satisfaction, the better educated and wealthier are also more
likely to voice and exit, as both activities have lower costs for them.
Generally speaking, individual voice activity has a lower cost than
relocation or exit to private providers because of the high costs of reloca-
tion and the extra charges that private provision entail. However, collective
voice might be more relatively expensive too, especially as it might have a
low probability of having a positive result. Collective voice might be more
expensive than moving to the private sector or even moving across juris-
dictions for the relatively wealthy. Exit across providers within the public
sector might be relatively costless.
C onc lusions 49
Loyalty
As we have seen, loyalty has proved the most criticized and troublesome
concept of Hirschman’s trio. In Chapter 3 we consider a modification of
Hirschman that introduces a fourth category, ‘neglect’, and reinterprets
loyalty in behavioural rather than psychological terms. We dismiss that
behavioural interpretation there not so much because it is inherently false
or inconsistent, but rather because we desire to examine Hirschman’s
hypotheses as he sets them out. As we argue, the exit, voice, loyalty and neg-
lect (EVLN) literature sets off on a rather different track from Hirschman.
The framework developed there produces hypotheses somewhat different
from Hirschman’s (though some are identical), and constitutes a rather
different mechanism. Thus we want to conceptualize loyalty in a manner
more in keeping with what we find in Hirschman though in a manner that
is empirically measurable.
To be sure, any intervening psychological effect of loyalty is bound to
have some observable behavioural consequences if it is to be measured,
but we reject loyalty as a separate behaviour from exit or voice. Rather,
following Hirschman, it is something that should make voice (at least in
some varieties) more likely than exit in response to any given level of sat-
isfaction. But this does mean that we need to be able to measure loyalty
in some manner. If it can be measured in a way that is independent of the
voice–exit trade-off itself, then loyalty is not the non-explanatory equation
filler that Barry claims.
Given that in Hirschman’s account loyalty is supposed to be a psycho-
logical concept and thus cannot be measured directly we need some inde-
pendent observable behaviour that theoretically should correlate with any
such concept. One possibility suggested by Hirschman (1970, pp. 77–8) is
attachment, and certainly in examples such as loyalty to one’s country it is
the person’s identification with the object of loyalty that creates that loy-
alty. Group loyalty depends upon one’s identification with the group itself,
which may flow from the individual’s investment in it. Our conception
of loyalty here is not simply an unobservable psychological disposition
but is behavioural (it is measured by past behaviour) and institutional (its
effects can be expected to vary across institutional settings). Loyalty in our
framework is supposed to do the job of loyalty as in Hirschman’s original
framework – to increase the probability of voice relative to exit.
Conclusions
We have shown that Hirschman’s inferences that making exit easier might
drive out voice and this might have a deleterious effect on public services
do follow from his framework. However, we have also argued that other
mechanisms might also operate that mean that competition can increase
quality in the public sector as it can in the private. We have argued, how-
ever, that empirically demonstrating the relationships from his simple
model is problematic since there are various types of exit and various types
of voice and we cannot expect them to interact in identical ways. We have
also argued that loyalty is best seen in a specific manner as loyalty to an
object removed from the quality inherent in the product itself, and loyalty
can in part be measured in terms of past voice activity. Before we turn to
examining the relationships we have identified through the survey work
we have conducted, we want to examine the exit, voice, loyalty and neglect
framework. We do so for two interrelated reasons. First of all because it
is an important framework in itself that uses empirical evidence – sur-
vey evidence over satisfaction of urban services – very similar to our own.
And second, because we see it as a conceptual exercise separate to that of
Hirschman and ourselves and we want to draw those important concep-
tual differences to the attention of those interested in Hirschman’s EVL
approach.
3
Exit, voice, loyalty and neglect
Introduction
No empirical study of Hirschman would be complete without an assess-
ment of the most influential model that has emerged to date, that of exit,
voice, loyalty and neglect (EVLN), developed by Caryl Rusbult, and sub-
sequently applied by many of her colleagues, notably David Lowery. This
Hirschman-like line of study is not only important for our theoretical ana-
lysis, it is also important for the context of our own study, the urban arena.
In spite of its unique contribution and powerful insights, we argue that it is
ultimately a different framework to Hirschman’s. In this chapter we review
some of the empirical literature on exit and voice in the urban context,
before turning to the EVLN approach itself.
Hirschman in the urban context

Our own data come from services directly provided by local governments;
and also from health and education, two services traditionally provided
at the urban level. There have been quite a few studies examining the
Hirschman framework in the urban context; indeed in political science
urban scholars have made most use of the Hirschman framework. In urban
studies, the EVL framework has generally been developed from the study
of exit as an extension of the Tiebout (1956) model. Charles Tiebout sug-
gests that if there are a large number of local jurisdictions in a single metro-
politan area, they could compete for households by offering different tax
service packages. Households, or in Tiebout’s phrase ‘citizen-consumers’
would then ‘vote with their feet’ by moving inter-jurisdictionally to the
51
52 E x i t, voice , l oya lt y a n d n e gl e c t
local government which offered the preferred tax service mix (for a review
see Dowding et al. 1994). Some urban scholars have moved beyond Tiebout
to consider the effects upon exit possibilities on the quality of voice. In
terms of urban politics, leaving the jurisdiction or dropping out of public
service provision for private provision constitutes exit, while engaging in
any form of collective action to improve services or making private com-
plaints to public officials constitutes voice.
Young (1976) in a study following directly from Hirschman’s account
argues that the exit–voice framework is relevant along two dimensions of
public policy. First, there is the spatial dimension, where citizen-customers
can either voice their opinion on the delivery of public services or exit by
moving to a different jurisdiction. Second, there is a vertical dimension,
which refers to the different possible arrangements in the delivery of pub-
lic services. These arrangements vary from total consolidation, where the
public authority produces and disburses the services, to full diversification
where there are many private firms competing to provide the service, with
the public authority having a regulatory role.
The vertical dimension is clearly relevant not only to national public pol-
icy but to local public policy as well. After studying the delivery of different
services with different ways of organizing delivery (Young 1972; Young
and Nelson 1973; Young 1974) proposed some crucial service characteris-
tics suggesting exit is likely to be inefficient when:
1. The quality of the output is difficult to observe.
2. The consumers of the service are difficult to identify.
3. There are economies of scale.
4. The costs of investigating alternatives are high.
5. The costs of switching suppliers are high.
6. Some citizen-consumers can afford to exit more easily than others.
Although the analysis of different forms of organizing service delivery has
become increasingly relevant politically attracting more attention since
the early 1970s, there has been little progress along the lines of Young’s
work. The focus has been on the efficiency of different organizational
forms of exit and voice separately, with little analysis of their interaction.
We attempt to analyse this interaction in terms of stated responses, but
few have tried to analyse these relationships in actual service areas theo
retically or empirically. In the analyses of different organizational forms
of exit and voice, most of the attention has focused on information and
incentive problems of the different agents (for examples of such analysis
for the provision of health services in Britain, see Glennerster et al. 1998
or Le Grand 2007). The complexity of the configurations that these organ-
izational forms can take is daunting. However, the fact that there has been
H i r s ch m a n i n t h e u r ba n c on t e x t 53
no progress in the comparative analysis of public services within the exit–

voice framework can only hamper research on the delivery of particular
services.
Although the six service characteristics that render exit ineffective are
also applicable along the spatial dimension, research within the exit–voice
framework moved away from Young’s approach of analysing structural
characteristics, and toward analyses of individual behaviour. These ana-
lyses attempt to model statistically individual decision-making or attitudes
toward exit and voice. Despite the large amounts of empirical research
conducted along these lines, there has not been a great deal of progress
in both our theoretical and our empirical understanding of the exit–voice
relationship and on how different exit–voice organizational configurations
affect the efficiency of urban government. Again it is not so much the fail-
ure of analysts but the complexity of the issues.
Most of the literature concentrated on the exit options follows in the
wake of Tiebout. Hirschman suggests that high-income, better educated
people will be more able to afford to exit, and these people can also most
effectively use voice. Thus exit reduces voice. Teske et al. (1993) argue that
the fact high-income households can exit strengthens their voice and they
are able to raise the standards of public service provision for everybody.
Exit might reduce voice, but governments can respond to the fact that exit
is possible. Whilst the game-theoretic analysis of Clark et al. (2007) dem-
onstrates this can be true (see Chapter 2, Table 2.2) the conditions of such
a response are highly dependent upon information conditions. Empirical
evidence is required to demonstrate that this possibility actually exists.
There are no studies that directly support this position. The evidence that
high-income households are more mobile and better informed about local
public finance than low-income households has never been connected
with improved local government performance. More to the point the evi-
dence that Teske et al. (1993) present, that high-income movers are more
likely to have engaged in voice activities, contradicts their main claim.
If high-income citizens are supposed to be so effective with their voice
because they are mobile, then why do they leave? And in any case, such
analyses after a move has taken place are very problematic as is argued
below. Dowding and Mergoupis (2003) have argued that despite evidence
of Tiebout moving, there is little evidence of the efficiency of the Tiebout
exit mechanism. Most studies are in urban environments far removed
from the institutional conditions (multi-service local providers) that are
required for such efficiency to emerge. In their empirical analysis of the
UK, larger urban districts where exit is more difficult are more efficient.
They suggest that voice mechanisms, which they do not model, might help
to explain why fiscal mobility does not lead to efficient exit effects.
Despite the rhetoric couched in the exit–voice framework, there have

been few attempts to seriously analyse the interrelationship between exit
and voice. In fact, despite the massive literature in political participation
and fiscal mobility, the only survey-based studies of the interrelation-
ship between exit and voice until the EVLN studies and our own are
those by Orbell and Uno (1972) and Sharp (1984, 1986). In particular,
Sharp’s 1984 study remains one of the most thorough. In it she analyses
the probabilities of having voiced, conditional on intentions to exit. The
analysis was on those citizens who considered the performance of their
local government problematic, and considered two types of voice. The
first was contacts with local government officials, which she calls ‘indi-
vidualistic’ participation, and the second was working with neighbours
to solve problems, which, following Verba and Nie (1972), she calls ‘com-
munal’ participation mode or what has been called ‘horizontal’ voice.
She stresses that definite intentions to move were associated with a large
drop in the ‘communal’ mode of participation among those with less
education. She concludes that ‘[t]he better-educated may indeed have
more mobility potential, but this does not interfere with their inclination
to use voice when local government violates their preferences’ (Sharp,
1984, p. 74).
There are two main problems with her study, undermining her con-
clusion. First, her cross-tabulations do not control for length of stay in
an address or in the community. This is problematic because those who
moved recently are both less likely to have expressed either of the types of
voice and more likely to move on because of the large percentage of fre-
quent movers among all movers. Second is the problem of ex post analysis
of participation behaviour. To examine the probabilities of having voiced,
given intentions to move, is something quite different from the problems
of personal calculus of exit and voice. The main interest of this calculus is
not what happens after voice has been exercised, but before; in other words,
how the possibility of exit affects whether voice will be exercised or not.
One proper methodology is to ask people at point t1 about their satisfaction
with local government, and in point t2, examine if they have moved, and
ask them about their current level of satisfaction, and whether they have
exercised voice. We perform this type of dynamic analysis in our survey
analysed in Chapters 4 and 5.
Devereux and Weisbrod (2006) directly examine the effects of satisfac-
tion on voice in the form of complaints and geographical exit with data
from a survey of Chicago residents. They find that complaints are responses
to stated dissatisfaction and the more dissatisfied complain more. They
also find that geographic exit is a strategy for some and that voice and exit
are substitutes rather than complements.
H i r s ch m a n i n t h e u r ba n c on t e x t 55
Progress in the analysis of exit and voice in the spatial dimension of

public policy is most promising in two directions. First is improved meth-
odology. Given the high expense of the longitudinal approach, the next
best approach is analysis of intended behaviours. Of course there is a crit-
ical gap between intentions and their realization but that can only be ana-
lysed when there is data on both the intention and the realization – which
brings us back to the longitudinal survey. Secondly, Sharp showed that
there are significant differences between how (past) private and collect-
ive voice relate to intentions to exit. This confirms Hirschman’s intuition
on the richness of voice activities. It is likely to be misleading therefore to
treat all these activities as one. Again we analyse more fully the differences
between individual and collective voice in this regard.
Moving back to Tiebout once again, there have been some studies look-
ing at tax competition and voice. The key is to add extra variables to the
aggregate analysis, often in an ingenious way. Feld (1997) looks at tax
competition between the twenty-six Swiss cantons in 1990, considering
the impact of referendums to control expenditure. He tests the hypoth-
esis that tax competition is greater in representative democracies where
voters do not have the opportunity of voting on tax policies, compared to
where there are referendums. He estimates a model to explain the shares
of taxpayers in different income groups, which he seeks to explain by the
local tax rate, their neighbours’ tax burdens and the tax rate in represen-
tative democracies. The results show that both tax competition and voice
are at work in explaining the number of taxpayers. The tax rates of repre-
sentative democracies reduce the numbers of taxpayers. Tax competition
is higher in a representative democracy. He also examines capitalization
of taxes into property prices, a common way to estimate tax competition.
Less voice should mean more capitalization. He tests by predicting rents
by covariates and the index of the tax burden and the index for the cantons
with representative democracies, i.e. without referendums. Here the tax
variable of the interaction terms is significant and negative as expected,
which means that fiscal competition and opportunity to represent trade off
negatively. Having no opportunity to voice reduces rents and house prices
because people want to move. In this study, voice drives out exit rather
than the other way round.
Another take on exit–voice mechanisms in Swiss cantons comes from
Schaltegger and Küttel (2002) testing the Leviathan thesis: that exit
may constrain the decisions of government, forcing it to mimic those
of its neighbours. Like Feld, they see representative institutions giving
less voice. Thus they will have mimicking behaviour in contrast to their
neighbours with direct democratic mechanisms. Schaltegger and Küttel
(2002) run equations to see whether revenue expenditure is predicted by
a combination of voice and exit (measured by fragmentation), and mim-

icking. They use annual expenditure and tax data from 1980 to 1998. They
find that voice decreases expenditure as predicted, as well as fostering its
mimicking behaviour, with an interaction between voice and the neigh-
bourhood factor. There is a strong finding for tax: policy mimicking goes
down when there is voice, confirming the exit–voice trade-off.
Hendrick et al. (2007) make inferences about the type of taxes to ascer-
tain voice and exit processes. They assume that property taxes indicate
voice whereas sales taxes are about tax competition. This is because exit is
a more expensive option for property taxes, but not for sales taxes – this
is an important assumption, which does not directly observe voice, so the
credibility of the article depends on it. They estimate a spatial regression
model for the municipal governments in the Chicago region, which aims
to reveal the extent to which the taxes in one jurisdiction depend on the
rates in the other, showing that competition exists. The significance of the
term for the spatial lag in the equations estimating property taxes shows
that they experience competition. There is no relationship for sales taxes
as predicted.
These studies show a relationship between satisfaction and exit or voice
activities. Some suggest that there is an exit–voice trade-off as Hirschman
suggests with exit intentions leading to less voice. They do not really grapple
with the different forms of voice, nor with different types of exit; and none
of them examine loyalty. In the urban service context the work of David
Lowery and colleagues does examine loyalty and exit–voice relationships.
They utilize the EVLN framework rather than that of Hirschman himself.
In the next section we explain the EVLN framework in some detail before
moving on to discuss the work of Lowery et al. in subsequent sections
where we also contrast the EVLN approach to that of EVL itself.
Exit, voice, loyalty and neglect

The EVLN approach is a social-psychological approach to human behav-
iour with a somewhat different research question from our own, but it
has been applied to similar questions about public service provision in
an important book, Lyons et al. (1992) and related articles (Lyons and
Lowery 1986; Lowery and Lyons 1989; DeHoog et al. 1990). We see the
EVLN approach as a somewhat separate research programme from that
of Hirschman as the basic concepts within the EVLN approach are for-
mulated and operationalized rather differently from those in Hirschman
and therefore do not entail the hypotheses he expresses. Despite this
rather different background some of their findings are similar to ours. We
critically examine the EVLN approach in this chapter and discuss Lyons
E x i t, voice , l oya lt y a n d n e gl e c t 57
et al. (1992), though we also compare and contrast their findings with
ours in later chapters.
Our criticisms of EVLN here are given in a comparative context and as
an application of Hirschman’s original ideas. Our critique is not meant
to suggest that the EVLN framework is incoherent or does not provide
insights in the empirical contexts to which it is applied. Rather we suggest
that the EVLN framework is not really a development of Hirschman even
though it is obviously inspired by it. Conceptually it is completely separate.
The terms exit, voice and loyalty are all conceptualized differently from
Hirschman, notably the term ‘loyalty’, and these conceptions lead to infer-
ences that are apparently at variance with those of Hirschman. By contrast
our own framework set out in Chapter 2 is intended to be a development
of Hirschman and directed at precisely his own research questions. We
aim to overcome problems with Hirschman’s original framework but to
do so within his own terms. Our critique of EVLN, therefore, is that it
is different and not necessarily a rival model to our own. Our model, we
argue, provides a framework more in keeping with the theoretical issues
of exit and voice, market and forum that motivated Hirschman’s original
book. In other words, whilst the EVLN framework was clearly motivated
by Hirschman’s work, its theoretical concepts are so different from his that
studies utilizing it are not really testing Hirschman’s ideas. Rather they are
testing related but not identical ones.
The EVLN framework was developed by the social psychologist Caryl
Rusbult and her colleagues and has been applied in a number of differ-
ent contexts, notably to personal love and marriage relationships (Rusbult
et al. 1982; Rusbult and Zembrodt 1983; Rusbult et al. 1986; Rusbult 1987;
Gaines et al. 1997; Rusbult et al. 1998; Vigoda 2000); workplace relation-
ships (Rusbult and Farrell 1982; Farrell 1983; Rusbult and Lowery 1985;
Rusbult et al. 1988a; 1988b; Withey and Cooper 1989; Turnley and Feldman
1999; Lee and Jablin 1992; Thomas and Au 2002; Naus et al. 2007); psycho-
therapy (Derlega et al. 1993); and ‘everyday life’ (Drigotas et al. 1995). It
was further applied in a very similar empirical context to our own empir-
ics in Chapters 4 and 5, namely to efficient service delivery in local gov-
ernment, by David Lowery and his colleagues in a series of articles (Lyons
and Lowery 1986; 1989; Lowery and Lyons 1989; DeHoog et al. 1990) and
an important book (Lyons et al. 1992), as well as in the context of urban
school systems (Matland 1995).
The social-psychological EVLN literature is both an analytic framework
and a predictive tool used to examine reactions to dissatisfaction in vari-
ous social relations at work, within and between groups, and in romantic
situations. The concepts employed are rather different from those intro-
duced by Hirschman though they bear a superficial resemblance; and the
operationalization of those concepts does not match the original descrip-

tions of Hirschman’s framework. The EVLN model consists of four dif-
ferent classes of behavioural responses to dissatisfaction which are then
tested in the sets of relationships mentioned above. The four behavioural
responses are the following – the first mentioned in relation to romantic
involvement, the later ones to organizational workplace studies:
Exit: leaving a relationship; or leaving an organization by quitting,
transferring to another organization, sabotage, or searching for
another job, or thinking about quitting.
Voice: discussing problems and seeking help; or actively and construct-
ively trying to improve conditions through discussing problems with
supervisors or co-workers, taking action to solve problems, suggest-
ing solutions, seeking help from outside agencies such a trade union,
or whistle blowing.
Loyalty: waiting for the situation to improve; or passively but optimis-
tically waiting for conditions to improve; giving support, both public
and private to the organization, waiting and hoping for improvement,
or practising good citizenship within the organization.
Neglect: putting in less effort and developing negative attitudes to the
partner or relation; or passively allowing conditions to deteriorate
through reduced interest or effort, chronic lateness or absenteeism,
lack of effort leading to errors.
We can note from the outset that every one of these concepts differs in
varying degrees from those in Hirschman. For example, exit is not simply
leaving an organization but thinking about leaving. Studies have shown
a strong relationship between intentions to exit and actual exit (Rossi
1980) and in our analysis we too use intentions to exit. But we note here
that intentions to exit and exiting are not identical. Considering leaving
an organization might have behavioural consequences for the would-be
exiter which might be different from actual exit. For example, someone
who has exited cannot voice (at least not in Hirschman’s original sense)
but a would-be exiter might have more reason to voice. We examine this
dynamic factor in our empirics in Chapters 4 and 5. These ‘noisy exiters’
play an important role in Hirschman’s original argument (Barry 1974).
However, this is not a possibility in the EVLN framework since exit and
voice are separate behavioural categories. In Hirschman the actual oppor-
tunities for exiting given a desire to exit are important as to whether some-
one is more or less likely to voice. Again this is not a possible inference in
the EVLN framework because voice and exit are separate states. People are
placed in only one of the exit or voice or loyalty or neglect categories. Being
dissatisfied and wanting to leave but being unable to do so encourages
E x i t, voice , l oya lt y a n d n e gl e c t 59
voice responses according to Hirschman. This is not a possible inference

in the EVLN framework.
Voice is seen only in positive terms in the EVLN framework. While
Hirschman importantly sees voice as having positive effects on organ-
izational output, it is often viewed as simple complaints about a prod-
uct. He makes clear that voice is a nuanced response that can emerge at
several levels and types. Voice is not necessarily a positive or construct-
ive response in Hirschman’s original framework. Loyalty is also viewed
somewhat differently, as we discuss at some length below. Exit and voice
are behavioural responses for Hirschman; loyalty, whilst having some
behavioural components, is more of an attitude or disposition of the
agent towards the organization or product. The important use of loyalty
in Hirschman is as a mediator between exit and voice, making the latter
more likely than it would otherwise have been in the absence of agent
loyalty. In the EVLN framework loyalty is a behavioural response on a
par with, and rival to, voice and exit. Neglect, of course, is a completely
new and separate behavioural category not seen within the Hirschman
framework.
These categories are seen as four behavioural responses to a person’s
situation in a relationship or workplace in response to some level of dis-
satisfaction. Thus each of the categories is a rival and responses can be
assigned to one or other of the categories. Given that the categories are
measured in the EVLN framework by a multitude of variables it is pos-
sible for someone to have elements of voice and exit in their responses.
However, the analysis essentially assigns individual response to one or
other of the categories. Everyone can be assigned a place in one of the
four categories dispositionally, and indeed in one study where individ-
uals’ responses were coded six years apart, their assignation to one of
these categories was relatively stable over time (Withey and Cooper 1989).
The conditions under which people respond to current dissatisfaction
are conditioned on previous responses allowing for some dispositional
change amongst agents.
Furthermore, according to Rusbult and her colleagues these four
classes of responses differ from each other along two evaluative dimen-
sions: a constructiveness/destructiveness dimension and activity/passivity
dimension (see Figure 3.1). Whereas exit and voice are considered to be
active responses to dissatisfaction; loyalty and neglect are deemed to be
passive responses. In addition, voice and loyalty are considered to be con-
structive responses, whereas exit and neglect are thought to be destructive.
It is in this way that the categories of exit and voice notably take on an
evaluative dimension that is much stronger than anything in Hirschman’s
original formulation.
Active
EXIT VOICE
Destructive Constructive
NEGLECT LOYALTY
Passive
Figure 3.1 EVLN responses to dissatisfaction
Active and passive; constructive and destructive

The underlying dimensions seen in Figure 3.1 are dispositional in nature.
It would appear that people respond to dissatisfaction either construct-
ively or destructively. And they also respond actively or passively. The
idea that there are dispositional natures, which tend to lead people to
certain responses to environmental conditions, is intuitively plausible,
and mapping out such personality traits is standard within the psycho-
logical and social-pyschological disciplines. However, capturing such
dispositions with the use of standard multiple regression techniques is
problematic. Such dispositions underlying behaviour cannot be directly
observed in the data but are assumed to underlie a series of indicator
variables. If they are not unobservable then one or more of the coded
variables would correlate with the observed behaviour. Together with
some theory those correlations would be used to infer that those vari-
ables partly structure the behaviour. Such underlying dimensions might
be inferred by correlating sets of variables: factor analysis is a favoured
technique for doing this. Factor analysis examines the variability among
a set of observable variables to model them as linear transformations
(with error terms) of a smaller set of unobservable variables or factors. It
is an essentially descriptive technique and does not allow causal inference
even though it may be useful in investigating the underlying categories
in the data – as we do at the start of the following chapter. However, inde-
pendent evidence of the features captured by the factors would enable
causal inference. Thus, if we were to identify some gene in individuals
who were dispositionally passive, then we could make the causal infer-
ence. Without such independent evidence all we have is redescription in
Ac t i v e a n d pa s si v e 61
terms of underlying variables. For example, sullenness, accepting decline

and absenteeism are a definition of ‘passivity’. Without some independ-
ent evidence of something – a gene for example – that is to be found
in people who display this behaviour we have not explained these three
responses in terms of ‘passivity’; we have simply labelled them. If we find
them associated with observables such as social class and education,
together with a theory why, say, lower social class and education lead to
passivity – through higher costs of voice, lower efficacy and so on – we
do have a purported explanation.
EVLN studies often utilize multidimensional scaling (MDS) to map
these underlying unobservable cognitive structures that might motiv-
ate behaviour. MDS provides a geometrical interpretation using space
coordinates to describe the assumed underlying cognitive structure. The
map then allows for estimates of the similarity or difference in respond-
ents’ perceptions of the possible response options put before them. As well
as spatial mapping, MDS allows for an interpretation in terms of con-
structive or destructive, passive or active behaviour. In order to operate
these two elements two groups of subjects are required; in one of these the
similarity–dissimilarity mapping is engendered, with the second group
being asked to interpret the behaviours. It is on these grounds that voice is
seen as constructive and exit as destructive, with the second group identi-
fying the normative criteria. It is the second group therefore that adds the
normative dimensions to the described behaviours.
In this process the categories of exit, voice, loyalty and neglect are out-
come variables and this part of the process is essentially descriptive. It
details how different types of response to immediate dissatisfaction can
be reasonably grouped into four categories with those titles. It is also
normative since the active–passive and especially the constructive–
destructive underlying variables are clearly described in ethical terms.
The normative elements of these descriptive behaviours are those likely
to be recognized in the normative dimension by the second, interpretive
group.
The active–passive dimension might be compared to Hirschman’s alert–
inert categorization of people. Hirschman suggests that when there is
quality decline alert people notice and will voice or exit. The inert will not
notice the decline. If the question is about noticing some objective decline
in quality then we would not expect to see the inert saying that they are
dissatisfied. However, inert people might be dissatisfied but be passive in
the EVLN sense. We might equate alert and active people, and inert and
passive people if each category is seen as a personality or character trait not
necessarily captured by observable personal characteristics such as educa-
tion or social class.
Use of EVLN
Once the mapping techniques described above are completed the infer-
ences that might affect different types of behaviour are made. The EVLN-
typology has received considerable empirical support in surveys classifying
responses to dissatisfying events in romantic relationships (Rusbult and
Farrell 1982; Rusbult et al. 1982; Rusbult and Zembrodt 1983; Rusbult et al.
1986; Drigotas et al. 1995; Gaines et al. 1997). These surveys generally ask
people to recall a time or situation in their relationship with which they
were dissatisfied. They are then asked to describe their reactions, both
via open (‘how did you respond?’) and closed answer questions (‘to what
extent did you consider leaving?’). The EVLN-taxonomy receives further
support in research using role-play methodologies, in which participants
have to respond to scenarios describing a particular dissatisfying relation-
ship (Rusbult et al. 1982). Beyond providing a taxonomy of reactions, these
studies have also examined the conditions under which the exit, voice,
loyalty and neglect responses are more or less likely to occur. In general,
dissatisfaction appears to be associated with destructive responses (exit
and neglect), while a large investment in the relationship (having bought
a house, having children) is linked to constructive responses (voice and
loyalty).
The EVLN framework has been used descriptively to define specific
types of responses to levels of satisfaction in the workplace or relation-
ship. Thus EVLN researchers infer that the higher the satisfaction of the
workforce prior to any specific problems that might arise which then lead
to dissatisfaction, the higher the likelihood that constructive voice or loy-
alty responses will occur. If there has been dissatisfaction for some time,
then further dissatisfaction will lead to negative neglect or exit behaviour.
Second, they argue that the better the outside alternatives for a worker
the more likely will be the active responses of exit or voice; and the lower
the likelihood of loyalty or neglect. Thirdly, they argue that the greater the
investment that a person has put into the job the more likely that current
dissatisfaction will lead to voice or loyalty. By investment they mean two
things. First, the time and resources that an employee has invested in a
firm, such as the amount of effort expended in the past, the longer they
have been in the job, security of tenure or non-portable human capital
such as training and expertise. Second, ‘outside’ resources such as satisfac-
tion with housing, ease of travel to workplace, friends in the workplace and
non-portable retirement funds contribute to positive behaviours.
We can see that these hypotheses are not entirely aligned with
Hirschman. The first hypothesis that prior satisfaction should lead to voice
or loyalty rather than to exit or neglect seems aligned (given the differences
Use of EV L N 63
in the constructs) with Hirschman’s framework. The third hypothesis

that the more someone has invested in a job the more likely they are to
be loyal or to voice also seems aligned with Hirschman (again allowing
for the differences in the construct). However, the second hypothesis that
better alternatives in the EVLN framework lead to either voice or exit is
not. For Hirschman, the better the alternative all things being equal, the
more likely people will exit than voice. Loyalty comes in to make voice
more likely relative to exit. In EVLN it is a separate (passive) alternative to
exit and voice. Here the difference in the constructs brings about a slightly
different hypothesis.
The various EVLN studies do not always provide identical findings
with regard to these main EVLN hypotheses, but we will not review the
findings in detail here. In general these three hypotheses have been con-
firmed though the first hypothesis does have some contrary evidence. In
some studies high levels of employee investment do not seem to stop exit.
The extraneous factors in the third hypothesis seem to inhibit exit behav-
iour but employee investment does not. This is important for our own
approach since we see social investments (including past voice activity)
as an important component of the construction of loyalty, which should
inhibit exit and make voice more likely, though, as we shall discuss later,
our notion might be closer to the extraneous component in the EVLN
framework. Withey and Cooper (1989) have a more subtle approach to
evaluating employee responses to dissatisfaction. They use some models
based on those of Freeman and Medoff (Freeman 1976; 1980; Freeman
and Medoff 1984) that seek to explain exit and voice by trade union mem-
bership controlling for wages, other measures of pecuniary rewards, age
and years of schooling, but also adopt the Rusbult et al. EVLN framework.
Using a longitudinal data set of employees’ responses to dissatisfaction
they find exiters are affected by the costs of their response. High satis-
faction predicts voice whereas exiters and loyalists have the lowest scores.
Whilst their voice responses are as predicted by theory, the evidence is
rather weak. To explain the unreliability of their results, they articulate
the need for more complex theoretical formulation: ‘voice is a complex cat-
egory which may have several sub-components’ (Withey and Cooper 1989,
p. 534) and ‘this category is complex, with acts differing on whether voicers
take action or talk about action and whether they work individually or col-
lectively’ (Withey and Cooper 1989, p. 535).
Leck and Saunders (1992) adopt a similar approach – also by using the
EVLN model. They examine 320 students’ responses to dissatisfaction,
finding that it was positively correlated with exit, neglect and what they
call ‘patience’ or the EVLN (or Birch) ‘loyalty’. They suggest that these
responses might be toward different types of dissatisfaction but do not
explore that issue further. They do not find dissatisfaction did not lead
to increased voice, rather satisfaction was positively related to voice. They
confirm Hirschman that loyalty is positively related to voice and negatively
related to exit. As with Withey and Cooper, voice performs weakly in their
model, suggesting that this is because their measures do not load together.
Though they do not discuss this issue, this problem implies that there are
different types of voice. This concept is central to our analysis.
EVLN has also been applied to research on marketing channels. A mar-
keting channel is an organized network of agencies and institutions which
carry out marketing. Ping (1993) examines the responses to dissatisfaction
with the channel using categories from Hirschman and the EVLN adap-
tation: exiting, voice (which he regards as constructive attempts to change
conditions), loyalty, opportunism (what he calls ‘self interest seeking with
guile’), and neglect (which is the desire to reduce contacts). So this is far
from a pure EVL test – and does not test the effect of loyalty on the exit–
voice trade-off. He tests his version of EVLN with 222 hardware retailers
using structural equation models. He finds that satisfaction predicts voice.
Alternative options predict exit but not voice. So this study is only a partial
test of EVL. Ping (2003) carries out similar analysis on a new sample of
hardware retailers but with two stages to the survey, again with structural
equation modelling. He finds that voice intention leads to satisfaction, but
not the other way round. Satisfaction decreases the attraction of alterna-
tives (exit) as well as the other way round – the attraction of alternatives
decreases satisfaction.
Ping (1999) seeks to discover if loyal behaviour is negatively associated
with voice, and voice is negatively associated with exiting, if satisfaction
moderates these associations, and if satisfaction is positively associated
with loyal behaviour and voice, and negatively associated with exiting.
He wants to discover if the cost of exit is positively associated with loyal
behaviour and voice, and negatively associated with neglect and propen-
sity to exit. Moreover, satisfaction should weaken the association between
cost of exit and loyal behaviour. This is again tested with structural equa-
tion models, though these fit poorly. He did not find a voice–exit trade-off,
but a positive path from loyalty to voice, a negative relationship from voice
to neglect, and a positive relationship from neglect to exit. Ping regards
these responses as confirmation of Hirschman: that as problems increase
so loyalty goes down; and as loyalty goes down so voice goes up. Rather
than confirming Hirschman, however, it seems to point to the opposite
conclusion. In Hirschman loyalty is supposed to increase voice relative to
exit, though at high levels of dissatisfaction voice might go down. However,
given the reconceptualization of the concepts it is not clear that this is a test
of Hirschman at all, although Ping’s further findings do seem to confirm
EV L N a n d ci t i z e n sat isfac t ion w i t h se rv ice s 65
Hirschman: ‘in addition, as offended retailers’ voice declined, their rela-

tionship neglect was more likely to increase than their exit-propensity …
as offended retailers’ voice declined, their exit-propensity increased, but
only by increasing their relationship neglect’ (Ping 1999, p. 232). Moreover,
the results ‘suggest that voice may hold exit at bay only when it is less costly
for an offended retailer to use voice than it is for them to neglect the rela-
tionship’ (Ping 1999, p. 232).
In a similar vein, Hibbard et al. (2001) investigate the responses of deal-
ers to suppliers’ destructive acts using a mix of EVLN and Hirschman
responses, particularly different kinds of voice, some of which are positive
and others are not. Much of this work is not concerned with Hirschman
as such, as it does not look at exit; it is rather about a more general set
of responses and their relationships to performance. The hypotheses were
tested on 742 dealers in one marketing channel, using path analysis. The
main finding is that constructive discussion (voice) has a positive relation-
ship to disengagement (which is regarded as a form of exit).
As we have seen, there are important differences in the EVLN as
opposed to the EVL framework. Placing those aside however, and draw-
ing out the empirical implication of EVLN studies to EVL we can see
that there is support for Hirschman. Dissatisfaction does lead to more
exit as we should expect, though satisfaction sometimes seems to lead
to greater voice. This might be explained by the context of some studies,
especially in the workplace where voice activity might require a degree of
trust and confidence themselves likely to lead to greater satisfaction. It has
also been suggested however that intentions to voice seem to lead to sat-
isfaction rather than satisfaction leading to voice. Investment, something
we associated with loyalty, does seem to make voice rather than intentions
to exit more likely.
The EVLN studies are important but not all those who study satisfaction
in the urban context use the EVLN approach. We consider some of those
approaches next.
EVLN and citizen satisfaction with services

Lyons et al. (1992) use the EVLN framework to examine citizen satisfaction
with services. The book is designed to explain why citizens become dissat-
isfied with public services, in the context of US urban governments, and
to examine the institutional effect on that satisfaction, notably the effect of
large-scale consolidated urban government versus fragmented urban gov-
ernment. In later chapters we will consider in detail some of their findings
in comparison with our own. Here we merely set out their study with some
conclusions which are framed within the EVLN approach. They slightly
Active
EXIT VOICE
e.g., e.g.,
leaving or considering contacting officials;
leaving jurisdiction; collective action;
opting for privatized signing a petition
services
Destructive Constructive
e.g., e.g.,
not following public trusting officials;
issues; defending the community;
believing there is no believing community problem
point complaining will be solved
NEGLECT LOYALTY
Passive
Figure 3.2 The Lyons et al. EVLN framework
Source: simplified from Lyons et al. (1992), p. 54
modify the EVLN framework to apply it to the context of citizen satisfac-

tion with public services, as shown in Figure 3.2.
Like the EVLN studies of organizations, Lyons et al. examine the four
responses in relationship to three aspects facing citizens: their prior satis-
faction with services, the investment that citizens have made in their com-
munity, and the alternatives that are available. In Table 3.1, reproduced
from their book, we see the expected relationships.
They expect exit behaviour – which in the Lyons et al. framework refers
to ‘Tiebout exits’ (exit-move): that is, citizens leaving a government juris-
diction because of dissatisfaction with services (Tiebout 1956; Dowding
et al. 1994), or to move to a private rather than a public service – when prior
satisfaction and social investment are low and there are viable alternative
government jurisdictions they can move to. Social investment in Lyons
et al. (1992) involves social ties with family and friends or psychological
ties in the form of not wanting to leave the area. Voice is expected where
prior satisfaction, investment and other alternatives are all high. Neglect is
expected where they are all low. They expect loyalty to occur where prior
satisfaction and investment are high but other alternatives are low.
In their empirical analysis they find that investment works as expected.
Prior satisfaction is also correctly signed for exit-move and neglect. Prior
satisfaction also works for loyalty, but not for voice where it seems it has the
opposite sign to that expected. For the alternatives they find nothing that
is significant. Lyons et al. (1992) suggest these results give moderate sup-
port for EVLN. The problematic finding is that past satisfaction does not
EV L N a n d ci t i z e n sat isfac t ion w i t h se rv ice s 67
Table 3.1 Hypothesized relationships between propensity to invoke

responses to dissatisfaction and the determinants of responses
Responses to dissatisfaction
Neglect
Determinants of response Exit Voice loyalty
Prior satisfaction – + – +
Investments – + – +
Alternatives + + – –
Source: Lyons et al. 1992, p. 58
encourage voice. The reason might be that past satisfaction is highly cor-
related with present satisfaction in which case we would not expect voice
activities. People are unlikely to voice in the form of complaints when they
are satisfied (though see Chapter 4). In order to test for that possibility
Lyons et al. (1992) respecify their framework, replacing prior satisfaction
with current dissatisfaction and find that voice is significant and correctly
signed as they expect.
The problem with the voice reaction points to a problem with the EVLN
literature in general. The four responses point to dispositional natures
given environmental conditions. Past satisfaction is supposed to engender
a specific type of response to current or episodic dissatisfaction. However,
capturing the nature of the current dissatisfaction in the context of past
satisfaction levels through the survey design produces multicollinearity
and current satisfaction is not well measured in much of the EVLN litera-
ture (Naus et al. 2007). In that sense it does not address the fundamental
concerns of Hirschman’s original framework.
The Lyons et al. framework, however, suffers from more serious prob-
lems related to the more general difficulties with their approach. First,
there is the general difficulty of attempting to explain alienation from the
political sphere using behavioural characteristics, when there is a substan-
tial body of evidence that shows that there are structural factors that are
responsible for it (for a good summary, see Howard et al. 1994). Secondly
and relatedly, any measures entering the model are bound to be flawed.
For example, the measure of neglect itself turns out to be a measure of
belief in the responsiveness of local government (Lyons et al. 1992, p. 196).
But a belief that a government will be unresponsive to voice may be based
on years of experience with political activism as well as on an attempt to
rationalize inactivity. Their model is unable to distinguish between the
two, and hence unable to distinguish structural from behavioural causes.

In short, neglect may be a useful category in the individual calculus of
personal relationships, but less so in the calculus of political participation.
Further, as we saw with EVLN above, the reconstruction of Hirschman’s
original framework in this context confuses psychological and behav-
ioural variables, and actions which may be complementary are represented
as exclusive (Dowding and John 1996).
Lyons et al. (1992) conclude that citizen satisfaction comes from a broad
range of public services, the amount of social investment citizens put into
their community and a sense of citizen efficacy. These lead them to support
consolidated rather than fragmented government. In essence they argue
that voice is enhanced with lower geographical exit opportunities but do
not do so on the Hirschman grounds that exit reduces voice, as they do
not consider the exit–voice trade-off directly. We can learn much from the
Lyons et al. (1992) approach but it suffers the problems we have identified
with the EVLN approach with regard to testing Hirschman directly.1
Loyalty: attitude or behaviour?

The major conceptual problem with the EVLN framework, from the view-
point of those interested in examining Hirschman’s hypotheses, is the use
of the concept of ‘loyalty’ as a specific form of behaviour. This differs from
the usage that Hirschman adopts when developing his explanation of the
interrelationship between exit and voice, though it does have some textual
justification from Hirschman.
Loyalty is without question a normative concept. Loyal people are
thought well of. But we might disparage someone who is seen as too loyal;
that is, because excess loyalty turns into slavish commitment with no crit-
ical faculty. Or the object or cause to which one is loyal might be thought to
be immoral, iniquitous or evil; loyalty to such objects might be criticized.
However, the loyalty of one’s enemies to their cause might be applauded if
the cause of one’s enemies is not considered immoral or evil. The converse
of loyalty – disloyalty – evokes the idea of treachery or betrayal (Graham
and Keeley 1992, p. 191), which is universally condemned.
Despite the universal appeal of loyalty, there are different views about
it. Hirschman, of course, sees loyalty as an important component in
1
To illustrate the limitations of EVLN in this manner, another study that uses it does not
test the relationship between exit and voice at all. Menahem (2001) surveys 970 residents’
attitudes to local services in Tel Aviv, reporting regression results of exit, voice, loyalty
and neglect, by the factors of satisfaction, attachment and covariates with largely intui-
tive results of the impact of dissatisfaction.
L oya lt y: at t i t u de or be h av iou r? 69
recuperative mechanisms to halt decline in organizations or products,

suppressing exit and encouraging voice. Everyone agrees that loyalty sup-
presses exit, but they disagree about its effect on voice. Some argue that
loyalty encourages voice (Barry 1974; LaPonce 1974; Evan 1975; Graham
1986; Spencer 1986; Boroff 1989). Others suggest that loyalty suppresses
voice (Birch 1975; Farrell 1983; Fisher and Locke 1992). Birch (1975, p. 75)
was the first to express this clearly: ‘loyalty means a disposition to accept
rather than a disposition to criticize’. We call this second view ‘Birch
loyalty’.
We can see how Birch’s notion might appeal to those who advocate radi-
cal change in an organization, precisely perhaps to turn back long-term
slow decline. They might view voice activities (at least in some forms) as
deleterious to the reform process. Voice in terms of criticism of change
might be seen as a threat to the radical changes that new managers might
consider vital to arrest long-term decline. In that sense managers who
advocate change would sooner have loyalists who keep quiet and stick by
the organization during upheaval. These managers would want loyalists
who trust the leadership to bring about change for the better. Loyalists in
this sense are those who have trust in their leaders and managers. There
might be a sense in which loyalty might be revealed through voice, voices
that support change, but voice itself does not denote loyalty. Chronic com-
plainers might be just that – grousers who do not help the organization
or the community. Voice has many qualities and can be used positively or
negatively. And of course one person’s positive contribution might be seen
by another as a negative response.
This conflict can be seen in the operationalization of loyalty in EVLN
studies. Cooper et al. (1990) and Withey and Cooper (1989; 1992) argue
that there is a construct validity problem with the concept of ‘loyalty’
as specified in some EVLN studies. Loyalty is defined as supporting the
organization but the items used to measure it are defined too narrowly to
conform to most people’s idea of support (such as ‘say nothing and assume
things will work out’). This seems to follow the Birch line that loyalty is
not connected to voice. Following this line of thought, Withey and Cooper
(1992) suggest that there are passive and active forms of loyalty, the lat-
ter being constructive, the former destructive. In their own study they
found ‘passive loyalists’ tended to be dissatisfied and uncommitted, but
active loyalists tended to be highly committed and satisfied. In Withey
and Cooper (1992) active loyalty is associated with investment (or what
they see as high psychological costs of exit) but low economic costs (that
is skill specificity and sunk costs in present employment) and low voice
costs. Passive loyalty is related to entrapment or ‘lock-in’ and, they argue,
it is found to be correlated with individual characteristics and was robust
with individuals over two studies six years apart. This seems to track an
underlying psychological characteristic that might be denoted as ‘active’
and ‘passive’ personalities along one of the EVLN underlying dimensions,
and ‘loyalty’ might emerge in either form. However, this does not fit the
general EVLN model of four distinct behaviours, and turns loyalty into a
hybrid disposition that includes saying nothing and encouraging voice.
It should be noted that secondary analysis of the terms used to describe
‘passive loyalty’ are not categorized as ‘loyalty’, suggesting that ‘active loy-
alty’ is what most people associate with the type of loyalty that Hirschman
originally identified.
There might be passive loyalty and active loyalty. It might be that
there are personality types that lead some to be passively loyal and
others actively so. These dispositions might be related to other features
or characteristics of individuals and be empirically specifiable. We do
not doubt that. However, in keeping with Hirschman’s aim in his frame-
work to explain the interrelationships between exit and voice, and how
different institutional forms and structures might affect that relation-
ship, loyalty is seen as important in the sense that it ‘holds exit at bay
and activates voice’ (Hirschman 1970, p. 78) and that is the sense that we
use loyalty in this work. We see a close connection between loyalty and
voice, to the extent that past voice can be seen as a measure of invest-
ment that is likely to correlate with degree of loyalty. Our aim, then,
is somewhat different from that of the EVLN approach. Rather than
trying to describe different reactions to organizational change or qual-
ity decline in relationships, and see how those correlated with psycho-
logical dispositions, we more directly examine Hirschman’s hypotheses
drawn from his framework. We do see the need to modify that frame-
work in ways we mentioned in Chapter 1 and specify in more detail in
Chapter 4. But our framework more directly follows that of Hirschman
than the EVLN approach, in part because our research questions are
more closely aligned to Hirschman’s than those of Caryl Rusbult and
her colleagues.
Hirschman’s own views are most clearly expressed in chapter 7 of his
original book entitled ‘A Theory of Loyalty’. He says there that a loyalist
is ‘the person who cares, who leaves no stone unturned before he resigns
himself to the painful decision to withdraw or switch’ (Hirschman 1970,
p. 83). It is clear from Hirschman’s Figure 1, page 87 in that chapter, that
Hirschman thinks that loyalty increases voice. In that figure voice activ-
ity is virtually zero where the non-loyalist would exit the organization,
and voice increases as discontent increases. He suggests that in organiza-
tions with severe initiation ceremonies voice activity will be lower, but as
L oya lt y: at t i t u de or be h av iou r? 71
loyalists consider the exit option seriously voice activity in these organiza-
tions will be greater than loyalists in ‘ordinary’ organizations.
Hirschman (1970, p. 77) says that ‘clearly the presence of loyalty makes
exit less likely, but does it, by the same token, give more scope to voice?’
Voice rather than exit varies with the trade-off of certainty of exit and
the uncertain effects of voice. If they think they might be more effica-
cious then they are more likely to voice: ‘with a given estimate of one’s
influence, the likelihood of voice increases with the degree of loyalty’
(Hirschman, 1970, p. 77). Those who are attached to an organization will
try to think of ways to make themselves influential; and those who are
influential are likely to be attached to the organization: ‘As a rule, then,
loyalty holds exit at bay and activates voice. It is true that, in the face
of discontent, an individual member can remain loyal without being
influential … but hardly without the expectation that someone will act
or something will happen to improve matters’ (Hirschman 1970, p. 78,
emphasis removed).
Loyalty can stop the most quality-conscious people exiting early, so
can perform a valuable social function. The paradox of loyalty accord-
ing to Hirschman is that loyalty becomes most useful when alternatives
are most similar. A loyal person might not exit an organization because
they feel their leaving might cause a decline (Hirschman 1970, p. 98),
especially if the person is influential. This effect might be seen in the
behaviour of former party leaders or influential members of political
parties when the party has clearly taken an ideological direction that is
not theirs. This effect might sometimes be seen with middle-class par-
ents who might be tempted to remove children from a poor school but
fear the school would decline further without their voice efforts. Such
parents might redouble efforts to improve the school before deciding
to exit if the situation does not get better. Such loyalists will continue
to care about the organization even after they have left it (Hirschman,
1970, p. 99).
According to Hirschman such ‘specialist loyalist behaviour’ occurs when
(1) exit will lead to (or cause) further decline, (2) the person continues to
care about the organization after they have left. Hirschman suggests that
the latter condition is evidence that full exit is not possible – caring about
something means you are still part of it somehow:
The concept of public goods makes it easy to understand the notion that in
some situations there can be no real exit from a good or an organization so
that the decision to exit in the partial sense which may be possible must take
into account further deterioration in the good that may result. What becomes
difficult to grasp, in fact, once the concept of public goods is introduced is how
even a partial exit from such goods is possible (Hirschman 1970, pp. 101–02).
For example, one might exit from state education to provide better school-
ing for one’s children, but one cannot avoid any social loss that might arise
from poor state education overall.
Hirschman is clearly not interested in ‘passive loyalty’ (it only seems
to occur when deterioration starts and the loyalist has not yet voiced) but
rather sees loyalty as a quality that can assist the reform of organizations.
Loyalists are willing to tolerate higher levels of dissatisfaction and become
increasingly active as deterioration continues. Hirschman (1970, p. 88)
clearly states that loyalty and voice are closely connected whilst Barry
(1974, p. 195; see also Graham and Keeley 1992) points out that voice ‘is
already built into this concept of loyalty’. We make use of these insights
in our development of Hirschman. We see that loyalty can be represented
as previous investment in an organization or community: loyalty can be
viewed in terms of past voice activity.
The EVLN literature uses Hirschman’s framework both as an ana-
lytic and a predictive tool to examine reactions to dissatisfaction in vari-
ous social relations at work, within and between groups, and in romance.
However, the concepts introduced by Hirschman have been used rather
differently and the operationalizations do not match the original descrip-
tions of Hirschman’s framework. The EVLN model has strengthened the
empirical basis of Hirschman’s original claims. There are, however, import-
ant differences between the EVLN model and the original EVL framework
which go beyond the addition of ‘neglect.’ First, loyalty (and neglect) is
treated as a behavioural response to dissatisfaction rather than a psycho-
logical condition moderating the relation between dissatisfaction and exit/
voice. Loyalty is considered an alternative to exit and voice. Not only is this
conceptually odd, but the various studies demonstrate little empirical sup-
port for this conceptualization.
Loyalty, for example, correlates highly positively with voice and highly
negatively with exit (for example Rusbult and Lowery (1985) report cor-
relations of 0.60 and -0.47). Moreover, whilst the multidimensional
scaling techniques discussed above clearly see responses along the con-
tinuum of the constructive–destructive dimension, the evidence for an
active–passive dimension is mixed (Rusbult and Zembrodt 1983). Thus,
there are doubts as to whether loyalty is really an alternative response.
This problem has been admitted by EVLN researchers who noted that
‘loyalty is less visible than the other responses’ (Drigotas et al. 1995,
p. 596). The constructive–destructive dimension might simply model the
exit–voice dimension and have the constructive–destructive elements
C onc lusions 73
only in the conditions of the empirical situations the EVLN analysts

study.
Exit and voice are clearly not exclusive of each other. The poles oppos-
ite voice and exit are silence and non-exit or stay rather than loyalty and
neglect. There might well be an active and passive dimension but this is
surely related more to personality type, which emerges under certain con-
ditions rather than in terms of the institutional relationship with which
Hirschman is primarily concerned. He identified this dimension with his
concept of ‘alert’ and ‘inert’ people, though these terms and his discussion
suggest that these types of people are those who become aware of a deteri-
oration in quality, rather than having a response to dissatisfaction which
requires notice of such deterioration.
Finally, it is not clear why voice is treated as constructive and exit as
destructive in the EVLN model. This normative addition does not fit at all
with Hirschman’s argument, for he recognized the possible constructive
effects of exit under, for example, market conditions. Similarly people can
voice their dissatisfaction in a considerate, friendly and constructive man-
ner, but can also use voice aggressively and destructively.
Setting conceptual issues aside, there are some methodological issues.
The EVLN studies have all used cross-sectional designs. It is therefore
impossible to establish any temporal differences in the use of exit, voice,
loyalty and neglect. For example people might first try to use voice, and
then resort to exit if their voice proves to be unsuccessful.
Conclusions
This chapter has reviewed the main empirical and theoretical applica-
tions of Hirschman, in particular in the literature on political science. The
EVLN model addresses the same problem as we do. How should schol-
ars utilize Hirschman’s insightful account of exit and voice in such a way
as to be useful in a systematic empirical project? That is, it is a genuine
empirical application rather than using Hirschman’s concepts simply as
metaphors for a set of historical examples. It generates hypotheses that
can be tested with survey or aggregate data. The EVLN model takes a par-
ticular approach to the problem, using psychological categories to generate
the main elements to the model, which can then be used in multimethod
survey-based design. Our review demonstrates the powerful and attractive
nature of this model, and has stimulated our own approach. In particu-
lar, it brings out an important dynamic component in Hirschman that we
attempt to address in Chapter 5. In the end, we wish to distinguish EVLN
from our own approach, mainly because we wish to present a framework
that is closer to Hirschman’s account. By setting out carefully Hirschman’s
propositions, theorizing about loyalty as a category in its own right and

setting out different aspects to voice and exit, we have developed a testable
model that does not rely on underlying attitudes such as active and passive
and constructive and destructive.
Having laid the ground, and reviewed the main alternative, we now pro-
ceed to the testing of our modified Hirschman framework.
4
The structure of UK public services
and some simple relationships
Introduction
The modification to Hirschman’s EVL framework introduced in Chapter 2
alters some of the usual expectations about the relationship of exit and
voice, partly because of the different circumstances under which the trade-
off arises. We do not expect all forms of voice to interact with dissatis-
faction in the same manner. For that reason we do not expect the same
trade-offs to exist between all forms of voice and all forms of exit. Our
conceptualization and operationalization of the concept of voice enable
us to measure the degree of loyalty and to draw out expectations simi-
lar to Hirschman’s. We expect investment in the community to increase
collective voice and it might have a similar impact upon individual voice.
We expect that such investment will reduce intentions to exit, though we
might not expect individual voice to reduce the intention to exit. To test
these expectations we use survey data collected from a sample of internet
users surveyed over a five-year period which we describe in more detail
below.
Hirschman suggests that if exit is made easier then more will exit and
these are likely to be of higher socio-economic status (SES) and hence take
their stronger voices with them. We cannot test for whether more people
exit as exit is made easier, but we can infer that if exiters share the personal
characteristics of voicers, then indeed making exit easier will reduce the
amount of voice left behind. As we have seen, Hirschman argues that some
people are more active than others. His discussion in this regard suggests
that alert people are more aware of the decline in real quality whilst the
inert do not notice. We do not attempt to compare levels of satisfaction
75
76 T h e st ruc t u r e of U K pu bl ic se rv ice s
with objective indicators of quality, so we cannot see if there are types

of people who are alert in this manner. However, the related dimension
discussed especially in the EVLN approach is between active and passive
personality types. We can attempt to examine if there are people who, out-
side of their social class, are more active than others. That is, at any given
satisfaction level and controlling for social class is it the case that some
people are more likely to both voice and exit than others? We can thus
attempt to see whether this active–passive dimension is purely based upon
SES, or whether, controlling for SES, there are some unobserved variables
corresponding to some psychological active–passive dimension. Finally,
social class may affect both exit and voice at the same time in that differ-
ent levels of income may cause individuals to trade off the two differently,
with richer people more willing to exit in the first place, creating a stronger
relationship between voice and exit; if you are poor you are more willing
to trade off the two.
Linked to this idea, there might be people who would like to exit into
the private sector or to another local government jurisdiction but cannot
afford to do so. Hirschman argues that if people become locked in to a
service because exit is too expensive for them, then once dissatisfied they
are more likely to voice that dissatisfaction. In our survey we asked people
to imagine that health insurance would cost them £600 per year and then
to imagine their income has increased by £1,000 per annum. We asked
them under these circumstances how likely it would be for them to take
out health insurance. We asked similar questions about private educa-
tion of parents with children of school age. (See Appendix C for the sur-
vey questions.) In this manner we have tried to see whether Hirschman’s
account of lock-in exists in practice, which allows us to examine its effect
on voice.
Loyalty is supposed to make voice more attractive in comparison to exit.
We measure loyalty by past voice activity, and by a number of social capital
variables, such as feelings of belonging to the neighbourhood and length
of tenure in the neighbourhood. In this way we test how social capital or
loyalty relates to voice. We believe that the greater the social investment or
social capital of a respondent the greater their voice activity.
We might think that once people have decided to exit geographically
they will have less of an interest in attempting to improve matters, so we
expect that those intending to exit will voice less than those who have no
such intention. However, those who are closely connected to a neighbour-
hood might be less inclined to reduce their involvement even if they are
intending to exit. Hence we suggest that intentions to exit will decrease
collective voice activity; but we also think that loyalists who intend to exit
will not reduce their collective voice activity.
T h e nat u r e of t h e data 77
These two implications seem to work in opposite directions. The edu-

cated and rich are more likely to have exit options available, thus decreas-
ing their propensity to voice. On the other hand, voice activity is cheaper
for them, so they are less likely to exit. In order to tease out whether these
relationships hold we need to examine some of these relationships dynam-
ically over time. And if we wish to examine the type of dynamic relation-
ship illustrated in Figure 1.2, then we also need a dynamic framework.
Panel data – behaviour observed of the same individual over successive
waves – enable us to model these relationships statistically; we do this
extensively in Chapter 5.
After explaining the nature of the data and the survey, the chapter
explains the context of our study, public services in the UK. The chap-
ter then reports some basic descriptive statistics such as how many people
exit and voice. It then starts to probe some of the descriptive relationships
between exit and voice, using cross-tabulations of changes over time. This
chapter aims to lay the ground for more complex analysis that follows in
Chapter 5.
The nature of the data

Our data are based on a five-year panel survey of internet users in the UK
conducted by the commercial pollsters, YouGov. YouGov recruits people
to its master panel via those who register by visiting its website (www.
yougov.com); through recruitment via professional third-party recruit-
ers; through specific survey panel research projects; and from a mem-
ber referral scheme. In 2005 YouGov sampled 9,500 users from the bank
of over 100,000, yielding 4,026 responses, a response rate of 42 per cent.
Each year the respondents from any previous year were contacted again,
and the panel topped up to reach our target for each wave by which time
YouGov had a bank of over 250,000 internet users. The numbers we sur-
veyed and our response rates appear in Table 4.1. Respondents are paid a
small amount to take part in each survey they do. The sample is weighted
by YouGov to reflect the UK population. We compare the distribution of
our survey over the five years to the general population in Table 4.2.
Online surveys have become popular with academics as well as indus-
try in the past decade as they are considerably cheaper than face-to-face or
random-digit dialling (RDD) telephone surveys (Dillman 2000). As well
as cost considerations online surveys also have the advantage that they
take place over a shorter time period than similar numbers of face-to-
face interviews. They have the advantage over RDD that longer question-
naires can be adopted. They can avoid certain biases as the respondents
fill the questionnaire in their own time, and are less likely to be led by
Table 4.1 Numbers and response rates during the panel
Response
Number Response rate Wave 1
Total Response from rate Wave 1 respondents
number rate (%) Wave 1 sample (%) (%)
2005 4,026 42 4,026 42
2006 2,619 2,619 27 95
2007 4,952 2,098 22 52
2008 3,468 1,688 17 42
2009 3,664 1,216 12 30
Responded 856 856 9 21
all waves
Total 8,248
Table 4.2 Characteristics of respondents in the panel and in the UK

population
All respondents Waves 1 to 5 Population1

Gender (female) 48.3 53.2 50.8
Average age 44.2 49.9 392
Race (non-white) 7.4 7.0 7.93
South East 15.2 15.5 14 4
No qualifications 8.3 8.8 135
Household annual £32,351 £30,565 £30,0006
income
Notes:
1
Source: Office for National Statistics, Key Population and Vital Statistics
2007 Series VS No 34, PPI No do, available at www.statistics.gov.uk/
downloads/them_population/KPVS2007.pdf.
2
Office for National Statistics, www.statistics.gov.uk/cci/nugget.asp?id=6.
3
Office for Census, April 2001, Office for National Statistics, www.
statistics.gov.uk/cci/nugget.asp?id=273.
4
Office for National Statistics, Social Trends 39, p. 4. www.statistics.gov.uk/
downloads/theme_social/Social_Trends39/Social_Trends_39.pdf.
5
Labour Force Survey, cited in Social Trends, ibid., p. 41.
6
2007/2008, Office for National Statistics, www.statistics.gov.uk/cci/
nugget.asp?id=334.
T h e nat u r e of t h e data 79
an interviewer, whilst anonymity allows for greater honesty overcom-

ing ‘social desirability’ biases where respondents give answers they think
their questioners want. Nevertheless, they have come in for considerable
criticism (Cooper 2000; Schoen and Faas 2005; Malhotra and Krosnick
2007) largely on grounds of sampling and bias. With no master list of
internet addresses it is not possible to draw probability samples and hence
respondents are unlikely to be representative of the population at large.
Despite increased coverage, large numbers of people are not online, and
internet access is not randomly distributed, with disproportionate num-
bers of the elderly and poor amongst those excluded. YouGov tries to
overcome some of these problems by its recruitment process and through
weighting – what used to be called quota sampling, where the survey
company allocates respondents to fill socio-demographic categories until
there is a sample that represents the population. Since it can pay a modest
amount per survey it can capture the poorer households who are online,
which helps representativeness.
Addressing these concerns, Sanders et al. (2007) conducted a survey
experiment as part of the 2005 British Election Study. They compared
the two-wave face-to-face national probability panel survey with a three-
wave internet survey conducted by YouGov. They discovered statistically
significant but not large differences in the characteristics of respondents;
however, when estimating parameters in their voting behaviour models
the face-to-face probability models and the online models reached very
similar results. One difference that might be significant for some of our
results is that reported turnout was higher in the internet surveys than the
face-to-face survey, though reported turnout is always higher than actual
turnout. However, the higher reported turnout might reflect higher real
turnout as the internet respondents seemed more interested in politics
than respondents in face-to-face probability samples.
The experiment of Sanders et al. (2007) enables us to be sanguine about
criticisms of online surveys, though their known defects should make us
cautious about generalizing any point predictions in our survey to that of
the general population of the UK, or anywhere else. It is not in fact our
intention in this book to make any claims about specific features of the
UK, but rather to examine the EVL relationships drawn from our modi-
fied Hirschman framework. That is, we do not make any claims that n%
of the British people are dissatisfied with the NHS or that m% would like
to be able to afford private health care. The only claims we make from our
quasi-experimental design concern whether specific relationships between
exit, voice and loyalty exist in the form Hirschman claims. In that regard
any minor sampling and bias problems within online surveys are largely
beside the point.
The full details of the survey instrument appear in Appendix C at the end
of the book. We asked questions about the use of various welfare services.
We asked questions about voice activities, such as whether a respondent
had made a complaint about a service (individual voice) and what response
the respondent received. We asked about collective voice, such as voting
and forms of participation: whether they had attended a public meeting or
rally, taken part in a public demonstration or protest, met with neighbours
to complain or lobby, or signed a petition. We asked about intended exit
activities. We put various social capital questions, such as group member-
ship and neighbourhood attachment, as well as the usual battery of SES
and personal questions, pertaining to a respondent’s income, education,
ethnicity and sex. We also had information about the location of respond-
ents over the five-year period so could examine actual geographical exit.
We asked some questions about whether people would exit to the private
sector if their income was higher, to examine the lock-in hypotheses. The
survey is designed to draw out in a quasi-experimental setting whether the
EVL framework does elicit the kind of relationships Hirschman suggests.
We do not make any general claims about the representativeness of the
UK as a whole, though we do take our quasi-experimental design to prod-
uce generalizable claims about the EVL framework and regard our survey
as broadly representative of the UK population. Nevertheless, the specific
institutional details of the relevant welfare services in the UK are import-
ant and we discuss these briefly in the next section.
The structure of services in the UK

Our questions largely concern satisfaction with public services in the UK.
Devolution of Scotland and Wales notwithstanding, the UK has a central-
ized welfare state even though many services are provided through elected
local governments. With the exception of Northern Ireland, and some
variations in Scotland, the system of local government is broadly the same.
England has the idiosyncrasy of a two-tier local government system in
some areas, where some functions, such as housing and planning, are car-
ried out by lower-tier district councils. The main services that we examine
are health, education and the local government services in general. These
latter include refuse collection, refuse disposal, street cleaning, local road
repairs (including pavements or sidewalks), housing, dog-catching and
local planning. We ask questions about local government services, first
about whether the respondent used the services, such as household ref-
use collection, environmental services, local culture and leisure services,
local planning services, local transport services, social services and coun-
cil housing. We ask about satisfaction with council services, local schools,
T h e st ruc t u r e of se rv ice s i n t h e U K 81
the local tax (council tax), public leisure facilities, public transport, and
various outcomes closely associated with local government, local pollution
and noise levels and crime.
The NHS is a popular welfare service – as demonstrated by the mas-
sive outpouring of support from British people following attacks on it by
Republicans during the early debate on President Obama’s health reforms
in the summer of 2009. The NHS was set up in 1948 as three services with
separate legislation for England and Wales, Scotland, and Northern Ireland.
From 1969 when powers were transferred to the Welsh office, four systems
have operated. The Welsh system is accountable to the Welsh Assembly
(Welsh Department of Health and Social Services), the Scottish to the
Scottish Government (Scottish Government Health Department), the
Northern Irish to the Northern Ireland Executive (Department of Health,
Social Services and Public Safety), with the English system accountable to
the UK government through the Department of Health. Notwithstanding,
patients are transferable across the systems.
Every UK resident is entitled to belong to a GP (General Practitioner)
practice, which can then refer patients to specialists and for hospital treat-
ment. The NHS has been reorganized several times since its inception,
largely involving increasing the geographical size of the organizational
units. In 1997 the government moved to release hospitals from the control
of regional health authorities and make ‘hospital trusts’ the major surgi-
cal and long-term care units with such trusts operating several hospitals
within cities or local areas. The Liberal–Conservative coalition elected in
2010 has sought to tear up the institutional framework once again, propos-
ing to abolish the trusts. However, there was no major reorganization of
the NHS during the five years of our study, though funding has increased
in real terms (from £44.5bn to £94.5bn, more than doubling in monetary
terms from 2001/02 to 2006/07; see Public Expenditure Statistical Analyses
2008, pp. 26–27); and several central-government inspired initiatives have
been introduced, notably giving greater financial control to GP practices
and increasing patient choice (Le Maistre et al. 2004; Farrington-Douglas
and Allen 2005; Newman and Vidler 2006; Propper et al. 2006; Le Grand
2007; Dowding and John 2009). The idea of increasing choice within the
health service is that patients, with the advice of their GPs, have the right
to choose from which hospital they can receive care. Information on wait-
ing lists and other rating scores are publicly available from a government
website (www.nhs.uk/Pages/HomePage.aspx).
In our survey we ask questions about whether respondents feel they have
been treated within an appropriate timeframe for an illness or injury, and
given a correct treatment for that injury – specific measures of satisfac-
tion with the health service. We ask the same questions of their relations
with their GP about expectations of the correct treatment. Although med-

ical services in the UK are largely provided by the state system, there is a
private system of treatment that runs alongside the public service, often
supplied by the same publicly employed practitioners working privately
for part of their time. So we ask whether the respondent or their employer
pays for private health care other than dental health treatment. We also
ask whether the survey respondent intends to pay for private health insur-
ance in the near future.
In the UK state education is free for all children from the age of five, and
is compulsory until the age of sixteen. The system in Scotland is different in
some ways from that of England and Wales, though not importantly so for
our education satisfaction questions. There are local education authorities
in all areas but these have been of decreasing importance for the past thirty
years, and schools are self-running to a great extent, with control over
appointments and budgets. The precise organization of schools (whether
they are middle schools, grammar schools, etc., see below), selection cri-
teria within central government guidelines and support services are pro-
vided by local education authorities. However, these education authorities
do not have the wide-ranging powers they once enjoyed.
During the period of our survey children faced four ‘key stages’: Key
Stage 1 for ages 5–7; Key Stage 2, 7–11; Key Stage 3, 11–14; Key Stage 4,
14–16. The first two stages constitute primary education; the second two
constitute secondary education. After the age of sixteen further educa-
tion is not compulsory but can be continued in some schools and BTECs,
and is necessary for higher education in universities and colleges (the UK
government has recently raised the statutory leaving age to seventeen).
We ask questions about primary schools (5–11) and secondary schools
(11–16) in our survey.1 Not all secondary schools are the same. In a few
areas examinations at age eleven split children into those that go to gram-
mar schools and those who go to less academic schools. Under the Labour
Government’s education initiative first announced in 2000, academies
have been set up which attract some private sponsorship. However, these
are not private schools and parents do not have to pay fees. Furthermore,
there are regular inspections of schools by the government inspectorate
Ofsted which then publishes the results allowing newspapers to compile
‘league tables’ of schools. Ofsted designates problem schools as ‘failing
1
Some local education authorities provide ‘pre-school places’ to under 5s and ‘rising 5s’
to a greater extent than others. In some areas, there are ‘primary schools’ (5–9), middle
schools (9–12) and secondary (high) schools 13–16. However, we chose to ask about pri-
mary and secondary schools as a default, as the Key Stages should be well understood by
all parents.
T h e st ruc t u r e of se rv ice s i n t h e U K 83
schools’ and these receive direct intervention to improve quality or face

closure.
Under school choice parents may apply to send their children to any
school in their local education department area. However, strict criteria
exist for rationing places in the most popular schools. These involve
catchment areas to give advantage to local children. State-funded and
run church (and latterly ‘faith’) schools exist, which may give precedence
to children of parents who profess the relevant faith; and schools may
specialize in certain areas, such as music, and give places to those tal-
ented in those specialities. Schools are not supposed to give precedence
to children based on general examination results (i.e. there is no gen-
eral ‘selectivity’). The private sector schools take a variety of forms, from
traditional ‘public’ schools like Eton and day colleges charging fees to
specialist schools. Schools in the private sector are not subject to regu-
lation by Ofsted and for that reason quality of education in the private
sector can vary massively.
In our survey we asked questions about satisfaction with secondary
and primary schools, which is where most people send their children. We
asked whether respondents currently or used to send their children to pri-
vate schools. Both with education and health we had questions on the use
of and making a complaint about the service and the respondents’ opinion
about response of the school, education authority, hospital and GP, and
whether their complaint was responded to adequately.
Although devolution has created a variation in the delivery of services
across the UK (see Keating and McEwan 2006), we argue that the basic
opportunities for exit and voice remain relatively constant, partly because
the basic pattern of local government is similar in England, Scotland and
Wales, the reach of the NHS extends across the UK and education has
similarities across England and Wales. The main difference is that England
instituted a performance measurement regime for monitoring services
and offered more choice for health services during our survey under the
Labour governments, which gave more opportunities for exit. The private
sector for health is more developed in England and the central govern-
ment used the private sector to help deliver public services. However, we
did not seek to measure these new choices. Moreover, data analysis on the
England-only data showed no difference in results (see Dowding and John
2011), which might be partly expected because it forms 86 per cent of our
sample, the same ratio as in the UK population. More surprisingly, there is
no large difference in results if the sub-samples of Scotland and Wales are
analysed separately in the same way despite the reduction in numbers and
the differences in administration. So, for both theoretical and practical
reasons, we report results for the whole of the UK.
Understanding the data

Our data form a panel, with many of the same respondents answering
the same questions each year over a five-year period, which has a num-
ber of advantages for data analysis and for understanding why people use
exit and voice in relation to a variety of circumstances, such as dissatis-
faction or the availability of exit. As we discussed in Chapters 1 and 2,
the analysis requires a dynamic analysis to understand what happens to
an individual over time when they experience different situations. Much
of the analysis we present seeks to discover how an individual changes
over time or how past actions such as past voice affect a current or future
action.
To understand these relationships, we present tables of how people change
their attitudes and behaviours over time. This is carried out by stacking the
five surveys so that we have cases for each individual’s response from time
t=1 to t=5, which can then be differenced across the years, t – t−1. Thus a
hypothetical individual might report turning out to vote in local elections
during the five successive annual surveys in the following manner: votes
in year one and year two, decides not to vote in year three, votes again in
year four and finally does not vote in year five. This would create a run of
differences 0, −1, +1, −1, which can be cross-tabulated with another vari-
able similarly differenced.
The other way we present changes is by averaging the responses for all
the respondents to the question. We present in a series of relationships
between three variables measured at different time periods and then
testing for the differences of the pairs of means. In this way, we can see
whether, for example, someone who remains dissatisfied decides to exit
following voicing in a previous time period. In other parts of this chap-
ter, and much more so in Chapter 5, we present regression results, which
also use the stacked data and allow for inferences of changes over time.
Regression allows us to control for some familiar determinants of voice
and exit, such as income, as well as test for the impact of the hypothesized
factors. These panel regressions present some estimation issues, which
we discuss later and in Appendix B. Of course, some relationships occur
just within one time period, and we report some of these in this chapter,
mainly in the text.2
2
For these, we use all the respondents together, pooling the waves, in the stacked version
of the data. This includes the first wave of respondents, then the respondents we added
subsequently. This makes best use of all the responses and saves having to present five sets
of results for each wave. We have checked whether it makes sense statistically by running
regression models with the two variables in the regression and respectively a dependent
and independent variable, controlling for year dummy and then interactions between
De s cr i bi ng e x i t a n d voice 85
Table 4.3 The incidence of exit and voice
Per cent Total

Intention to exit
Intend to move 28.5 18,022
Intend to exit jurisdiction if were to move home 46.3 3,881
Intent to use private health 6.8 17,092
Exit
Moved in year surveyed 5.6 18,562
Use private education 2.6 11,704
Use private health 15.7 18,717
Lock-in
Lock-in to education 27.2 11,397
Lock-in to health 24.7 17,621
Individual voice
Made a complaint during the year of the survey 28.2 18,402
Made a complaint about education 2.9 18,402
Made a complaint about health 7.2 18,402
Collective voice
Local voting 69.2 17,899
Participating in a petition, rally, lobby or 26.7 18,402
demonstration
Describing exit and voice

Before we explore the relationships between exit, voice, loyalty and sat-
isfaction, in Table 4.3 we present some frequencies of the variables that
represent them. These are the total proportion of the total responses indi-
cating whether respondents have carried out an action, like voice or exit,
over the five surveys. This gives an idea of the numbers of people who
decide to exit, those who intend to exit and those who voice, both indi-
vidually and collectively. We then explore how much these relationships
change across the panel dataset in Table 4.4.
Table 4.3 shows the sizable number that report exiting or voicing – or
intended to do so. It needs a few words of explanation. The per cent column
each year and the independent variable (with a reference category). We run a regres-
sion with complaining as the dependent variable, with dissatisfaction, year dummies and
interactions between year and dissatisfaction for years 2, 3, 4 and 5, which shows that
none of the interactions is statistically significant, which suggests we can present the
pooled results in the tables.
Table 4.4 The dynamics of exit and voice
Decrease % No change % Increase % Total

Intention to exit
Intend to move 21.7 56.6 21.7 9,102
Intend to exit 6.9 87.4 5.7 4,876
jurisdiction if
were to move
home
Intent to use 3.6 93.3 3.2 8,585
private health
Exit
Moved in year 94.4 5.7 11,704
surveyed
Use private 0.8 98.5 0.6 5,981
education
Use private health 6.8 87.4 5.8 9,718
Lock-in
Lock-in to education 9.1 82.7 8.2 5,787
Lock-in to health 12.2 78.7 9.1 8,790
Individual voice
Made a complaint 14.8 74.2 11.0 9,589
during the year of
the survey
Made a complaint 2.2 96.3 1.5 9,589
about education
Made a complaint 5.1 90.6 4.4 9,589
about health
Collective voice
Local voting 9.6 82.0 8.4 9,142
Participating 11.5 75.3 13.0 9,589
in a petition,
rally, lobby or
demonstration
shows proportions of the total number of answers across the five years.
Most are based on totals of about 18,000 responses, varying because of
some missing cases across the years or where ‘don’t knows’ are coded as
missing. ‘Intend to exit jurisdiction’ is based on a much smaller number of
respondents because it is only asked of those who intend to move. The use
of private education figure is smaller because the question is asked only of
respondents with children.
De s cr i bi ng e x i t a n d voice 87
The overall findings show that over a quarter of responses reveal an

intention to exit from the respondent’s current residence. If respondents
were to move home, about 40 per cent of them would cross a local gov-
ernment jurisdiction. Much lower numbers report using private health
services, which probably reflects the cost and advantage of using free uni-
versal provision of health care in the UK. This pattern is similarly reflected
in the relatively low numbers intending to exit to private health or educa-
tion, with half as much for education (at about 3 per cent) than for health
(at about 6 per cent). In contrast with the numbers intending to move, only
about 6 per cent moved in the year of the survey. The actual moving fig-
ures contrast with those who feel constrained by their situation, what we
call lock-in; in health and education about 25 per cent of the population
are locked in.
Individual voice is a more frequent activity than exit, but still a major-
ity of respondents do not use it. Over the five years, over a quarter made
a complaint to some public body or another. There are 27 per cent acts of
collective voice, with a majority with claims to have voted. So, from these
statistics, the survey shows that both exit and voice are important activities,
which reflect a set of intentions and frustrations wider than the exit action.
Voice is more common, perhaps because it is less costly and because citi-
zens will choose voice first before exiting. In fact, we test this in Chapter 5.
Important for our discussion of Hirschman is the sense that exit and
voice are not constant over time, but appear in reaction to other factors,
such as dissatisfaction, and in relationship to each other. Table 4.4 shows
the changes in the numbers of respondents indicating whether they have
carried out an action, like voice or exit. This gives an idea of the average
how much, over the five surveys, individuals decide to exit or voice or to
stop doing so. This may be contrasted with the individuals who do not
change their behaviour over successive years.
The decrease column shows the percentages of cases where the respond-
ent stopped doing the activity (or stopped intending to do something or
believing they were locked in). The middle column represents no change
(whether doing the activity or not) and the final column are people who
started doing the activity or changed their attitudes. Most frequencies are
based on totals of about 10,000 responses, varying because of some miss-
ing cases across the years or where ‘don’t knows’ are coded as missing.
Overall the figures show that most people tend not to change their
behaviours over time, with the no-change option ranging from 56.6 per
cent for the intention to move to 98.5 per cent for private education. This is
not surprising because people tend to carry on doing what they do in the
past because their circumstances do not change, the costs of change are
too high or simply from inertia. This also occurs because voicing and exit
may take more than one year. Thus people may intend to exit for a number
of years, but not get round to moving; they may take out private insurance,
which is costly to revoke; or they may put their children into a private
school, where it would be difficult to move them back to the state sector
quickly. People also get into the habit of certain activities, such as voting,
so do not change their behaviour much over time.
Nonetheless, with the exception of change of use of private education,
which is partly explained by the small numbers overall using it, there
are substantive changes occurring across our data. There are also large
changes, such as over intentions to move and the sense of lock-in shown
by the respondents. Changes in individual voice in the form of complaints
show high numbers across the panel. Thus amongst our respondents are
individuals who are changing their use of voice and exit, mostly likely in
response to changes in their circumstances, such as dissatisfaction with
public services, as we shall see later. While the numbers do not domin-
ate the sample – as we expected – they form an important minority of
respondents. These people may give signals to those in public authorities
and who run those services, and marginal changes in the numbers who
exit and voice may have implications for the viability and legitimacy of
those public services.
The dimensions of exit and voice

The variables we discuss above form part of the general categories of voice
and exit, which in turn subdivide respectively into collective and individ-
ual voice and different kinds of exit. We expect the data to reflect these
categories, so if one person tends to use one form of collective voice they
will use another closely related to it. So someone who petitions is more
likely to lobby than someone who does not. Likewise someone who buys
private health is also likely to pay for private education. There is also likely
to be a relationship, if less close, across the subcategories, so that someone
who collectively voices is probably likely to voice individually too. Given
that exit is a different category, we expect less of a correlation between it
and voice, partly because – as Hirschman argues – people may trade off
between the two and they are very different kinds of responses to dissatis-
faction. Even though there will be a subset of the population who will both
exit and voice on average we do not expect these variables to co-vary when
seeing the whole population in the round.
Now of course it is entirely possible that the world is not like that. It
may be the case that underlying both voice and exit are some general
characteristics of individuals that affect their responses to dissatisfac-
tion so they exit and voice in proportion to each other. This disposition
T h e di m e nsions of e x i t a n d voice 89
might be thought of as a dimension of passive and active we discussed

in Chapter 3: passive individuals would prefer to do nothing, avoiding
voice and exit, but remaining dissatisfied with services as a result; active
individuals tend to both voice and exit, but are more satisfied as a result
of their activities.
It is hard to measure this disposition directly – that would need a bat-
tery of questions that could not be included in our survey. Instead we seek
to observe it indirectly, from both voice and exit, to see if there is an under-
lying dimension to the data and whether one underlying factor (active–
passive) represents both voice and exit. The alternative is to say that our
specification is the better one: that exit and voice either form two dimen-
sions to the data, or more than two because of the different categorizations
of voice and exit. To get at this issue we use a familiar technique of data
reduction, factor analysis, to examine the relationship between the vari-
ables in the data. The idea is to construct latent variables from the correla-
tions in the data, which could reflect an active–passive dimension. Here we
use principal components analysis, which produces the underlying factor
by linear combinations of the variables. We present an analysis of our key
measures of voice and exit: voting, going to a rally, lobbying, demonstrat-
ing, petitioning, complaining, intending to move, intending to use private
health, going to a private school and using private health.
Table 4.5 presents the rotated factor scores, and shows the main factors
that load highly on unobserved variables, Factors 1, 2 and 3. Factor 1 shows
that there is strong loading on a good proportion of the variables with a
high eigenvalue of 1.97. However, it is only the voice factors that load heav-
ily on it, both individual and collective voice terms. Voting does not load
highly on Factor 1, which confirms our view that this kind of collective
voice is different from the others. In contrast, individual voice does load
on Factor 1 in the form of complaining. The second factor loads on the
exit variables, on forms of private exit. It does not load on intentions to
move, which confirms our idea that moving between services and moving
between jurisdictions are different kinds of activity. Indeed, the third fac-
tor does load on the intention to move. This factor loads on not much else,
except negatively on the variable of local vote, which is just the type of col-
lective voice that does not load well on Factor 1. In fact, Factor 3 uses voice
and exit variables that do not figure in Factors 1 and 2. Perhaps this factor
indicates how individuals trade off between aspects of exit and voice, just
as Hirschman predicted. This relationship becomes important when we
discuss these variables later in this book, especially in Chapter 5.
If a univariate active–passive dimension had been behind the data, we
might have expected one factor to drive the variables since they would all
load on one or two factors. Instead, the results show that exit and voice
Table 4.5 Rotated factor loadings for voice and exit using principal
components (N =9944)
Factor 1 Factor 2 Factor 3

Local vote 0.1835 0.1211 −0.6527
Rally 0.6014 0.0776 −0.0995
Lobby 0.7163 0.0372 0.0229
Demonstrate 0.5260 0.0450 −0.1505
Petition 0.6829 −0.0545 −0.0293
Complain 0.5280 −0.0152 0.2557
Intend to move 0.0905 0.1038 0.7493
Intend to use private health 0.0514 0.7289 0.1423
Use private health −0.0228 0.7762 −0.1045
Use private education 0.0231 0.4393 0.0695
Eigenvalue 1.97 1.35 1.11
Scores greater than 0.4 are highlighted in bold
work in different ways. Of course this does rule out the application of an
active and passive dimension to attitudes as this itself might have different
dimensions and express itself in different ways for exit and voice. As we do
not directly measure active and passive we cannot rule this account out.
Another way to look at the different kinds of respondents is to parse
them according to whether they use exit or voice, and then see how the dif-
ferent kinds of respondents relate to other kinds of activity and passivity,
such as group membership, being retired (though this is not necessarily a
case of passivity) and being unemployed (which could be seen as enforced
passivity). Table 4.6 shows the divisions of the sample into these categories
(it does not consider those who voted in elections as this is a different kind
of voice and takes up the bulk of the sample).
Comparing combinations of voice and exit

Table 4.6 shows that the largest group is the voicers who do not exit, which
is consistent with our hypothesis that this is a lower-cost form of activ-
ity than exiting, which few people do, but which at least allows people to
respond to problems. The do-nothings are about 40 per cent of the sample,
which is a large number of people. (Note here these are not necessarily pas-
sive or inert since they might all not be dissatisfied.) The exit-only people
is a smaller category than the voice-only people, which reflects the costs of
leaving rather than voicing.
C om pa r i ng c om bi nat ions of voice a n d e x i t 91
Table 4.6 Voice and exit compared
Not voice, not Voice, not Not voice, Voice

exit exit exit and exit
% of sample in 40.7 28.7 16.3 14.2
category (4132) (2912) (1654) (1449)
% dissatisfied 22.9 31.7 29.6 38.3
(4016) (2877) (1610) (1435)
% retired 27.1 30.7 17.1 17.1
(4132) (2912) (1654) (1449)
% unemployed 16.4 18.1 14.3 15.7
(4132) (2912) (1654) (1449)
% member of one 50.1 73.8 56.4 78.1
or more groups (4132) (2912) (1654) (1449)
Raw numbers in parentheses
Table 4.6 shows the four possible types of activity between exit, voice,
both and neither. Exit is defined as those people who did one or more of
the following activities: intended to move, intended to use private health
and used private education; voice comprises those who had gone to a pub-
lic meeting or rally, took part in a public demonstration or protest, met
with neighbours to complain or lobby, signed a petition or made a com-
plaint. The first row is the cell percentages, that is the proportions of the
total of each category of exit and voice. This summarizes the information
contained in Table 4.6, and shows that those doing neither are the lar-
gest category of 40 per cent. The next largest is voice and not exit or want
to exit, which again shows the importance of voice mechanisms occur-
ring without the threat of exit. Exiters, who either voice or not voice, are
a smaller proportion of the total. The next four rows show the propor-
tions within each category, so we can see if there are strong differences
across the subgroups. The figures show there is not much difference in the
categories whether someone is unemployed, dissatisfied or retired (with
the exception that the retired are less likely to exit). There are large dif-
ferences according to whether the respondent is a member of a group or
not, which relates to the link between group membership and voice. In
fact adding these variables to the factor analysis carried out for Table 4.5
finds group membership to load on the first factor, which gives some sup-
port to the idea that Factor 1 might be linked to a form of the active–
passive dichotomy or that group membership is similar to collective voice.
Being unemployed and being retired load negatively on Factor 2, the one
Table 4.7 Factor analysis: voice and exit compared
Variable Factor 1 Factor 2 Factor 3 Uniqueness

Local vote 0.2518 –0.4091 0.1162 0.7558
Rally 0.6332 –0.0938 –0.0790 0.5841
Lobby 0.6239 0.0024 –0.1113 0.5984
Demo 0.6053 –0.0895 –0.0967 0.6162
Petition 0.4972 0.0235 –0.0635 0.7482
Complain 0.4846 0.2350 –0.2020 0.6692
Intend move 0.0606 0.4880 0.0417 0.7565
Go private 0.1797 0.2534 0.6435 0.4894
Private health 0.1779 0.1074 0.7281 0.4266
Private school 0.1001 0.1879 0.3949 0.7987
Group member 0.6259 –0.0563 –0.0420 0.6033
Retired –0.0001 –0.6750 0.0497 0.5420
Unemployed –0.0512 0.4912 –0.4095 0.5884
Dissatisfaction 0.0769 0.4781 –0.0477 0.7633
associated with exit, which probably reflects the constraints on exit rather
than the active–passive dichotomy.3
Some simple relationships

In the rest of this chapter we move from describing voice, exit and sat-
isfaction to some of the simpler relationships we have drawn from the
Hirschman model and then examine the main hypotheses in Chapter 5.
We first look at the most obvious relationship: that dissatisfaction does
indeed increase voice.
Individual voice (for any service) will increase as dissatisfaction

increases (for that service)
If we do not find any relationship between voice and dissatisfaction in
our data then there would be little point in analysing it further! We do,
as one would expect, find a strong relationship between individual voice
and dissatisfaction. For general complaints and overall satisfaction levels
those who are very dissatisfied are more than twice as likely (47 per cent)
3
Adding the other factors in Table 4.6 to the factor analysis from Table 4.5 gives Table 4.7
(constrained to three factor scores).
S om e si m pl e r e l at ionsh i ps 93
Table 4.8 Change in satisfaction by change in individual voice (complaints)
More More
Complain satisfied No change dissatisfied Total
Decrease in voice 15.7 14.6 14.7 14.9
No change 73.3 75.1 72.4 73.9
Increase in voice 11.0 10.3 12.9 11.2
Total 100.0 100.0 100 100.0
N 2,353 3,998 2,250 8,500
Pearson chi = 12.3, p. = 0.015
to self-report individual voice in the previous year than those who are very
satisfied (22 per cent). We find similar results for specific issues, with those
unhappy with secondary schooling five times more likely to voice com-
plaints (10 per cent) as those who are satisfied (2 per cent). Similarly with
national health services, those who are dissatisfied with the timeliness of
treatment are five times more likely to have individually voiced (21 per
cent against 4 per cent) and those dissatisfied with the nature of the treat-
ment six times more likely (25 per cent versus 4 per cent). We also find that
as dissatisfaction increases, one is more likely to complain: 40 per cent of
dissatisfied people make a complaint compared to 24 per cent of satisfied.
Across the panel we see a similar relationship. A change in satisfaction
leads to change in individual voice. Table 4.8 reports this relationship,
showing that as people become more dissatisfied over time, so they are
more likely to complain. Here 11 per cent of those who become more sat-
isfied increased their voice, compared to 13 per cent of those who become
more satisfied, which is not a large difference, but may act as a signal to the
policy-makers if they monitor satisfaction carefully.
Collective voice may show any relationship to general

levels of satisfaction
Collective voice is that type of voice that people do more jointly and is
hence subject to the collective action problem. Furthermore as we have
extensively argued people might use collective voice to defend services
they like as much as to complain about poor services. Hence our hypoth-
esis is almost empirically contentless. Without specifics about the likeli-
hood of defence as opposed to offence in voice activities we have no specific
expectations.
We measure collective voice in two ways. First we measure it by vote:

collective voice vote (CVV); second by participation in collective efforts,
collective voice participation (CVP), indicated by participation in demon-
strations, petitions or lobby groups. Since both CVV and CVP are likely
to be directed at collective goods and such goods may not be salient ones,
that is, they might be ones whose preferred characteristics might be dis-
puted by different groups of people, there might be incentives for people to
collectively voice even if they are satisfied with the services they currently
receive. For example, parents might be very happy with the local school,
but lobby hard if it is threatened with closure. One might be very happy
with the local services but be motivated to vote since one is concerned that
a potential change in the party control of the council might lead to tax cuts
threatening local services. Thus we suggest that in contrast to individual
voice, the amount of collective voice might not be correlated with levels of
satisfaction.
We examine this hypothesis in a number of ways. For example the more
dissatisfied a person is the less likely they are to vote. Thus 76 per cent of
the survey respondents who indicated they were very satisfied said they
voted in a local election compared to 68 per cent of people who were dis-
satisfied. This result might be driven by SES. It might be driven by the dis-
satisfied who feel alienated and hence less likely to vote. We examine these
possibilities below. There is no relationship over time, however, as a change
in satisfaction of respondents over the waves of the survey does not lead to
a change in voting. Both the attitudes and the behaviour are settled rather
than adjust in relationship to each other.
For CVP we examine whether those who are more satisfied and dis-
satisfied are more likely to have lobbied than those neither satisfied nor
dissatisfied, with the satisfied in fact lobbying more. It is well known that
mobilization is easier if interests are threatened than if one is trying to
change things for the better (Hardin 1982, pp. 82–9; Walker 1983; Marwell
and Oliver 1993), indeed David Hume (1740/1978, Bk III, Prt II Sec I, pp.
482) wrote in A Treatise on Human Nature ‘men generally fix their affec-
tions more on what they are possess’d of, than on what they never enjoy’d’.
It is not clear why people seem, in some sense, to lose more utility from
moving from one indifference curve to a lower one than they gain in mov-
ing from the lower to the upper but it is a well-known fact that has led
to prospect theory with its famous ‘S’ shaped utility curves (Tversky and
Kahneman 1974; Kahneman and Tversky 1979). We find similar results
for demonstration, rallies and petitions though not always of a U shape,
but rather linear as satisfaction increases collective voice activities. For
participation in general we find this linear relationship. As dissatisfac-
tion increases CVP, and the more dissatisfied the greater the participation.
S om e si m pl e r e l at ionsh i ps 95
Table 4.9 Change in dissatisfaction by change in intention to exit

geographically
Less More
Intention to exit dissatisfied No change dissatisfied Total
Less likely 23.2 20.9 21.3 21.6
No change 55.3 58.3 55.6 56.8
More likely 21.5 20.8 23.1 21.6
Total 100.0 100.0 100.0 100.0
N 2,263 4,082 2,273 8,618
Pearson chi = 10.56, p. = 0.032
There is not much difference however. We find that 30 per cent of those
who are dissatisfied across the waves tend to participate compared to 28
per cent who do. But confirming the U-shaped relationship once again,
25 per cent of those who are neither satisfied nor dissatisfied participate,
less than at the extremes of satisfaction. There is no relationship across
the panel, however, as there is no link between changes in satisfaction and
changes in participation.
Hence in our survey we seem to find that in some forms of collective
voice, the satisfied voice more; in other forms of collective voice those
who are dissatisfied voice to a greater extent. However, we do not find a
dynamic relationship as the cross-tabulations of changes are not statistic-
ally significant.
Each type of exit will increase with dissatisfaction.

We have suggested that even if one is satisfied with a public service such as
health care, one might still take out private health insurance if one thinks
it provides a superior service. Similarly, of course, one might geographic-
ally exit, or intend to do so, for reasons entirely unrelated to the quality of
local government services. Nevertheless, we expect to see a relationship
between level of dissatisfaction and both private and geographical exit
(including Tiebout) exit.
In Table 4.9 we find that a change in dissatisfaction increases inten-
tions to move and intentions to move across local government jurisdic-
tions, suggesting that dissatisfaction with local government services
encourages geographical exit. Unlike participation, this is a dynamic
relationship as a change in satisfaction increases the intentions to exit
over the panel.
Table 4.10 Change in dissatisfaction by change in intention to exit

jurisdiction
Intention to exit Less More

jurisdiction dissatisfied No change dissatisfied Total
Less likely 9.0 5.3 7.1 6.7
No change 85.8 89.4 86.5 87.7
More likely 5.2 5.3 6.5 5.6
Total 100.0 100.0 100.0 100.0
N 1,258 2,195 1,219 4,672
Pearson chi = 20.32, p. = 0.000
We found that more than half of our respondents who were very dis-
satisfied wanted to move jurisdictions. In Table 4.10 we report attitudes to
moving given changes in satisfication.
Those who become more dissatisfied are more likely to say they will
exit the jurisdiction. But the difference is about two percentage points,
with only a small subsection of the sample who have both changed
their level of satisfaction and have changed their willingness to exit
the jurisdiction. The reality is that exit of a jurisdiction is not a realis-
tic option for many people in the UK. But the differences support our
hypothesis and are consistent with Tiebout exit as we have found in
previous research (John et al. 1995; Dowding and John 1996; Dowding
and Mergoupis 2003; Dowding 2008) in the UK and with other Tiebout-
survey research.
In Table 4.11, which shows the changes in expectations of timely treat-
ment across the panel, we find again that those who become dissatisfied
with the waiting times in the NHS are more likely to want to shift to pri-
vate health care than those who are satisfied.
The numbers changing their wish to go private remain small, but of
course private health care is expensive and the NHS is a comprehensive
health care system. We get similar numbers for those who have been dis-
satisfied with the nature of the health care they have received through the
NHS. This is confirmed when we examine satisfaction in relation to those
who feel locked into the NHS care since they cannot afford private health
care. Those who feel locked in are nearly six times more likely to feel dis-
satisfied with NHS care. In Table 4.12, which shows the changes across
the panel, we find that decreases in the expectations of a timely treatment
lead to an increase in their ascription of lock-in. The change at the extreme
points is only three percentage points, however.
Su m m i ng u p on sat isfac t ion 97
Table 4.11 Cross-tabulation of change in expectation of timely treatment

for an illness or injury by change in intention to private exit
One One Two

Two increase decrease decreases
Intention increases in No in in
to exit in expectation expectation change expectation expectation Total
Decrease 8.0 3.6 3.3 3.6 4.9 3.6
No change 88.3 92.9 93.8 93.0 88.2 93.3
Increase 3.7 3.4 2.8 3.4 6.9 3.1
Total 100.0 100.0 100.0 100.0 100.0 100.0
N 162 2,083 4,260 1,589 102 8,196
Pearson chi = 17.98, p. = 0.021
Table 4.12 Change in expectation of timely treatment for an illness or

injury by change in locked into the NHS
Two One One Two

increases increase increase decreases
in in No in in
Locked in expectation expectation change expectation expectation Total
Become not 16.9 13.7 11.6 11.5 13.3 12.2
locked in
No change 69.7 77.7 80.0 77.9 70.8 78.7
Become 13.5 8.6 8.4 10.6 15.9 9.1
locked in
Total 100.0 100.0 100.0 100.0 100.0 100.0
N 178 2,166 4,413 1,688 113 8,558
Pearson chi = 29.99, p. = 0.000
There are similar results for our other measure of health dissatisfaction,
expectation for the correct treatment for an illness or injury. We also find
similar results in education with parents of both primary and secondary
(high) school children prepared to go private if dissatisfied. Primary school
parents are 6.5 times more likely to want to go private if dissatisfied and
four times more likely for secondary (high) schools.
Summing up on satisfaction
As a way of rounding off the discussion on satisfaction, we seek to com-
pare the responses to dissatisfaction, whether exit or voice. We do this by
Table 4.13 Dissatisfaction, voice and exit
Logit coefficient
of the response to Semi-elasticity
Response to dissatisfaction1 dissatisfaction d(lny)/d(x)
CV vote –.117** –.026**
(.036) (.008)
CV participate .022 .027
(.026) (.031)
IV complain .391*** .430***
(.026) (.029)
IV complain to NHS .571***² .1.086***
(.036) (.067)
IV complain education .014*** .826***
(.002) (.140)
Intend to move .253*** .233***
(.034) (.031)
Intend to use private health .017*** .688***
(.002) (.112)
p. < .01, ** p. < .01 *** p. < .05

Notes:
1
This compares response to health dissatisfaction rather than general
satisfaction.
2
This – and the regression for intent to use private health – is not a
mixed model because the initial values are not feasible.
carrying out a mixed-effects logit regression of dissatisfaction on voice

controlling for household income, where we can compare the coefficients.
Table 4.13 presents the results, both for current dissatisfaction, and for
past dissatisfaction, controlling for current dissatisfaction. We devote
Appendix B to a discussion of statistical methods. Here we note that we
model both the fixed parts of our data (the characteristics of respondents
that change over time, such as the amount of exit, voice and satisfaction)
and the random parts (the aspects of respondents that do not change dur-
ing the panel, which can include covariates like age which can be expected
to affect individual voice, satisfaction and exit). As with the tables we
report, the regressions take account of the panel structure of our data but
they include respondents who did not respond in some waves and those
whom we added to the panel after wave two. Thus the mixed models make
full use of the panel rather than just those who responded in subsequent
S o ci a l cl a s s a n d voice 99
waves and who answered all relevant questions. We also report the semi-
elasticity, which gives us the extent to which a unit of the independent
variable affects the dependent, or more precisely proportional change in y
for a unit change in x, in this case from zero to one. This gives us a measure
of the extent to which voice and exit constitute a response to satisfaction.
From Table 4.13 we see confirmed that individual voice is the strongest
response to dissatisfaction, but stronger for service-specific complaints
about health. The dissatisfied are less likely to vote, but the standard-
ized estimates show the effect is less strong. We did not have any specific
hypothesis with regard to voting and dissatisfaction, expecting it could go
either way given the nature of collective goods. However, as we suggest,
the alienated are also more likely to be dissatisfied and also the least likely
to vote. The table also shows a weak non-significant coefficient for collect-
ive voice participation.
The results for exit are similar, though less strong. Current dissatisfac-
tion drives both intentions to geographically exit and to exit privately,
though past dissatisfaction does not have that effect on geographical exit.
Intentions to exit for private health care show the effect of both current
and past dissatisfaction.
Social class and voice

It is well known that participation in politics is correlated with social
class, with both the better educated and the better-off tending to par-
ticipate more than those with less education and lower incomes. Such
findings go back to the beginning of survey research, as seen in Almond
and Verba (1963), and form the bedrock of the classic studies of political
participation (Milbrath 1965; Milbrath and Goel 1997). In recent years,
this has influenced Verba et al.’s (1995) civic voluntarism model, wherein
differences in social and economic status affect the resources that individ-
uals bring to politics.
With regard to individual voice activity the evidence is more mixed.
There is less of an automatic connection between complaining and socio-
economic status than in the voting or general participation cases (Thomas
and Melkers 1999). Whilst higher socio-economic status groups might be
more able to complain, they use fewer public services, and those they tend
to use are less likely to generate problems. Females and older people are
hypothesized to complain more because they are in the frontline of ser-
vice provision to a greater extent. The better educated and better-off tend
to be less satisfied, though the evidence base is mixed or hard to interpret.
Ross and Mirowski (1984) suggest that lower-class groups tend to acqui-
esce when responding to surveys. Other researchers think that because
the middle classes tend to get higher quality services, they will be satisfied
(Van Ryzin et al. 2004).
In some services, however, real numbers of complainants are higher
among lower social classes, almost certainly reflecting their greater
take-up of these services. In our survey we find that greater voice activity
is associated with higher levels of education and social class: 69 per cent
in households which earn less than £35,000 a year claim to vote compared
with 72 per cent in households in higher income groups. Similarly, 40 per
cent of lower income households participate in at least one act of civic voice
compared to 45 per cent of higher income households. There is a diffe-
rence with individual voice: 30 per cent of individuals in the lower income
households complained compared to 27 per cent of those in higher income
households. This may be because richer households do use the principal
social services. But even if we remove the non-users, the same relationship
exists with 34 per cent of lower income groups (£5,000 and under) mak-
ing a complaint, compared to 27 per cent of the highest income earners
(£75,000 a year and over) – with a linear decrease in complaining for each
higher category of income.
Geographical exiting is also more likely to occur with high income
groups. Twenty-six per cent of people in households with income less than
£35,000 say they are intending to move, compared to 30 per cent of indi-
viduals in households with greater income – not a large difference but there
nonetheless. The relative costs of moving can be lower for high income
groups, though low income groups in privately rented housing often have
much higher mobility rates. For low income groups in publicly assisted
housing (whether council houses or housing trust dwellings) geographical
mobility can be problematic, requiring complex series of ‘swaps’. Richer
people are also more prepared to exit state education and state-funded
health services than poorer people: 4 per cent of people in households
earning £35,000 or less per year compared to 10 per cent of people in richer
households. These results are not surprising, of course: richer people can
afford to exit.
Conclusions
We have shown that the dissatisfied use individual voice as expected, and
that changes in dissatisfaction lead to changes in voice. The relationship
with collective voice is also as expected. We had no particular hypothesis
with regard to the nature of the correlation between collective voice vote
and dissatisfaction. The negative correlation probably shows that the alien-
ated are more dissatisfied and less likely to vote. Collective voice participa-
tion is correlated with dissatisfaction, though not as strongly as individual
C onc lusions 101
voice, again as expected. Collective voice participation is subject to the

collective action or free-rider problem so we should not expect to see such
a strong correlation as with individual voice. As Hirschman argues, dis-
satisfaction is the driver of much public action. Citizens become dissatis-
fied and their first impulse is to voice. We might suppose exit is a further
response to dissatisfaction. We show it is importantly related to dissatis-
faction. We show these dynamic aspects of voice and exit much further in
Chapter 5. Here we have demonstrated the main drivers behind the behav-
iour that form a key part of the EVL model. Dissatisfaction is the motor
behind voice and exit. The task of the next chapter is to show how voice
and exit, and then loyalty, interact with each other.
5
Evidence of the major EVL relationships
Introduction
In this chapter we examine the major EVL relationships. Thus far we have
found that the level of satisfaction that individuals have is associated with
higher voice activities and/or exit or intention to exit. We have examined
some of the determinants of voice to see how these correlate with social
class. We have also probed to see whether there are two personality types –
the active and the passive – with the former having some unobservable
personal characteristics that lead them to be more likely to voice and/or
exit than others whom we can deem passive. These unobservable charac-
teristics are outside of the other determinants of exit or voice behaviour
such as dissatisfaction or social class. In contrast to the EVLN studies we
have found little evidence of such unobservables that lead to both exit and
voice and cannot be reduced to social class characteristics.
In this chapter we will examine specific relationships that Hirschman
hypothesizes would be the case. Here we also examine the dynamic rela-
tionships over time, which so far have been under-examined in the litera-
ture applying Hirschman empirically. In this chapter we will seek to test
out Hirschman’s ideas more comprehensively than has been done previ-
ously. In particular, we see whether exit does in fact drive out voice, exam-
ining the change in relationships over time.
The exit–voice trade-off

The primary claim of Hirschman’s book is that people will trade off
exit and voice. Whilst he recognizes that exit makes voice efficacious in
102
I n t roduc t ion 103
many circumstances, there is often a decision about whether to exit or to

voice. One of his main claims is that making exit easier will reduce voice
through two processes. One is that if it is only rational to exit then no one
would voice and the art of voice will atrophy. We ignore this aspect. The
second, more interesting, claim is that those who are most likely to voice
are also most likely to exit. Active people are more likely to do either of
the positive responses. And Hirschman suggests that objective character-
istics of people, such as their education and socio-economic status, will
determine who is likely to be active and who will find it easiest both to
voice and to exit.
With our data we cannot directly examine whether increasing exit oppor-
tunities decreases voice. We do not have reliable and accurate measures of
changes in exit opportunities in education, health or geographical mobil-
ity over our five-year time frame by which we could see if our respondents
are more inclined to exit. Certainly it is plausible that if the relative costs of
exit decline in relationship to voice then more people will choose the exit
option, and several studies suggest as much (Foreman-Peck 1981; Fornell
and Bookstein 1982; Maute and Forrester 1993; Langston 2002; McKee
et al. 2006; Pettit 2007; Barakso and Schaffner 2008). However, with our
data we can see at any given level of satisfaction whether those who choose
to exit are less likely to voice than those who choose not to exit. And we
can see whether the active can be predicted by objective personal charac-
teristics. We can see whether those who are dissatisfied and choose to voice
and remain dissatisfied are more likely to exit than those whose dissatis-
faction led them to voice and then become satisfied. We can see whether
those who have already decided they are going to exit, those intending to
exit, are less likely to use voice than those who have no such intention.
We can also see, at any given level of satisfaction, whether those who are
unable to exit, those who are locked in to a public service, are more likely
to voice than those who have an exit option even if they do not take it.
The effects of loyalty

We can also examine the effects of loyalty on the exit–voice trade-off. As it
is a psychological variable we cannot measure it directly. However, we have
argued that loyalty is formed through long-term associations with place,
with social investment in institutions, and that it is therefore coupled with
social capital. The longer someone has lived in a community the more con-
nected they will feel with it; if people invest in their community, in their
schools and even in national institutions such as the NHS, they will be
more loyal towards them. Past social investment, such as past collective
voice activity, will therefore also be a measure of loyalty, so we can see if
104 Ev i de nce of t h e m ajor EV L r e l at ionsh i ps
past voice activity is likely to increase current voice activity. Other social
capital variables, such as trust and the number of social groups one belongs
to, will also be associated with loyalty. We can also measure loyalty more
directly by asking people if they feel they belong to their neighbourhood
or support national institutions. We examine these relationships in the fol-
lowing sections.
Voice, satisfaction and exit

One of Hirschman’s claims is that encouraging exit will mean that those
who would voice – whom he calls ‘the alert’ but, following Rusbult, we
term ‘the active’ – are more likely to exit leaving behind the inert or pas-
sive. We cannot directly measure whether making exit easier has this effect
with our data, but we can try to see whether those who would voice are also
more likely to exit under some conditions. We first see whether those who
individually voice are less likely to exit than those who choose to exit at
any given level of satisfaction.
Individual voice and exit

We first see if those who are currently dissatisfied and voice are less likely
to intend to exit than those who are satisfied. We find from the survey that
the dissatisfied are less likely to change to intend to exit if they decide to
voice whereas there is no effect for the satisfied respondents. Among the
dissatisfied respondents, 25 per cent of those who shifted from complain-
ing to not complaining were more likely to intend to move compared to
20 per cent who decided to complain (p. > .1). However, the difference is
only significant at the 10 per cent level and there is no statistically sig-
nificant difference between satisfied and dissatisfied citizens. People both
voice and exit, but not in relation to dissatisfaction. The result might show
some evidence of the active–passive dimension though we do not control
for observable personal characteristics at this stage.
Table 5.1 shows the relationship between voice and exit may be different
for the service-specific forms of action, in this case complaining about edu-
cation. Whilst there is choice in schools, being within the catchment area
of a given school massively increases the probability of being able to send
one’s children there. Hence we might expect to see a relationship between
dissatisfaction, change to individual voice and changes in intention to exit
geographically for schools. Indeed there is a large literature demonstrating
the capitalization of good schools into house prices, which demonstrates
that geographical mobility does respond to school quality in the UK and
elsewhere (Bogart and Cromwell 1997; Clark and Herrin 2000; Cheshire
Voice , sat isfac t ion a n d e x i t 105
Table 5.1 Change in intention to exit geographically by change in

individual voice to schools and satisfaction
Stop Start
Intend to move complaining No change complaining Total
Dissatisfied
Decrease 26.9 22.5 34.4 23.1
No change 36.5 55.8 53.1 54.7
Increase 36.5 21.7 12.5 22.2
Total 100.0 100.0 100.0 100.0
N 52 916 32 1,000
Satisfied
Decrease 25.7 20.6 22.1 20.8
No change 52.3 58.3 46.8 58.0
Increase 22.0 21.1 31.2 21.3
Total 100.0 100.0 100.0 100.0
N 109 4,533 77 4,719
Dissatisfied: Pearson chi = 11.85, p. = 0.018

Satisfied: Pearson chi = 7.38, p. = 0.117
and Sheppard 2003; Rosenthal 2003; Reback 2005; Brasington and Haurin
2006; Clapp et al. 2008).
Table 5.1 is a three-way table, but it is just like the two-way tables we
presented in Chapter 4. It can be read as two sets of relationships between
changes in individual voice and exit, shown first for dissatisfied consumers
and then, in the lower half of the table, for the satisfied. We can compare
the impact of changes in complaining on intentions to move at different
levels of satisfaction by comparing the two halves of the table. In education
the relationship between complaining and intentions to exit is mediated by
satisfaction. For the dissatisfied, we find that those who start complaining
are less likely to decide to move (12.5 per cent as against 36 per cent). There
is not a relationship among the satisfied. If anything, the satisfied are less
likely to intend to move or to voice individually as we would expect, but
the satisfied who have voiced are just as likely to intend to move as those
who are dissatisfied and have not voiced. Again this might simply be pick-
ing up more active people. The satisfied here are what Hirschman called
‘noisy exiters’, who appear quite a lot in our survey as we shall see.
We would not expect to see such incentives to exit geographically within
health care. Whilst health care does vary across regions in the UK, as
regional authorities have some discretion over what resources they direct
to different areas of health care, the new choice agenda within the NHS
means that patients can choose in which hospitals they can have opera-
tions. They also have some discretion over which GP surgery they can sign
up to. Thus geographical exit for better health care would generally imply
moving rather large distances in order to achieve significantly different
health care. Moving out of a large metropolis, like London, to the prosper-
ous Home Counties or further afield might perhaps lead to better options
in GP services and hospital care or moving from a region where GPs are
oversubscribed might improve matters. But those moving short distances
are unlikely to have very different choices in health care. Given that there
are much weaker incentives to move in health care, it is not surprising that
we find no relationship between changes in individual voice and intentions
to exit geographically in health care similar to that for education.
What we would expect in the NHS is a relationship between changes
in private exit and voice in health care. However, the numbers changing
their private voice and intentions to buy health care are small (twenty-
six respondents for the dissatisfied and forty-two for the satisfied).
Nevertheless even with these small numbers it is possible to observe that
among satisfied respondents there is an increased probability of intending
to use private health if the person had started to make a complaint, which
again points to the noisy exit hypothesis (3.8 per cent of those who decided
to stop a complaint compared with 6 per cent of respondents who started
a complaint at p. <.1). There is no relationship among dissatisfied respond-
ents, who may have been more stoical about their fate.
These findings are broadly consistent with Hirschman, but what we
really want to see is the dynamic perspective when citizens actually make
the decision to exit. Do they trade off different types of voice in different
ways? We find that those who have moved house in the previous twelve
months are less likely to have used individual voice over local government
services than those who have not moved in the previous twelve months.
Though we do not report the results here, there is the same relationship
with regard to education though not in health issues. One reason might
be that people Tiebout-exit from a given local authority or from a given
school catchment area (John et al. 1995; Dowding and John 1996). If they
have chosen to move for these reasons then they are more likely to be satis-
fied with the new services – either because they have chosen wisely or per-
haps because problems can take some time to emerge. Often when people
have complaints about a service or a school, these same issues keep arising
over time. However, they might also take time to emerge. We might not see
this pattern in health because as we have suggested above, households are
less likely to Tiebout-exit for health care reasons.
Table 5.2 Satisfaction by geographical exit in current year

and income
Not moved in Moved in

current year current year Total
Low income
Satisfied, neither 71.6 71.5 71.6
Dissatisfied 28.4 28.5 28.4
Total 100.0 100.00 100.0
N 7,383 418 7,801
High income
Satisfied, neither 72.5 77.3 72.7
Dissatisfied 27.5 22.7 27.3
Total 100.0 100.0 100.0
N 6,763 419 7,182
Low income: chi = .002, p. = .96

High income: chi = 4.7, p. = .03
Another reason might simply be that the newly moved have not had
time to complain about anything. The person may be happier with the new
service provider and this may cause her or him to be satisfied, and thus less
likely to complain as a result. At the cross-sectional level we do not find
that having just moved increases satisfaction; however, there is an effect
that is statistically significant when controlling for household income.
Table 5.2 shows this impact using a three-way table: individuals in richer
households become more satisfied after moving, poorer ones do not.
In the survey we only have a few measures that show the transition
between intentions to exit geographically and actual geographical exit. In
the rest of our dynamic analysis of EVL relationships we examine what
people do over different time periods, using lags in the years of the data.
These lags allow us to observe behaviour changes over time periods in
response to changing circumstances. For example, are those who voice fol-
lowing dissatisfaction more likely to intend to move at a later time if they
are still dissatisfied than those who have voiced but have become satisfied?
We analyse these effects by looking at what happens to responses in our
data in one year and compare with responses in the subsequent year. This
process allows us to run tests of difference of means (t tests) to show the
relationship between exit and voice over successive time periods respond-
ing to changes in another variable such as satisfaction for example. Instead
of tables, we present figures to show these dynamic relationships. The
GE means Order Predicted order

D
IV 2.515504 1 1
2.280142 3 3
D
2.45759 2 2
S
~IV
2.219987 4 4
t1 t2
Figure 5.1 Intentions to exit geographically following previous individual voice.
Difference in means of intent to move between satisfied and dissatisfied for indi-
vidual voice is −.235 (t=4.03, p.=0.0); for non IV is −.233 (t=5.3, p.=0.0)
lines in the figures indicate how individuals differ between initial categor-
ies, and then how exit and voice vary in subsequent time periods. These
figures represent the dynamic relationships first described in Chapter 1
(Figure 1.2).
With this format, we can ask whether those who individually voice are
less likely to want to exit if they become satisfied in comparison to those
who are still dissatisfied after they voice. It is this dynamic aspect we are
interested in where there is an interaction between different actions or
intended actions happening at different points in time. We can also ask
questions about the nature of the relationship of these two sets of people
with those who did not use individual voice in the past.
Figure 5.1 shows the mean score of intentions to exit geographically
(GE) of those who used individual voice (IV) and those who did not (~IV)
at time t1. Then one year later at time t2 they were either now satisfied (S) or
were still dissatisfied (D). The predicted order of the means is given on the
right-hand side of the figure (1 is highest, 4 lowest), the actual order next
to the mean scores. We expect those who voice and are still dissatisfied a
year later to be most likely to want to exit geographically. The next likely
to intend to exit geographically are those who are dissatisfied but did not
individually voice the previous year. Of those who are satisfied in year t2
we expect those who voiced in year t1 to be more likely to intend to exit
geographically than those who did not voice in year t1. The second set of
predictions, that the satisfied who did voice also intend to exit, we expect
because we think that voice activity is also a sign of being active or alert,

D
IV 0.116 1 1
0.0757 3 3
D
0.0917 2 2
S
~IV
0.0513 4 4
t1 t2
Figure 5.2 Intentions to exit privately following previous individual voice,
NHS. Difference in means of intent to go private between satisfied and dissatis-
fied for complainers is -.038 (t=2.48, p.=0.013) and for non-complainers is −.043
(t=5.94, p.=0.0)
and such people are more likely to be thinking about moving than the pas-
sive. By inspecting the means we can see that the intentions to exit are as
predicted by the EVL framework.
This is a strong result, especially since our measures of individual
voice cover a range of services in local government, and dissatisfaction
for some people need not mean overall dissatisfaction, though we would
expect such a correlation. With our measures of health, however, we might
expect a stronger result, and indeed we get it. This effect can be seen more
dramatically in Figure 5.2, which concerns complaints about state health
services and intentions to exit to private health schemes. Those who have
individually voiced and are dissatisfied a year later are much more likely to
intend to go to private health care than others, followed by those who did
not individually voice and are still dissatisfied. Voicers who are satisfied
are more likely to want to go private than non-voicers who are satisfied,
which does not necessarily fit expectations, but again might be evidence
of the active–passive dimension. Voicers are active people and so are exit-
ers. Some people might do neither. An analysis of those who have actually
have gone private in the second time period shows a similar strong pattern
though we do not report those results here.
Figure 5.3 shows parents’ exit decisions to private education. It indi-
cates that those who have individually voiced the previous year and are
currently dissatisfied are most likely to intend to send their children to
private schools, with non-voicers currently dissatisfied second most likely

D
IV 0.070 1 1
0.019 2 2
D
0.054 3 3
S
~IV
0.021 4 4
t1 t2
Figure 5.3 Private school if previously complained and did not like state sec-
ondary school. Difference in means of intent to go private between satisfied and
dissatisfied for complainers is −.051 (t=2.01, p.=.022) and for non-complainers is
−.032 (t=5.11, p.=0.0)
to intend to send their children to private school. This is not unexpected:

it might simply show that some people are dissatisfied but do not bother to
try to sort the problem out through voice procedures. It might be the prob-
lems are so great or the problem so immediate they remove their children.
The third most likely to exit are non-voicers who are satisfied, followed by
satisfied voicers. This is not as expected, in the sense that we expect voic-
ers to also be more likely to exit on the grounds that they are the active
or alert, whereas the non-voicers are less active; again, however, what we
might be picking up is people who are not unsatisfied with state schools
(they are ‘satisfactory’) but feel their children would be better off in a pri-
vate school. The voicers are people who are both satisfied and care about
state education, so they get involved in school activities and are the least
likely to send their children to private schools.
Responses to individual voice

We have examined whether those who have become satisfied in a subse-
quent year, when controlling for their individual voice activity are less
likely to intend to exit than those who are still dissatisfied, and discovered
that this is indeed so. However, there is another aspect we can examine,
whether those who have voice believe that the response of those to whom
they complained was adequate. We asked a question (Appendix C) whether
respondents felt that the issue over which they had complained had been
sorted out to their satisfaction.
Table 5.3 Change in intention to exit geographically by change in

individual voice sorted out or not
Change in intention Recently not Recently

to move sorted out No change sorted out Total
Decide not to move 21.4 21.7 22.8 21.7
No change 53.7 56.7 58.7 56.6
Decide to move 24.9 21.6 18.5 21.7
Total 100.0 100.0 100.0 100.0
N 957 7,414 731 9,102
Pearson chi = 10.4346, p. = 0.034
We divided the sample into those who had individually voiced and those
who had not, creating a binary variable. Table 5.3 shows the impact of a
complaint not being sorted out on the intentions to move. Here 24 per cent
of people who had a complaint recently not sorted out decided to move
compared with 18.5 per cent of people whose complaint had just been
sorted out. One problem with Table 5.3 is that many factors may influ-
ence the intentions to move other than a complaint not being dealt with
satisfactorily. Indeed, in earlier work we demonstrated that services are a
bigger factor in where households intend to move to (pull into a location)
than in the intention to move out (push from a location) (John et al. 1995).
There are a large number of factors in a decision, such as level of income,
satisfaction more generally, sense of attachment to a neighbourhood and
various kinds of social investment. To control for these effects we use mul-
tiple regression to predict intentions to move (see Appendix B for more
details about the estimation).
In Table 5.4 we predict intentions to move by household income, satis-
faction with services, knowledge of the neighbours and a sense of belong-
ing to their neighbourhood, for those respondents who individually voice
during the year in question. To this we add a variable called complaints
resolved, which takes positive values for complaints resolved satisfactor-
ily and negative values for complaints not sorted out, which leads to the
prediction that it should have a negative and significant sign with com-
plaints resolved satisfactorily depressing intentions to move, and unsatis-
factorily resolved complaints increasing intentions to move. It should be
noted that that substantive examination of the impact of complaint reso-
lution on the intention to exit is weak as can be seen by the lower coeffi-
cient. As we showed in the last chapter, we can express this in terms of
Table 5.4 Multilevel mixed-effects linear

regression on intention to exit geographically
(voicers only)
Variables Intend to move

Fixed effects
Household income 4.80e-06***
(1.25e-06)
Dissatisfaction 0.0446*
(0.0203)
Know neighbours −0.194***
(0.0423)
Belong to neighbourhood −0.328***
(0.0235)
Complaint resolved −0.0947*
(0.0420)
Constant 3.839***
(0.137)
Random effects
Age 1.97e-12***
9.27e-13
Constant 1.0763***
(.0250)
Log restricted-likelihood −8002.4
Observations 4638
Number of groups 3038
Standard errors in parentheses

*** p<0.001, ** p<0.01, * p<0.05
s emi-elasticity – the percentage change in y resulting from a unit change

in x. For complaint resolution the elasticity is −.01 compared to .477 for
belonging to a neighbourhood.
Now we note that Table 5.4 concerns individual voicers only – there is
no general effect on mobility of complaint resolution. The findings are as
expected: dissatisfaction predicts exit; household income and age both
predict intentions to exit (richer and older people are more likely to intend
to exit). We see that knowing one’s neighbours and belonging to the neigh-
bourhood both predict lower intentions to exit. We see the variable reso-
lution behaves as expected: getting your complaint resolved satisfactorily
reduces intentions to exit; not receiving satisfaction increases intentions
to exit. We have discussed in Chapter 2 and will discuss further below the
effects of loyalty seen as social capital. Here it appears that loyalty in the
form of social capital does make exit less likely as Hirschman predicts.
Furthermore, we see that a complaint being dealt with satisfactorily also
tends to prevent intention to exit. The effect is a strong one and compares
well with the predictors.
These results using different sorts of data and analysis suggest that there
is indeed a relationship between exit and voice. We have shown further-
more, that those who voice and remain dissatisfied are more likely to exit
than those who have not voiced and are dissatisfied. Demonstrating this
dynamic relationship is an important addition to evidence for Hirschman.
Both groups are more likely to exit than those who become satisfied; how-
ever, among the latter group voicers are still more likely to exit, in keeping
with Hirschman’s idea and the results of the EVLN empirics, which found
an active–passive dimension in responses to dissatisfaction.
Collective voice and exit

We do not expect those who collectively voice to be more likely to exit
than those who do not. Collective voicers might be defending services as
they are and so there might be no particular relationship between satisfac-
tion and voice. We can look exclusively at those who are dissatisfied and
collectively voice, as we have done for individual voice. However, we can-
not be assured, as we can with individual voice, that collective voice vote
(CVV), or collective voice participation (CVP) is actually correlated with
any specific dissatisfaction with local government services, health or edu-
cational issues. We did report in Chapter 5 that both CVV and CVP are
positively correlated with dissatisfaction, but people vote and participate
for a variety of reasons and not just because they are dissatisfied. (Though
we note again that the framing literature and prospect theory of social
psychology suggests that people are more likely to be motivated to col-
lectively act to defend interests that are threatened rather than to promote
interests. [See references in Chapter 4.] Empirical evidence on pressure
groups seems to confirm the experimental results in social psychology.
Defending interests that are threatened is not the same as being dissatis-
fied but conceptually does not seem far removed.) There might also be a
positive relationship simply because collective voicers are more active than
non-voicers and because they are active are more likely to exit due to their
personal characteristics. When we look at collective voice and exit we do
indeed find mixed results. We see in Table 5.8 that the dissatisfied are more
likely to intend to exit collectively when dissatisfied, as expected; however,
those who have collectively voiced appear more likely to intend to exit than
those who have not, which is not expected. However, the results are not
statistically significant.
Whilst collective voice is not thought to have any effect on exit or exit
intentions, we might think the opposite relationship might hold. Those
intending to exit a community might have less of an incentive to col-
lectively voice since they will not enjoy any of the benefits of their voice
activity. As long as such voice activity is itself a cost and not a benefit (as
Hirschman suggests it sometimes is) we might expect those intending
to exit geographically to be less likely to participate in political activities
designed to benefit the community. Our general participation measures
are not fine-grained enough for us to expect to be able to measure this
effect on CVP. However, we might expect this effect on local voting (CVV).
If someone is intending to leave a jurisdiction and therefore not expecting
to use its local services, then we should expect them to be less likely to vote
in local elections. Note here our questions are on self-reported voting in
the local elections, which will have occurred a few weeks before the survey
took place. The important relationship is with the current intention to move
affecting the decision made a few weeks earlier rather than the survey on
whether or not to vote. Basically those who intend to move are less likely
to vote in local elections. We see that of those who intend to exit geograph-
ically only 58 per cent self-report voting in the last local election against
75 per cent who have no intention of moving, a difference of 17 per cent
or about 25 per cent higher turnout for those with no intention to move.
We can see the fall is roughly linear as intentions to move strengthen (see
also Dowding and van Hees 2009). We do not find these effects with other
collective voice activities and intentions to move, which is not surprising.
We devote a part of our final section on exit, voice and loyalty to exploring
this relationship in a multivariate context.
Locked in
We would expect to see, at any given level of satisfaction, that those who
are unable to exit – those who are locked into a public service – are more
likely to use voice than those who have an exit option even if the latter
do not actually use it. We would expect this effect to occur with individ-
ual voice for state services. To the extent that some people are less able to
geographically move, we might expect to see higher levels of IV, CVP and
CVV than among those who are able to move. However, since the ability
to exit geographically is highly correlated with social class we might not
expect to be able to tease out these effects.
We also expect that those who feel unable to exit state education or the
NHS are more likely to individually voice (IV) than those who are not
locked in. This is the case very marginally as 4.8 per cent of those who are
locked in are likely to make a complaint compared to 3.5 per cent of those
E f f e c ts of l oya lt y on t h e e x i t–voice t r a de - of f 115
who are not (p.=.003). But the difference is small and there is no effect of
the change in relationships. Similarly, in health 6.6 per cent of those who
are not locked in made a complaint compared to 10 per cent of those who
are locked in (p.=0.00), a larger difference but with still no impact across
the panel. However, it appears that those who feel locked into a public ser-
vice are more likely to voice individually than those who do not feel so
locked in.
Effects of loyalty on the exit–voice trade-off

We have argued that loyalty is correlated with past investment, social cap-
ital and, for local issues, the length of time someone has lived in a com-
munity. We now examine these relationships and also question whether a
self-reported sense of belonging affects the exit–voice trade-off in the man-
ner Hirschman hypothesized. We first examine whether loyalty affects the
probability of exit. We next see if it affects the probability of voice activity.
We then see if we can perceive any marginal effects of loyalty on the exit–
voice trade-off.
Loyalty and geographical exit

First we examine whether the length of time someone has been living in a
community affects the probability that they will intend to exit geograph-
ically. Examining the relationship between ‘intending to exit geographic-
ally’ with length of time that the respondent has resided in a community
shows a strong relationship, with those having lived more than five years at
their current residence being half as likely to intend to move as those who
have not lived there for so long. This relationship remains strong and sig-
nificant among both older (>42 years) and younger age groups (<43), with
younger groups still twelve points less likely to intend to move if they have
lived more than five years in their current home. This suggests that living
in a community does affect a sense of belonging and intentions to exit.
Loyalty and private exit

We have seen that loyalty affects intentions to exit geographically. Does it
have similar effects on private exit? In Table 5.5 we use regression to exam-
ine various effects on private exit in health care.
Table 5.5 demonstrates that dissatisfaction with health drives private
exit as we have seen previously and also that income increases the likeli-
hood of such private exit for the obvious reason that going private is more
feasible for higher income households. Social capital terms do not seem to
Table 5.5 Multilevel mixed-effects linear regression on intention to exit and

actual exit in health care
Intend to use
Variables private health Use private health
Fixed effects
Household income 1.43e-06*** 5.71e-05***
(1.36e-07) (2.66e-06)
Lack of confidence of timeliness of .0178*** 0.272***
treatment (.002)
(0.0372)
Group membership .00276* 0.0800***
(.0014) (0.0244)
Know neighbours .00495 0.0299
(.0046) (0.0852)
Belong to neighbourhood .00111 0.135***
(.0027) (0.0509)
Constant –.0355* –6.573***
(.014) (0.292)
Random effects
Constant .1598 2.612
(.0031) (.0884)
Log likelihood 87.02571 –5230.6
Observations 13304 14508
Number of groups 6,027 6358

*** p<0.001, ** p<0.01, * p<0.05
have any effect upon private exit intentions. Associational links and local
ties do not have any effect upon the intention to privately exit from national
health care. We can see that such ties do not affect private exit in the same
way as geographical exit. This is not surprising. Local ties give people rea-
sons to remain within a community even if they are unhappy with many
aspects of it. Social attachment, even a strong belief in a national health
service, does not necessarily translate into staying within the public sector
for one’s own health care if one is disappointed with the care one receives.
Someone might even vote for a party committed to higher taxes for a
national health service at the same time as paying for private health care,
though obviously privately exiting does give incentives to care less about
the NHS. Group membership, however, is linked to actual exit.
Table 5.6 Multilevel mixed-effects linear regression on use of

private education
Variables Private education

Fixed effects
(4.60e-06)
Poor personal rating of secondary schools 0.166*
(0.0820)
Poor personal rating of primary schools 0.357***
(0.0915)
Group membership 0.0476
(0.0476)
Know neighbours 0.102
(0.184)
Belong to the neighbourhood 0.203
(0.112)
Constant –8.534***
(0.598)
Random effects
Constant 1.8033***
(.0646)
Log likelihood –810.5
Observations 7,446
Number of groups 3,514

*** p<0.001, ** p<0.01, * p<0.05
Are these effects the same for private exit in education? Table 5.6 reports
a regression using satisfaction with primary and secondary schools as the
satisfaction variables.
Again we see that income and the two forms of dissatisfaction predict
private exit in education as we would expect. Belonging to the neighbour-
hood also predicts private exit (at p. < .1), though the other social capital
terms do not. If voice does not work for state schools then parents have
only geographical exit to the catchment area of another school or private
exit. If one is attached to one’s neighbourhood then one might sooner
not move house but rather choose private education. Given that the extra
worth of local schools is capitalized into house prices, liking one’s current
neighbourhood gives a strong reason for staying rather than chancing that
one would not feel so at home in a different part of the city or county.
Now we have examined the effects of loyalty on exit behaviour we shall
see what effect it has on voice activity.
Loyalty and voice

We expect to find that our measures of loyalty should increase voice activ-
ity. We begin by examining whether collective voice in terms of voting
(CVV) is increased by feelings of belonging to the neighbourhood and
other social capital items and then whether collective voice participation
(CVP) is affected by such social investment.
We find that if you are a member of a number of groups then you are
more likely to vote than if you are not; there is the same effect for partici-
pation in political activities such as signing petitions, demonstrating and
taking part in rallies and marches. In other words both kinds of collective
voice activities are encouraged by social capital, a finding that is consist-
ent with a great deal of research on participation in politics. We also find
a similar, though weaker relationship between social investment and indi-
vidual voice. Similar findings are found for those who know their neigh-
bours and feel they belong to the neighbourhood: such people are more
likely to vote, and to participate in other ways.
One traditional measure of social capital has been the degree to which
people say they trust others. We find that trust is correlated with self-reported
turnout and participation: 75 per cent of those who are trusting self-report
voting whereas only 64 per cent of the non-trusting vote. Similarly, 45 per
cent of those who are trusting participate in activities whilst only 38 per cent
of the non-trusting so participate. However, when it comes to individually
voicing, the opposite relationship holds. Here 32 per cent of non-trusting
people have individually voiced in the previous year, whereas only 26 per
cent of the trusting people have individually voiced. Of course, how much
you trust others will be affected by how one has been treated and being badly
treated by public authorities might make one less trusting. (Again we need to
control for satisfaction here.) However, this might constitute some evidence
of ‘Birch loyalty’, that is, where loyal people are less likely to (individually)
voice. Birch (1975), remember, suggests that loyalty might simply result in
people saying nothing as quality declines. Loyalty, in this sense, might have
some relationship to those who trust others and feel less inclined to indi-
vidually voice when they face problems believing that these problems are
not the fault of others and assume that organizations will eventually correct
errors themselves. The trusting, however, seem very willing to take part in
collective acts that are designed to improve social life.
Those who are less trusting seem to complain more to the NHS. It has
long been known that lack of trust in politicians and government reduces
turnout (Miller 1974; Miller and Listhaug 1990; Hetherington 1999; Levi
and Stoker 2000; Bélanger and Nadeau 2005) and there is some evidence
that less trusting people overall are less likely to vote and collectively partici-
pate more generally (Putnam 2000, pp. 251–54; Dekker and Uslaner 2001).
From our evidence it seems that trust, whilst it increases collective voice as
the social capital literature maintains, decreases individual voice. Trust is
thought to increase collective voice because people who have faith in others
are more likely to contribute to collectively beneficial arrangements. The
positive effects of trust on collective participation reach back to a literature
on the social bases of participation (Lane 1959) and in the literature on the
civic culture (Almond and Verba 1963; Verba et al. 1995). Individuals look-
ing out for their own interest, on the other hand, complain because they
have less trust that matters would be put right. Lack of trust entails people
need to look after themselves if there are problems. However, those of a
trustful nature might be more likely to trust that others will put problems
right without their input. Again this demonstrates the importance of our
distinction between individual and collective voice. The motivations for
individual and collective voice might be related (a desire to improve service
quality) but are not identical, and extraneous factors such as degree of trust
in others will affect which different forms of voice will be utilized.
It might seem that with regard to collective voice trust acts like loyalty
in the Hirschman/Barry sense; whereas with regard to individual voice
it acts more like Birch loyalty. It might well be that loyalty in the former
sense is viewed as interventions that are for the common or collective
good. Interventions in the second sense are viewed as for personal gain
and hence loyal people remain silent for the collective good. We suggested
in Chapter 2 that managers in organizations might tend to see interven-
tions as being disloyal perhaps because they are viewed as voices raised to
protect personal or sectional interests. Those raising their voices might see
such interventions as being in the collective good of the organization as a
whole. The distinction therefore between individual and collective motiv-
ations is often difficult to judge and perhaps it is what drives the different
interpretations of voice activity as being loyalist. We do not explore these
ideas further, but again suggest that the individual and collective voice
distinction is a key one. We should also note that there is some controversy
over interpreting responses to ‘trust’ questions in surveys and we will not
explore the trust issue further here either (Glaeser et al. 2000; Fine 2001;
Uslaner 2002; Newton 2007).
We have suggested that past collective voice activity demonstrates a
commitment to community. Thus local voting in the past is likely to lead
to greater voting in the future. We do indeed find this, though the effect
does not really show that past collective voice does lead to future voice. The
correlation might just as well show habit (see Gerber et al. 2003; Franklin
2004) or that the population can be broken down into active and passive
personalities, the former tending to vote and the latter tending not to
(Denny and Doyle 2008).
We found that those who had decided to exit geographically were less
likely to vote in local elections. Would loyalty in the form of social invest-
ment reduce this effect? If a person has decided to exit geographically from
a community, then the results of the local election would not affect them
so much once they have moved. Material self-interest suggests they are less
likely to vote. However, if they care about the community they are leaving,
still feeling loyalty towards it despite their intended move, then they might
still vote even though the local election results will not affect them person-
ally. We find that indeed people are more likely to vote if they belong to a
number of groups. In fact in Dowding et al. (forthcoming) we show that the
number of groups is important for the marginal increase in propensity to
vote up to a limit of three. We argue there that since benefits provided by
local government will not be enjoyed personally by those intending to move,
they have less incentive to vote unless they care about their community in
itself. We demonstrate that the greater the social investment, as measured
by the number of community organizations a person belongs to, the more
likely they are to vote even if they are intending to move. We have shown
that the intention to move will reduce the probability that someone will
vote, suggesting that people take into account the benefits consequent upon
their vote when deciding whether to cast a vote. Those intending to move
are less likely to gain benefits from the representatives for whom they are
voting if they think they will not be in the community in the near future.
We further demonstrate that the effect on turnout of intending to move
tends to wash out, the greater the social investment of an individual. We
suggest that this occurs because the more people invest in society the more
they care about it, and hence have an interest in voting even if they will no
longer be part of the community where they vote. These results work for
intending to vote in both general elections and for local elections, but much
more strongly for the latter, which is consistent with the argument.
Voters do take into account the costs and benefits of voting, but those
benefits need not simply be self-interested ones; and the more one invests
in a community the more one cares about what happens to it. These results
are consistent both with rational actor models of voting broadly under-
stood and with social capital accounts of political participation understood
in terms of participation as a social investment leading to social capital
returns and greater future social investment.
Whilst the fall-off is greater for those belonging to a number of groups

(10 per cent as opposed to 7 per cent) that might merely reflect the higher
starting point, or it might reflect the fact that not all groups that people
belong to are neighbourhood ones. Group membership increases voting as
the social capital literature shows, but there is still a self-interested compo-
nent, which hits more directly for those most likely to vote when their vote
will not affect them personally.
We can examine the local loyalty issue more directly with our question
about neighbourhood attachment. Using knowledge of neighbours as the
loyalty variable, we find the trade-off is as we and Hirschman expect. For
those who know only a few or no neighbours the drop in turnout associated
with intentions to move is twelve percentage points, whereas for those who
know many neighbours the drop is eight percentage points. Both differences
are statistically significant. Looking across the two sub-tables we can show
that difference to voting of knowing many neighbours, holding intentions
to move constant, is also statistically significant. The difference to voting
of knowing neighbours is .13 (t=7.6, p.=0.0) for no intentions to move and
.09 if the respondents are intending to move (t=11.2, p.=0.0). For belonging
to the neighbourhood the impact of knowing neighbours on the vote–exit
trade-off is about the same – approximately ten percentage points.
Exit, voice and loyalty

In this final section we seek to put together the analysis of exit, voice and
loyalty, discuss the impact of different kinds of exit on the different voices
and understand these results within the Hirschman framework. Here we
use multiple regression to control for different aspects of the relationship.
We seek to predict when citizens voice over time, so we present the results
for collective voice vote and collective voice participation, then for individ-
ual voice in terms of complaints. Again we use mixed models measuring
the fixed and random parts of the model. The results show how the differ-
ent factors affect voice.
We first examine local vote as in our investigations above. The regres-
sion seeks to control for the negative relationship between collective voice
vote and geographical exit by accounting for other time-varying factors,
such as household income, dissatisfaction with local services, then the
intention to move and the three social investment variables of group mem-
bership, knowledge of neighbours and belonging to the neighbourhood.
Table 5.7 presents these results. They confirm the hypothesis that inten-
tions to exit geographically reduce voting, which was a feature of the data
in our earlier publications (Dowding and John 2008). All the other vari-
ables work satisfactorily, except that it appears that dissatisfaction does not
Table 5.7 Multilevel mixed-effects logistic

regression: local voting
Variables Local vote

Fixed effects
(2.21e-06)
Dissatisfaction –0.0120
(0.0365)
Intend to move –0.152***
(0.0249)
Group membership 0.286***
(0.0247)
Know neighbours 0.568***
(0.0737)
Belong to the neighbourhood 0.278***
(0.0445)
Constant –0.825**
(0.256)
Random effects
SD (race, gender) 1.13
(.069)
Corr (race, gender) .729
(.170)
Observations 13,772

*** p<0.001 ** p<0.01, * p<0.05
cause exit in this model. This contradicts the finding in the bivariate ana-
lysis reported in Chapter 4 and is also inconsistent with earlier findings of
ours (John et al. 1995; Dowding and John 1996; Dowding and Mergoupis
2003; see also Dowding et al. 1994) where we found that Tiebout-exit does
occur, though only marginally on the push (dissatisfaction with current
services) side, and slightly more to pull people into more efficient jurisdic-
tions. What we show here is that any effects of dissatisfaction are wiped
out by other more important factors in decisions to move. As we discuss in
Chapter 6 geographical exit is not a good signal for service providers over
service quality (see also Dowding and Mergoupis 2003).
Table 5.8 Multilevel mixed-effects Poisson

regression: participation
Variables Participate
Fixed effects
Household income –1.50e-06
(9.24e-07)
Dissatisfaction 0.0739***
(0.0163)
Intend to move 0.0293**
(0.0113)
(0.00787)
Know neighbours 0.139***
(0.0329)
Belong to neighbourhood 0.0990***
(0.0201)
(0.115)
Random effects
Constant .790***
(.022)
Observations 14,022

*** p<0.001, ** p<0.01, * p<0.05
So the Hirschman model works in dynamic perspective with the local vot-
ing decision, which is an important if distinctive aspect of collective voice.
What about other forms of collective voice? Here we take all the different
forms – lobbying, petitioning, rallying and demonstrating – and combine
them into one variable that can range between zero and four. This is a count
variable which can be analysed in a Poisson model. We include the same
fixed variables as for the local vote decision, so in many respects the interpret-
ation of Table 5.8 is the same as for Table 5.7, but with a different dependent
variable and estimator. Also the random effects variables – the demographic
terms – were not significant in this case so they are not included.
As we should expect, dissatisfaction predicts voice: citizens get involved
when they are not happy with public services and this is the driver for this
form of collective action that is directed more to specific goals. We also
find the social investment variables are all positive and significant again
as we should expect from the social capital literature. What might be sur-
prising is that intending to move predicts this form of collective voice.
People who are intending to exit geographically are also people who are
more likely to be involved in collective voice activities. If our measures of
participation were all of local groups this might be surprising. We might
expect that people would be less likely to get involved in local issues if they
are intending to leave the area, much as we find they are less likely to vote.
However, participation is more of a collective activity than voting, and is
likely to have both greater personal selective benefits (Olson 1971): first, it
can be fun as well as helping to improve the world (Hirschman 1982), and
second such participation is more face-to-face than voting and so social
pressures are more likely to operate with such participation than with
voting. Furthermore, our measures of participation do not only include
participation in local forums. Hence moving might have less of an effect
on the incentive to stop participating. What this regression does suggest
evidence of, however, is that some people are more active than others: they
are more likely to both voice and exit than passive people. Hirschman was
concerned that making exit easier would mean the active voicers would
also be more likely to exit, leaving behind the passive or inert with a social
loss of voice in a community.
Finally we turn to individual voice. Table 5.9 reports a multilevel mixed-
effects linear regression that brings together the main relationships dis-
cussed so far in the chapter. It again takes a similar form to the other
regressions and is again a Poisson model because the variable counts the
complaints each individual reports making to public authorities. From this
model, we see that being locked in positively predicts individual voice, as we
have seen before, whilst intentions to move and participation also predict
individual voice. Again these latter might signal more active citizens who
are prepared both to use voice and to consider exit if voice does not bring
results.
Our penultimate set of regressions concerns individual voice and the
NHS where we have a range of measures of exit to match against individ-
ual voice. As before, we find that these relationships stand up in the multi-
variate context. Tables 5.10 and 5.11 present regressions seeking to predict
individual voice much in the same way as Table 5.9.
The model works in line with our expectations. Income negatively affects
individual voice because the rich have other options, not least exit to the
private sector. Dissatisfaction affects individual voice as expected. Social
capital resources and group membership also predict individual voice. We
do not necessarily expect such social capital to affect individual voice, but
again might be finding active people who get involved socially and who
also are prepared to complain individually when things go wrong. We also
C onc lusions 125
Table 5.9 Multilevel mixed-effects Poisson regression: individual voice
Variables Complain Complain with lock-in

Household income –7.25e-06*** –9.14e-06***
(1.03e-06) (1.35e-06)
Dissatisfaction 0.256*** 0.253***
(0.0163) (0.0200)
Intend to move 0.0283* 0.0559***
(0.0115) (0.0148)
Group membership 0.162*** 0.184***
(0.00850) (0.0105)
Know neighbours 0.112*** 0.0852*
(0.0326) (0.0419)
Belong to neighbourhood –0.0196 -0.0396
(0.0192) (0.0239)
Education lock-in 0.107*
(0.0445)
Health lock-in 0.179***
(0.0440)
Constant –2.356*** –2.238***
(0.116) (0.146)
Log likelihood –12833.7 –8181.0
Observations 14,022 8,546
Number of groups 6,232 3,877

*** p<0.001, ** p<0.01, * p<0.05
find that in line with the bivariate relationships neighbourhood attachment

reduces individual voice. Again we might explain this as a form of Birch
loyalty that might reduce complaining in some circumstance. Lock-in pre-
dicts individual voice as expected.
Table 5.11 concerns complaints about education. Here we find the vari-
ables behaving as expected but that lock-in does not affect education even
though it does affect individual voice more generally.
Conclusions
We have sought to examine Hirschman’s relationships more fully and
dynamically in this chapter. We have used a variety of different ways to
examine and present the empirical data of our five-year panel study and we
see that most of our results are robust across all techniques. In the main we
Table 5.10 Multilevel mixed-effects logistic regression: individual

voice to the NHS
Complain to
NHS with
Variables Complain to NHS lock-in
Fixed effects
Household income –8.49e-06*** –9.39e-06***
(2.43e-06) (2.46e-06)
No confidence in timely 0.570*** 0.553***
treatment
(0.0368) (0.0375)
Group membership 0.204*** 0.202***
(0.0229) (0.0230)
Know neighbours 0.180* 0.166
(0.0858) (0.0869)
Belong to neighbourhood –0.129** –0.129**
(0.0480) (0.0485)
Locked into health 0.329***
(0.0943)
Constant –5.181*** –5.128***
(0.277) (0.280)
Random effects
Race 1.542*** 1.3889***
(.3219) (.3314)
Constant 1.638 1.627
(.0977) (.0986)
Log likelihood –3441.2 –3352.3
Observations 14,509 13,984
Number of groups 6,358 6,228

*** p<0.001, ** p<0.01, * p<0.05
have confirmed the expected relationships between various forms of exit

and voice in our modified Hirschman framework.
Table 5.12 gives a summary of the relationships found. The types of
exit and then types of voice in the left-hand column can be read as the
independent variable, with the effects of the dependent variables on the
independent written in the second two columns.
We have found many of the relationships expected in our modified
Hirschman framework. We have found that intending to exit geographically
C onc lusions 127
Table 5.11 Multilevel mixed-effects logistic regression: individual

voice in education
Complaint to
Variables education
Fixed effects
Household income –7.79e-06
(4.44e-06)
Poor personal rating of secondary schools 0.310***
(0.0767)
Poor personal rating of primary schools 0.0576
(0.0689)
(0.0368)
Know neighbours 0.145
(0.156)
Belong to neighbourhood –0.332***
(0.0842)
Locked into education 0.0698
(0.168)
(0.502)
Random effects
Constant 1.8090***
(.1784)
Observations 7,219

*** p<0.001, ** p<0.01, * p<0.05
reduces the propensity to vote in local elections. We have found that this
effect is mitigated somewhat if respondents are loyal in the sense that they
have ties to the community, know their neighbours, have lived there a long
time and have collectively voiced in the past (both CVP and CVV). Voice
is traded for exit, but loyalty mitigates some of that trade. This effect is
not uniform for all kinds of voice: intentions to exit increase voice when
it involves participation in acts such as petitions and protest and when
it involves individual acts of complaining. Here is an example of noisy
exit, where people participate even more before they eventually exit, espe-
cially in activities where they can see an individual benefit rather than the
Table 5.12 Summary of main EVL relationships
Response to
dissatisfaction Effect of voice or exit Effect of loyalty
Vote (CV) Intend to move (–ve) +ve
Other collective Intend to move (+ve)_ +ve
Complain (IV) Intend to move (+ve) +ve
Lock-in (+ve)
Complain (IV) to NHS Lock-in (+ve) +ve (group
membership), –ve
(neighbours)
Complain (IV) in Lock-in (n/s) +ve (group), –ve
education (neighbours)
Intend to move Individual voice –ve (know
responded to (–ve) neighbours, belong
Previous dissatisfaction to neighbourhood)
and voice (+ve)
Intend to use private Previous dissatisfaction n/s
health and voice (+ve)
Use private health +ve (groups
and belong to
neighbourhood)
Use private education Previous dissatisfaction +ve (belong to
and voice (+ve) neighbourhood)
or n/s
+ve = a positive relationship, –ve = a negative relationship, n/s = non-significant

relationship
general collective voice of voting. Whereas it does not make sense to vote
once you intend to leave a jurisdiction, it makes more sense to complain
and agitate. In this way, we confirm the distinctions we have made between
different kinds of voice, and pull out voting as where the Hirschman trade-
off works in its classic manner.
We find loyalty can reduce exit and encourage voice. However, it applies
largely to collective voice and not individual voice. Indeed the social cap-
ital aspects of loyalty we have identified are associated with increased geo-
graphical and private exit as well as with individual voice. With regard to
individual voice we seem to be identifying active people who both exit and
voice. Of course, such people, if they truly exit, will not voice in the future.
Someone unhappy with a local school will voice and intend to geograph-
ically or privately exit even as they are voicing, but once their child has
C onc lusions 129
left the school they will no longer exercise individual voice with regard to
that school. Not all relationships with voice and exit work as we expected:
loyalty is associated with exit from the NHS (though not for intentions to
exit).
Thus we have shown that exit, voice and loyalty do affect each other
as Hirschman hypothesized, albeit in more complex ways than he sug-
gested, due to the greater complexity of the concepts of exit, voice and
loyalty that we have established. Hirschman argued that these relation-
ships have important policy consequences. If we make exit easier then
it will impact upon voice. His main claim was that the quality of state
schools could deteriorate if better educated middle-class people took their
children from local state schools to others or to the private sector. In our
rewriting of Hirschman, we find that an important individual voice affect-
ing the local quality of schools would indeed be lost; and an important
collective voice of pressure on politicians through voting and through
other means of participation would be lost as well because parents would
have less incentive to care about the quality of state schools. Our evidence
does not speak directly to this claim; we have only shown that the forces
Hirschman hypothesized are established in our evidence. In Chapter 6 we
examine our modified framework in relation to work in other areas and
take up some of the policy implications of our results to see what we can
conclude about Hirschman’s claims about public policy.
6
Exit, voice and welfare
Applying Hirschman
The ideas in Hirschman’s Exit, Voice and Loyalty have a wide application
to current issues in contemporary political science as well as to many other
disciplines in the social sciences. His framework has been applied to citi-
zens’ reactions to state policy (how do citizens react to state policy in the
form of voting, protest and migration?); members’ reactions to leadership
within political parties (they can voice individually, or collectively through
factions, can leave the party or form a new one); consumers’ reactions to
product change (they can individually voice to the firm, collectively voice
to the firm, as we saw with our opening Coca-Cola example, or simply exit
to other products; where an industry fails, consumers can individually or
collectively voice not only to the industry or firms therein, but to govern-
ment to demand industry regulation, the latter because exit from a whole
industry might be impossible or very costly). Hirschman has also been
applied to relationships within a firm or bureaucracy, to how employees
react to managers’ policies. Most of the extant applications use the EVLN
model we have discussed at length if not here then in Dowding et al (2000).
And of course Hirschman can be applied to specific public services as we
have done in this book. We suggest that the insights that we have provided
in the modified Hirschman account, our three-exit, three-voice and social
investment framework, can be applied in all these areas as well as to urban
and public services.
The EVL framework is able to assess choices made by citizens, organiza-
tions and political entities as they decide what to do in the face of dissat-
isfaction or a problem. The key insight of Hirschman is that there is often
130
A pply i ng H i r s ch m a n 131
not just one response to a change in personal circumstances, but two: voice
and exit. These responses differ in their attributes both in the nature of
their costs and in the character of the benefits they might bring. People
weigh up the relative benefits and time horizons of voice and exit, trading
one for the other both immediately and over time. For Hirschman, it is
the trade-off and how loyalty affects it that is important. In many applica-
tions of Hirschman’s framework the trade-off is forgotten as social scien-
tists only consider exit or voice independently (Dowding et al. 2000). We
have suggested that each of these two types can be further subdivided.
We argue that these subdivisions are not only important because in them-
selves they bring different costs and expected benefits, but because they
also change the relationship between exit and voice, and the role of loyalty,
which we see as social investment.
For Hirschman the purported interrelationship between exit and voice
has important normative implications for how we view democratic and
bureaucratic processes, and how we want to organize our service provision
for citizens. Loss of voice has implications for the representation of interests
and the efficiency of organizations. Increasing exit opportunities might
reduce voice, affecting the efficiency of public sector organizations and, in
so far as our concepts of collective voice are part-and-parcel of democratic
processes, reducing political efficacy and the value of democracy.
In spite of the power of the insights from Hirschman’s framework and
the large amount of literature that does utilize it in some way, scholars
have not realized its full potential. We believe this fact results from the
problems inherent in the original account that we have tried to rectify,
and hope that our extension will also be applicable in other spheres. The
aim of this book has been to restate Hirschman’s ideas in such a way that
they may more easily have wider application. We hope the readers of this
book will be able to see the elements of the framework in a clearer per-
spective and use Hirschman in their own studies. We went back to first
principles, to re-examine the concepts of the original framework in the
terms with which Hirschman defended them. In Chapter 1 we restated this
argument clearly, drawing attention to the complex relationship between
satisfaction, voice, exit and loyalty. We devoted some attention to the last
troublesome concept, partly because more clarity about it can help us
understand what its relationship is to the other factors. We have shown it
to be the most problematic and controversial aspect of Hirschman’s the-
sis, partly because ‘loyalty’ may be understood in several manners. In the
end, our modest innovation to the EVL model conceives of loyalty as a
form of investment into communities and groups that reduces the trade-
off between exit and voice. In that way loyalty can be objectively measured
in behavioural terms.
132 E x i t, voice a n d w e l fa r e
Part of the reason Hirschman has not been empirically applied as widely
as he might (as opposed to used more casually) is that his framework is
too simple. In Chapter 2 we introduced our main innovation, which is to
expand the concepts of voice and exit to differentiate between numerous
different situations that citizens and other decision-makers commonly
face, especially as the world of public services in the twenty-first century
is more complex than that of 1970 when Hirschman’s book was first pub-
lished. We showed that collective voice processes are different from indi-
vidual ones, and should interact with exit processes rather differently. In
some ways individual voice has more forms of expression now than in the
1970s. For example, there are more articulated procedures for individual
voice through complaints procedures than there were, and there are more
ways citizens may make contact with public authorities – face-to-face, let-
ter, telephone, email and through dedicated websites where citizens can
upload comments. Public authorities around the world, but notably in the
UK in the past decade, have worked hard to make complaints procedures
easier, targets for responses have been set and public authorities regularly
run surveys of citizen satisfaction. It is not surprising then to find the fig-
ure of 28 per cent complaining during our survey, a figure that would have
surprised citizens from previous generations (though there is a positive
relationship between age and complaining in our survey).
We suggested that some aspects of Hirschman’s original framework are
problematic. We have argued that there are important differences between
individual voice (IV) and collective voice (CV), and divided the latter into
two aspects, collective voice vote (CVV) and collective voice participa-
tion (CVP). Collective voice involves collaboration with others, through
petitions and various forms of protest. Whilst some forms of collective
action have reduced since the 1970s, notably through trade unions, new
social movements and the proliferation of pressure group organizations
have seen an increase in collective voice directed at government. Collective
voice is more difficult to achieve due to the collective action problem that
Hirschman never clearly recognized in his work. It also tends to be directed
at different types of problem than those addressed by individual voice. The
costs of collective voice participation have been reduced, at least in its ini-
tial stages, through technical advance. Horizontal voice is easier through
social networking sites, Twitter and through the ubiquitous use of mobile
or cell phones. Recent events around the world have demonstrated how the
ease of communication in personal ways between people separated geo-
graphically can help mobilize collective action, massively decreasing the
costs of such mobilization in its early stages. As horizontal voice mobilizes
people it enables them to collectively voice vertically through protests,
marches and other forms of collective action.
We distinguish collective voice in the form of collective action from vot-

ing. Voting is, of course, a collective act in that each person’s vote is effec-
tual only in aggregation with every other vote. It too, famously, suffers
from the collective action problem (Dowding 2008). However, it is more
of an individual effort and therefore might interact with exit differently
from more communal forms of collective voice. Empirically we show this
is the case.
We have argued that theoretically we would not expect to see the same
relationship between dissatisfaction and IV as between dissatisfaction and
either CVV or CVP. Empirically we have shown this to be the case. We
also distinguished between two sorts of exit: geographical exit, possible
when services are provided locally; and private exit from the public sector
to the private sector. We have shown that dissatisfaction does not always
lead to exit in the same manner for each of these types of exit. Importantly
we have shown that IV and CV do not trade off with geographical exit
and private exit in precisely the same manner in all circumstances. We
also introduced the idea of internal exit where consumers move from one
public provider to another, though we do not have the quality of data to
empirically analyse this exit in relation to voice and loyalty.
We also suggest that exit comes in several varieties that need to be dis-
tinguished, for they are likely to differ with respect to how they are traded
with the different forms of voice. Geographical exit is costly and will only
be undertaken for service reasons (what we call Tiebout exit) if the benefits
are high. Exiting from the public to the private sector might also be costly
but the costs and benefits might be easier to quantify for consumers, whilst
internal exits might be almost costless. Each of these sorts of exit might
affect voice in different ways. Internal exits will take individual voice away
from trying to improve specific providers, but might have no effect on col-
lective voice through lobbying or voting in relation to the overall service
provider. Geographical exit might also have no effect on collective voice
if voters remain within the same jurisdiction; and have no effect on the
national voice. The more complex forms of public service delivery have
given greater individual choice over exit from the providers as public agen-
cies have decentralized and given autonomy to the delivery units within
organizations.
Our modification of Hirschman’s account of loyalty was designed to fol-
low Hirschman’s own hypothesis closely – that is, that loyalty will affect
the exit–voice trade-off more in favour of voice – we have argued that loy-
alty should be seen as a form of social investment in a community or a way
of life. Thus we have associated loyalty with standard measures of social
capital, including past voice activities (especially CVV and CVP), and sug-
gested that the object of loyalty will not be the product itself, but rather
something associated with it. So loyalty will be to the local community,

not local collective goods or local government services; to the NHS, not to
a specific service one receives from it, to state education and local schools,
not to the specific educational service one receives locally.
With exit, voice and loyalty so defined, we are able to test hypotheses
on data, using the urban context that had been the context also for the
previously most influential study of Hirschman’s ideas, the EVLN model.
The urban context, as others have discovered (Orbell and Uno 1972; Sharp
1984; Lyons et al. 1992; Lowery and Lyons 1989), provides the ideal labora-
tory in which to apply Hirschman’s ideas, and we use it to test the refor-
mulation of the framework. Citizens in urban areas in the UK at the time
of our study have ready access to a variety of voice opportunities, and can
choose easily between jurisdictions as well as between public and private
and between providers. We learned much from the EVLN approach, in
particular the close attention to the categories in the data, even though
we resist the EVLN typology since, as we argued, it departs too far from
Hirschman’s original framework. We have used insights from formal
models that also point to the categorizations we propose.
Our case study of public services in the United Kingdom in the 2000s
provided a period when citizens observed much change and more choice
available, as well as several services from which to observe these rela-
tionships: health, education and local government. Our dataset collected
over five years allows us to observe our hypotheses in dynamic fashion,
addressing a weakness of many studies. As we argue, exit and voice are not
always rivals. Citizens can choose first to voice and then to exit if voice, in
any guise, does not bring satisfaction. Or indeed, if they choose to exit they
can still choose to collectively voice if they want to re-enter public services,
for example.
We have shown that dissatisfaction drives both private voice and exit,
and is associated with collective voice. We have shown that those who
intend to exit are less likely to collectively voice than those without such
intentions and that once they have exited they are less likely to voice. We
have shown there do seem to be some people who are both more likely
to voice and to exit. These are people we identify as ‘active’ (using a term
from the EVLN studies), who correspond to Hirschman’s notion of the
‘alert’. If these active people leave, they leave behind those who are less
likely to voice – the passive or ‘inert’ – which will mean less voice overall.
We identify active people by objective characteristics such as being more
educated and of higher socio-economic status. We also argue that previ-
ous social investment in the form of voice activity is associated with future
voice activity. Social investment leads to activity, an argument familiar to
those who work on social capital.
The costs of voice and exit do not fall on all equally. Some people are pas-
sive not necessarily because of some personality trait, but simply because
the costs of voice are greater for them. The educated feel more able to voice
complaints, and the costs of collective action might be relatively lower for
the richer and more educated. Wealth can also affect the ability to exit, in
different ways. Moving house is very expensive for home owners, cheaper
for renters, but often very difficult for certain categories of council tenants.
Similarly, some simply cannot afford to move from public sector providers
of health or education to private providers. We identify those who would
like to be able to exit, but are unable to do so for cost reasons. These people
are ‘locked in’ to a given service. We show that this feeling of being locked
in increases both individual and collective voice activity. This finding is
potentially important. The active exiters reduce voice activity because they
take their voices with them; but we show that the dissatisfied who cannot
exit do not give up and stop voicing. Those who feel locked in continue to
be active.
We also show that individual voice activity is strongly related to sat-
isfaction. Not only are the dissatisfied more likely to voice, as one would
expect, but those who have voiced and who are satisfied are more likely not
to intend to exit in the future, and also more likely to voice than those who
have never voiced. All of these findings endorse Hirschman’s apprehension
that increasing the ease of exit, or in the terms of contemporary public pol-
icy ‘choice’, will have a deleterious effect upon voice. Our findings certainly
show that individual voice (IV) reduces when exit is easier for people, but
that collective voice participation (CVP) and voting (CVV) reduces less.
As we argue, people have an incentive to vote even when satisfied, and
whilst the incentive to collectively engage in political activity is lessened
when people exit, past social investment (past collective voice) still incites
people to care and to continue to voice.
We show that loyalty affects the exit–voice trade-off, though in complex
ways. Local loyalty does make CVV more likely and geographical exit less
likely as a response to poor local services. Social capital also makes CVP
more likely and is associated with IV, but it does not make private exit
less likely. Hence geographical exit and CVV do trade off, but CVP and
IV are not traded for private exit. Loyalty does make CVP more likely to
occur, but has the opposite effect on IV. Because there is no trade-off, and
social capital is a mark of active people, such people are also more likely to
exit privately. In this manner we have confirmed Hirschman’s analysis and
addressed his concerns.
We have demonstrated the importance of different types of voice and exit.
CVP participation is not affected by geographical exit largely because many
forms of this participation continue within broader geographical areas, and
because voice can be used for positive as well as negative reasons. Intentions
to move affect CVV – once people have decided to move house they do not
have the same incentives to vote locally since they will be untouched by the
benefits or failings of the government once they are in a new area.
We expect future studies to confirm and extend many of the relation-
ships we observe. In spite of a lot of early academic attention making
incorrect specifications of loyalty, as in the EVLN model, and too simple
conceptions of voice, we suggest in Chapter 2 and in Appendix A (where
we provide a comprehensive list of the studies) that the recent wave of
studies, published since our last review in 2000, is more inclined to test for
changes over time (for example, see Pfaff and Kim 2003), use a more elabo-
rated definition of voice (such as Luchak (2003) and Okamoto and Wilkes
(2008) who use our 2000 formulation), and explore different definitions
of loyalty. In our view, armed with the modifications we have elaborated,
scholars working in this field are ready to carry out tests in a manner that
tests out the core propositions of the framework. We hope our empirical
work shows what kinds of study can be done.
The main limitation upon our study is that most of the effects we
uncover are not large. For example, whilst individual voice reduces once
individuals intend to exit, this effect is small when compared to other fac-
tors affecting voice. The addition of exit is a useful complement to existing
understanding of citizen reactions to dissatisfaction, but the usual covari-
ates of age, education and income do explain more of the variance in terms
of voice than do exit possibilities.
Despite the weak effects, we have tested the modified Hirschman frame-
work, and shown that most relationships we expect exist. The question for
policy-makers, however, is that given these relatively weak effects, how
much need they be concerned about exit–voice trade-offs? Citizens will
voice, both individually and collectively, to the extent they have exit inten-
tions and opportunities. However, Hirschman did not just wish to under-
stand the reactions of citizens. He believed the choices of exit and voice
would send a signal to policy-makers. Exit gives a crude signal that services
are failing; voice gives a more nuanced response, but provides fewer incen-
tives for public servants to respond. Given the introduction of competi-
tion and the breaking up of public monopolies, will exit come to provide a
clearer signal than voice? Our research does not answer that question, but
we can comment on the small numbers in our survey who intend to exit
and the large numbers who are locked in to the services they receive, even
in this era of greater choice of public services.
About 27 per cent of our respondents are locked into health and edu-
cation, whereas the numbers who intend to exit are low. It appears that
the signals provided by voice are still much stronger than those provided
by exit. In terms of intentions to exit, only those intending to move house
T h e c osts of e x i t opp ort u n i t i e s 137
form a sizable proportion at 29 per cent; whilst 16 per cent use private
health care and only 3 per cent use private education. The closest direct
measure of individual voice is complaints and 28 per cent of the cases
report a complaint over the five years. We also know that 43 per cent of
complaints were believed by the complainant to have been resolved to their
satisfaction. This figure suggests a reasonable level of efficiency for individ-
ual voice, given that we cannot make judgements on the reasonableness of
the complaints or how they could expect to be satisfied. Indeed it seems
relatively high, given that many complaints may be hard for public author-
ities to respond to with their resource and legal constraints. We also know
that those who had their complaint sorted out were much less likely to be
dissatisfied (28 per cent) than those whose complaint had not been sorted
out (40 per cent). We know the impact of exit on individual voice is not
a direct one. Intentions to exit predict individual voice (but have a more
complex relationship to collective voice) as dissatisfied individuals both
voice and intend to exit.
Here the context of our study might be part of the explanation of weak
incidence of exit. The period of the 2000s was one of a very large increase
in spending on the services we review here – education, health and local
government. The numbers wishing to exit were probably not high. In
addition, satisfaction was high. It makes more sense for consumers to
think about exit, voice individually and get a good response from their
complaint. In that sense the threat of exit may be part of the tactics of
consumers when they are dissatisfied. After the end of the survey, the
UK and other states entered a period of austerity where it is less likely
that consumers will get the services they want. The CVV result is small
in effect and the noise associated with it given the complexity of issues
during elections gives a far weaker signal to policy-makers than individ-
ual voice. Voice mechanisms operating within local authorities are much
stronger. Thus we know that voters will punish an incumbent for pro-
ducing poor services (Boyne et al. 2009), which is a direct way in which
voice may impact on efficiency. Inefficiency leads to removal from power,
which is a powerful incentive. So long as the mechanisms are in place to
sustain collective voice, then local service provision does have a powerful
motor to efficiency. If local government services become fragmented and
citizens opt for more private provision of services, such accountability
mechanisms are likely to weaken.
The costs of exit opportunities

The relatively weak incidence of choice in our survey is an important lesson
for governments who wish to roll out implementation of policies dependent
on greater choice, such as possibilities for patients to choose their hospital,
and increasing choice of parents over schools which has been an import-
ant policy of governments in the UK since 1980. Other countries have
expanded choice too. In spite of the advantages of matching supply with
demand, are the citizens there in great numbers to take advantage of these
choices? Moreover, as governments have increased the exit/choice oppor-
tunities, have those opportunities really brought welfare gains? Often
adding extra alternatives may also exclude some; those excluded might be
the most valued. Second, market and quasi-market choice might reduce
options (as some providers go out of business) as well as increase them.
What is the real objective – efficiency or wider choice? Third, the most
important aspect of the choice agenda is implementation. Patient choice is
about offering real alternatives that bring advantages to people, not simply
a menu of alternatives. For that reason ‘soft’ expansion in the form of med-
ical practitioners talking through potential procedures and letting patients
have their say in their treatment may well bring much greater benefits
than ‘hard’ expansion of targeted numbers of alternatives (Dowding and
John 2008). Soft expansion requires training and providing incentives for
‘knightly’ behaviour (Le Grand 2003), but is much more important than
the hard expansion which might benefit only a few people. Encouraging
knightly behaviour necessitates ensuring, at the very least, that it is not
discouraged by simplistic targets. GPs and hospitals need to be encouraged
to spend time with their most difficult patients, so data on the social and
medical backgrounds of patients are required. Similarly, judging schools
simply on the basis of their examination results without factoring in the
socio-economic backgrounds of children encourages cream-skimming –
the selection of unproblematic citizens rather than those from poor and
disadvantaged backgrounds.
Fourth, the costs and benefits need to be addressed. Would the money
spent on implementing the choice agenda be better spent elsewhere? More
importantly, what are the costs and benefits to consumers? Choice can be
stressful. In some areas – such as pension policy or major surgery – the ben-
efits of choice to some might be outweighed by the stress it causes others.
Of course, patients always need to give consent for treatment, but doctors
might judge that directing some is preferable to leaving decisions in their
hands. Again the choice agenda might best be left soft, where street-level
bureaucrats are given discretion to decide how best it can be implemented.
The idea that we should not have a public service where ‘one size fits all’
can be applied to the process of offering choice, as well as to the idea of
offering more than one alternative.
Choice cannot be simply judged by the number of options available
(Dowding and van Hees 2009). Policies that simply lead to a menu, enab-
ling the government to state that consumers now have a choice of possible
T h e be n e f i ts of l oya lt y 139
alternatives where once they had none, may add little to public welfare.
Indeed, given the costs of implementation there may be welfare loss.
Rather, choice must be viewed more broadly in terms of the welfare benefits
it might bring, efficiency gains through competition and information, and
the feelings of autonomy it enhances. These are all to be encouraged, but
on the other hand one must be aware of the costs that implementing choice
brings about. Doctors spending more time with patients costs money and
means lower throughput. And choice can be costly for the chooser too.
Choice does not come free and those costs need to be examined carefully.
However, with a fully rounded analysis and the right preconditions in
place, the choice agenda might well enhance the British welfare state in the
years to come.
The benefits of loyalty

At the same time that governments, notably the British government, have
been pressing for greater choice and promoting market solutions to col-
lective problems, they have also been concerned about falling interest and
participation in social life and political processes. For those who have
read Hirschman this is somewhat ironic, since his argument is precisely
that pushing for the first might damage the second. Encouraging people
to relocate geographically to find what they want from public services is
not likely to encourage social investment and loyalty within communities.
Suggesting that parents should have the right to send their children to any
school rather than the local school is hardly likely to help cement local
community spirit. Yet, as our evidence demonstrates, such social invest-
ments encourage collective voice activity without damaging individual
voice, or even affecting intentions to exit geographically. If schools, health
services and local governments provide more equal and better services,
then the incentives to want to internally or geographically exit reduce;
high-quality public services will reduce private exit. For those unable to
afford to privately exit, making it more difficult to exit will make them feel
locked in to the services they use. To the extent there are problems lock-in
encourages individual voice, and lock-in encourages collective voice under
all circumstances.
There is a wealth of evidence that social investment within communities
increases political participation in both collective voice vote (CVV) forms
and collective voice participation (CVP) forms (for example, Verba et al.
1995; Newton 1997; Milner 2001; Stolle and Hooghe 2005). Our survey evi-
dence supports these findings. It is also consistent with earlier work (Van
Vugt et al. 2003) showing that those who are more dependent upon their
community are more likely to use voice. Dependency is the other side of
the dice from investment, but it shares the feature that the services matter
more to the individual. Moreover, we have confirmed Hirschman’s hypoth-
esis that both these aspects of loyalty reduce the desire to exit in all forms.
Social investment does indeed encourage collective voice activities at the
expense of exit. Such social investment does not encourage individual voice
as such, though those who are most socially engaged are also most likely to
individually voice and there is no reason to think that encouraging social
engagement will not enable individual voice too. Hirschman suggests that
voice will atrophy if it is not practised, and engagement in social activities,
especially those directed at social ends (as opposed to mere personal sat-
isfaction or enjoyment), will make people feel more empowered to take
up their personal grievances individually. There is evidence not only that
education and social class affect feelings of efficacy and voice activity, but
also that those who are engaged in social activities (controlling for those
effects) feel more empowered and voice more (Verba et al. 1995).
Exit, voice and welfare

The survey evidence on public attitudes to greater exit opportunities is
favourable but seems based upon people getting the choice they want. A
survey of parents found that 90 per cent were satisfied with the outcome
and 70 per cent were satisfied with the process of school choice (Flatney
et al., 2001, p. 15). Variance in satisfaction depends upon the complexity
of the process as implemented by different local education authorities.
And lower rates of satisfaction were found among parents in London areas
where there are shortages of secondary school places. Similarly in health
care, whilst there is some satisfaction based on greater choice, this pro-
ceeds more from discussions with doctors and information about health
needs.
People do value community: they would prefer good local providers to
having to go further afield. They do see exit as a response to poor quality
rather than something to be desired in itself. As Hirschman argues, exit is
a rather crude binary response to poor quality. Where there many products
in the market and each is composed of only a few characteristics, where exit
is cheap – such as simply reaching for a different item on a supermarket
shelf – and where there are many consumers, then exit provides the most
efficient signal for producers. But where goods and services are complex –
such as comprehensive health care coverage, local government services,
and to a lesser extent education – then exit is harder to interpret. Local
governments cannot judge whether they are providing good services sim-
ply by examining the in- and out-flow of households. Seeing private health
E x i t, voice a n d w e l fa r e 141
care take-up in itself provides a government with only limited information

about whether its expenditure priorities on the NHS satisfy demands.
Where exit is costly, and many other factors enter into exit, which cer-
tainly occurs for geographical exit, the signals provided are noisy and dif-
ficult to interpret. In these cases voice activity is required. Governments,
through party political processes, deliberative forums, referenda and sur-
veys, can discover what the public would like from services. Encouraging
individual voice through efficient complaints procedures can also help
satisfy demands if clients feel empowered. Providing information both
through online and other services, and encouraging doctors, teachers
and front-line bureaucrats to provide advice and information will enable
voice.
We have shown that dissatisfaction drives both private voice and exit,
and is associated with collective voice. We have shown that those who
intend to exit are less likely to collectively voice than those without such
intentions and once they have exited do not voice. We have shown there
do seem to be some people who are more likely both to voice and to exit,
and if these active people leave then those left behind will overall voice
less. Active people in this sense are predicted by objective characteristics
such as level of education and socio-economic status. We have shown that
those who feel locked in are more likely to voice than those that do not.
This is strong evidence that increasing exit opportunities will reduce voice.
It is not the case that public services should respond only to exit or only
to voice. The two processes can mutually support each other. It is not the
case that without the threat of exit voice will not work, since voice proce-
dures do, under British democracy, constitute a threat. Collective voice
vote can throw politicians and parties out of power if they do not deliver.
Collective voice participation puts politicians and bureaucrats under
stress, and individual voice, as long as the complaints procedures follow
due process, can threaten individuals working in the public sector. Voice
works without the need for exit, but that is not to say that exit is irrelevant.
However, voice does provide more nuanced information for public provid-
ers to respond to, whilst to enable exit at the expense of voice will not be
efficient. Furthermore, all forms of exit are less costly for the better off and
the better off also voice more. To increase exit will be at the expense of
voice, and that will be at the expense of the poorest and most vulnerable
people in society.
Appendix A
Summary of the empirical literature
testing EVL
Much of the applied work on EVL neither properly considers the inter-
relationships between the three variables nor examines Hirschman’s
hypotheses. In this appendix we review some examples of empirical
applications that test these hypotheses. In part this appendix is a precis
of our longer review (Dowding et al. 2000), but also includes discussion
of literature published subsequent to 2000. The appendix may be read in
conjunction with Table A.1 below, which summarizes the relationships
being tested, the dependent variables and the authors’ details. This can
lead to the same study being cited twice, where they test a different rela-
tionship within the same study design. Table A.1 should be read with
some caution because the ways in which the concepts of exit, voice and
loyalty have been theoretically constructed and then operationalized in
empirical research means that sometimes findings that seem contradict-
ory may in fact not be so.
Employer–employee relationships
Hirschman’s model has been straightforwardly applied in labour econom-
ics and management studies, boiling the framework down to the simple
exit–voice trade-off. Exit is the decision to quit a job when an employee is
dissatisfied; voice is the articulation of grievances through a trade union.
It is argued that the availability of voice causes workers to exit less fre-
quently. Moreover, the existence of trade unions creates a more negotiated
system of management in firms, which enhances and increases voice activ-
ity. With relatively little theory, and no discussion of the different forms of
voice and exit, an early generation of labour economists developed a simple
142
Su m m a ry of l i t e r at u r e t e st i ng EV L 143
yet powerful model of labour turnover. This model allows investigators to

appraise factors other than wages that cause employees to quit.
Higher dissatisfaction is found to lead to higher exit of workers (Withey
and Cooper 1989; Leck and Saunders 1992) and greater intentions to exit
amongst the workforce. Greater satisfaction of workers seems, however,
to lead to higher voice activities, not, as might be expected, lower ones
(Withey and Cooper 1989; Leck and Saunders 1992). This result might be
due to a conflation of private and collective voice in the analyses, or the
fact that employees only feel able to voice when they trust higher man-
agement – which itself might be expected to be strongly associated with
satisfaction. Greater voice activity is associated with lower intentions to
exit (Lee and Whitford 2008), conforming to expectations from the simple
Hirschman framework.
The third variable, loyalty, is generally considered separately in relation-
ship to voice and to exit. Greater loyalty (generally measured by feelings of
loyalty and perceptions of congruence between the values of the individual
and organization (Davis-Blake et al. 2003)) is associated with lower inten-
tions to exit (Lee and Whitford 2008), lower dissatisfaction (Boroff and
Lewin 1997) and higher voice (Hoffman 2006), both directly and through
representatives (Luchak 2003). However, Boroff and Lewin (1997) find
greater loyalty is associated with lower voice, though they code ‘silence’ as
a separate category from voice and do not consider the effects of exit on
voice, so those dissatisfied and lacking loyalty might simply be voicing and
leaving.
The earlier studies of Freeman and Medhoff (1984) showed that union-
ization (which can be seen as providing potential for voice) leads to lower
exit (see also Miller and Mulvey 1991). Higher levels of pay also, not sur-
prisingly, lead to lower exit (Freeman and Medhoff 1984; Miller and Mulvey
1991) and lower intentions to exit (Lee and Whitford 2008). Unionization
also leads to lower absenteeism (Allen 1984). More complex relationships
are examined by Bendor and Sloane (1998) who argue that bad relation-
ships between workers and management are associated with lower levels
of unionization and lower satisfaction. Lee and Whitford (2008) show that
workers on permanent contracts have higher satisfaction, which is then
associated with lower exit and higher voice activity. Similarly Davis-Blake
et al. (2003) demonstrate that casual contracts lead to lower satisfaction
and then to higher exit and decreasing voice.
Consumers and producers

Given Hirschman’s original setting, it is perhaps surprising that so little
related work is located in the marketing literature. Maute and Forrester
144 Su m m a ry of l i t e r at u r e t e st i ng EV L
(1993) examine consumer complaints within an experimental set-up, set-

ting three variables of the magnitude of dissatisfaction, barriers to exit and
the attractiveness of alternatives created by different air travel experiences.
They find that the greater the dissatisfaction of consumers the greater
their propensity to exit and the less they voice, with more alternatives also
leading to greater exit and lower voice. They also find that loyalty leads to
greater exit and lesser voice, though they measure loyalty as doing noth-
ing in a similar way to the notion of ‘neglect’ in the EVLN literature we
discussed in Chapter 3. Examining perceived efficacy, McKee et al. (2006)
find that there is a positive effect of efficacy on voice and a negative rela-
tionship between exit and voice. In rather different work, narrating the
history of the Rover car company during the inter-war years, Foreman-
Peck (1981) suggests that more competitive markets lead to higher exit and
lower voice.
Social relationships
Outside of the EVLN literature there are studies inspired by the social
dependence literature (Kelley and Thibaut 1978; Rusbult and Van Lange
1996). Van Vugt et al. (2003) use data from a dedicated survey on satis-
faction with local services and assume exit and voice strategies will be
determined by levels of satisfaction and degree of dependence. Exit here
is intention to move and dependence is dependence on community and
services. They find dissatisfaction leads to higher exit, whilst those who are
locked in to their communities display higher voice. Hence dissatisfaction
predicts exit and also voice; people feeling their exit is restricted since they
are forced to stay in the community to fulfil their needs will lead to greater
voice and weaker intentions to move. They find that exit is the dominant
strategy for those who are weakly dependent and voice is used more by
those who are highly dependent. In both cases they found a relatively mod-
est proportion of variance explained in both exit and voice strategies
Comparative politics
Within comparative politics Hirschman’s framework has been applied to
protest and migration, especially from authoritarian regimes (O’Donnell
1986; Scott 1986; Hirschman 1995; Evans 1998; Pfaff and Kim 2003;
Ådnanes 2004; Pfaff 2006; Okamoto and Wilkes 2008). The only one of
these studies with clear findings relevant to Hirschman’s hypotheses is
Okamato and Wilkes, who use the distinction between individual and
collective voice we introduced (Dowding et al. 2000). They operation-
alize these as differences in costs versus opportunities. Okamoto and
Wilkes (2008, p. 351) argue that ‘fight (rebellion) will be a substitute for
flight (emigration) when the costs of exit are high or when the opportun-
ities for exit are few. Conversely, flight (emigration) will be a substitute
for fight (rebellion) when the costs of voice are high and the opportun-
ities are low.’
The EVL framework has also been applied to the behaviour of mem-
bers of political parties (Eubank et al. 1996; Kato 1998; Whiteley and Seyd
2002; Pettit 2007). Kato (1998) finds voice is used by members prior to exit
and Eubank et al. (1996) that loyalty leads to lower levels of exit. Pettit
shows that making exit harder for members increases their voice. Whiteley
and Seyd (2002) use Hirschman’s ideas to examine party members exiting
in relation to high levels of voice – their ‘high intensity participation’. The
same or rather similar factors leading to dissatisfaction cause high-intensity
members to voice even more whilst low-intensity members choose to exit.
Weakening the incentives to participate leads to less efficacy, causing some
to participate less and others to exit.
In another context Barakso and Schaffner (2008) look at the mem-
bership of interest groups and their varying opportunities for voice.
Contrary to most other studies, they show that making exit easier
increases voice, though the context in their study is that where there are
barriers to exit there is also lower democracy, which itself should affect
voice opportunities.
Summing up
We provide a tabular summary of the relationships found in the literature
we have mentioned in this appendix and in other chapters of the book in
Table A.1. It demonstrates clearly that the relationships that Hirschman
identified do exist in numerous fields of academic endeavour. Having said
that, however, the literature on Hirschman has not generally examined the
threefold relationship between exit, voice and loyalty in any great depth or
interrogated his specific hypotheses. The literature has demonstrated, in
completely different contexts, that shifting the costs and benefits of voice
versus exit results in changing the propensity of the dissatisfied to use one
rather than the other. It has also shown that loyalty does tend to be associ-
ated with non-exit and higher voice. The dynamic element of voice and
then exit was generally ignored till the more recent literature, and there is
no doubt that the study of EVL is becoming more sophisticated. Some of
the weak results and counter-intuitive findings might result from the more
complex sets of relationships between the different sorts of voice and exit
we have identified and these might be usefully employed in all fields. The
collective/individual voice distinction is especially important in the fields
Table A.1 Main empirical relationships found in the literature
Independent variable Dependent variable Study

Employer–employee
EVL inferences
Higher dissatisfaction → higher exit (Withey and Cooper 1989;
Leck and Saunders 1992)
Higher dissatisfaction → higher intentions to exit (Lee and Whitford 2008)
Higher satisfaction → higher voice (Withey and Cooper 1989;
Leck and Saunders 1992)
Higher voice → lower intentions to exit (Lee and Whitford 2008)
Greater loyalty → lower intentions to exit (Lee and Whitford 2008)
Greater loyalty → lower dissatisfaction (Boroff and Lewin 1997)
Greater loyalty → higher direct voice (Luchak 2003)
Greater loyalty → higher voice (Hoffman 2006)
Greater loyalty → lower voice (Boroff and Lewin 1997)
Lesser loyalty → higher representative voice (Luchak 2003)
Related inferences
Unionization → lower exit (Freeman and Medhoff 1984;
Miller and Mulvey 1991)
Higher pay → lower exit (Freeman and Medhoff 1984;
Miller and Mulvey 1991)
Higher pay → lower intentions to exit (Lee and Whitford 2008)
Unionization → lower absenteeism (Allen 1984)
Bad relations → lower unionization → lower satisfaction (Bendor and Sloane 1998)
Permanent contracts → higher satisfaction → lower exit and higher voice (Lee and Whitford 2008)
Casual contracts → lower satisfaction → higher exit and decreasing (Davis-Blake et al. 2003)
voice
Consumers–producers
EVL inferences
Higher dissatisfaction → higher exit, lesser voice (Maute and Forrester 1993)
More alternatives → higher exit, lesser voice (Maute and Forrester 1993)
Higher loyalty → higher exit, lesser voice (Maute and Forrester 1993)
Higher exit → lesser voice (McKee et al. 2006)
Lower exit → higher voice (McKee et al. 2006)
Monopoly → higher voice (Fornell and Bookstein 1982)
More competitive market → higher exit (Fornell and Bookstein 1982)
More competitive market → higher exit, lesser voice (Foreman-Peck 1981)
Related inferences
Barriers to exit → lesser exit (Maute and Forrester 1993)
Social relationships
EVL inferences
Dissatisfaction → higher exit (Van Vugt et al. 2003)
Lock-in → higher voice (Van Vugt et al. 2003)
Related inferences
Weak dependence → higher exit (Van Vugt et al. 2003)
Strong dependence → higher voice (Van Vugt et al. 2003)
Exit barriers → higher dissatisfaction (Ellemers et al. 1993)
Comparative politics
EVL inferences
Higher loyalty → lower exit (Eubank et al. 1996)
Higher voice → exit, BUT non-linear as exit
increases
→ voice increases then (Okamoto and Wilkes 2008)
decreases
Table A.1 (cont.)
Independent variable Dependent variable Study

Voice used prior to exit (Kato 1998)
Exit harder → higher voice (Pettit 2007)
Exit easier → higher voice (Barakso and Schaffner 2008)
Related inferences
Barriers to exit → no effect (Okamoto and Wilkes 2008)
Barriers to exit → less democracy (Barakso and Schaffner 2008)
Citizen satisfaction in urban
context
EVL inferences
Higher dissatisfaction → higher voice (Devereux and Weisbrod
2006)
More voice → lower exit (Feld 1997)
Related inferences
Less voice → lower capitalization (Feld 1997)
Higher voice → lower state expenditure (Schaltegger and Küttel 2002)
Higher voice → higher property taxes (Schaltegger and Küttel 2002)
Higher exit → higher sales taxes (Schaltegger and Küttel 2002)
Notes: We make no distinction between studies that use continuous and those that use dichotomous scales. The terms ‘higher’
and ‘lower’ voice and exit, and ‘greater’ and ‘lesser’ for loyalty should be read as higher and lower, or greater and lesser, levels of
exit, voice and loyalty amongst people.
of employer–employee relations, party politics and within civil society–

state relationships. We should not expect that individual exit from parties
or firms should have the same relationship to either individual or collect-
ive voice that migration might have. The nature of the costs and benefits
might differ substantially. We should also expect that the nature of the
constraints on exit opportunities will affect voice differently.
Appendix B
Note on statistical methods
The data for this book come from a five-wave panel. This allows us to
observe how individuals sampled for the survey both voice and exit – and
display other behaviours and attitudes – over the period 2005–2009. As
well as tables of cross-tabulations, we use regression analysis to under-
stand how these relationships change. Recent work by statisticians suggest
that mixed models are appropriate for relatively short panels, where there
is not much variation over time, as is the case with this data, but where
models can take account of individual responses at different time periods
(see Singer and Willett 2009, chapters 4 and 5; Zimmerman et al. 2009).
These mixed models treat time-invariant factors as random effects and
time-varying covariates as fixed.
We estimated these models using the xtmixed suite of commands in
Stata 11.1 (StataCorp 2009), also using mixed logit and Poisson models.
We ran likelihood ratio tests that showed that mixed models are a better fit
than pooled and random effects equivalents.
We tried alternative specifications of these mixed models, in particular
models that had age and education as random effects, the usual predictors
for voice (see Verba et al. 1995) and important for exit. But these terms
were not always statistically significant because the fixed part of the model
had explained much of the variance. Most models use the variables race
and gender.
We sought to model for the error variance because independence was a
strong assumption: the errors might be affected by correlations across the
panel. We tested for four alternative error structures. The lowest BIC value
was for exchange models. For example, Table B.1 summarizes the results
150
Not e on stat ist ic a l m et hods 151
Table B.1 Summary of BIC/AIC values for the variance–

covariance for models of intentions to vote
Independence Exchange Identity
BIC 15843.65 14023.07 14029.22

AIC 15790.47 13955.3 13968.9
Order BIC/AIC 3 1 2
for Table 5.7 in Chapter 5 (the unstructured model did not converge). All
the regressions reported in the book assume independence.
The data is to be stored in the ESRC Data Archive at Essex. The Stata do file
to reproduce the tales is available from Peter John, either at peterchjohn@
gmail.com or from his current work location.
Appendix C
The survey instrument
Exit and Voice as a Means of Enhancing Service Delivery (EVMESD)

Survey
Questionnaire
This research is being conducted on behalf of The London School of
Economics and is designed to identify differences between the regions
of Great Britain.
Respondents will be aggregated by virtue of the region in which they
live. We would like you to confirm your postcode so that we can do this
accurately. All the information you provide is completely confidential
and non-attributable.
What is your postcode?
What year did you move into your current home?
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
152
T h e su rv ey i nst ru m e n t 153
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
Before 1980
Don’t know
Which of the following best describes how you occupy your home?
Own the leasehold/freehold outright
Buying leasehold/freehold on a mortgage or loan
Pay part rent and part mortgage (shared ownership)
Rented from council/local authority
Rented from private landlord
Rented from a housing association
Live rent-free (including rent-free in relative’s/friend’s property)
Other
Don’t know
How likely are you to move from your current home to a new home in
the next two years?
Very likely
Fairly likely
Neither likely nor unlikely
Fairly unlikely
Very unlikely
Don’t know
If you were to move home, would you move within your current city,
district or borough council or to a new local council?
Within current council
To a new council
154 T h e su rv ey i nst ru m e n t
Don’t know
If you were to move home, approximately how far do you think you
would move?
Less than 1 mile
1 mile and less than 5 miles
5 miles and less than 20 miles
More than 60 miles
I would leave the UK
Don’t know
We also need a few details about your previous fixed residence. If you
previously had more than one fixed residence (e.g. a student living away
from home), please answer about the residence at which you spent most
time.
What was the postcode of your previous home? [Please type the full
postcode below, ‘don’t know’ if you cannot remember, or ‘never moved’
if appropriate]
When you moved to your current home from where did you move?
[Please type the name of the last city, town or village in which you
used to live, ‘don’t know’ if you cannot remember, or ‘never moved’ if
appropriate]
If you know the name of your PREVIOUS city, district or borough coun-
cil, please type it in below, otherwise leave it blank. [Optional]
Approximately, what is the distance between your current residence and
your previous residence?
Less than 1 mile
1 mile and less than 5 miles
More than 60 miles
From outside UK
Don’t know/Not applicable
Would you say that you know the names of many, a few, or none of your
neighbours?
Many
A few
None
Generally speaking, would you say that most people can be trusted or
that you need to be very careful in dealing with people?
Most people can be trusted
Need to be very careful in dealing with people

Don’t know
How strongly do you feel you belong to your neighbourhood?
I feel strongly that I belong
I tend to feel that I belong
I neither feel I belong nor do I feel I don’t belong
I tend to feel that I do not belong
I feel strongly that I do not belong
Don’t know
And, how strongly do you feel you belong to the local authority/London
borough area?
I feel strongly that I belong
I tend to feel that I belong
I neither feel I belong nor do I feel I don’t belong
I tend to feel that I do not belong
I feel strongly that I do not belong
Don’t know
Please tell us how much you like each of the following about your neigh-
bourhood – using a scale of 1 to 5 where 1 means ‘do not like at all’ and
5 means ‘like very much’.
Local crime rate
Council services
Local schools
Council tax
Commuting to work
Public transport
Neighbours
Private leisure facilities
Public leisure facilities
Access to schools
Local pollution and noise levels
Access to countryside
1 – Do not like at all
2
3
4
5 – Like very much
An important part of this research is about people’s resources, including
financial. We need to ask about your income. We would like to remind
you that everything you tell us is completely confidential.
Firstly, which of these applies to you?

Working full time (30 or more hours per week)
Working part time (8–29 hours per week)
Working part time (less than 8 hours per week)
Full time student
Retired
Unemployed
Other (not working)
We would like to know both your personal income – that is the amount
of income you are individually responsible for – and, if you pool your
resources with a partner or other family member(s), your household
income – that is the amount of your combined incomes.
Gross personal income is an individual’s total income received from all
sources, including wages, salaries, or rents and BEFORE tax and contri-
butions to national insurance are deducted. Into which of the following
bands does your gross personal income fall?
£1 to £9,999 per year (£1 to £199 per week approximately)
£10,000 to £19,999 per year (£200 to £399 per week approximately)
£50,000 to £59,999 per year (£1,000 to £1,199 per week approximately)
£70,000 a year or more (£1,400 or more per week)
Don’t know
Prefer not to answer
Gross household income is the combined money income of all those
earners in a household including wages, salaries, or rents and BEFORE
tax and contributions to national insurance are deducted. Into which of
the following bands does your gross household income fall?
£1 to £9,999 per year (£1 to £199 per week approximately)
£70,000 a year or more (£1,400 or more per week)
Don’t know
Do you have any children in the following age brackets? [Please tick all
that apply][MULTICODE]
0 to 2 years
3 to 5 years
6 to 10 years
11 to 16 years
17 to 18 years
19 to 23 years
24 years or older
Not applicable – I have no children
From what you know, please indicate how you rate the overall quality of
PRIMARY schools in your area – using a scale of 1 to 7 where 1 means
‘very poor’ and 7 means ‘very good’.
1 –very poor
2
3
4
5
6
7 – very good
Don’t know
From what you know, please indicate how you rate the overall quality
of SECONDARY schools in your area – using a scale of 1 to 7 where 1
means ‘very poor’ and 7 means ‘very good’.
1 – very poor
2
3
4
5
6
7 – very good
Don’t know
BASE: To parents
Are you or have you ever sent any of your children to a fee paying day
school? [Please do not include universities where fees are paid]
Yes – I am currently sending one or more of my children to a fee paying
school
Yes – I used to send one or more of my children to a fee paying school
No – as yet I have not sent my child/any of my children to a fee paying
school
BASE: Parents who are not currently sending a child to a fee paying school
Good private day schools (i.e. not boarding), sometimes called inde-
pendent schools, typically charge fees of £10,000 per child per year.
Imagine you have one child aged eleven years old who is soon to start
secondary school and that you are living where you are now with the
same choice of state schools as now. Imagine also that your household
income increases by £20,000.
Under these circumstances would you send your child to a private school
at a cost of £10,000 per year or would you send them to one of the state
schools local to you?
Private school
State school
Don’t know
Still imagining you have an eleven year old but this time your household
income increases by £10,000.
Private school
State school
Don’t know
Now imagine that your income increases by £5,000.
Private school
State school
Don’t know
BASE: Only to those who would send their child with a £5,000 increase
Now imagine that your household income DECREASES by £5000 …
under these circumstances would you send your child to a private school
Private school
State school
Don’t know
How confident are you that if you were ill or injured and you went to
your local NHS hospital for treatment you would be …
treated within an appropriate timeframe for that injury?

Very confident
Fairly confident
Neither confident nor unconfident
Fairly unconfident
Very unconfident
Don’t know
Given the correct treatment for that injury?
Very confident
Fairly confident
Neither confident nor unconfident
Fairly unconfident
Very unconfident
Don’t know
hospitals in your area – using a scale of 1 to 7 where 1 means ‘very poor’
and 7 means ‘very good’.
1 – Very poor
2
3
4
5
6
7 – Very good
Don’t know
your GP – using a scale of 1 to 7 where 1 means ‘very poor’ and 7 means
‘very good’.
1 – Very poor
2
3
4
5
6
7 – Very good
Imagine private health insurance, which would cover 90 per cent of
health care costs for private GPs and hospitals, would cost you £600 per
year. Major operations, if you chose to have them done privately, would
cost several thousand pounds.
Imagine also that your income increased by £1,000. Under these cir-
cumstances how likely are you to take out private health insurance?
Very likely
Fairly likely
Fairly unlikely
Very unlikely
Don’t know
Imagine also that your income increased by £600. Under these circum-
stances how likely are you to take out private health insurance?
Very likely
Fairly likely
Fairly unlikely
Very unlikely
Don’t know
Imagine also that your income increased by £300. Under these circum-
stances how likely are you to take out private health insurance?
Very likely
Fairly likely
Fairly unlikely
Very unlikely
Don’t know
Imagine also that your income DECREASED by £300. Under these cir-
cumstances how likely are you to take out private health insurance?
Very likely
Fairly likely
Fairly unlikely
Very unlikely
Don’t know
Is your dental care provided by the National Health Service (NHS) or do
you pay privately?
NHS (I usually pay nothing)
NHS (I usually pay some but not all of the charge)
Private
Do not go to a dentist
Don’t know
Do you or your employer pay for private health care and provision other
than dental treatment outside of the NHS? This could include physio-
therapy, osteopathy or alternative treatments.
Yes
No
Don’t know
Do you intend to start paying for private health insurance for treat-
ments other than dental in the near future?
Yes
No
Don’t know
Overall, how satisfied are you with the performance of your local author-
ity services?
1 – very dissatisfied
2
3
4
5
6
7 – very satisfied
Don’t know
Which, if any, of the following services have you personally made use of
within the past 12 months? [Please tick all that apply][MULTICODE]
Local education services (this includes provision of schools for children
and community education)
Household refuse collection (this concerns the collection of domestic
waste)
Local environmental services (this includes street lighting, clean-
ing of streets, environmental health, council-run tips and recycling
centres)
Local cultural and leisure services (this includes parks, sports and leisure
facilities, council libraries, museums, galleries and cultural facilities)
Local planning services (this includes planning land use, approving
planning applications and implementing building regulations)
Local transport services (this includes maintaining roads, bus subsidy
schemes, car parking schemes)
Social services: adult (this includes council day care, home support
(including meals on wheels) and residential care)
Social services: children (this includes family support, child protection,
fostering and adoption services)
Council housing services (this includes management and maintenance
of council-owned homes)
Hospital (this includes emergency, outpatient or any other services)
GP (this includes any visit for check-up, vaccination or any other service)
None of the above
Have you complained about any of the following public services in the
last 12 months? [Please tick all that apply][MULTICODE]
Local hospital (including anything to do with the way in which you or
your family received medical treatment)
Local GP (including anything to do with the way in which your GP
treated you or your family)
Educational service (including anything to do with school services)
Council housing services (this includes management and maintenance
of council-owned homes)
Local education services (this includes provision of schools for children
and community education)
Household refuse collection (this concerns the collection of domestic
waste)
Housing services (this includes helping the homeless, dealing with prop-
erties in disrepair (including providing grants) but does not include
council housing services)
Local environmental services (this includes street lighting, clean-
ing of streets, environmental health, council-run tips and recycling
centres)
Local cultural and leisure services (this includes parks, sports and
leisure facilities, council libraries, museums, galleries and cultural
facilities)
Local planning services (including land use, development) (this includes
planning land use, approving planning applications and implement-
ing building regulations)
Local transport services (this includes maintaining roads, bus subsidy
schemes, car parking schemes)
Social services: adult (this includes council day care, home support
(including meals on wheels) and residential care)
Social services: children (this includes family support, child protection,
fostering and adoption services)
Payment of Housing Benefit and Council Tax Benefit (this includes
administration and payment of these benefits to claimants)
None of the above
BASE: who had complained about ‘local hospital (including anything to do
with the way in which you or your family received medical treatment)’
You have mentioned that you complained about the local hospital …
who did you contact to complain about ‘local hospital (including any-
thing to do with the way in which you or your family received medical
treatment)’? [Please tick all that apply][MULTICODE]
Local councillor
Member of Parliament (MP)
Public official working for local council
Government official
Elected member of the Greater London Assembly
The Mayor of London
Official working for the Greater London Assembly
Teacher or head of school
Official from a public transport company
Other
None of the above
Were you satisfied with the way your complaint was handled?
Yes
No
Has the problem you addressed now been sorted out to your
satisfaction?
Yes
No
Who had complained about ‘local GP (including anything to do with the
way in which your GP treated you or your family)’
You have mentioned that you complained about the local GP …
who did you contact to complain about ‘local GP (including anything to
do with the way in which your GP treated you or your family)’? [Please
tick all that apply][MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘educational service (including anything
to do with school services)’
You have mentioned that you complained about the educational service
…
who did you contact to complain about ‘educational service (includ-
ing anything to do with school services)’? [Please tick all that apply]
[MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘council housing services (this includes
management and maintenance of council-owned homes)’
You have mentioned that you complained about the council housing
services …
Who did you contact to complain about ‘council housing services (this
includes management and maintenance of council-owned homes)’?
[Please tick all that apply][MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘local education services (this includes
provision of schools for children and community education)’
You have mentioned that you complained about the local educational
services …
who did you contact to complain about ‘local education services (this
includes provision of schools for children and community education)’?
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘household refuse collection (this concerns
the collection of domestic waste)’
You have mentioned that you complained about the household refuse
collection …
who did you contact to complain about ‘household refuse collection
(this concerns the collection of domestic waste)’? [Please tick all that
apply][MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘housing services (this includes helping the
homeless, dealing with properties in disrepair (including providing grants)
but does not include council housing services)’
You have mentioned that you complained about the housing services
…
who did you contact to complain about ‘housing services (this includes
helping the homeless, dealing with properties in disrepair (includ-
ing providing grants) but does not include council housing services)’?
Local councillor
Government official
The Mayor of London

Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘local environmental services (this includes
street lighting, cleaning of streets, environmental health, council-run tips
and recycling centres)’
You have mentioned that you complained about the local environmen-
tal services …
who did you contact to complain about ‘local environmental services
(this includes street lighting, cleaning of streets, environmental health,
council-run tips and recycling centres)’? [Please tick all that apply]
[MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘local cultural and leisure services (this
includes parks, sports and leisure facilities, council libraries, museums, gal-
leries and cultural facilities)’
You have mentioned that you complained about the local cultural and
leisure services …
Who did you contact to complain about ‘Local cultural and leisure ser-
vices (this includes parks, sports and leisure facilities, council libraries,
museums, galleries and cultural facilities)’? [Please tick all that apply]
[MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
Has the problem you addressed now been sorted out to your satisfaction?
Yes
No
Those who had complained about ‘local planning services (including land
use, development) (This includes planning land use, approving planning
applications and implementing building regulations)’
You have mentioned that you complained about the local planning ser-
vices …
who did you contact to complain about ‘local planning services (includ-
ing land use, development) (this includes planning land use, approving
planning applications and implementing building regulations)’? [Please
tick all that apply][MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘local transport services (this includes
maintaining roads, bus subsidy schemes, car parking schemes)’
You have mentioned that you complained about the local transport ser-
vices …
Who did you contact to complain about ‘local transport services
(this includes maintaining roads, bus subsidy schemes, car parking
schemes)’? [Please tick all that apply][MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘social services: adult (this includes coun-
cil day care, home support (including meals on wheels) and residential
care)’
You have mentioned that you complained about social services: adult
…
who did you contact to complain about ‘social services: adult (this
includes council day care, home support (including meals on wheels)
and residential care)’? [Please tick all that apply][MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
Has the problem you addressed now been sorted out to your satisfaction?
Yes
No
Those who had complained about ‘social services: children (this includes
family support, child protection, fostering and adoption services)’
You have mentioned that you complained about social services: chil-
dren …
who did you contact to complain about ‘social services: children (this
includes family support, child protection, fostering and adoption ser-
vices)’? [Please tick all that apply][MULTICODE]
Local councillor
Government official
The Mayor of London

Other
None of the above
Yes
No
satisfaction?
Yes
No
Those who had complained about ‘payment of Housing Benefit and Council
Tax Benefit (this includes administration and payment of these benefits to
claimants)’
You have mentioned that you complained about payment of Housing
Benefit and Council Tax Benefit …
who did you contact to complain about ‘payment of Housing Benefit
and Council Tax Benefit (this includes administration and pay-
ment of these benefits to claimants)’? [Please tick all that apply]
[MULTICODE]
Local councillor
Government official
The Mayor of London
Other
None of the above
Yes
No
satisfaction?
Yes
No
Moving on now …
Did you manage to vote in the last LOCAL election in your area?
Yes
No
Can’t remember/Don’t know
Did you manage to vote in the last GENERAL election?
Yes
No
Can’t remember/Don’t know
If there were a LOCAL election tomorrow, which party would you vote
for?
Conservative
Labour
Liberal Democrat
Green
National Front/British National Party
United Kingdom Independence Party (UKIP)
Independent
Scottish National
Plaid Cymru
Some other party
Would not vote
Don’t know
If there were a GENERAL election tomorrow, which party would you
vote for?
Conservative
Labour
Liberal Democrat
Green
National Front/British National Party
United Kingdom Independence Party (UKIP)
Independent
Scottish National
Plaid Cymru
Some other party
Would not vote
Don’t know
In the last 12 months, that is since the end of May last year, have you
done any of the following? [Please tick all that apply][MULTICODE]
Attended a public meeting or rally
Taken part in a public demonstration or protest
Met with neighbours to complain or lobby
Signed a petition
None of the above
Which, if any, of the following groups, clubs or organizations have you
taken part in, supported or helped, over the last 12 months? [Please tick
all that apply][MULTICODE]
Children’s education/schools
Youth/children’s activities (outside school)
Education for adults
Sports/exercise (taking part, coaching or going to watch)
Religious (e.g. church)
Political (e.g. a group fighting for a particular cause or issue, political
party)
Trade union
The environment, animal rights
Justice and Human Rights
Citizens’ Groups
Local community or neighbourhood groups
The elderly
Health, disability and social welfare
Safety
First Aid
Hobbies/recreation/arts/social clubs
Internet group or chatroom
Other
None of the above
How much interest do you generally have in what is going on in politics?
A great deal
Quite a lot
Some
Not very much
Not at all
Don’t know
How much should government try to ensure a fair distribution of income
and wealth in this country?
A lot
A little
Not at all
Don’t know
On a scale from 0 to 10, where 10 means ‘a great deal of influence’ and
0 means ‘no influence’, how much influence do you have on politics and
public affairs?
0 – No influence
1
2
3
4
5
6
7
8
9
10 – A great deal of influence
Don’t know
Finally, a couple of questions to help us classify your answers …
To which of these groups do you consider yourself to belong?
White British
Any other white background
White and Black Caribbean
White and Black African
White and Asian
Any other mixed background
Indian
Pakistani
Bangladeshi
Any other Asian background
Black Caribbean
Black African
Any other black background
Chinese
Other ethnic group
Which of the following best applies to you?
I own a car
I do not own a car but I have use of one
I neither own nor have use of a car
Bibliography
Ǻdnanes, Marian. 2004. ‘Exit and/or Voice? Youth and Post-Communist Citizenship
in Bulgaria.’ Political Psychology 25:795–815.
Allen, Steven G. 1984. ‘Trade Unions, Absenteeism, and Exit-Voice.’ Industrial and
Labour Relations Review 37 (3):331–45.
Almond, Gabriel A., and Sidney Verba. 1963. The Civic Culture: Political Attitudes and
Democracy in Five Nations. Princeton University Press.
Aronson, J. Richard. 1974. ‘Financing Public Goods and the Distribution of Population
in Metropolitan Areas: An Analysis of Fiscal Migration in the US and England.’
In Economic Policy and Social Goals, ed. A. Culyer. Oxford: Martin Robertson.
Barakso, Maryann, and Brian F. Schaffner. 2008. ‘Exit, Voice, and Interest Group
Governance.’ American Politics Research 36:186–209.
Barry, Brian. 1974. ‘Review Article: “Exit, Voice and Loyalty”.’ British Journal of
Political Science 4:79–107.
Barzeley, Michael. 2001. The New Public Management: Improving Research and Policy
Dialogue. Berkeley, CA: University of California Press.
Bélanger, Eric, and Richard Nadeau. 2005. ‘Political Trust and the Vote in Multiparty
Elections: The Canadian Case.’ European Journal of Political Research
44:121–46.
Bendor, Keith A., and Peter J. Sloane. 1998. ‘Job Satisfaction, Trade Unions and Exit–
Voice Revisited.’ Industrial and Labour Relations Review 51 (2):222–40.
Bentley, Arthur. 1908. The Process of Government. Harvard: Belknap Press.
Bevan, Gwyn and Christopher Hood. 2006a. ‘Have Targets Improved Performance in
the English NHS?’ British Medical Journal 332:419–22.
2006b. ‘What’s Measured Is What Matters: Targets and Gaming in the English
Health Care System.’ Public Administration 84 (3):517–38.
Birch, A. H. 1958. The Theory of Committees and Elections. Cambridge University Press.
1975. Economic Models in Political Science: The Case of “Exit, Voice and Loyalty”.’
British Journal of Political Science 6:463–82.
175
176 Bi bl io gr a ph y
Black, D. 1948. ‘On the Rationale of Group Decision-making.’ Journal of Political

Economy 56:23–34.
Blanz, Mathias, Amélie Mummemendey, Rosemarie Mielke and Andreas Klink. 1998.
‘Responding to Negative Social Identity: A Taxonomy of Identity Management
Strategies.’ European Journal of Social Psychology 28:697–730.
Block, Fred. 1977. ‘The Ruling Class Does Not Rule: Notes on the Marxist Theory of
the State.’ Socialist Revolution 33:6027.
Bogart, William T., and Brian A. Cromwell. 1997. ‘How Much More Is a Good School
District Worth?’ National Tax Journal L:215–32.
Boroff, Karen E. 1989. ‘Loyalty – A Correlate of Exit, Voice, or Silence?’ In Industrial
Relations Research Association Series: Proceedings of the Forty-Second
Annual Meeting, ed. J. F. Burton. Madison, WI: Industrial Relations Research
Associations.
Boroff, Karen E., and David Lewin. 1997. ‘Loyalty, Voice, and Intention to Exit a Union
Firm: A Conceptual and Empirical Analysis.’ Industrial and Labor Relations
Review 51 (1):50–63.
Boyne, George, Oliver James, Peter John and Nicolai Petrovsky. 2009. ‘Democracy and
Government Performance: Holding Incumbents Accountable in English Local
Governments.’ Journal of Politics 71:1–12.
Brasington, David, and Don R. Haurin. 2006. ‘Educational Outcomes and House
Values: A Test of the Value Added Approach.’ Journal of Regional Science 46
(2):245–68.
Brekke, Kurt, and Lars Sørgard. 2006. ‘Public Versus Private Health Care in a National
Health Service.’ Health Economics 16:579–601.
Burgess, Simon, Carol Propper and Deborah Wilson. 2005. Will More Choice Improve
Outcomes in Education and Health Care? The Evidence from Economic Research.
University of Bristol: The Centre for Market and Public Organization.
Cheshire, Paul, and Stephen Sheppard. 1998. ‘Estimating Demand for Housing, Land
and Neighbourhood Characteristics.’ Oxford Bulletin of Economics and Statistics
60:357–82.
2003. ‘Capitalizing on the Value of Free Schools: The Impact of Constraints on
Supply and Uncertainty.’ Paper to European Regional Science Society www.ersa.
org/ersaconfs/ersa03/cdrom/papers/8.pdf.
Clapp, John M., Aanupam Nanda and Stephen L. Ross. 2008. ‘Which School Attributes
Matter? The Influence of School District Performance and Demographic
Composition of Property Values.’ Journal of Urban Economics 63 (2):451–66.
Clark, David E., and William E. Herrin. 2000. ‘The Impact of School Attributes on
Home Sale Prices in California.’ Growth and Change 31 (3):385–407.
Clark, William Robert, Matt Golder and Sonia Golder. 2007. ‘The Balance of Power
between Citizens and the State: Democratization and the Resource Curse.’
In http://polisci.fsu.edu/~polisci/graduate/prospectives/doctoral/documents/
clark_golder_golder.pdf.
Coleman, James S. 1990. Foundations of Social Theory. Cambridge, MA: Belknap Press
of Harvard University Press.
Cooper, Mick P. 2000. ‘Web Surveys: A Review of the Issues and Approaches.’ Public
Opinion Quarterly 64:464–94.
Bi bl io gr a ph y 177
Cooper, William H., L. Dyke and P. Kay. 1990. ‘Developing Act Frequency Measures of
Organizational Behaviors.’ In Academy of Management Best Paper Proceedings,
ed. L. R. Jauch and J. L. Wall. San Francisco, CA: Academy of Management.
Curtice, John, and Heath, Oliver. 2009. ‘Do People Want Choice and Diversity of
Provision in Public Services?’ In British Social Attitudes, the 25th Report, ed.
Alison Park, John Curtice, Katarina Thomson, Miranda Phillips, Elizabeth Clery.
London: Sage.
Davies, Howard. 1982. ‘Fiscal Migration and the London Boroughs.’ Urban Studies
19:143–54.
Davis-Blake, Alison, Joseph P. Broschak and Elizabeth George. 2003. ‘Happy Together?
How Using Nonstandard Workers Affects Exit, Voice, and Loyalty among
Standard Employees.’ Academy of Management Journal 46:475–85.
DeHoog, Ruth Hoogland, David Lowery and William E. Lyons. 1990. ‘Citizen
Satisfaction with Local Governance: A Test of Individual, Jurisdictional and
City-specific Explanations.’ Journal of Politics 52:807–37.
Dekker, Paul, and Eric Uslaner (ed.). 2001. Social Capital and Participation in Everyday
Life. London: Routledge.
Denny, Kevin, and Orla Doyle. 2008. ‘Political Interest, Cognitive Ability and
Personality: Determinants of Voter Turnout in Britain.’ British Journal of Political
Science 38 (2):291–310.
Derlega, Valerian J., Barbara A. Winstead, Robin J. Lewis and James Maddux. 1993.
‘Clients’ Responses to Dissatisfaction in Psychotherapy: A Test of Rusbult’s
Exit-Voice-Loyalty-Neglect Model.’ Journal of Social and Clinical Psychology 12
(3):307–18.
Devereux, Paul J., and Burton A. Weisbrod. 2006. ‘Does “Satisfaction” with Local
Public Services Affect Complaints (Voice) and Geographic Mobility (Exit)?’
Public Finance Review 34 (2):123–47.
Dillman, Don A. 2000. Mail and Internet Surveys: The Tailored Design Method. New
York: John Wiley.
Dowding, Keith. 1991. Rational Choice and Political Power. Aldershot: Elgar.
1996. ‘Public Choice and Local Governance.’ In Rethinking Local Democracy, ed. D.
King and G. Stoker. London: Macmillan.
2008. ‘A Pandemonium of Confusions: Kay and Marsh on Tiebout.’ New Political
Economy 13 (3):335–48.
Dowding, Keith, and Patrick Dunleavy. 1996. ‘Production, Disbursement and
Consumption: The Modes and Modalities of Goods and Services.’ In Consumption
Matters, ed. S. Edgell, K. Hetherington and A. Warde. Oxford: Blackwell.
Dowding, Keith, and Martin van Hees. 2009. ‘Freedom of Choice.’ In Oxford Handbook
of Rational and Social Choice, ed. P. Anand, K. Suzamura and P. K. Pattanaik.
Oxford University Press.
Dowding, Keith, and Peter John. 1996. ‘Exiting Behavior under Tiebout Conditions:
towards a Predictive Model.’ Public Choice 88:393–406.
1997. ‘Fairytale Critiques and Political Science: A Reply to Kenneth Newton.’ British
Journal of Political Science 27 (1):152–55.
2008. ‘The Three Exit, Three Voice and Loyalty Framework: A Test with Survey
Data on Local Services.’ Political Studies 56 (2):288–311.
2009. ‘The Value of Choice in Public Policy.’ Public Administration 87 (2):219–33.

2011. ‘Exit, voice and loyalty in the UK National Health Service: Using Hirschman
to Understand Responses of Citizens to Dissatisfaction.’ Public Administration,
89 (4):1403–18.
Dowding, Keith, Peter John and Stephen Biggs. 1994. ‘Tiebout: A Survey of the
Empirical Literature.’ Urban Studies 31 (4–5):767–97.
Dowding, Keith, Peter John, Thanos Mergoupis and Mark Van Vugt. 2000. ‘Exit,
Voice and Loyalty: Analytic and Empirical Developments.’ European Journal of
Political Research 37 (4):469–95.
Dowding, Keith, Peter John and Daniel Rubenson. Forthcoming. ‘Geographic
Mobility, Social Connections and Turnout.’ Journal of Elections, Public Opinion
and Parties.
Dowding, Keith, and Thanos Mergoupis. 2003. ‘Fragmentation, Fiscal Mobility and
Efficiency.’ Journal of Politics 65 (4):1190–207.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.
Drigotas, Stephen M., Gregory A. Whitney and Caryl E. Rusbult. 1995. ‘On the
Peculiarities of Loyalty: A Diary Study of Responses to Dissatisfaction in
Everyday Life.’ Personality and Social Psychology Bulletin 21 (6):596–609.
Dunleavy, Patrick, and Christopher Hood. 1994. ‘From Old Public Administration to
New Public Management.’ Public Money and Management 14:9–16.
Ellemers, Naomi, Henk A. M. Wilke and Ad F. M. Van Knippenberg. 1993. ‘Effects of
the Legitimacy of Low Group or Individual Status on Individual and Collective
Identity Enhancement Strategies.’ Journal of Personality and Social Psychology
64(5):766–88.
Eubank, William Lee, Arun Gangopadahay and Leonard B. Weinberg. 1996. ‘Italian
Communism in Crisis: A Study in Exit, Voice and Loyalty.’ Party Politics 2
(1):55–75.
Evan, William M. 1975. ‘Power, Conflict and Constitutionalism in Organizations.’
Social Science Information 14 (1):53–80.
Evans, Geoffrey. 1998. ‘Ethnic Schism and the Consolidation of Post-Communist
Democracies.’ Communist and Post-Communist Studies 31 (1):57–74.
Farrell, Dan. 1983. ‘Exit, Voice, Loyalty, and Neglect as Responses to Job
Dissatisfaction: A Multidimensional Scaling Study.’ Academy of Management
Journal 26:596–607.
Farrington-Douglas, Joe, and Jessica Allen. 2005. Equitable Choice for Health. London:
IPPR.
Feld, Lars P. 1997. ‘Exit, Voice and Income Taxes: The Loyalty of Voters.’ European
Journal of Political Economy 13:455–78.
Ferlie, Ewan. 1996. The New Public Management in Action. Oxford University Press.
Fine, Ben. 2001. Social Capital Versus Social Theory. London: Routledge.
Fisher, C. D., and E. A. Locke. 1992. ‘The New Look in Job Satisfaction Research and
Theory.’ In Job Satisfaction, ed. C. J. Cranny, P. C. Smith and E. F. Stone. New
York: Macmillan.
Flatley, John, Helen Connolly, Vanessa Higgins, John Williams, John Coldron, Kathy
Stephenson, Angela Logie and Nicola Smith. 2001. Parents’ Experiences of the
Process of Choosing a Secondary School. London: DfES.
Foreman-Peck, James. 1981. ‘Exit, Voice and Loyalty as Responses to Decline: The
Rover Company in the Interwar Years.’ Business History 23:191–207.
Fornell, Claes, and Fred L. Bookstein. 1982. ‘Two Structural Equation Models: LISREL
and PLS Applied to Consumer Exit–Voice Theory.’ Journal of Marketing Research
19:440–52.
Franklin, Mark N. 2004. Voter Turnout and the Dynamics of Electoral Competition in
Established Democracies since 1945. Cambridge University Press.
Freeman, Richard B. 1976. ‘Individual Mobility and Union Voice in the Labor Market.’
American Economic Review 66:361–8.
1980. ‘The Exit–Voice Tradeoff in the Labor Market: Unionism, Job Tenure, Quits
and Separations.’ Quarterly Journal of Economics 94 (3):643–73.
Freeman, Richard B., and James L. Medoff. 1984. What Do Unions Do? New York:
Basic Books.
Gaines, Stanley O., Harry T. Reis, Shandra Summers, Caryl E. Rusbult, Chante L.
Fox, Michael O. Wexler, William D. Marelcih, and Gregory J. Kurland. 1997.
‘Impact of Attachment Style of Reactions to Accommodative Dilemmas in Close
Relationships.’ Personal Relationships 4:93–113.
Gehlbach, Scott G. 2006. ‘A Political Model of Exit and Voice.’ Rationality and Society
18 (4):395–418.
Gerber, Alan S., Donald P. Green and Ron Shachar. 2003. ‘Voting May Be Habit-
Forming: Evidence from a Randomized Field Experiment.’ American Journal of
Political Science 47 (3):540–50.
Glaeser, Edward L., David I. Laibson, Jose A. Scheinkman and Christine L. Soutte.
2000. ‘Measuring Trust.’ Quarterly Journal of Economics 115 (3):811–46.
Glennerster, Howard, Anna Cohen and Virginia Bovell. 1998. ‘Alternatives to
Fundholding.’ International Journal of Health Service 28 (1):47–66.
Graham, Jill W. 1986. ‘Principled Organizational Dissent: A Theoretical Essay.’
Research in Organizational Behavior 8:1–52.
Graham, Jill W., and Michael Keeley. 1992. ‘Hirschman’s Loyalty Construct.’ Employee
Responsibilities and Rights Journal 5 (3):191–200.
Hardin, Russell. 1982. Collective Action. Baltimore, MD: Johns Hopkins
University.
Hendrick, Rebecca, Yonghong Wu, and Jacob Benoy. 2007. ‘Tax Competition
among Municipal Governments: Exit Versus Voice.’ Urban Affairs Review 43
(2):221–55.
Hetherington, Marc J. 1999. ‘The Effect of Political Trust on the Presidential Vote,
1968–1996.’ American Political Science Review 93:311–26.
Hibbard, Jonathan D., Nirmalya Kumar, and Louis W. Stern. 2001. ‘Examining the
Impact of Destructive Acts in Marketing Channel Relationships.’ Journal of
Marketing 38:45–61.
Hirschman, Albert O. 1967. Development Projects Observed. Washington: Brookings
Institute.
1970. Exit, Voice and Loyalty: Responses to Decline in Firms, Organizations and
States. Cambridge, MA: Harvard University Press.
1975. ‘Exit,Voice and the State.’ World Politics 31 (1):90–107.
1982. Shifting Involvements. Oxford: Martin Robertson.
1986. ‘Exit and Voice: An Expanding Sphere of Influence.’ In Rival Views of Market
Society and Other Essays, ed. A. O. Hirschman. Harmondsworth: Penguin.
1990. ‘Interview.’ In Economics and Sociology, ed. A. Swedberg. Princeton University
Press.
1995. ‘Exit, Voice and the Fate of the German Democratic Republic.’ In A Propensity to
Self-Subversion, ed. A. O. Hirschman. Cambridge, MA: Harvard University Press.
Hoffman, Elizabeth A. 2006. ‘Exit and Voice: Organizational Loyalty and Dispute
Resolution Strategies.’ Social Forces 84:2313–30.
Howard, Christopher, Michael Lipsky, and Dale Rogers Marshall. 1994. ‘Citizen
Participation in Urban Politics: Rise and Routinization.’ In Big-City Politics,
Governance and Fiscal Constraints, ed. G. E. Peterson. Washington, DC: Urban
Institute.
Hume, David. 1740/1978. A Treatise on Human Nature. Translated by P. H. Nidditch.
Oxford University Press.
John, Peter. 1997. ‘The Policy Implications of Tiebout Effects.’ Local Government
Studies 23 (2):67–79.
John, Peter, Keith Dowding and Stephen Biggs. 1995. ‘Residential Mobility in London:
A Micro-level Test of the Behavioural Assumptions of the Tiebout Model.’ British
Journal of Political Science 25:379–97.
Jud, G. Donald, and James M. Watts. 1981. ‘Schools and Housing Values.’ Land
Economics 57:459–70.
Kahneman, Daniel, and Amos Tversky. 1979. ‘Prospect Theory: An Analysis of
Decision under Risk.’ Econometrica 47 (2):263–91.
Kato, Junko. 1998. ‘When the Party Breaks Up: Exit and Voice among Japanese
Legislators.’ American Political Science Review 92 (4):857–70.
Keating, Michael, and Nicola McEwan. 2006. Devolution and Public Policy. London:
Routledge.
Kelley, Harold H., and John W. Thibaut. 1978. Interpersonal Relations: A Theory of
Interdependence. New York: John Wiley & Sons.
Lane, Robert. 1959. Political Life. New York: The Free Press.
Langston, Joy. 2002. ‘Breaking Out Is Hard to Do: Exit, Voice, and Loyalty in
Mexico’s One-Party Hegemonic Regime.’ Latin American Politics and Society 44
(3):61–88.
LaPonce, Jean. 1974. ‘Hirschman’s Voice and Exit Model as Spatial Archetype.’ Social
Science Information 13 (3):67–81.
Laver, Michael. 1976. ‘“Exit, Voice and Loyalty” Revisited: The Strategic Production
and Consumption of Public and Private Goods.’ British Journal of Political
Science 6:483–2.
Le Grand, Julian. 2003. Motivation, Agency and Public Policy: Of Knights and Knaves,
Pawns and Queens. Oxford University Press.
2007. The Other Invisible Hand: Delivering Public Services through Choice and
Competition. Princeton University Press.
Le Grand, Julian and Bartlett, Will (eds.). 1993. Quasi-Markets and Social Policy.
Basingstoke: Macmillan.
Le Maistre, Naomi, Rachel Reeves, and Angela Coulter. 2004. Patients’ Experience of
CHD Choice. Oxford: Picker Institute.
Leck, Joanne D., and David M. Saunders. 1992. ‘Hirschman’s Loyalty: Attitude or
Behavior?’ Employee Responsibilities and Rights Journal 5 (3):219–30.
Lee, Jaesub, and Fredric M. Jablin. 1992. ‘A Cross-cultural Investigation of Exit, Voice,
Loyalty and Neglect as Responses to Dissatisfying Job Conditions.’ Journal of
Business Communication 29 (3):203–28.
Lee, Soo-Young, and Andrew B. Whitford. 2008. ‘Exit, Voice, Loyalty and Pay:
Evidence from the Public Workforce.’ Journal of Public Administration Research
and Theory 18 (4):647–71.
Levi, Margaret, and Laura Stoker. 2000. ‘Political Trust and Trustworthiness.’
In Annual Review of Political Science 3, ed. N. W. Polsby. Palo Alto: Annual
Reviews.
Lowery, David, and William E. Lyons. 1989. ‘The Impact of Jurisdictional Boundaries:
An Individual-level Test of the Tiebout Model.’ Journal of Politics 51:73–97.
Luchak, Andrew A. 2003. ‘What Kind of Voice Do Loyal Employees Use?’ British
Journal of Industrial Relations 41 (1):115–34.
Lyons, William E., and David Lowery. 1986. ‘The Organization of Political Space and
Citizen Responses to Dissatisfaction in Urban Communities: An Integrative
Model.’ Journal of Politics 48:321–46.
1989. ‘Citizen Responses to Dissatisfaction in Urban Communities: A Partial Test
of a General Model.’ Journal of Politics 48:321–46.
Lyons, William E., David Lowery and Ruth Hoogland DeHoog. 1992. The Politics of
Dissatisfaction: Citizens, Services and Urban Institutions. Armonk, NY: M. E.
Sharpe.
Malhotra, Neil, and Jon A. Krosnick. 2007. ‘The Effect of Survey Mode and Sampling
on Inferences about Political Attitudes and Behavior: Comparing 2000 and
2004 ANES to Internet Surveys with Nonprobability Samples.’ Political Analysis
15:286–323.
Marwell, Gerald, and Pamela Oliver. 1993. The Critical Mass in Collective Action: A
Micro-Social Theory. Cambridge University Press.
Matland, Richard E. 1995. ‘Exit, Voice, Loyalty, and Neglect in an Urban School
System.’ Social Science Quarterly 76 (3):506–12.
Maute, Manfred F., and William R. Forrester, Jr. 1993. ‘The Structure and Determinants
of Consumer Compliant Intentions and Behavior.’ Journal of Economic Psychology
14:219–47.
McKee, Daryl, Christina Simmers and Jane Licata. 2006. ‘Customer Self-Efficacy and
Response to Service.’ Journal of Service Research 8 (3):207–20.
McLaughlin, Kate, Stephen P. Osborne, and Ewan Ferlie (eds.). 2002. New Public
Management: Current Trends and Future Prospects. London: Routledge.
Menahem, G. 2001. ‘Urban Civic Worlds: A Conceptual and Empirical Exploration.’
Local Government Studies 27:37–60.
Milbrath, Lester W. 1965. Political Participation: How and Why do People Get Involved
in Politics. Chicago University Press.
Milbrath, Lester W., and Madan L. Goel. 1997. Political Participation. 4th edn. Chicago
University Press.
Miller, Arthur H. 1974. ‘Political Issues and Trust in Government, 1964–1970.’
American Political Science Review 68:951–72.
Miller, Arthur H., and Ola Listaug. 1990. ‘Political Parties and Confidence in
Government: A Comparison of Norway, Sweden and the United States.’ British
Journal of Political Science 29:357–86.
Miller, Gary J. and Moe, Terry M. 1983. ‘Bureaucrats, Legislators and the Size of
Government.’ American Political Science Review 77:297–322.
Miller, Paul, and Charles Mulvey. 1991. ‘Australian Evidence on the Exit/Voice Model
of the Labor Market.’ Industrial and Labour Relations Review 45 (1):44–57.
Milner, Henry. 2001. ‘Social Capital, Civic Literacy and Political Participation:
Explaining Differences in Voter Turnout.’ In Challenges to Democracy, ed. K.
Dowding, J. Hughes and H. Margetts. Basingstoke: Macmillan.
Miranda, Rowan and Allan, Lerner. 1995. ‘Bureaucracy, Redundancy and the
Privatization of Services.’ Public Administration Review 55 (2):193–200.
Moore, Mark H. 1995. Creating Public Value: Strategic Management in Government.
Cambridge, MA: Harvard University Press.
Morton, Rebecca B. 1999. Methods and Models: A Guide to the Empirical Analysis of
Formal Models in Political Science. Cambridge University Press.
Mueller, Dennis C. 1993. The Public Choice Approach to Politics. Aldershot: Edward
Elgar.
Naus, Fons, Ad van Iterson and Robert Roe. 2007. ‘Organizational Cynicism: Extending
the Exit, Voice, Loyalty and Neglect Model of Employees’ Relations to Adverse
Conditions in the Workplace.’ Human Relations 60 (5):683–718.
Newman, M., and E. Vidler. 2006. ‘Discriminating Customers, Responsible Patients,
Empowered Users: Consumerism and Modernization of Health Care.’ Journal of
Social Policy 35:193–209.
Newton, Keith. 2007. ‘Social and Political Trust.’ In The Oxford Handbook of Political
Behavior, ed. R. J. Dalton and H.-D. Klingemann. Oxford University Press.
Newton, Kenneth. 1997. ‘Social Capital and Democracy.’ American Behavioral Scientist
40 (5):575–86.
Niskanen, William A. 1994. Bureaucracy and Public Economics. Aldershot: Edward
Elgar.
O’Donnell, Guillermo. 1986. ‘On the Fruitful Convergence of Hirschman’s Exit, Voice
and Loyalty and Shifting Involvements: Reflections from the Recent Argentine
Experience.’ In Development, Democracy and the Art of Trespassing: Essays in
Honor of Albert Hirschman, ed. A. Foxley, M. McPherson and G. O’Donnell.
University of Notre Dame Press.
Ogawa, Rodney T., and Jo Sargent Dutton. 1997. ‘Parent Involvement and School
Choice: Exit and Voice in Public Schools.’ Urban Education 32 (3):333–53.
Okamoto, Dina G., and Rima Wilkes. 2008. ‘The Opportunities and Costs of Voice
and Exit: Modelling Ethnic Group Rebellion and Emigration.’ Journal of Ethnic
and Migration Studies 34 (3):347–69.
Olson, Mancur. 1971. The Logic of Collective Action: Public Goods and the Theory of
Groups. 2nd edn. Cambridge, MA: Harvard University Press.
1982. The Rise and Decline of Nations: Economic Growth, Stagflation, and Social
Rigidities. New Haven, CT: Yale University Press.
Orbell, John M., and Toru Uno. 1972. ‘A Theory of Neighbourhood Problem Solving:
Political Action vs. Residential Mobility.’ American Political Science Review
66:471–89.
Osborne, David, and Ted Gaebler. 1992. Reinventing Government: How the
Entrepreneurial Spirit Is Transforming the Public Sector. New York: Penguin.
Ostrom, Vincent, Charles M. Tiebout and Robert Warren. 1961. ‘The Organisation of
Government in Metropolitan Areas – A Theoretical Enquiry.’ American Political
Science Review 55:831–42.
Percy, Stephen L., and Brett W. Hawkins. 1992. ‘Further Tests of Individual-level
Propositions from the Tiebout Model.’ Journal of Politics 54:1149–57.
Percy, Stephen L., Brett W. Hawkins and Peter E. Maier. 1995. ‘Revisiting Tiebout:
Moving Rationales and Interjurisdictional Relocation.’ Publius 25:1–17.
Pettitt, Robin T. 2007. ‘Challenging the Leadership: The Party Conference as a
Platform for Dissenting Membership Voice in British and Danish Parties of the
Left.’ Scandinavian Political Studies 30:229–48.
Pfaff, Steven. 2006. Exit–Voice Dynamics and the Collapse of East Germany: The Crisis
of Leninism and the Revolution of 1989. Durham, NC: Duke University Press.
Pfaff, Steven, and Hyojoung Kim. 2003. ‘Exit–Voice Dynamics in Collective Action:
An Analysis of Emigration and Protest in the East German Revolution.’ American
Journal of Sociology 109 (2):401–44.
Ping, Robert A. 1993. ‘The Effects of Satisfaction and Structural Constraints on
Retailer Exiting, Voice, Loyalty, Opportunism, and Neglect.’ Journal of Retailing
69 (3):320–52.
1999. ‘Unexplored Antecedents of Exiting in a Marketing Channel.’ Journal of
Retailing 79:237–48.
2003. ‘Antecedents of Satisfaction in a Marketing Channel.’ Journal of Retailing
79:237–48.
Prest, John. 1990. Liberty and Locality: Parliament, Permissive Legislation, and
Ratepayers’ Democracies in the Mid-Nineteenth Century. Oxford: Clarendon
Press.
Propper, Carol, D. Wilson, and Steven Burgess. 2006. ‘Extending Choice in English
Health Care: The Implications of the Economic Evidence.’ Journal of Social Policy
35:537–57.
Przeworski, Adam, and Michael Wallerstein. 1988. ‘Structural Dependence of State on
Capital.’ American Political Science Review 82:11–29.
Putnam, Robert D. 1995. ‘Bowling Alone: America’s Declining Social Capital.’ Journal
of Democracy 6 (1):65–78.
2000. Bowling Alone: The Collapse and Revival of American Community. New York:
Simon and Schuster.
Putnam, Robert D., with Robert Leonardi and Raffaella Y. Nanetti. 1993. Making
Democracy Work: Civic Traditions in Modern Italy. Princeton University Press.
Reback, Randall. 2005. ‘House Prices and the Provision of Local Public Services:
Capitalization under School Choice Programs.’ Journal of Urban Economics 57
(2):275–301.
Rosenthal, Leslie. 2003. ‘The Value of Secondary School Quality.’ Oxford Bulletin of
Economics and Statistics 65 (3):329–55.
Ross, Catherine E., and John Mirowsky. 1984. ‘Socially-desirable Response and
Acquiescence in a Cross-cultural Survey of Mental Health.’ Journal of Health and
Social Behavior 25:189–97.
Rossi, Peter H. 1980. Why Families Move. London: Sage.
Rusbult, Caryl E. 1987. ‘Responses to Dissatisfaction in Close Relationships: The
Exit-Voice-Loyalty-Neglect Model.’ In Intimate Relationships: Development,
Dynamics, and Deterioration, ed. D. Perlman and S. Duck. Newbury Park, CA:
Sage.
Rusbult, Caryl E., Victor L. Bissonnette, Ximena B. Arriaga and Chante L. Cox.
1998. ‘Accommodation Processes during the Early Years of Marriage.’ In The
Developmental Course of Marital Dysfunction, ed. T. N. Bradbury. Cambridge
University Press.
Rusbult, Caryl E., and Dan Farrell. 1982. ‘A Longitudinal Test of the Investment Model:
The Impact on Job Satisfaction, Job Commitment, and Turnover of Variations
in Rewards, Costs, Alternatives and Investments.’ Journal of Applied Psychology
68:429–38.
Rusbult, Caryl E., Dan Farrell, Glen Rogers, and Arch G. Mainous, III. 1988b. ‘Impact
of Exchange Variables on Exit, Voice, Loyalty and Neglect: An Integrative Model
of Responses to Declining Job Satisfaction.’ Academy of Management Journal
31:599–627.
Rusbult, Caryl E., Dennis J. Johnson, and Gregory D. Morrow. 1986. ‘Determinants and
Consequences of Exit, Voice, Loyalty and Neglect: Responses to Dissatisfaction
in Adult Romantic Involvements.’ Human Relations 39:45–63.
Rusbult, Caryl E., and David Lowery. 1985. ‘When Bureaucrats Get the Blues:
Responses to Dissatisfaction among Federal Employees.’ Journal of Applied Social
Psychology 15:80–103.
Rusbult, Caryl E., David Lowery, Michael L. Hubbard, Orly J. Maravankin and Michael
Neises. 1988a. ‘Impact of Employee Mobility and Employee Performance on the
Allocation of Rewards under Conditions of Constraint.’ Journal of Personality
and Social Psychology 54:605–15.
Rusbult, Caryl E., and Paul A. M. Van Lange. 1996. ‘Interdependence Processes.’
In Social Psychology: Handbook of Basic Principles, ed. E. T. Higgins and A.
Kruglanski. New York: Guilford.
Rusbult, Caryl E., Isabella M. Zembrodt, and Lawanna K. Gunn. 1982. ‘Exit, Voice,
Loyalty and Neglect: Responses to Dissatisfaction in Romantic Involvements.’
Journal of Personality and Social Psychology 43 (6):1230–42.
Rusbult, Caryl E., and Isabella M. Zembrodt. 1983. ‘Responses to Dissatisfaction
in Romantic Involvements: A Multidimensional Scaling Analysis.’ Journal of
Experimental Psychology 19 (3):274–93.
Salkeld, G. 1998. ‘What Are the Benefits of Preventive Health Care?’ Health Care
Analysis 6 (2):106–12.
Samuelson, Paul A. 1954. ‘The Pure Theory of Public Expenditure.’ Review of Economics
and Statistics 36:387–89.
Sanders, David, Harold D. Clarke, Marianne C. Stewart and Paul Whiteley. 2007. ‘Does
Mode Matter for Modelling Choice? Evidence from the 2005 British Election
Study.’ Political Analysis 15 (3):257–85.
Sandler, Todd. 1992. Collective Action: Theory and Applications. Hemel Hempstead:
Harvester-Wheatsheaf.
Savas, Emanuel S. 1987. Privatization: The Key to Better Government. Chatham, NJ:
Chatham House.
Schaltegger, C., and D. Kiittel. 2002. ‘Exit, Voice, and Mimicking Behavior: Evidence
from Swiss Cantons.’ Public Choice 113:1–23.
Schoen, Harald, and Thorsten Faas. 2005. ‘When Methodology Interferes with
Substance: The Difference of Attitudes towards e-campaigning and e-voting in
Online and Offline Surveys.’ Social Science Computer Review 23:326–33.
Scott, Rebecca J. 1986. ‘Dismantling Repressive Systems: The Abolition of Slavery in
Cuba as a Case Study.’ In Development, Democracy and the Art of Trespassing:
Essays in Honor of Albert Hirschman, ed. A. Foxley, M. McPherson and G.
O’Donnell. Notre Dame Press.
Sen, Amartya. 2002. ‘Goals, Commitment and Identity.’ In Rationality and Freedom,
ed. A. Sen. Cambridge, MA: Harvard University Press.
2005. ‘Why Exactly is Commitment Important for Rationality?’ Economics and
Philosophy 21:5–13.
2006. Identity and Violence: The Illusion of Destiny. New York: W. W. Norton and
Co.
Sharp, Elaine. 1984. ‘Exit, Voice and Loyalty in the Context of Local Government
Problems.’ Western Political Quarterly 7:67–83.
1986. Citizen Demand-making in the Urban Context. Birmingham, AL: University
of Alabama Press.
Singer, Judith D., and John B. Willett. 2009. Applied Longitudinal Data Analysis:
Modeling Change and Event Occurrence. New York: Oxford University Press.
Spencer, Daniel G. 1986. ‘Employee Voice and Employee Retention.’ Academy of
Management Journal 29 (3):488–502.
StataCorp. 2009. Stata Statistical Softward: Release 11. College Station, TX: StatCorp
LP.
Stoker, Gerry. 2006. ‘Public Value Management: A New Narrative for Networked
Governance?’ American Review of Public Administration 36 (1):41–7.
Stolle, Dietland, and Marc Hooghe. 2005. ‘Inaccurate, Exceptional, One-Sided
or Irrelevant? The Debate about the Alleged Decline of Social Capital and
Civic Engagement in Western Societies.’ British Journal of Political Science 35
(1):149–67.
Teske, Paul, Mark Schneider, Mark Mintrom and Samuel Best. 1993. ‘Establishing Micro
Foundations of a Macro Theory: Information, Movers, and the Competitive Local
Market for Public Goods.’ American Political Science Review 87 (3):702–13.
Thane, Pat. 1982. The Foundations of the Welfare State. London: Longman.
Thomas, David C., and Kevin Au. 2002. ‘The Effect of Cultural Differences on
Behaviorial Responses to Low Job Satisfaction.’ Journal of International Business
Studies 33 (2):309–26.
Thomas, John Clayton, and Julia Melkers. 1999. ‘Explaining Citizen-initiated Contacts
with Municipal Bureaucrats: Lessons for the Atlanta Experience.’ Urban Affairs
Review 34 (5):667–90.
Tiebout, Charles M. 1956. ‘A Pure Theory of Local Expenditures.’ Journal of Political
Economy 64:416–24.
Turnley, William H., and Daniel C. Feldman. 1999. ‘The Impact of Psychological
Contract Violations on Exit, Voice, Loyalty, and Neglect.’ Human Relations 52
(7):895–922.
Tversky, Amos, and Daniel Kahneman. 1974. ‘Judgement Under Uncertainty:
Heuristics and Biases.’ Science 185:1124–31.
Uslaner, Eric. 2002. The Moral Foundations of Trust. New York: Cambridge University
Press.
Van Ryzin, Greg G., Douglas Muzzio, Stephen Immerwahr, Lisa Gulick and Eve
Martinez. 2004. ‘Drivers and Consequences of Citizen Satisfaction: An
Application of the American Customer Satisfaction Index Model to New York
City.’ Public Administration Review 64:321–41.
Van Vugt, Mark, Keith Dowding, Peter John and Eric Van Dijk. 2003. ‘The Exit of
Residential Mobility or the Voice of Political Action? Strategies for Problem
Solving in Residential Communities.’ Journal of Applied Social Psychology 33
(2):321–38.
Verba, Sidney, and Norman Nie. 1972. Participation in America. New York: Harper
and Row.
Verba, Sidney, Kay Lehman Schlozman and Henry E. Brady. 1995. Voice and Equality:
Civic Voluntarism in American Politics. Cambridge, MA: Harvard University
Press.
Vigoda, Eran. 2000. ‘Organizational Politics, Job Attitudes, and Work Outcomes:
Exploration and Implications for the Public Sector.’ Journal of Vocational
Behavior 57 (2):326–47.
Walker, Jack. 1983. ‘The Origins and Maintenance of Interest Groups in America.’
American Political Science Review 77:390–406.
Whiteley, Paul, and Patrick Seyd. 2002. High-Intensity Participation: The Dynamics of
Party Activism in Britain. Ann Arbor: University of Michigan Press.
Withey, Michael J., and William H. Cooper. 1989. ‘Predicting Exit, Voice, Loyalty and
Neglect.’ Administrative Science Quarterly 34:521–39.
1992. ‘What’s Loyalty?’ Employee Responsibilities and Rights Journal 5 (3):231–40.
Wood, B. Dan, and Richard W. Waterman. 1994. Bureaucratic Dynamics: The Role of
Bureaucracy in a Democracy. Boulder, CO: Westview.
Wright, Joseph. 2007. ‘When Does Economic Crisis Lead to Democratization?’ In
www.yale.edu/cpworkshop/papers/Wright.pdf.
Young, Dennis R. 1972. How Shall We Collect Garbage: A Study in Economic
Organizations. Washington, DC: The Urban Institute.
1974. ‘Exit and Voice in the Organization of Public Services.’ Social Science
Information 13 (3):49–65.
1976. ‘Consolidation or Diversity: Choices in the Structure of Urban Governance.’
American Economic Review 66 (2):378–85.
Young, Dennis R., and Richard R. Nelson. 1973. Public Policy for Day Care of Young
Children: Organization, Finance, and Planning. Lexington, MA: D. C. Heath.
Zimmerman, Dale L., Vicente A. Nunuez-Anton, Richard L. Smith and Valerie Isham.
2009. Antedependence Models of Longitudinal Data. London: Chapman and
Hall.
Index
active people see alert or active people collective goods see public goods
alert or active people 8, 10, 61, 73, 75–76, 102, collective voice (CV) 9, 16, 17–18, 23, 44–45,
104, 108–09, 110, 120, 134–35 46, 75, 89, 91, 130, 132
attachment 2, 13, 20, 49, 68, 80, 111, 116 collective voice activity 46, 76, 93, 103, 119,
neighbourhood 76, 80, 104, 111, 112, 135, 139
117–18, 121, 125 participation see collective voice partici-
see also group loyalty pation (CVP)
Audit Commission (UK) 28 voting see collective voice vote (CVV)
dissatisfaction, and see satisfaction
ballot box see voters/voting lobby group 1, 7, 44, 46, 47, 80, 88, 89, 91,
Birch ‘loyalty’ 21, 63, 69, 118, 119, 125 94, 123, 133
brand loyalty 2, 20, 21, 27, 59 petition 9, 15, 16, 41, 44, 45, 46, 80, 89, 91,
British Election Study (2005) 79 94, 118, 123, 127
pressure group 1, 16, 41, 44, 46, 113, 132
campaigns/campaigning 1, 2, 46 protest 1, 9, 10, 16, 80, 91, 127, 130,
see also pressure group 132, 144
choice 137–39 public demonstration 44, 80, 91, 94
NHS 137–39 public rally 17, 18, 86, 89, 91
state education 138 satisfaction, and see satisfaction
voice, and 33–37, 50, 130 see also ‘Coca-Cola Drinkers of America’;
church, loyalty to 20 horizontal voice; individual voice;
citizenship and state policy 2, 10, 18, 24, 26, vertical voice
31–33, 130, 132 collective voice participation (CVP) 44, 46,
Coca-Cola 1–2, 4, 5, 14, 16, 130 80, 132, 141
Coca-Cola Classic, reintroduced as 1, 5, 16 EVL framework, and 123–24, 127
New Coke, introduction of 1, 5, 16, 19 exit, and 113–14, 132, 135
Coca-Cola II, rebranded as 1 locked in 114–15, 135
see also Pepsi loyalty, and 118–21, 135, 139
‘Coca-Cola Drinkers of America’ 1, 4, 16, 130 Poisson model 123, 124, 150
collective action problem 15–18, 27, 45, 46, satisfaction, and 94–95, 99, 100–01
93, 132 see also trust
188
I n de x 189
collective voice vote (CVV) 28–29, 44, 46, 80, efficiency 6, 7, 28, 34, 37, 42, 48, 53, 131, 137,
132, 133, 137, 141 138, 139
EVL framework, and 121–23, 127–28 see also competition
exit, and 113–14, 132, 135 electoral process
locked in 114–15, 135 generally see voters/voting
loyalty, and 118–21, 135, 139 local elections 84, 94, 114, 120, 127
satisfaction, and 94–95, 99, 100–01 empirical literature (EVL) 142–49
see also trust; voters/voting comparative politics 144–45
competition 6, 18, 28, 34, 38, 39, 42, 50, 136, consumers and producers 143–44
139 employer–employee relationships 142–43
choice and voice 33–37, 50, 130 social relationships 144
Nigerian railways study 6–7, 13 summing up 145–49
tax competition 55–56 employment status see socio-economic status
see also efficiency (SES)
complaint/complaining ‘equation filler,’ loyalty as 19–22
collective see collective voice (CV) EVL framework (Hirschman) 15–24
personal or individual see individual voice analysis of 121–29
(IV) CVP 123–24
complete exit 29, 37–38 CVV 121–23
consumers IV 124–25
alert see alert or active people application of 130–39
inert see inert or passive people conclusions 22–24
production and consumption, relationship criticisms and extensions 15–22, 26
between 35 collective action problem 15–18, 27, 45,
contracting out 18–19, 33 46, 93, 132
loyalty as ‘equation filler’ 19–22
democracy 13, 32, 55, 131, 141, 145 product ‘quality’ 18–19
tax competition, and 55 elements of see exit; loyalty; voice
demonstration see public demonstration empirical literature see empirical literature
Department of Health, Social Services and (EVL)
Public Safety (Northern Ireland) 81 game-theoretic model 29–30, 53
Department of Health (UK) 81 major relationships see major EVL
direct voice 44 relationships
disbursement 35–36 neglect, and see EVLN framework
dissatisfaction see satisfaction (Rusbult)
non-formal model, as 26–29
education 7, 13, 35, 41, 46, 51, 72, 80 original argument 2–6, 131
choice and voice 34 original puzzle 6–8
complete exit 38 publication of 2
internal exit 38–39 research using 25–26
private education see private education see also empirical literature (EVL)
private exit, and 39–40, 117–18 separate behaviour model 31–33, 49
satisfaction, and 47 simple relationships see simple
state education see state education relationships
Tiebout exit 43 urban context, examination in 51–56
educational attainment 34, 43, 48, 53, 54, 61, see also EVMESD Survey
77, 99, 103, 134–35, 140, 141 EVLN framework (Rusbult) 23, 49, 50, 51–74,
voice, and 99–100 102, 134
see also social class; socio-economic status active/passive dimension 59, 60–61, 72–73,
(SES) 76, 89–90, 91–92, 102, 113, 134–35
190 I n de x
EVLN framework (Rusbult) (cont.) complete exit 29, 37–38

conclusions 73–74 costs of opportunities 12, 14, 17, 20, 30, 48,
constructive/destructive dimension 59, 52, 64, 87–88, 90, 103, 135, 137–39,
60–61, 62, 72–73 141, 144, 145
everyday life 57 dissatisfaction, and see satisfaction
examination of 56–59 economic response, as 5, 9
exit 58–59, 61, 62–65, 66–68 educational attainment 34, 43, 48, 53, 54,
exit−voice framework see exit−voice 61, 77, 99, 102, 134–35, 140, 141
framework efficiency of services, and see competition;
Hirschman in the urban context 51–56 efficiency
introduction 51 elements of see exit (elements)
local government services, satisfaction EVL mix 15
with 57, 65–68 EVLN framework see EVLN framework
loyalty 58–59, 61, 62–65, 66–68, 136 (Rusbult)
attitude or behaviour, whether 68–73 EVMESD Survey see EVMESD Survey
marketing channels 64–65 geographical see geographical exit
multidimensional scaling 61, 72 high income. and 34, 43, 48, 53, 77, 99,
neglect 58–59, 61, 62–65, 66–68 100, 135
psychotherapy 57 Hirschman’s original idea 2–6, 27
romantic relationships 57, 58, 59, 62, 70, 72 individual voice, and 104–10, 135, 137
urban school systems 57 internal exit 38–39, 43, 133
use of 62–65 neglect, and see EVLN framework
voice 58–59, 61, 62–65, 66–68 (Rusbult)
workplace relationships 57, 58, 59, 62–63, noisy exit 25–26, 58–59, 105, 106, 127
70, 72 object of loyalty, and see exit, voice and
EVMESD Survey 75, 77, 80–101, 134 object of loyalty
conclusions 100–01 private exit see private exit
EVL framework, against 125–29, 134–37 separate behaviour model 31–33, 49
education socio-economic status 75, 76, 99, 103, 134,
private 76, 88, 89, 91, 97, 109–10 138, 141
state 80, 82–83, 93, 97, 109–10 Tiebout exit see Tiebout exit
health trade-off see exit−voice trade-off
NHS 81–82, 83, 93, 95, 96–97, 109 voice, relationship with see exit−voice
private 82, 83, 87, 88, 89, 91, 96–97, 99, 109 framework
nature of data 77–80 welfare, and see exit, voice and welfare
questions asked see survey instrument below see also non-exit
satisfaction, summing up on 97–99 exit (elements) 8–9
simple relationships see simple interpretation of signal 8–9
relationships speed 8
social class and voice 99–100 exit, voice and loyalty see EVL framework
statistical methods 150–51 (Hirschman)
survey instrument 76, 77, 80–81, 152–74 exit, voice and object of loyalty 20–22, 25–50,
understanding the data 84–85 133
voice and exit choice and voice 33–37, 50, 130
comparing combinations of 90–92 conclusions 49–50
describing 85–88 formalizing Hirschman 29–33
dimensions of 88–90 game-theoretic model 29–30, 53
exit separate behaviour model 31–33, 49
binary response, as 9, 45, 140 Hirschman’s framework 26–29
Coca-Cola study 2 introduction 25–26
I n de x 191
loyalty 49, 50 high income, and 100

satisfaction see satisfaction loyalty, and 115
three exits 27, 37–40, 44, 50, 130 NHS, and 105–06
geographical see geographical exit Tiebout exit see Tiebout exit
internal exit 38–39, 43, 133 German Democratic Republic, exit of people
private see private exit from 26
three voices 43–46, 50, 130 government see local government services;
collective action see collective voice state
participation (CVP) GP practice see General Practitioner (GP)
collective vote see collective voice vote practice
(CVV) group loyalty 20–22, 49, 91, 104, 120, 121, 124
individual see individual voice (IV) church 20
exit, voice and welfare 140–41 family 14, 20, 66
exit−voice framework 7, 8, 9–12, 13, 17, local community 21, 22, 42, 66, 68, 72, 75,
52–56, 70, 75, 136, 142 103, 115, 116, 119, 120, 127, 133, 134,
analysis of interrelationship 54–56 139, 140
communal participation 54 organization 20, 21, 22
game-theoretic model 29–30, 53 state 20, 21, 22, 31, 49
individualistic participation 54 tribe 20
intended behaviours 55, 56, 75, 76, 86–87, see also attachment; social investment
89, 108–09
loyalty as mediator 14, 59, 127 health care 7, 34, 35, 41, 45, 46, 51
mimicking behaviour 55–56 choice and voice 36–37
optimal mix 15 health insurance 76, 88, 95
spatial dimension 52–53, 55 internal exit 38, 39
trade-off see exit−voice trade-off private see private health
vertical dimension 52–53 private exit 40, 106
voice as complement to exit 10 public see National Health Service (NHS)
voice as residual of exit 10 (UK)
exit, voice, loyalty and neglect see EVLN Tiebout exit 43
framework (Rusbult) see also National Health Service (NHS)
exit−voice trade-off 17, 18, 47, 48, 56, 64, 68, (UK)
75, 89, 102–03, 128, 131, 136 high income 34, 43, 48, 53, 77, 80, 134–35
effect of loyalty on 103–04, 115–21, 127, geographical exit, and 100
128, 131, 133, 135 voice, and 99–100
geographical exit 115 see also social class; socio-economic status
loyalty and voice 118–21 (SES)
private exit 115–18 Hirschman, Albert O. 1–22, 25, 26–29, 44, 47,
53, 70–73, 101, 102–4, 115, 124, 130–4,
family 43, 45, 48 136–7
loyalty to 14, 20, 66 horizontal voice 16–17, 43–44, 46, 54, 132
focus groups 1 vertical voice, distinguished from 16,
43–44
game-theoretic model 29–30, 53 hospitals see health care
General Practitioner (GP) practice 39, 81, 82,
83, 106, 138 individual voice (IV) 4, 9, 15–16, 44–45, 46,
geographical exit 40, 54, 68, 76, 80, 95, 99, 48, 75, 80, 87, 88, 89, 130, 132, 135
104–08, 121, 122, 124, 126–27, 133, dissatisfaction, and 92–93, 99, 100–01, 133,
135–36 135, 137
education, and 104–05, 106 EVL framework, and 124–25, 128
192 I n de x
individual voice (IV) (cont.) Coca-Cola case 2

exit, and 104–10, 135, 137 CVP, and 118–21, 135, 139
locked in 114–15, 125, 135 CVV, and 118–21, 135, 139
responses to 110–13 degree of 21
see also collective voice description of 13
inert or passive people 8, 10, 61, 73, 75, 87, 90, elements see loyalty (elements)
103, 104, 120, 124, 134–35 ‘equation filler,’ as 19–22
internal exit 38–39, 43, 133 EVL mix 15
EVLN framework see EVLN framework
‘knaves’ and ‘knights’ 27–28 (Rusbult)
exit−voice trade-off, effect on 103–04,
lobby group/lobbying 1, 7, 33, 44, 46, 47, 80, 115–21, 127, 128, 131, 133, 135
88, 89, 91, 94, 123, 133 geographical exit 115
see also petition/petitioning; pressure group; loyalty and voice 118–21
public demonstration; public rally private exit 115–18
local community, loyalty to 21, 22, 42, 66, 68, game-theoretic model 29–30, 53
72, 75, 103, 115, 116, 119, 120, 127, 133, group, to see group loyalty
134, 139, 140 Hirschman’s original idea 2–6, 27
local elections 84, 94, 114, 120, 127 neglect, and see EVLN framework
local government services 18, 19, 34, 47, 51, (Rusbult)
80–81, 95, 106, 113, 132, 134, 137, 140 object of see exit, voice and object of loyalty
choice and voice 36–37 passive 69–70, 72
council tax 81 personal history, and 21–22
crime 81 private education, and 116–17
culture and leisure 80, 81 private health, and 115–16
dog-catching 80 psychological effect 49
environmental services 80 separate behaviour model 31–33, 49
EVLN framework, application of see EVLN social capital see social capital
framework (Rusbult) social investment, as see social investment
housing 80 specialist loyalist behaviour 71–72
planning 80 trust, and 65, 69, 104, 118–19, 143
pollution and noise 81 voice, effect on 69–73, 75, 76
refuse collection 17, 18, 36, 41, 45, 46, 80 see also exit−voice trade-off
road repairs 45, 80 see also non-exit; silence
satisfaction, and 47–48, 109 loyalty (elements) 13–14
see also EVMESD Survey inefficient results 14
sewerage 41 mediator between exit and voice, as 14,
street cleaning 80 59, 127
street lighting 45 rational behaviour, as 14
Swiss cantons, studies in 55–56
transport 6, 7, 81 major EVL relationships 102–29
UK, history in 41–42, 53 conclusions 125–29
US, studies in 23, 56, 65–68 exit, voice and loyalty 121–25
water supply 41 exit−voice trade-off see exit−voice trade-off
loyalty introduction 102–04
active 69–70 voice, satisfaction and exit see voice,
attachment see attachment satisfaction and exit
benefits of 139–40 market research 2, 5
Birch ‘loyalty’ 21, 63, 69, 118, 119, 125 focus groups 1
brand loyalty 2, 20, 21, 27, 59 Marxism 21, 32
I n de x 193
MDS see multidimensional scaling (MDS) disbursement 35–36

migration see exit individual voice 45
mobile phones (cell phones) 132 public goods, distinguished from
monopolies 6, 7, 8, 10, 13, 27, 136 18, 45
transport 6, 7, 81 private health 39, 95
see also education; health care; welfare EVMESD Survey 82, 83, 87, 88, 89, 91,
services 96–97, 99, 109
Mullins, Gary 1, 4, 5, 16 loyalty, and 115–16
multidimensional scaling (MDS) 61, 72 private voice see individual voice (IV)
privatization 19, 33
National Health Service (NHS) (UK) 23, 28, product
79, 103, 116, 119, 124, 129, 134 brand loyalty 2, 20, 21, 27, 59
choice 137–39 production and consumption, relationship
establishment of 38, 81 between 35
EVMESD Survey 81–82, 83, 93, 95, 96–97, quality see product quality
137 product quality 18–19, 31
geographical exit, and 105–06 bakery example 2–5, 8, 14, 18
locked in 114–15, 135, 136 protest 1, 9, 10, 16, 80, 91, 127, 130,
private exit, and 115–16 132, 144
reorganization of 81 see also lobby group/lobbying; petition/
neglect see EVLN framework (Rusbult) petitioning; pressure group; public
NHS see National Health Service (NHS) (UK) demonstration; public rally
Nigerian railways 6–7, 13 public demonstration 44, 46, 80, 89, 91, 94,
noisy exit 25–26, 58–59, 105, 106, 127 118, 123, 132
non-exit 4, 31, 45, 73, 145 see also lobby group/lobbying; petition/
Northern Ireland Executive 81 petitioning; pressure group; protest;
public rally
passive people see inert or passive people public goods 17–18, 19, 29, 34–35, 40, 41–43,
Pepsi 1, 2, 5, 14 44, 71–72, 99
Coca-Cola, rivalry with 1, 2 definition
petition/petitioning 9, 15, 16, 41, 44, 45, 46, ‘jointly supplied’ 18, 40
80, 89, 91, 94, 118, 123, 127 ‘non-excludable’ 18, 40
see also lobby group/lobbying; pressure disbursement 35–36
group; protest; public demonstration; individual voice 45
public rally local collective goods problems 40–43
preference revelation 41, 42, 44 preference revelation 41, 42, 44
pressure group 1, 16, 33, 41, 44, 45, 46, 113, 132 private goods, distinguished from 18, 45
see also lobby group/lobbying; p etition/ pure public goods 18, 41
petitioning; protest; public Tiebout exit see Tiebout exit
demonstration; public rally true value 41
private education 34, 47 public rally 17, 18, 80, 86, 89, 91, 94, 118, 123
EVMESD Survey 76, 88, 89, 91, 97, 109–10 see also lobby group/lobbying; petition/
loyalty, and 117–18 petitioning; pressure group; public
private exit 39–40, 43, 89, 99, 106, 128, 133, demonstration
135, 139 public services see local government services;
loyalty and 115–18 state service provision
private goods 18, 19, 40, 41
definition rally see public rally
‘excludable’ 18 referendums 55, 141
‘rival’ 18 representative voice 44
194 I n de x
satisfaction 46, 47–48, 54, 102–04, 130, 131 state education 34, 47, 129, 134
CVP, and 94–95, 99, 100–01, 133 choice 138
CVV, and 94–95, 99, 100–01, 133 EVMESD Survey 80, 82–83, 93, 97, 109–10,
education 47 137
see also EVMESD Survey geographical exit, and 104–05, 106, 128
EVLN framework see EVLN framework locked in 114–15, 135, 136
(Rusbult) private exit 117–18, 128
exit and dissatisfaction 95–97, 101 state service provision 132
expectation, and 48, 71, 96–97 contracting out and privatization 18–19, 33
individual voice and dissatisfaction 92–93, locked in 13, 76, 87, 96–97, 103, 114–15,
99, 100, 133, 135 124, 125, 135, 136, 139, 141, 144
local services 47–48 market failure 34–35
see also EVMESD Survey see also education; health care
summing up 97–99 surveys 9, 28, 50, 54, 55, 62, 64, 67, 73, 132
voice, satisfaction and exit see voice, face-to-face 77, 79
satisfaction and exit online 77–79
Scottish Government Health Department 81 telephone,random-digit dialling (RDD)
SES see socio-economic status (SES) 77–79
silence 4, 5, 10, 11, 30, 73, 143 see also EVMESD Survey; focus groups;
simple relationships 92–97 market research
CVV and CVP, satisfaction and 93–95
individual voice and dissatisfaction 92–93 Tiebout exit 40–43, 47, 51–52, 53, 55, 66, 95,
types of exit and dissatisfaction 95–97 106, 133
social capital 22, 76, 80, 103, 112–13, 115–16, forms of 43
117, 118, 120, 121, 124, 133, 134, 135, trade unions 44, 58, 63, 132, 142
139–40 transport 6, 7, 81
social class 48, 61, 76, 140 tribe, loyalty to 20
voice, and 99–100 trust 65, 69, 104, 118–19, 143
see also socio-economic status (SES) Twitter 132
social investment 22, 63, 66, 68, 75, 76,
103–04, 111, 118, 120, 121, 124, 130, UK public services 75–101
131, 133, 134, 139, 140 education see state education
future investment 120 NHS see National Health Service (NHS)
past investment 103, 135 (UK)
social networking (Internet) 132 structure of 80–83
social-scientific model 26–27 survey see EVMESD Survey
framework, distinguished from 27
socio-economic status (SES) 48, 75, 76, 80, 94, vertical voice 16–17, 43–44, 132
99, 103, 134–35, 138, 141 horizontal voice, distinguished from 16,
civic voluntarism model 99 43–44
voice, and 99–100 voice
see also educational attainment; high choice, and 33–37, 50, 130
income; social class Coca-Cola study 2
state collective see collective voice (CV)
choice and voice 33–37, 50, 130 definition of 9
loyalty to 20, 21, 31, 49 direct voice 44
voice and exit dissatisfaction see satisfaction
game-theoretic model 29–30, 53 educational attainment 34, 43, 48, 53, 54,
separate behaviour model 31–33 61, 77, 99, 103, 134–35, 140, 141
see also local government services elements of see voice (elements)
I n de x 195
EVL mix 15 voice (elements) 9–13

EVLN framework see EVLN framework collective see collective voice (CV)
(Rusbult) individual see individual voice (IV)
EVMESD Survey see EVMESD Survey negative returns 10
exit, trade-off with see exit−voice trade-off nuances of behaviour 9, 45, 59
framework see exit−voice framework nuances of delivery 13
high income 34, 43, 48, 53, 77, 99, residual of exit 10
100, 135 voice, satisfaction and exit 104–15
Hirschman’s original idea 2–6 collective voice and exit 113–14
individual see individual voice individual voice and exit 104–10
loyalty, effect of 69–73, 75, 76 locked in 13, 76, 87, 96–97, 103, 114–15,
see also exit−voice trade-off 124, 135, 136, 139, 141, 144
neglect, and see EVLN framework responses to individual voice 110–13
(Rusbult) voters/voting 17, 28, 32, 33, 34, 41, 44, 46,
noisy exit, and 25–26, 58–59, 105, 106, 127 89, 130
non-voicers 109, 110, 113 see also collective voice vote (CVV)
object of loyalty, and see exit, voice and
object of loyalty wealth see high income
political activity, as 7, 9, 10, 13, 19 welfare services 7, 21, 30, 32, 38, 40, 41, 46,
representative voice 44 80, 81, 138, 139
satisfaction, and see satisfaction agencies and departments 28, 29
separate behaviour model 31–33, 49 see also education; exit voice and welfare;
social class, and 99–100 health care; ‘knaves and knights’;
socio-economic status 75, 76, 99, 103, 134, transport
138, 141 Welsh Department of Health and Social
tax competition, and 55 Services 81
welfare, and see exit, voice and welfare
see also silence YouGov 77, 79

(Theories of Institutional Design) Keith Dowding, Peter John - Exits, Voices and Social Investment - Citizens' Reaction To Public Services-Cambridge University Press (2012)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Theories of Institutional Design) Keith Dowding, Peter John - Exits, Voices and Social Investment - Citizens' Reaction To Public Services-Cambridge University Press (2012)

Uploaded by

Copyright:

Available Formats

E x i t s , Voic e s a n d S o c i a l I n ve s t m e n t

k e i t h d ow di ng is Professor of Political Science in the School of Politics and

pe t e r joh n is Professor of Political Science and Public Policy, School of Public

Other books in the series

© Keith Dowding and Peter John 2012

This publication is in copyright. Subject to statutory exception

First published 2012

Printed in the United Kingdom at the University Press, Cambridge

Library of Congress Cataloguing in Publication data

ISBN 978-1-107-02242-3 Hardback

Cambridge University Press has no responsibility for the persistence or

List of figures page vi

1 Hirschman’s original idea 1

1.1 The voice–exit decision tree page 11

1.1 Responses to decline in quality page 11

4.11 Cross-tabulation of change in expectation of timely

Services: Exit and Voice as a Means of Enhancing Service Delivery’ RES-

Exit, voice and loyalty

conceptualized theoretically to appear in some circumstances rather than

to complain and these complaints will put pressure on government to

improving public services should be obvious, as new public management

Elements of the three elements

Hirschman (1970, p. 31) argues that there might be negative returns to

Table 1.1 Responses to decline in quality

complaining and simply shifts to another product. Finally there is a cat-

Figure 1.2 Dynamic relationship between voice and exit

Voice can be delivered in a variety of nuanced ways. Hirschman sug-

Hirschman defends loyalty as rational in the sense that exit can be a

Table 1.2 Decline primarily arouses

Modified from Hirschman 1970, p. 122

The EVL mix

Criticisms and extensions of the EVL framework

Collective action problems

beneficial each horizontally voiced concern will be in overcoming the col-

1. A good is ‘jointly supplied’ (or ‘non-rival’) if one person’s consump-

privatizing service provision is controversial. Some people believe contract-

Loyalty as an ‘equation filler’

a conference when talking to rival firms. Or a US citizen might be highly

Chapter 4 reports some basic descriptive statistics and examines a few

farewell celebration. And there is an interesting line of research that takes

as a response to the politicians themselves. We see it as a signal to politi-

# Equilibrium Required conditions Outcome

Choice and voice

In many regards these initiatives have, according to objective measures

Table 2.3 Types of goods provision

Production Disbursement Consumption

distribution associated with the market provision of some goods such as

host of other possibilities. Public subsidy of private transactions; private

The idea in both of these circumstances is that competition provides the

Table 2.4 Types of exit in public provision

Type of exit Description Disbursement

one’s children to school (and not providing any educational replacement

The distinct property of a market mechanism from the viewpoint of effi-

Table 2.5 Types of voice in public provision

Type of voice Description

We introduced the individual and collective distinction between voice

Hirschman in the urban context

no progress in the comparative analysis of public services within the exit–

Despite the rhetoric couched in the exit–voice framework, there have

Progress in the analysis of exit and voice in the spatial dimension of

a combination of voice and exit (measured by fragmentation), and mim-

Exit, voice, loyalty and neglect

operationalization of those concepts does not match the original descrip-

List of figures page vi

1 Hirschman’s original idea 1

1.1 The voice–exit decision tree page 11

1.1 Responses to decline in quality page 11

p. < .01, p. < .01 * p. < .05