Susana Mourato, David Pearce, Ece Ozdemiroglu The Price of Virtue The Economic Value of The Charitable Sector

The Price of Virtue
There is no possibility of making a numerical estimate of the total scale of

philanthropic action. Nor indeed would much be gained by attempting this. Indi-
vidual philanthropic agencies differ from one another so widely that they cannot
be made the subject of useful statistical summary.
W. Beveridge, Voluntary Action (George Allen and Unwin, London, 1948).
The Price of Virtue
The Economic Value of the Charitable Sector
Vivien Foster
The World Bank, Washington, DC, USA and CSERGE,
University College London, UK
Susana Mourato
T.H. Huxley School of Environment, Earth Sciences and
Engineering, Imperial College London and CSERGE, University
College London, UK
David Pearce
CSERGE, University College London, UK
Ece Özdemiroğlu
Economics for the Environmental Consultancy Ltd (EFTEC),
London, UK
Edward Elgar
Cheltenham, UK • Northampton, MA, USA
© Vivien Foster, Susana Mourato, David Pearce, Ece Özdemiroğlu 2001
All rights reserved. No part of this publication may be reproduced, stored in a

retrieval system or transmitted in any form or by any means, electronic, mechanical
or photocopying, recording, or otherwise without the prior permission of the pub-
lisher.
Published by
Edward Elgar Publishing Limited
Glensanda House
Montpelier Parade
Cheltenham
Glos GL50 1UA
UK
Edward Elgar Publishing, Inc.

136 West Street
Suite 202
Northampton
Massachusetts 01060
USA
A catalogue record for this book

is available from the British Library
Library of Congress Cataloguing-in-Publication Data
The price of virtue : the economic value of the charitable sector / Vivien Foster …
[et al.].
Includes bibliographical references and index.
1. Charities—Economic aspects—Evaluation. 2. Voluntarism—Economic
aspects—Evaluation. 3. Fund raising—Evaluation. I. Foster, Vivien, 1968–
HV48.P75 2001
338.4'33617—dc21
00–062287
ISBN 1 84064 485 0 (cased)
Typeset by Manton Typesetters, Louth, Lincolnshire, UK.

Printed and bound in Great Britain by Bookcraft (Bath) Ltd.
Contents
List of figures vi
List of tables viii
A note on contributions x
Acknowledgements xi
Preface xii
PART I MEASURING THE ECONOMIC VALUE OF THE

CHARITABLE SECTOR
1 Conceptual foundations 3
2 The benefits of charities to the general public 19
3 The benefits of charities to users: the homeless 72
4 The benefits of volunteering 101
5 The aggregate benefits of the charitable sector: summary 114
PART II CAPTURING THE ECONOMIC VALUE OF CHARITIES
6 Providing fiscal incentives for giving 123

7 Choosing fundraising methods 147
8 Targeting donors 167
PART III POLICY AND SOCIAL CAPITAL
9 On social capital 193
10 Conclusions and policy implications 204
References 209
Index 219
v
List of figures
1.1 Social benefits from charities 12

1.2 Social costs of charities 17
2.1 Importance of charities 33
2.2 Attitudes towards donating 34
2.3 Attitudes towards charitable donations: scope 38
2.4 Attitudes towards charitable donations: foresight 39
2.5 Attitudes towards charitable donations: commitment 40
2.6 Attitudes towards volunteering: motive 44
2.7 Alternatives to volunteering 46
2.8 WTP distribution in CVALL and CVHH 50
3.1 Sources of income 80
3.2 Current expenditure breakdown 81
3.3 Percentage frequency distribution of current weekly income
and expenditure 81
3.4 Duration of homelessness 83
3.5 Percentage of respondents using general hostel facilities 84
3.6 Percentage of respondents using counselling and support
services 85
3.7 Satisfaction with the quality of general hostel facilities and
services 86
3.8 Satisfaction with the quality of hostel counselling and support
services 87
3.9 Frequency distribution of housing preferences across different
time horizons 88
3.10 Answers to the question ‘Would you like this idea to actually
happen?’ 89
3.11 Distribution of WTA compensation amounts 92
3.12 Predicted expenditure breakdown after the proposed scenario 96
4.1 Breakdown of differences between valuation methodologies
for males in 1993 111
4.2 Breakdown of differences between valuation methodologies
for females in 1993 111
6.1 Percentage frequency distribution of tax price 134
6.2 Percentage frequency distribution of efficiency price 136
vi
List of figures vii
6.3 Percentage frequency distribution of time-value price 137

7.1 Percentage frequency distribution of average donations for
different fundraising methods 157
7.2 Marginal effect of an approach on the probability of a gift
against number of approaches 164
8.1 Positioning of some of the UK’s major environmental groups in
characteristics space 173
8.2 Fitted probability of group membership against continuous
explanatory variables 178
8.3 Marginal cost–revenue ratio against optimal number of
targeting variables for the ‘AND’ strategy 184
List of tables
2.1 Bid vectors for CVALL and CVHH 23

2.2 Payment ladder used in the general public survey (CV version)
and choices of a hypothetical respondent 24
2.3 Sample question used in the general public survey – CR version 26
2.4 Socioeconomic characteristics of different subsamples 29
2.5 Comparison of participation rates in different forms of
donations 31
2.6 Summary of existing donations of time and money 32
2.7 Correlation between different motives for monetary donations 36
2.8 Correlation between different approaches to monetary
donations 41
2.9 Correlations between different motives for volunteering 45
2.10 Percentage of respondents in each response category 47
2.11 Yearly WTP estimates for different models and uncertainty
assumptions 48
2.12 Yearly WTP for CVALL and for CVHH from payment ladder 50
2.13 Distribution of total WTP across different areas of charitable
activity 52
2.14 Description of explanatory variables 54
2.15 Valuation functions for CVALL and CVHH: payment ladder
results 56
2.16 Results of contingent ranking version 59
2.17 Summary of debriefing questions across subsamples 60
2.18 A comparison of WTP estimates across models 62
2.19 Relative importance of different charitable subsectors 64
3.1 Definition of services provided by different types of hostels 77
3.2 List and characteristics of selected hostels 78
3.3 Summary statistics of selected socioeconomic variables 79
3.4 Housing history 82
3.5 Correlation between attitudes towards the proposed scenario
and attitudes towards hostel services 90
3.6 WTA compensation to give up hostel services 91
3.7 Description of explanatory variables 93
3.8 Valuation function: WTA compensation for hostel closure 94
viii
List of tables ix
3.9 Reasons for not answering the valuation question 97

4.1 Summary of Heckman selectivity models used to fit net hourly
wage 104
4.2 Estimated value of volunteer time using the opportunity cost
approach 105
4.3 IGS volunteer categories against equivalent NES occupational
categories with corresponding average male and female hourly
wage rates for 1993 107
4.4 Estimated value of volunteer time using the replacement cost
approach 108
5.1 The net social value of housing and homelessness charities,
1997 117
5.2 The net social value of all charities, 1997 118
6.1 Summary of literature on price and income elasticities of giving 126
6.2 Contrast between tax treatment of charitable donations in the
USA and the UK 127
6.3 Summary of literature on price and income elasticities of
volunteering 130
6.4 Monthly philanthropic participation rates against different
price variables 138
6.5 Mean monthly donations against different price variables 139
6.6 Heckman selectivity models for giving and volunteering 142
6.7 Price and income elasticities for different types of
philanthropic activity 144
7.1 Cross-tabulation of fundraising approaches and giving
behaviour 154
7.2 Summary statistics for different fundraising methods 155
7.3 Ordinary least squares regression models for the number of
fundraising approaches 159
7.4 Selectivity models for all philanthropic gifts with controlling
for fundraising effort 162
7.5 Estimated marginal effects of fundraising 163
8.1 A comparative profile of some of the UK’s major
environmental groups 172
8.2 Environmental group membership patterns from BSAS 174
8.3 Summary of coefficient estimates for logit models of group
membership 177
8.4 P and N against number of variables used in targeting 183
8.5 Overlap ratios for target population segments between groups
for the ‘AND’ strategy 188
A note on contributions
Chapters 1–5 are the result of teamwork between EFTEC (Economics for the
Environment Consultancy) and CSERGE (Centre for Social and Economic
Research on the Global Environment) on a project funded by the Charities
Aid Foundation. Contributing authors are Vivien Foster, Susana Mourato,
Ece Özdemiroğlu, Steve Dobson, David Pearce and Tannis Hett.
Chapters 6–8 have been prepared by Vivien Foster under an award from
the UK Economic and Social Research Council (ESRC). Vivien wishes to
thank the ESRC for their support.
Chapters 9–10 were drafted by David Pearce, Vivien Foster and Susana
Mourato.
x
Acknowledgements
Our overwhelming debt is to Cathy Pharoah, the Head of Research at the

Charities Aid Foundation, and Michael Brophy, Chief Executive of the Chari-
ties Aid Foundation. It was an idea of Michael Brophy’s that sparked off the
partnership between CSERGE, EFTEC and the CAF. Michael was convinced
that there had to be a better way of determining the social value of the
voluntary sector. Over a London lunch several of us rashly agreed that what
we had been doing for years in the sphere of environmental economics could
be applied to the voluntary sector. We would have an answer in months.
Several years later this book has emerged, much of it containing the results of
the work that was funded by the CAF. We are grateful to the CAF for that
sponsorship, but more grateful still for the trust Cathy and Michael had in us
to do the work. We are aware that what we had done is partly controversial,
but science makes no forward moves without controversy.
A major part of the work has been contributed by Vivien Foster and Part II
is adapted from Vivien’s doctoral thesis for the Department of Economics at
University College London.
Finally, we wish to thank Steve Dobson, who was involved in the initial
stages of the CAF work when he was at EFTEC. He helped shape the ideas
and ‘models’ that we subsequently applied.
VF, SM, DWP, EÖ
London, August 1999
xi
Preface
Economists have paid some, but not much, attention to the charitable sector in
today’s economies. Also known as the voluntary or ‘non-profit’ sector, the
activity of giving time and money for the benefit of other people, the environ-
ment and cultural assets has grown to such proportions that it effectively forces
detailed scrutiny. Giving takes many forms, and this presents problems of
defining just what the voluntary sector is. A major international research project
– the Johns Hopkins Comparative Non-profit Sector Project – classifies volun-
tary organizations according to the following characteristics: formal activity in
the sense of having rules, self-governing and independent of government (though
not financially), acting primarily as a non-business, not distributing profits, and
being voluntary in terms of donations of time or money or both (Kendall and
Knapp, 1995). On this basis, the voluntary sector accounts for around 2 per
cent of total employment in Italy and Japan, 4 per cent in the UK and France,
and 6 per cent in the USA (Kendall, 1996). Across eight countries (UK, USA,
Sweden, France, Germany, Hungary, Italy and Japan) the operating expendi-
tures of the non-profit sector accounted for an average of 4.6 per cent of those
countries’ GNP in 1990, an absolute magnitude of $614 billion (Salamon et al.,
1995). By any yardstick, the voluntary sector is huge, and it is growing.
The problem with prevailing measures of the size of the non-profit sector is
that they do not measure the true ‘social value’ of the sector. Social value must
somehow reflect the output of the sector. Yet the sum of donations and grants
does not measure output; it measures input, that is, the cost of supplying
charitable services. In fact it does not even fully measure inputs, since the value
of volunteer time is not taken into account. Similarly, contributions to GNP are
not measured by social value but, again, by the costs of supplying the services.
This approach to GNP measurement is familiar: many economic activities are
not bought and sold in the marketplace, so there are no ‘revenues’ to observe.
In such circumstances it is commonplace to measure the contribution to GNP
in terms of the costs of providing the service, as with public education and
public health services. But the resulting measures are imperfect and potentially
misleading. One aim of the current volume, then, is to pursue the idea of
measuring the economic value of the charitable sector by looking at measures
based on output, not input. Put another way, we ask the question: what are
people willing to pay for the services provided by the charitable sector?
xii
Preface xiii
To our knowledge, the answers to this question, reported in Chapters 2–6

of this volume, represent the first attempt ever to measure the value of the
charitable sector in terms of willingness to pay. In case this looks like a straw
man, inventing a measure that has no particular rationale, willingness to pay
is precisely the measure that is used to measure the output of the marketed
sector of the economy. Willingness to pay (hereafter WTP) reflects individu-
als’ preferences for a good or service, whether that be the contents of a
supermarket trolley, the conservation of a historic building, wildlife preserva-
tion, or the provision of care for the aged. We therefore treat charitable
services just like any other economic good. The interest lies in the fact that
those services are not directly marketed, and hence we have to resort to
techniques of ‘non-market valuation’ to find out the WTP for them.
Non-market valuation techniques involve discovering what people would
be willing to pay if only there were a market. Broadly, two techniques are
involved. The first looks for existing markets and asks if they embody in
some way the value of the associated good or service we are interested in.
Suppose the problem is the economic value of cleaning up air pollution. We
do not buy and sell pollution, but we do buy and sell houses and we know that
house prices reflect the neighbourhood amenities surrounding those houses,
including the quality of the air. This ‘revealed preference’ approach, then,
looks for an associated, or complementary, market and estimates WTP from
observations in that market. The second technique is familiar to anyone who
has ever carried out, or been the subject of, market research: we ask people
for their WTP. Sophisticated questionnaires are constructed with the aim of
either asking directly for WTP (what are you willing to pay?), or asking
whether respondents are willing to pay a particular price (are you willing to
pay X? – yes/no), or asking for individuals’ rankings of alternative options
where there is a link to the cost of providing the option. In the last case, WTP
is inferred rather than stated directly. These contingent valuation and contin-
gent ranking approaches have become very powerful in recent years and are
widely used in environmental economics and health economics. This is the
first time they have been applied to charitable services.
Because the detail of the valuation study is extensive, we have presented a
summary of the results of the study in Chapter 5. Chapters 1–4 explain the
analytical foundations for the study and the questionnaire results. Three
sources of economic value are identified: first, the WTP of the general public
to ensure that charitable services continue to be provided; second, the WTP
of the beneficiaries of the services to maintain those services; third, the
benefits to volunteers from the opportunities provided by charities.
A questionnaire-based approach is used to establish the WTP of the gen-
eral public. But we treat the beneficiaries’ WTP separately. Because there are
so many different groups of beneficiaries, we cannot possibly survey them
xiv The price of virtue
all. We deliberately chose a group of users where we would anticipate serious

potential problems of using our valuation techniques: the homeless. Our
reasoning was that if our approaches worked in this complex and sensitive
case, they would probably work in many others.
We make no apology for treating charitable services like other economic
commodities, but we fully appreciate that some people will find the approach
questionable. For example, the very reason we have a voluntary sector ap-
pears to be precisely because the market system does not provide those
services. Why does the market system fail? It fails because many of the goods
and services do not have apparent markets – recall the air pollution example.
It also fails because market systems operate through prices and those prices
may well exceed the ability to pay of the vulnerable groups who are the very
targets of many charities. But great care needs to be exercised when using
these failures of the marketplace to criticize the economic valuation ap-
proach.
First, we have already seen that there are often markets in the non-market
service or benefit: the example of the housing market for air pollution can be
extended. Historic houses and archaeological sites have implicit market value
because people spend money travelling to see them. Their costs of travel
provide a clue to finding their ‘price’.
Second, even if markets are absent, it does not follow that people would
not be willing to pay if there were a market. If we can find that WTP, then it
may be possible to ‘capture’ at least part of it by establishing an institution
that translates the WTP into actual payments. Charges for entry to parks,
conservation areas, cathedrals and so on are all examples of ‘capture’.
Third, charitable services are rarely provided by charities alone. Care of
the elderly is provided by charities, by local government and by the private
sector. There are potentially comparable markets out there. But this rationale
has to be treated with caution. The users of charitable services are sensitive to
who provides them, and there are distinct preferences for provision by one
agent rather than another. Additionally, while the service may appear superfi-
cially the same, there are often marked qualitative differences. None the less,
there is substitutability across a range of charitable services.
Fourth, as Chapter 5 concludes, charities compete for scarce resources.
Money received from donations could have been used elsewhere. Money
received from government grants could have been used to provide other
public services. It is important therefore to ensure that charities are the best
way of providing those services. But we cannot know that unless we know
what the ratio of their output is for every pound or dollar they receive. It may
be more efficient to provide a service through local government rather than
via a charity, or vice versa. Cost-effectiveness indicators exist, but they have
limited validity once it is recognized that outputs vary in the range of benefits
Preface xv
they provide and in the quality of service provided. We will not know more
about comparative efficiency until we have some broader calculus of cost-
effectiveness which, at the moment, does not exist on any widespread scale.
We hope our work begins the process of achieving that.
Finally, the techniques we have used are familiar in the world of environ-
mental economics. When environmental economics began there were concerns
that the environment was being turned into ‘commodity’, and that this was
illicit because the environment is ‘beyond price’, and somehow not to be
brought within the measuring rod of money. Some critics still argue that. But
the simple logic of opportunity cost – that whatever we do uses resources that
could have produced some benefit elsewhere – remains. Anyone who argues
that costs and benefits are irrelevant to social decision-making has to explain
how cost and benefit can be ignored. All kinds of moral arguments can be
invoked for doing so, but the fact that charitable services use resources to
provide those services remains. This always means that those resources could
have been used to provide some other service, which may just as easily serve
some moral purpose. Morality is not irrelevant, but it must account ad-
equately for opportunity cost.
Part I of the book is concerned with demonstrating the economic value of
charities. Part II focuses on the next logical question: if people are willing to
pay more than they actually pay for charities, how can this extra WTP, this
surplus, be captured and turned into flows of income for charities?
Chapter 6 addresses the question of the role that government fiscal policy
can play in stimulating the flow of resources to the charitable sector. It has
long been argued that donations can be far more effective if they are ‘tax-
efficient’, that is, more donations are generated if tax allowances on giving
are provided. While it may seem obvious that giving tax incentives should
increase donations, this is not borne out in practice. A higher rate of marginal
tax in a context where there are tax allowances should increase giving be-
cause more tax is written off for each pound or dollar given. But higher tax
rates also mean less after-tax income, so there is an income effect, which
depresses giving. The two forces work in opposite directions. The empirical
evidence reported in Chapter 6 indicates that the net effect of tax incentives
on giving is positive for the UK: giving is higher in the presence of tax
incentives than it would otherwise be. However, tax incentives do not provide
a very large stimulus to giving. Indeed, the extra donations stimulated by
fiscal incentives are not large enough to offset the associated loss of tax
revenues to the Treasury. A possible reason for the limited impact of tax
incentives is the relatively restrictive scope of mechanisms for tax-efficient
giving in the UK.
In practice it is the efforts of fundraisers, rather than any tax incentives,
which provide the greatest stimulus for philanthropic giving. Chapter 7 ex-
xvi The price of virtue
amines the relative efficacy of different fundraising methods, in particular

those which involve direct face-to-face contact between fundraiser and poten-
tial donor versus those which rely on remote methods such as telephone,
television, mailshots and so forth. The evidence shows that people are signifi-
cantly more likely to give when they are approached face to face. However,
the average size of donations received tends to be larger via remote fundraising
methods. Overall, face-to-face methods present charities with the highest
return to each fundraising approach. However, people’s generosity dimin-
ishes markedly as they are approached an increasing number of times.
Chapter 8 considers the extent to which charities should target the popula-
tion from which donations are sought. It is well known that many charities do
this. But what exactly is the return to the charities from this targeting?
Chapter 8 finds that targeting does indeed elicit more donations than ‘scatter-
gun’ approaches, although some targeting variables are shown to be much
more effective than others. But it also shows that larger charities have less to
gain from targeting than smaller charities. Since targeting involves reducing
the size of the population that is ‘trawled’ for donations, the effect of target-
ing for large charities is to lose out from the reduction of scale that comes
with targeting and to gain from the targeted population. There is a trade-off.
For small charities, the scale effect is less important than the targeting effect:
they stand to gain most from targeting. Overall, Chapter 8 offers the elements
of a theory of optimal fundraising.
Part III raises an issue for discussion: can our measure of the economic
value of charities be used as a measure of social capital? There is now a
substantial interest in the notion of ‘social capital’ – the ‘glue’ that holds
society together and without which there is mistrust and social enmity that
interferes with the smooth workings of the economy. Indicators of social
capital are scarce. There are numerous negative indicators – measures of
inequality, corruption and social breakdown such as crime, divorce and fam-
ily disruption – but there are few positive indicators based on broad
questionnaires about feelings of trust, or on measures of civil and political
liberties. Yet the most obvious indicator of the extent to which human beings
care for each other, the relationship of giving, has been ignored in the social-
capital literature. We offer the observation that, in measuring the ‘output’ of
the charitable sector, we are contributing to that measure of social capital.
Chapter 9 raises the possibility of using the value of the charitable sector as at
least a component part of the definition of social capital, a departure from the
existing, and rapidly growing, literature on the subject.
Chapter 10 completes the book with some summary conclusions on the
size of the charitable sector and on the mechanisms for capturing the surplus
value we claim we have identified.
PART I
Measuring the Economic Value of the Charitable

Sector
1. Conceptual foundations
1.1 PURPOSE OF THE BOOK

The charitable sector uses inputs such as land, capital and paid and unpaid
labour to produce outputs such as care of the elderly, shelter for the homeless,
environmental amenity and so on. In this respect the sector is just like any
other productive activity in that it transforms resources into something that
society values.
However, unlike many other productive sectors, the goods and services
produced by charities are not openly traded on the market. When goods and
services are bought and sold on the market, the price at which they are traded
provides a real indication of the value consumers place on the corresponding
outputs. In the case of charities, services are usually provided at zero, or
highly subsidized, cost to specific beneficiary groups who typically lack the
means to secure the services in any other way. The price therefore does not
provide any indication of the value which beneficiaries place on the service
provided, making it difficult to put a monetary figure on the output of the
charitable sector.
The absence of any measure for the value of the services provided by
charities is problematic for a number of reasons. First, as already noted
above, if output could be measured, it would enable an economic calculation
of the value added of charitable organizations, thereby providing a more
accurate economic measure of their size. This would permit them to be
integrated into the national accounts on an equal and consistent basis with the
private sector.
Second, there is a great deal of interest in the concept of social capital, the
‘glue’ that holds society together – such as family and other bonds, the
assumption of trust in contractual relationships and socially responsible be-
haviour. From the economist’s standpoint, one reason for this interest is that
societies where there are strong social bonds – good endowments of social
capital – may develop faster than those where there is less social capital
(Fukuyama, 1995; Putman, 1993). However, measuring social capital is prob-
lematic. Some insight into an indicator of social capital may be obtained by
considering the size of the charitable sector relative to the market sectors of
an economy. Participation in volunteer work is a sign of social concern. It
3
4 Measuring the economic value of charities
should be noted, however, that there is a risk that it might equally well be
argued that the bigger the charitable sector, the more evidence there is of
breakdown of social capital. This is because charitable services are often
remedial, for example drug rehabilitation or working with young offenders.
Third, without a measure of output, there is no satisfactory measure of the
efficiency of the sector. Cost-effectiveness measures such as cost per unit of
service provided might be appropriate, but even this requires some idea of
output such as lives saved, illness avoided and so on. Cost-effectiveness
measures are valuable, but they only permit answers to a limited set of
questions. Broader measures of efficiency, or real output (in terms of value
added) divided by input, should throw some light on whether voluntary
agencies are more or less efficient than private or government agents at
providing social services. In the longer term, it might enable some compari-
son of the relative efficiency of the charities themselves.
Fourth, output measures require a focus on the users of charitable services,
the customers, whether as visitors to a National Trust property or as homeless
people. Adopting non-market valuation procedures serves as a reminder that
the important social groups influencing the structure and size of the voluntary
sector are the users themselves, whether the general public, the target group,
or both.
Finally, and of immediate policy relevance, a true analysis of charitable
output serves to highlight the role that government funding should play in the
financing of charities, an issue that will be taken up further in Chapter 5.
The purpose of this book is to develop and illustrate a methodology for
valuing the output of the charitable sector. This introductory chapter lays the
conceptual foundations for this valuation exercise. The chapter begins by
reviewing existing approaches to measuring the scale of charitable activity,
and explains why these fall short of providing a value for the outputs of the
sector. Thereafter, the methodological foundations for a new approach to
measuring the value of the output of the charitable sector are derived from
standard welfare economics. The discussion goes on to explain how this
approach can be implemented by adapting a number of valuation techniques,
originally developed in the environmental economics field, to the charitable
context. These methodologies permit the estimation of monetary values for
goods and services which are not traded in the marketplace. Finally, the
remainder of Part I of this book illustrates a complete application of the
proposed methodology.
Conceptual foundations 5
1.2 EXISTING INDICATORS OF THE SIZE OF THE

CHARITABLE SECTOR
To motivate the need for a new methodology to measure the value of the
output of the charitable sector, this section reviews a number of measures that
are currently used to measure the significance of the charitable sector and
explains why they cannot be regarded as reliable measures of output.
1.2.1 Income Measures
The total operating income to a charity or group of charities from all sources
can be summarized as private earned income plus private giving plus income
from government (Kendall and Knapp, 1995). Kendall and Knapp (1995) find
that the total operating income of the ‘broad voluntary sector’ in the UK is
£29.5 billion and of the ‘narrow voluntary sector’ £12.3 billion. The broad
voluntary sector is defined by these authors as comprising all organizations
which meet the core definition criteria, which are that they are (1) formal, (2)
independent of government and self-governing, (3) not profit-distributing and
primarily non-business, and (4) voluntary. The narrow definition is a subset
of the above organizations, excluding those which are regarded as too exclu-
sive, not independent enough, insufficiently altruistic or not particularly
oriented to providing public benefits. Amongst the organizations excluded
from this narrow definition are those providing recreation, for example sports
clubs, primary, secondary and higher education, trade unions and business
associations. Total operating incomes of the top 500 fundraising charities are
also published annually by the Charities Aid Foundation (the latest at the
time of writing is CAF, 1999).
It is important to recognize that, though informative, such statistics do not
give us any clear indication of the value of charities to society. There is no
certainty that a charity with a higher operating income yields greater social
benefits, since this will depend on how effectively the income is used in the
provision of services. Indeed, such statistics merely reflect the ability of
charities to raise funds from all sources, but not the social value of the use of
such funds. Although fundraising is a very important aspect of charities’
activities, the interest here is the ability of charities to add to the value of
their income by increasing social welfare.
None the less, it could be argued that income provides a reasonable proxy for
the social value of a charity, inasmuch as the income of charities derives from
voluntary contributions which in and of themselves reflect the social value
which the different donor groups place on the services provided by charities.
However, there are a number of problems with this argument. A more detailed
consideration of the economic context of charitable contributions from each of
the different sources, undertaken below, illustrates that the link between in-
come and social value is more tenuous than may at first sight appear.
Private earned income

Kendall and Knapp (1995) show that private earned (commercial) income
can be separated into four main subcategories. These are user fees, sale of
goods, membership dues and investment income. Since membership dues are
in many cases a form of donation, and investment income is interest on past
income, attention here focuses on user fees and sale of goods.
As noted above, charities rarely charge users full cost rates for the services
they provide, since they are usually targeting disadvantaged groups who for
various reasons may be unable to pay full market rates for the services
provided. Consequently, user fees are likely to underestimate the value of the
services provided to beneficiaries.
The issue of the income from sales is even more problematic. When buying a
good from a charity shop or a ticket to a charity fundraising event, for example,
the consumer will be valuing both the good and the output of the charity in
taking the decision to buy. In this sense, the revenue from the sale overesti-
mates the value which the purchaser places on the services provided by the
charity. A further complication arises because, in some cases, the goods and
services sold by charities for fundraising purposes represent donations on the
part of third parties. For example, consider a situation in which a charity shop
sells a donated second-hand coat for a price of £10. In this case, the coat
represents a donation in kind reflecting the value that its original owner places
on the activities of the charity. However, the value of the second-hand coat to
the original owner may be very different from the price at which the charity
succeeds in selling on the item to the eventual purchaser. Hence the sale value
that appears in the charity accounts is not a reliable indication of the value that
the original donor places on the services provided by charities.
For all of these reasons, it is very difficult to draw any conclusions about
the social value of the output of charities from data on sales revenues. In the
case of user fees, revenues may understate the value of the services received
by the beneficiaries, while, in the case of sales of goods, revenues are likely
to overestimate the value that the purchasers of those goods place on the
services provided by charities.
Private giving
Kendall and Knapp (1995) categorize private giving into three main forms:
individuals, companies and grant-making trusts. It seems likely that such
donations reflect to some degree donors’ perceptions of the social value of the
services provided by charities. However, there are several reasons to believe
that private giving may not be an accurate reflection of social value.
First, there is a well-known tendency for people to free-ride on the generos-

ity of others. That is to say that many people who value the output of charities
may none the less fail to give because they believe that the services will be
financed anyway by other donors even if they do not give themselves. Thus
free-riders seek to benefit from the services provided by charities without
incurring any of the costs. In some cases, free-riders may avoid giving alto-
gether, while, in others, they may simply give less than the true value that they
place on the service. Laboratory experiments suggest that free-riding may
depress the voluntary contributions by as much as 50 per cent relative to their
full potential value (Foster, 1999). Consequently, voluntary giving may seri-
ously underestimate the full value that people place on charitable services. The
issue of free-riding by donors will be addressed in some depth in Chapter 2.
Second, a significant component of charitable giving is in kind, most
notably by volunteers who give time rather than money, but also by private
companies which may donate their services or private individuals who some-
times give second-hand goods. Estimates of charities’ income rarely attempt
to place a value on donations in kind, largely because of the methodological
difficulties of valuing such donations. This omission is particularly serious in
the case of volunteering. For example, using average hourly wage rates to
value volunteer time, the Volunteer Centre UK (1995) finds an economic
value for volunteering of £41 billion. While Foster (1999) shows that, even
taking the conservative approach of valuing volunteer time at the minimum
wage (that is currently £3.50 per hour), the aggregate value of time volun-
teered each year in the UK (at £3.5bn for 1992) exceeds the aggregate value
of private monetary donations (at £2.9bn for 1992). A detailed discussion of
alternative methods for valuing volunteer time is provided in Chapter 4.
Third, charitable giving entails certain transaction costs which may dis-
suade donors, particularly where relatively small gifts are concerned. Examples
of transaction costs include identifying suitable recipient charities, making
telephone calls, setting up covenants, writing cheques and so forth. The
problem of transaction costs, together with the strategies that charities may
use to overcome them, will be further explored in Chapter 7.
Fourth, charitable giving may not always be a reflection of the value that
donors place on the activities of the charity. It may instead be dictated by
social pressures or norms. For example, Salamon et al. (1995) identify the
problem of philanthropic paternalism, whereby the priorities of the voluntary
sector come ‘to be shaped by the preferences not of the community as a
whole but of its wealthy members’ due to their greater ability to donate. The
question of motivations for philanthropic giving will be investigated more
fully in Chapter 2.
For all of these reasons, private giving is unlikely to provide an accurate
estimate of the full value that society places on the services provided by
charities. The problems of free-riding, donations in kind and transaction costs

suggest that donations will underestimate the value of charitable services,
while the problem of motivations for giving suggests that donations could be
an overestimate.
Income from government

In general, voluntary contributions will fail to provide the socially optimal
level of funds due to the free-rider problem. Therefore, it could be argued that
the addition of government funds could theoretically be regarded as an at-
tempt to redress the shortfall of voluntary donations to charitable organizations
relative to their true social value. In practice, however, government expendi-
ture decisions are rarely made on the basis of these sorts of criteria. Government
assistance is most likely to go to those causes that have majority support due
to the constraints of the political system. This would not ensure that funds are
allocated to those organizations yielding greatest social value, as different
individuals may place very different values on different outputs.
1.2.2 Measures of Expenditure
In addition to income information, the Charities Aid Foundation also pub-

lishes figures giving the total expenditure of the top 500 charities and how
this is divided amongst charitable expenditure, fundraising, administration
and general expenditure. Charitable expenditure means expenditure directly
on the provision of charities’ services, or in other words the cost of producing
these services. It is thus a measure of the part of income which is spent
directly on providing benefits to the charity’s target group. As in the case of
income, expenditure measures are unlikely to provide an accurate reflection
of the value of charitable output.
First, there is no reason to expect that the value of charities’ output should
be commensurate with the costs of production. Indeed, to the extent that
charities add value to the resources they use, the value of outputs should
exceed the cost of the associated inputs, although the value of output could
conceivably be less than expenditure if resources are being used inefficiently.
A charity that spends more than another does not necessarily produce more
output; it may simply be less efficient.
Second, expenditures exclude the contribution of volunteers who, by defi-
nition, are not paid for the services they provide. In this sense, published
expenditure figures are not even an accurate reflection of the use of resources
by the charitable sector. This omission creates a significant distortion in
making comparisons between charities. Thus one charity could spend consid-
erably less than another but produce much more output if it is relying to a
greater extent on volunteers than on paid employees.
1.3 MEASURING THE ECONOMIC VALUE OF THE

CHARITABLE SECTOR
The previous section explained why traditional income and expenditure meas-
ures of the scale of charitable activity fail to provide a reliable measure of the
value of the goods and services which charities produce. The purpose of this
section is to develop an alternative approach which meets this objective.
The proposed approach builds upon the foundations of standard welfare
economics, and takes advantage of techniques developed in the field of envi-
ronmental economics in order to permit the application of these concepts to
the charitable sector.
1.3.1 Foundations in Welfare Economics
Welfare economics is concerned with the level of well-being or welfare of

society. Economic theory assumes that human well-being is determined by
the fulfilment of people’s preferences. A benefit is defined as anything that
increases human well-being and a cost as anything that reduces human well-
being. The intensity of a person’s preferences can be measured by their
maximum willingness to pay (WTP) for a benefit (or for the avoidance of a
cost) or their minimum willingness to accept (WTA) compensation for toler-
ating a cost (or foregoing a benefit). These measures correspond to the value
which a person places on changes in the quantity or quality of a good or
service. This is because they reflect the extent to which that person is willing
to sacrifice alternative uses of their limited income in order to secure that
particular outcome.
The well-being of society in turn is assumed to depend on the well-being
or utility of the individuals comprising society. A social welfare function
provides a way of aggregating the well-being of individuals to obtain the
overall well-being of society. The choice of welfare function is a subjective
one and will reflect ethical judgements about the relative importance of the
well-being of different individuals.
A generalized social welfare function can be stated as follows:
n
W (U1 , U2 , …, Un ) = ∑ aiUi ,
1
where U is the utility level of individual i from 1 to n and Σ indicates

summation. The multipliers a1, …, an indicate the weight accorded to each
individual’s well-being in the overall welfare of the society. These multipliers
have sometimes been used to give more weight to the well-being of individu-
als with relatively low income, in order to address issues of equity in the
evaluation of social welfare. However, this practice has fallen into disuse.
This is partly in response to the subjectivity entailed in specifying such
distributional weights. Moreover, it also reflects the view that income distri-
bution is best dealt with as a separate issue in its own right, and not brought
into welfare analysis of other issues.
Consequently, a popular form of the social welfare function is to set a1 = a2
= … = an = 1. The resulting Benthamite utility function is a simple sum of
individuals’ utility level, and thus effectively treats the well-being of each
person in society as equally important. This implies that society is indifferent
or unprejudiced between similar increases in well-being for different indi-
viduals; that is, a unit of well-being is equally valued regardless of to whom it
accrues.
One of the main uses of social welfare functions is to measure changes in
the welfare of society. The value of a welfare change, if we apply the Bethamite
welfare function, will be equal to the sum of the changes in the utilities of
individuals:
Change in total social welfare = ∆W = Σ∆Ui

= total social benefits – total social costs
This change in total social welfare can also be referred to as the net social or
economic value. One of the fundamental decision rules in economics is that
activities should only be undertaken when their net social value is positive.
Following the terminology developed above, the total social benefits of a
charity’s output are the sum of all the benefits that accrue to members of
society as a result of the existence of the charity, which they would not
otherwise receive. The provision of charitable services also entails substantial
costs. For example, the donations and grants that charities receive could
alternatively be spent elsewhere in the economy to provide other social
benefits. These foregone alternative benefits (or opportunity costs) represent
the total social costs of providing charitable services. Therefore the net social
value of charities is equal to the total social benefits minus the total social
costs, or in other words the difference between the overall level of social
welfare with and without the existence of the charity. To summarize:
● Total social value (TSV) of charities = total social benefits that accrue
to individuals from the existence of the charity.
● Total social cost (TSC) of charities = total foregone benefits (or oppor-
tunity costs) from alternative uses of resources devoted to the provision
of charitable services.
● Net social value (NSV) of charities = total social benefits – total social
costs of the charity.
1.3.2 Total Social Value of Charities
The analysis above indicates that to value the contribution that charities make
to social welfare, all the benefits that accrue to different members of society
as a result of the existence of the charity must be taken into account. This
requires the identification of the various groups of society who benefit from
the services provided by charities. In general terms, society can be divided
into two groups of beneficiaries: the target group (or the direct users) and the
rest of the society (or the indirect users).
Consider first the benefits to the target group of a charity, that is to say those
people who actually use the charity’s services. Direct recipients could include,
for example, handicapped people, the homeless, cancer sufferers and so on. In
many cases, the target group will be well defined and selected by the charity.
However, in other cases, the direct users or beneficiaries may simply be those
members of society who choose to take advantage of the services offered by the
charity, for example visitors to a museum or National Trust site.
Next there are the benefits to the rest of society, meaning those who are not
direct users of the charities’ services. These indirect benefits take a variety of
forms.
First, there are altruistic benefits. People may derive benefits simply from
observing or knowing that a charity’s services are provided to the target
group, because they feel altruistically towards that group. It is important to
note that altruistic preferences may sometimes be paternalistic, in the sense
that they do not necessarily coincide with the perceived self-interest of the
target group. For example, people may think that it is best for the homeless to
stay in overnight shelters even when the homeless themselves may prefer to
sleep rough on the streets.
Second, there are external benefits. Some of the services provided by
charities provide spin-off benefits to those who do not form part of the target
group. For example, the existence of a charity providing youth clubs may
have a positive effect on the local area in terms of reduced crime and more
peaceful streets, over and above the direct benefits to the young people
themselves. Local residents will consequently benefit from the existence of
the charity, even if they do not feel altruistically disposed to the young people
of the neighbourhood.
Third, there are option benefits. Even though people may not be direct
beneficiaries of charities’ services at the present time, it is possible that they
might require such services in the future. For example, someone who is
perfectly well at the present time might in the future become severely de-
pressed and require the services of the Samaritans. A person in this situation
may value the existence of the Samaritans as an insurance policy against their
own future need.
Fourth, there are ‘warm-glow’ benefits. Such benefits are due to what
Andreoni (1990) terms impure altruism, whereby donors gain satisfaction
purely from the knowledge of their own act of generosity as donors over and
above any altruistic sentiments they have towards their beneficiaries. The
‘warm glow’ may be a private feeling of moral well-being, or may reflect the
social recognition or status attached to being a donor.
Fifth, those who volunteer may obtain private benefits in the form of moral
satisfaction, social contacts, work experience, acquisition of new skills or
opening up of career opportunities. In fact, in some cases, volunteers may
even be members of the charities’ target group and have therapeutic or reha-
bilitative reasons for volunteering.
CHARITIES
Indirect benefits Direct benefits
Warm glow, Volunteer, altruistic, External, altruistic

altruistic, option, external benefits benefits
external benefits
Donors Volunteers Target groups
Rest of society
Figure 1.1 Social benefits from charities
Figure 1.1 provides a graphical representation of this typology of the total

social benefits from the charitable sector.
In order to estimate the total social benefits of the charitable sector accord-
ing to the framework developed above, it is necessary to aggregate the change
in individuals’ well-being that results from the provision of charitable serv-
ices compared with a situation where charities do not exist.
Regarding the benefits to volunteers, these can be estimated in terms of the
opportunity cost of the time that the volunteer gives up to the charity. The
reason is that volunteers could be expected to give time up until the point
where the marginal benefit of an hour of volunteering is equal to the marginal
benefit spent in the volunteer’s next most preferred activity. This argument
and the procedure for valuing volunteer benefits on this basis is further
developed in Chapter 4.
In the case of the target group and the wider beneficiaries in the rest of
society, a lower-bound estimate of welfare already exists in the form of the
actual fees paid for charitable services as well as the donations voluntarily
contributed. In these cases, the problem lies in estimating the additional
welfare that is not captured through these actual payments.
There are two possible methods for measuring this additional well-being.
The first is to consider the maximum amount of money that the individual is
willing to pay to avoid the loss of the charities’ services (their WTP). The
second is to estimate the minimum amount of money that the individual is
willing to accept in compensation for the loss of the charities’ services (their
WTA).
In both cases, the change in the person’s well-being as a result of the
existence of charities is being valued in terms of the equivalent amount of
money that would need to be paid by (WTP) or given to (WTA) the indi-
vidual in order to bring about a similar change in their well-being.
In the context of charitable services, WTP will not always be a very
appropriate measure of welfare. The reason is that WTP clearly depends upon
ability to pay, and charitable services are often provided precisely because
the target groups are not able to pay for the basic services that they require.
This is particularly true of charities operating in the social services sector.
Thus, in such circumstances, the WTP of the target groups will be negligible
even though the benefits they obtain from the service may be very large.
Consequently, it can be argued that, for the target group of many charities,
the correct measure of value is not WTP for the provision of the service, but
the WTA compensation for the loss of the service. Freeman (1993) discusses
the choice of the correct welfare measure and shows that it depends on
whether there is an implied property right in the status quo. He argues that
WTA compensation is an appropriate welfare measure in circumstances where
the beneficiaries are perceived to have a right to the services they receive.
This would appear to be applicable in the case of charities, in that they
provide basic services to which the beneficiaries could be argued to have a
basic right.
As far as indirect benefits to the rest of society are concerned, the status
quo is that many people do pay for the provision of charitable services,
whether directly through donations or indirectly through government grants.
The implication is that WTP is the appropriate welfare measure for these
indirect benefits.
WTP is nothing more than the demand for a good or service. When a
good or service is provided in a market context, it is relatively easy to
estimate this demand curve and thus the total WTP for the associated good
or service, which is equivalent to the area underneath the demand curve.
Even when this does not prove to be possible, the market price at least gives
a lower bound for each individual purchaser’s WTP, as it is clear that if an
individual’s WTP is lower than the market price, they will choose not to
purchase.
In the case of charitable services it is not so straightforward to estimate
WTP, since these services do not tend to be traded on the market, but rather
provided at zero or highly subsidized cost. Furthermore, there is often also a
public-good component in charities’ services, in that some members of soci-
ety receive benefits from the provision of such services, whether or not they
contribute to their provision. However, these difficulties are already well
known to economists working in environmental and health economics, and a
variety of techniques has been developed for estimating WTP in the absence
of direct market prices.
The principal methods that may be used to value the services provided by
charities in the absence of markets are briefly described below. They can be
divided into two broad categories: the indirect approaches and the direct
approaches.
The first category of ‘indirect approaches’ aims to find some indirect link
between the services provided by charities and services provided in the
marketplace. Where such a link exists, the market price can be used to
approximate the value of the charitable activity. These methods include the
private sector equivalent, the opportunity cost approach and the production
function approach.
Private sector equivalent

In many cases charities provide services that are also provided in the private
sector. Examples include care for the elderly, hostels for the homeless, and
education and health services. In these contexts, the rationale for charitable
provision is that the charity aims to provide coverage to a target group that
the private sector would not find attractive to serve. Where this situation
arises it may be possible to use the prices of the equivalent services charged
in the private sector as a means of valuing the services provided by the
charity. An important assumption underlying this approach is that there are
no differences in the quality of services provided by charities and by the
private sector. However, this assumption may not always be appropriate,
thereby invalidating the use of this method.
Opportunity cost approach

In a more limited number of cases, it may be possible to determine the value
of a charity’s services by estimating the opportunity cost of not providing
those services. For example, some day-care services enable the carers to go
to work. Thus these services could be valued in terms of the extra earnings
that the carers are able to obtain by going to work. However, clearly this
valuation would not take into account the emotional relief that a carer may
experience from being able to make use of a day-care centre.
Production function approach

Charitable services often boil down to the production of certain outputs
whose economic values are already known, either because they are traded in
the marketplace or because they have already been estimated in other con-
texts. For example, some charities’ principal output may be to save lives or to
extend life expectancy. There is already an extensive literature in the health
economics field on the economic value of additional life-years. If the rela-
tionship between the charities’ activities and the life expectancy outcomes
could be accurately estimated, it would then be possible to apply the corre-
sponding economic values to obtain a monetary measure of benefits.
Where it is not possible to make any link between the services provided by
charities and other services provided in the marketplace, a second category of
‘direct approaches’ may be used. These techniques, also known as stated
preference techniques – which include contingent valuation, choice model-
ling techniques and related variants – involve using a questionnaire within
which respondents are asked directly to reveal their WTP or WTA.
Contingent valuation method

The contingent valuation method (CV) is a survey-based technique. By means
of an appropriately designed questionnaire, a hypothetical market is described
where the good in question can be traded (Mitchell and Carson, 1989). This
contingent market defines the good itself, the institutional context in which it
would be provided and the way it would be financed. A random sample of
people are then asked directly to express their maximum WTP or minimum
WTA for a hypothetical change in the level of the provision of the good. A
critical assumption underlying this method is that respondents’ behaviour in the
hypothetical market will be no different than it would have been in an equiva-
lent real situation. The empirical evidence on this point is mixed, with some
studies finding that hypothetical WTP exceeds real WTP (see Foster et al.,
1997). Furthermore, a number of factors may systematically bias respondents’
answers. These factors are not specific to CV but are common to most survey-
based techniques and are mostly attributable to survey design and implementation
problems. Mitchell and Carson (1989) provide an extensive review. It is not
straightforward to assess the validity of the estimates produced by CV studies

for the obvious reason that actual WTP is unobserved. Nevertheless, certain
aspects of validity can be tested by indirect means (see EFTEC, 2001).
Choice modelling techniques

Choice modelling, on the other hand, applies to a family of survey-based
methods that model preferences for bundles of characteristics of goods and
isolate the value of individual product characteristics typically supplied in
combination with one another. There are many variations of choice model-
ling; however, they all share the same basic characteristics. Respondents are
given a choice set comprising two or more alternatives. For example, the
choice set may comprise three different types of hospice services. Each
alternative has a certain number of characteristics, but differs in terms of the
level of those characteristics. For example, the characteristics might be the
cost of using the hospice services, the quality of the rooms provided and the
availability of support services. Some of the hospices in the choice set may
be more expensive than others but offer higher-quality rooms and/or a wider
range of services. Respondents are asked to express their preferences over the
items contained in the choice set. This might be by choosing their most
preferred alternative (choice experiments), ranking the alternatives in order
of preference (contingent ranking), or simply giving each alternative a score
on some predetermined scale (contingent rating). With the aid of statistical
techniques, it is possible to infer WTP or WTA for the characteristics pre-
sented from the choices reported in the survey.
1.3.3 Total Social Cost of Charities
Total social costs are the opportunity costs of the services provided by the
charity, that is the value of the total social benefits foregone by allocating
resources to the charitable sector as opposed to their best possible alternative
uses. As with the total social benefits, the total social costs can be broken
down into a number of distinct categories.
First, if the target group has to pay a price for each service that the charity
provides, then the corresponding opportunity cost is the total value of the
(maximum) benefits that they would gain from using the money (and possi-
bly the time) spent on the charity’s services on alternative activities, in the
absence of the charity. This can be captured by the value of the fee revenue
paid.
Second, the opportunity cost to donors of the charity’s provision of services
is the total value of the (maximum) benefits that they could gain from spending
their donation on alternative goods and services, in the absence of the charity.
This can be measured in terms of the total financial value of donations.
Third, in the case of volunteers, one way of valuing the resources absorbed
by the use of volunteer time is to consider the costs of replacing volunteers
with paid employees on the open market. This approach is explained in
greater depth in Chapter 4.
Fourth, if the government also contributes grant finance to the charity, then
there is an additional opportunity cost of the charity’s provision of services.
This is equal to the (maximum) value of social benefits that could be gained
from reallocating the grant to some alternative area of public expenditure, in
the absence of the charity. The opportunity cost of government grants is not
straightforward to measure either, since it would depend on the alternative
uses to which such funds were put. Thus, in common with standard practice
in cost–benefit analysis, the financial value of the grant is used as an approxi-
mation to the opportunity cost.
Figure 1.2 provides a graphic representation of this typology of the total
social cost of the charitable sector.
Rest of society
Donors Volunteers Government Target groups
Donations Time, skills Grants Fees, price
CHARITIES
Figure 1.2 Social costs of charities
1.3.4 Net Social Value of Charities
In conclusion, this introductory chapter has argued that traditional measures

of the size of the charitable sector – such as the income and expenditure of
charities – do not provide any meaningful indication of the value of the
output of charities. In order to solve this problem it is necessary to measure
the net social value of charitable activities, that is the difference between the
social benefits generated by charities and the opportunity cost of the re-
sources that they absorb.
The services provided by charities are beneficial not only to the immediate
target group; they also generate a range of indirect benefits for society as a
whole. These benefits are not straightforward to value because they are not
traded in the marketplace like conventional goods. None the less, using a
range of techniques already tried and tested in the environmental and health
economics literatures, it is possible to put a monetary value on these benefits.
Resources used by charities have an opportunity cost in that they can no
longer be put to work in alternative uses. These resources include fees paid
by direct users, gifts made by donors, time given by volunteers and grants
from the public sector. By and large, these social costs are more straightfor-
ward to measure than the social benefits.
The net social value of the charitable sector is nothing but the difference
between these benefits and costs. The remaining chapters of Part I undertake
the steps outlined here in greater detail for one particular example, that of
charities working in the field of housing and homelessness. Chapter 2 applies
contingent valuation and choice modelling techniques to the problem of
measuring indirect benefits to society at large of charities in housing and
other sectors. Chapter 3 uses contingent valuation techniques to value the
benefits to users of hostels for the homeless in London, using a welfare
measure based on willingness to accept compensation. Chapter 4 values the
opportunity cost of volunteer time. All these elements are brought together in
Chapter 5, which performs a calculation of net social value.
2. The benefits of charities to the general
public
2.1 INTRODUCTION
A significant number of attempts have been made to estimate the size of the
voluntary sector, as measured by employment, income and expenditure. How-
ever, these studies provide only incomplete measures of the value of the
sector as they focus on the costs rather than the benefits, that is, the value
added of the sector.
The experiment reported in this chapter uses data from a study that, for
the first time, applied stated preference techniques to the valuation of the
output of the charitable sector in the UK, with special reference to the
housing and homelessness charities. Broadly speaking, charities can be
seen as providing use value to a relatively small group of beneficiaries and
non-use, option and indirect use values to society at large. This chapter
deals with the estimation of the latter type of values from the perspective of
the general public.
The purpose of investigating the value of charities to the general public in
the UK is twofold. The first objective is to measure the benefits which the
charitable sector provides to society at large, over and above the benefits
received directly by the target groups. As mentioned in Chapter 1, these
benefits could potentially be motivated by a number of considerations. Peo-
ple may benefit indirectly from the charities’ activities, for example by a
reduction in the number of rough sleepers they encounter on city streets. Or
they may feel that the charity provides a safety net, which could be of use to
them at some future date, for example research into a cure for cancer. Alter-
natively, for moral reasons, people may simply value the fact that the basic
needs of others are taken care of by charitable organizations.
The second objective is to measure the benefits to society at large which
are specifically associated with the activities undertaken by housing and
homelessness charities. A survey of the direct beneficiaries of these charities
has also been undertaken in a separate investigation (see Chapter 3) in order
to uncover the direct use values of these philanthropic activities. The summa-
tion of the two values obtained from each of the surveys permits an estimation
of the overall social value of such charities (see Chapter 5).
19
To carry out these objectives, two stated preference techniques were used
in parallel: contingent valuation (Mitchell and Carson, 1989) and contingent
ranking (Beggs et al., 1981). These techniques have been widely used to
value public goods and services in the areas of environment, culture and
health. However, they have rarely, if ever, been applied to the valuation of
charitable sector services.
In order to apply the stated preference approach to the problem of estimat-
ing the social value of the charitable sector, it is necessary to construct a
scenario which will enable people to think about how much the sector is
worth to them. In the context of the present study, this was done by describ-
ing a hypothetical situation where one or more sectors of charitable activity
were threatened with shutdown due to a funding crisis. The respondents were
then told that the government could prevent the shutdown by giving the
charities an emergency grant, but that this could only be achieved at the
expense of raising everybody’s tax bill. Following on from this, various
means were used to establish how much (if any) additional tax the respondent
was willing to pay to fund the emergency grant. It was stressed that this
payment would be over and above any existing donations that the respondent
was already making.
In theory, a scenario of this kind should be able to elicit from respondents
the desired value for the charitable sector. By using taxation as the means of
payment and presenting respondents with the prospect of losing the charita-
ble sector altogether, the scenario should avoid the free-rider problems
associated with voluntary donations. By asking for willingness to pay over
and above existing contributions, the survey should induce respondents to
reveal the extent to which their current contributions understate their true
valuation of the sector.
This chapter summarizes the design characteristics of the general public
survey and presents the results.
2.2 SURVEY DESIGN

For the purposes of the survey, the UK charitable sector was classified into
four subsectors: (1) housing and homelessness charities, which provide emer-
gency short-term accommodation, counselling and support services for
homeless people in hostels and night shelters, for example Shelter, Crisis,
Centrepoint and the Salvation Army; (2) social services charities, which work
to improve the lives of particularly needy groups of people such as the
elderly, the physically and mentally handicapped, disabled people such as the
blind and the deaf, and children from troubled backgrounds, for example Age
Concern, Dr Barnardo’s and the NSPCC; (3) health and medical research
Benefits of charities to the public 21
charities, which fund scientific research into presently incurable diseases as

well as providing hospice care for the terminally ill, for example the Imperial
Cancer Research Fund, the Multiple Sclerosis Society and the British Heart
Foundation; and (4) a residual category comprising charities active in the
areas of culture, environment and overseas aid, for example Oxfam, Save the
Children Fund, Friends of the Earth, the RSPCA and the National Trust.
The overall objectives of the general public survey necessarily required the
separate identification of housing and homelessness charities. The social
services and health categories were chosen because they represent a large
proportion of overall charitable activity. The final category was necessarily
something of a catch-all. However, overseas aid, environment and culture
were explicitly identified as the major charitable activities in this subcategory.
Educational and religious charities, understood to refer to those that exist to
benefit schools and churches, as well as political parties were explicitly
excluded from the survey.
The survey of the general public consisted of a combined contingent valua-
tion (CV) and contingent ranking (CR) split-sample experiment. In particular,
three different versions of the survey were administered to different subsamples
of the population:
1. Contingent valuation of housing and homelessness charities (CVHH) (sam-

ple size: 282): this version elicited the value of housing and homelessness
charities only using a CV approach;
2. Contingent valuation of all four charitable sectors (CVALL) (sample size:
279): this version elicited the value of all the four sectors of charitable
activity described above using a CV approach; and
3. Contingent ranking (CR) (sample size: 290): this version permits an
elicitation of the value of the whole charitable sector and of each of the
four subsectors separately by means of a CR approach.
In a preliminary section, common to all versions, respondents were intro-

duced to each of the four charitable sectors described above and asked to
think about their relative importance. Information was then collected on each
respondent’s current donations of time and money to each of these areas of
philanthropic activity. Owing to the free-rider problem, the magnitude of
existing contributions can only be regarded as a lower bound on people’s
valuation of charities.
As mentioned before, the purpose of the survey was to establish how much
more people would actually be willing to pay over and above their existing
donations. This was determined by creating a hypothetical scenario in which
all or some of the charities faced the prospect of shutting down as a result of a
shortfall in funding. Respondents’ WTP to prevent this eventuality was then
elicited using increases in taxation as a payment vehicle. In this context, the

WTP measure is an equivalent variation welfare measure, that is, the maxi-
mum amount respondents are willing to pay to avoid the closure of one or
more sectors of charitable activity.
An example of the hypothetical scenario used in the CVHH version is
presented below. There were only very minor adjustments to the wording
across the different versions of the questionnaire.
Please imagine that, due to a financial crisis, all the charities in the country
dealing with housing and homelessness were facing the prospect of shutting down
for a whole year. (Don’t worry, this is definitely not going to happen! But, even so,
we would like you to think about how you would feel about this if it were to
happen.)
Clearly, if these charities were to shut down, many of the people whom they
currently help would be left with no alternative but to sleep rough on the streets.
Now suppose that the government was considering making an emergency grant
to these charities, so as to prevent them from having to shut down. The only way in
which the government could fund this emergency grant would be by raising the
taxes we all pay – for example, taxes on income and the sale of goods.
Following the description of the hypothetical scenario, part of the sample

received a dichotomous choice CV treatment, while another part received a
CR treatment. The next sections present a detailed description of the valua-
tion questions presented in both the CV (CVALL and CVHH) and CR surveys.
Technical details of the theoretical and statistical procedures involved in the
estimation of WTP measures from these elicitation methods can be found in
the Statistical Appendix to this chapter.
2.2.1 Contingent Valuation
The CVALL subsample received a double-bounded dichotomous choice CV

question (Hanemann et al., 1991) which dealt with the potential closure of all
charities in all four of the sectors identified; while the CVHH subsample
received a double-bounded dichotomous choice CV question which dealt
with the potential closure of all charities operating in the housing and home-
lessness sector only.
In particular, respondents were asked whether or not they would be willing
to pay a specific tax amount £X, to which they might answer ‘yes’ or ‘no’.
The tax amount £X was varied across respondents. In addition, they were
asked a follow-up payment question that depended on the response to the first
tax level: if the respondent accepted the initial bid, they were asked to pay a
higher bid; if the answer to the first tax level was ‘no’ then the respondent
was presented with a lower amount. This procedure is known in the literature
as a ‘double-bounded dichotomous choice elicitation’. As an illustration be-
low is the exact wording of the initial valuation question used in the CVHH
version.
Suppose that the funding of the emergency grant would cost everyone, including
you personally, £X each month throughout the coming year – which would add up
to £12X over the year. This would be over and above any contributions of time and
money you already make towards charities dealing with housing and homeless-
ness. Would you be prepared to pay this extra amount to prevent these charities
from shutting down? Please think carefully about how much you can really afford
and where the additional money would come from and try to be as realistic as
possible.
The actual amounts of money used in the question varied from £0.20 to
£15.00 per month in CVHH and from £0.20 to £25.00 per month in CVALL.
Respondents were also reminded of the yearly implications of a monthly
payment. Table 2.1 shows the tax vectors that were used in the dichotomous
choice approach. The bid levels were chosen based on the results of the pilot
open-ended valuation questions.
Table 2.1 Bid vectors for CVALL and CVHH (£ per month and £ per year)
All charities (CVALL) Housing and homelessness

charities (CVHH)
Bid levels Bid levels
Low Initial High N Low Initial High N

M A M A M A M A M A M A
I 0.2 2.4 0.5 6 1 12 54 0.2 2.4 0.5 6 1 12 56

II 0.5 6 1 12 3 36 58 0.5 6 1 12 2.5 30 56
III 1 12 3 36 6 72 56 1 12 2.5 30 5 60 54
IV 3 36 6 72 12 144 57 2.5 30 5 60 10 120 58
V 6 72 12 144 25 300 53 5 60 10 120 15 180 58
Notes: M: monthly; A: annually; N: number of respondents; I–V: bid vectors.
Following the recommendations of the NOAA Panel (Arrow et al., 1993),

respondents were also offered a ‘don’t know’ option after each valuation
question. This implies that there were three possible answers to each bid level
presented: ‘yes’, ‘no’ and ‘don’t know’. Wang (1997) has recently shown
how such answers can be incorporated into the choice model for the case of
single-bounded dichotomous choice. In this chapter, the two extreme cases
are reported: the case where ‘don’t knows’ are treated as a rejection of the bid
level (providing a lower bound on true WTP) and the case where ‘don’t
knows’ are treated as an acceptance of the bid level (providing an upper

bound on true WTP).
As an internal test of consistency, after the dichotomous choice questions,
respondents were asked another WTP question, this time under the open-
Table 2.2 Payment ladder used in the general public survey (CV version)
and choices of a hypothetical respondent
£/month (£/year equivalent) ✔ or ✘
nothing (nothing) ✔
10p (£1.20) ✔
25p (£3) ✔
50p (£6) ✔
75p (£9) ✔
£1.00 (£12) ✔
£1.50 (£18) ✔
£2.00 (£24) ✔
£2.50 (£30) ✔
£3.00 (£36) ✔
£3.50 (£42)
£4.00 (£48)
£4.50 (£54)
£5.00 (£60)
£6.00 (£72) ✘
£7.00 (£84) ✘
£8.00 (£96) ✘
£9.00 (£108) ✘
£10.00 (£120) ✘
£12.50 (£150) ✘
£15.00 (£180) ✘
£17.50 (£210) ✘
£20 (£240) ✘
£25 (£300) ✘
£30 (£360) ✘
£40 (£480) ✘
£45 (£540) ✘
£50 (£600) ✘
£75 (£900) ✘
£100 (£1200) ✘
Over £100 (over £1200) ✘
ended format. Open-ended questions can be unaided (‘what is your maxi-

mum WTP?’) or aided, whereby respondents are presented with a card
containing several money amounts (a ‘payment card’) and are asked to choose
the figure that best approximates their maximum WTP. Given that the latter
approach is thought to simplify the valuation task for the respondent, a
version of the payment card was adopted.
Specifically, respondents were presented with a ladder of values identical
to that shown in Table 2.2. The ladder was completed by asking respond-
ents to begin with the lowest values and put a tick against those amounts
that they were ‘almost certain that they would be willing to pay’. Subse-
quently respondents were asked to turn to the highest values and put a cross
against those amounts that they were ‘almost certain that they would not be
willing to pay’. Thus the example given in Table 2.2 indicates that the
respondent was almost certain that she would be willing to pay as much as
£3 per month and equally certain that she would not be willing to pay as
much as £6 per month. Between those two values, the respondent was
unable to mark either a tick or a cross, thereby indicating that WTP was
uncertain over this range. The results obtained from the ladder complement
those obtained from the dichotomous choice questioning and are particu-
larly valuable in revealing the degree of confidence that respondents had in
stating their WTP.
In the CVALL version of the questionnaire, following the payment ladder,
respondents were asked to apportion their overall WTP between the four
different sectors of philanthropic activity considered. This exercise provides
an alternative ‘top-down’ welfare measure for the services provided by the
charities dealing with housing and homelessness issues which can be used as
a point of contrast with the single sector welfare measure administered sepa-
rately to the CVHH subsample.
2.2.2 Contingent Ranking
In this version of the survey, respondents were asked to rank in order of

preference a series of scenarios which involved different tax increments and
different patterns of closure of charitable sector activities. The valuation
question was described along the following lines:
The amount of additional tax that you would pay will depend on which charities
and how many charities are rescued from shutdown. The following showcard lists
a number of rescue options.
Option 1 is where all of the 4 types of charities shutdown and you pay NO extra
money to rescue any of them.
Option 2 is where you pay an extra £2.50 each month for a whole year (which
would add up to £30 over the year) to rescue the social services charities ONLY.
Option 3 is where you pay an extra £5 each month for a whole year (which
would add up to £60 over the year) to rescue the health and medical research
charities ONLY.
If you were faced with a choice between these three options, which one would
you choose? Please think carefully about whether you can really afford the extra
payment, and where the additional money would come from. Remember that this
would be over and above any contributions of time and money you already make
towards these charities.
If you were faced with a choice between the two remaining options, which one
would you choose?
Each respondent was given three different sets of options to rank. The
actual amounts of money used in these questions varied across different
versions of the questionnaire, but were chosen to lie in the same range as
those offered in the CV questions. An illustrative example of one of the CR
choice sets is provided in Table 2.3. These questions were administered as a
sequential choice of the most preferred alternative, first from the full set of
three options and then from the set of two options remaining after the first-
best option had been removed. This meant that respondents were effectively
required to provide a complete ranking of all the options in the choice set.
Table 2.3 Sample question used in the general public survey – CR version
Option 1 Option 2 Option 3
Housing and homelessness charities ✘ ✔ ✔

Social services charities ✘ ✘ ✔
Health and medical research charities ✘ ✘ ✔
Overseas aid, environment and culture ✘ ✘ ✔
Additional payment each month for a year none 20p 50p
(Corresponding annual amount) (none) (£2.40) (£6)
Notes: ✘: charity shut down and ✔: charity kept running.
Three design features of the CR version were identified as critical in

allowing rigorous comparisons to be made with the parallel CV. First, all
choice sets should include an alternative which involves no additional tax
payment but which entails the closure of all of the charities (the corre-
sponding situation in the CV version is a ‘no’ answer to the proposed bid
level). Second, respondents receiving the CR version should be faced both
with options which avoided the closure of a single sector of charitable
activity, such as the housing and homelessness sector, and with options
which avoided the closure of all charities. This would permit the estimation
of welfare measures equivalent to those defined for the CV version. Third,
the same set of bid levels that were attached to the two versions of the CV
questions should be attached to the corresponding alternatives in the CR
choice set.
The need to meet these three criteria for full comparability with CV neces-
sarily entailed a departure from the conventional principles of fractional
factorial design (Louviere, 1988) for the choice of the set of alternatives to be
presented in the questionnaire. In this context it is important to note that the
purpose of fractional factorial design is to ensure that ranking alternatives are
orthogonal, because this is a requirement of the traditional monotonic analy-
sis of variance applied to this kind of experimental data. When analysis is
undertaken in a logit regression framework, as in the present case, such
orthogonality continues to be desirable but is no longer essential for the
method to work satisfactorily.
Thus, in order to construct the chosen three-element CR choice sets, the
following procedure was adopted. First, the six alternative patterns of charity
closure of interest were identified: all charities are shut down, each of the
four subsectors alone is maintained, and all four of the subsectors together
are maintained. Second, it was established that in order for each possible
pair-wise combination to be presented to respondents alongside the zero
payment baseline, ten different triplets had to be presented in the survey
(given the constraint that the baseline must appear in each of the triplets,
there are C25 = 10 different pairs that can be formed from the remaining five
alternatives). Third, the bid levels used in the dichotomous choice CV were
allocated to the options appearing in the triplets so that each bid level ap-
peared at least once. This involved some repetition of the triplets, bringing
the total number up to 18. Finally, the 18 triplets were grouped into six sets of
three to be administered to different subsamples of the CR version. The series
of three were constructed in such a way that each respondent faced a wide
range of bid levels and one alternative involving the preservation of all of the
charity sectors.
Thus, in all of these ways, the CR questions were designed to mimic the
CV questions to the greatest possible degree, so that the only difference
between the two was the method by which the response was elicited.
2.2.3 Overview
The main objective of the experimental design described in this section was
to obtain valid and reliable estimates of the value that the general public
attaches to the charitable sector in the UK in general and to the charities
dealing with housing and homelessness issues in particular. But the survey
also had two secondary objectives.
The first was to test for robustness of results to the method. The literature
on economic valuation indicates that the results obtained can be highly sensi-
tive to the actual method chosen to elicit WTP from the survey population.
Since none of the methods can unambiguously be demonstrated to give the
‘right’ answer, it is consequently desirable to use a variety of different meth-
ods in parallel. This makes it possible to examine how robust the results are
to the methodology used. Where substantial variations across methods arise,
these can be used to place bounds on true WTP. The experimental design
adopted permits a systematic comparison of results obtained from dichoto-
mous choice CV and CR.
The second was to test for sensitivity of respondents to scope. The litera-
ture on economic valuation indicates that respondents are often not very
sensitive to the scope of what is being valued. In this context, scope refers to
whether respondents are being asked to value all charitable activities or just
one particular subsector of those activities, such as housing and homeless-
ness. In particular, the expectation is that, when asked to value housing and
homelessness charities alone, people will be willing to pay considerably
more than if they are asked to value all charities and then state what propor-
tion of that value is attached to the housing and homelessness subsector. By
including these two different approaches in different versions of the question-
naire, it is possible to examine to what extent this problem is impinging on
the results and to investigate whether, as has been suggested in the literature
(Hanley et al., 1998), CR avoids the problem of ‘part–whole bias’ which has
been documented in the case of CV.
Assuming that all other differences between questions have been adequately
controlled for, the experimental design adopted makes it possible to investi-
gate these secondary objectives.
The survey instrument was extensively piloted. The main survey took
place in the spring and summer of 1997 and covered a random sample of the
UK population composed of 851 respondents across 20 sampling points. The
survey was administered by means of a 30-minute face-to-face interview that
took place in the respondent’s home. Just under half of the interviews were
conducted after the death of Diana, Princess of Wales. A number of factors
about this event were important for the study including the depth of the
public reaction, the princess’s high profile role in the charitable sector and the
volume of donations into her special memorial fund. Since the dates of
individual interviews were recorded, it was possible to examine to what
extent this event had a significant impact on attitudes towards charities and,
in particular, WTP for their preservation.
2.3 THE SURVEY RESULTS
2.3.1 Socioeconomic Characteristics
Table 2.4 undertakes a comparison of each of the three subsamples receiving

different versions of the questionnaire. The table states the mean values of
each variable.
Analysis of the figures in Table 2.4 reveals that there are no significant
differences between the various subsamples. Thus any differences in estima-
tion arising from each of the valuation methods employed can be attributed to
the method itself rather than to any idiosyncrasies of the sample population to
which it was applied.
Table 2.4 Socioeconomic characteristics of different subsamples
CVHH CVALL CR Overall
Males (%) 41 51 46 46
Average age (years) 43.5 45.7 44.3 44.5
Socioeconomic group (%)
AB 20 18 18 18
C1 28 27 28 27
C2 25 24 27 26
DE 28 31 27 29
Married 60 65 66 64
Children (%)
Pre-school age 24 20 26 23
School age 64 64 62 63
Age completed full-time education (years) 16.6 16.2 16.5 16.5
Employment (%)
Full-time 46 41 49 45
Part-time 18 16 17 17
Unemployed 4 3 3 3
Monthly income (£) 945.6 997.2 950.4 965.3
Notes: CVHH: contingent valuation for housing and homelessness charities; CVALL: contingent
valuation for all four charitable sectors; CR: contingent ranking.
2.3.2 Current Donations
The questionnaire collected information on existing donations of time and

money made by respondents to the four charitable subsectors of interest.
These data are useful for two reasons: first, when combined with the results
of the valuation question they permit the estimation of an overall value for
the charitable sector; second, they can be used to check for consistency with
the results of earlier studies of philanthropic behaviour, in particular the
Individual Giving Survey (IGS) (Halfpenny et al., 1992, 1993, 1994).
Although it was not the purpose of the present survey to collect detailed
information on modes of charitable giving, a list of possible ways of giving
was used in the questionnaire in order to help respondents to remember how
much they had actually given. This list was loosely based on the classification
of payment modes used in the IGS. For the purposes of comparison, Table 2.5
gives the proportion of respondents who stated that they had given by each
method, both for this survey and for the last year of the IGS. The broad
pattern of giving is fairly close across the two surveys and is not significantly
different for most comparable classifications. However, the survey does ap-
pear to report substantially lower levels of participation than the IGS for
collection boxes, sponsorship and raffles. A substantially higher level of
participation is recorded in the case of credit cards which benefit charities.
Turning to the actual contributions made, Table 2.6 summarizes monthly
and annual averages for cash donations and hours volunteered to the four
charitable subsectors. As a point of comparison, the equivalent figures from
the 1993 IGS are also quoted, adjusted to 1997 prices where relevant.
The mean annual donation and hours volunteered is less than 12 times the
mean monthly donation and hours volunteered. In fact, the annual values are
only about 60 per cent of the grossed-up monthly values. There are two
possible explanations for this. First, people may have found it difficult to
recall over an annual period and thus may not have been able to record all the
contributions they actually made. Second, people may have mistakenly attrib-
uted to the past month contributions which they actually made further back in
the past; this phenomenon is known as telescoping. Unfortunately, it is not
possible to distinguish between these two hypotheses; indeed, it is likely that
elements of both could be occurring.
The mean monthly cash donations and hours volunteered which emerge
from the present survey are comparable to those obtained from the IGS,
although significantly lower with their confidence intervals not quite overlap-
ping.
On average, people contribute most money to health and medical research
charities, followed by overseas aid, environment and culture, then social
services and finally housing and homelessness charities. However, many of
the confidence intervals for donations to each of these subsectors overlap,
indicating that the differences are not statistically significant. In particular,
the category of overseas aid, environment and culture has a very wide confi-
dence interval, indicating that behaviour varies considerably across individuals.
Table 2.5 Comparison of participation rates in different forms of donations
Current survey IGS 1993
Tax-efficient
● donating by covenants, gift aid, payroll 0.054 (0.039–0.069) 0.096 (0.078–0.114)
● giving from a CAF account 0.019 (0.010–0.028)
Non-tax-efficient
(a) philanthropic
● TV, telephone, mail or newspaper appeal 0.043 (0.029–0.057) 0.061 (0.046–0.076)
● collection box in street, shop, pub, at door, work 0.417 (0.384–0.450) 0.607 (0.577–0.637)
31
● sponsoring somebody 0.135 (0.112–0.158) 0.224 (0.198–0.250)
● donating via a church or school 0.052 (0.037–0.067) 0.137 (0.116–0.158)
● donating to a church or school 0.074 (0.056–0.092)
(b) purchases
● goods from a charity shop, catalogue or sale 0.202 (0.175–0.229) 0.230 (0.204–0.256)
● attending a fundraising event 0.054 (0.039–0.069) 0.065 (0.050–0.080)
● raffle or lottery ticket (excluding National Lottery) 0.182 (0.156–0.208) 0.302 (0.274–0.330)
● using a credit card which benefits charity 0.025 (0.015–0.035) 0.007 (0.002–0.012)
● subscription to a charitable organization 0.025 (0.015–0.035) 0.045 (0.032–0.058)
● The Big Issue 0.085 (0.066–0.104)
(c) other 0.066 (0.049–0.083) 0.012 (0.005–0.019)
Note: confidence intervals in parentheses.

Table 2.6 Summary of existing donations of time and money
Cash donations (£) Hours volunteered (hrs, mins)
Monthly Yearly Monthly Yearly
Housing and homelessness 0.70 5.56 0,07 0,34

(0.50–0.90) (4.50–6.60) (–0.01–0.15) (0.11–0.58)
Social services 1.14 9.26 0,41 5,33
(0.89–1.4) (7.38–11.14) (0.33–0.49) (3.04–7.62)
32
Health and medical research 2.36 15.10 0,20 2,43
(1.69–3.03) (11.92–18.28) (0.10–0.30) (1.51–3.35)
Overseas aid, environment and culture 1.49 11.47 0,21 3,01
(0.63–2.35) (5.41–17.53) (0.03–0.39) (1.01–5.01)
Total 6.32 42.98 1,48 11,43
(4.73–7.91) (34.9–51.1) (1.03–1.93) (8.35–14.5)
IGS 1993 9.51 n/a 4,07 n/a
(8.58–10.44) (3,19–4,55)

On average, people contribute most time to social services charities, fol-

lowed by overseas aid, environment and culture, then health and medical
research, and finally housing and homelessness charities. However, as before,
these differences are not always significantly different in the statistical sense.
2.3.3 Attitudes towards Donations
The preliminary section of the survey contained a considerable number of

attitudinal questions which were intended to make respondents explore their
personal thoughts on philanthropic issues as a preparation for responding to
the valuation question. In addition, these questions were designed to reveal as
much as possible about the underlying motives for philanthropic donation, so
as to aid in the interpretation of the valuation responses.
The opening attitudinal question asked respondents about the degree of
importance of the contribution which charitable organizations made to soci-
ety. Figure 2.1 indicates that over 70 per cent of the sample thought that
charities were either ‘important’ or ‘very important’. However, it is worth
noting that these attitudes may be biased upwards by the fact that respondents
already knew that charities formed the primary focus of the survey.
not important at all
of some importance
important
very important
0 0.1 0.2 0.3 0.4 0.5 0.6
‘How important a contribution do you think charities make to our society?’
Figure 2.1 Importance of charities
A large number of attitudinal questions were posed with respect to mon-

etary donations and can be grouped into those aimed at uncovering the
motives for giving and those aimed at understanding the process of giving.
Figure 2.2 presents the results for the questions relating to underlying
motive. They are organized along a spectrum from primarily selfish to in-
creasingly altruistic motives.
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(a) ‘I often give because I feel too embarrassed to say “No” when someone
asks’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(b) ‘I give to charities because I like the feeling of being generous’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(c) ‘I give to charities because I or my family may personally benefit from
them at some stage’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(d) ‘I give to charities because they help to create a better society for
everyone by reducing the level of social problems’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(e) ‘I give to charities because I want to support the good causes for which
they work’
Figure 2.2 Attitudes towards donating
On one end of the spectrum is an embarrassment motivation: a person is

ashamed to appear to be ungenerous and gives simply to escape from an
embarrassing situation. Figure 2.2(a) indicates that the majority of respond-
ents seemed to resent this characterization of their philanthropic motives,
with over 60 per cent disagreeing or strongly disagreeing. Nevertheless, a
substantial minority of 25 per cent admitted that embarrassment was often a
determining factor in their decision to give.
Another type of selfish motivation for charitable giving is what has become
known in the literature as warm glow (Kahneman and Knetsch, 1992): a person
enjoys the feeling of their own generosity. Figure 2.2(b) reveals a similar
pattern to the preceding case, with nearly 60 per cent of respondents denying
that they exhibit this tendency, but over 20 per cent admitting that they do.
Moving along the spectrum we find option value: a person sees their
charitable donation as a kind of insurance premium to finance a safety net
service from which they or their relatives may personally benefit at some
future stage. As shown in Figure 2.2(c), the question of option value split the
sample more or less down the middle: 40 per cent identified with this motiva-
tion and 40 per cent rejected it.
Indirect use values may also motivate philanthropic giving: a person feels
that they benefit in an indirect way from the charity’s services because they
serve to strengthen the social fabric. Figure 2.2(d) shows that there was a
strong tendency to identify with this motivation, which met with agreement
from over 60 per cent of the sample.
Finally, at the altruistic end of the spectrum of motivations we find exist-
ence value: a person thinks that services provided by charities are worthwhile
for their own sake, irrespective of any personal spin-offs they may generate.
The pattern of results revealed in Figure 2.2(e) is very similar to the preced-
ing case but with an even higher degree of identification, corresponding to
over 80 per cent of the sample.
Overall, these results indicate that people tend to view their philanthropic
donations in strongly altruistic terms. None the less, there is a significant
subset of the population which gives for reasons more closely aligned with
self-interest.
It is interesting to enquire to what extent these different motivations over-
lap at the level of individual respondents. Table 2.7 reports the correlation
coefficients between each pair of attitudinal variables and reveals a number
of interesting points (none of the correlations reported is particularly high in
absolute terms, although most are statistically significant).
There is a particularly strong correlation (0.33) between people motivated
by embarrassment and people motivated by the warm glow. Indeed, 72 per
Table 2.7 Correlation between different motives for monetary donations
Warm Option Indirect Existence

Embarrassment glow value use value
Embarrassment 1
Warm glow 0.33 1
Option value 0.14 0.27 1
Indirect use 0.08 0.23 0.22 1
Existence value –0.06 0.10 0.23 0.55 1
cent of the sample consistently either agreed or disagreed with both of these
statements. Similarly, there is an even stronger correlation (0.55) between
those motivated by indirect use concerns and those motivated by existence
value. Indeed, 86 per cent of the sample consistently agreed or disagreed with
both of these statements. There is a very weak negative correlation (–0.06)
between those motivated by embarrassment and those motivated by existence
value considerations. Indeed, only 32 per cent of respondents consistently
agreed or disagreed with both statements.
Hence there seems to be a consistent thread in respondents’ answers sug-
gestive of the existence of two broadly distinct groups of respondents: a
majority that is mainly driven by altruistic motivations and a minority that is
motivated primarily by more selfish considerations.
The survey also examined the process of giving. Three distinct areas were
investigated: the scope of the donations, that is the extent to which giving was
focused on one particular subsector or spread widely across different areas of
charitable activity; the level of foresight, that is the extent to which giving
was spontaneous or carefully planned in advance; and the degree of commit-
ment, that is the extent to which giving was regular over time or tended to be
simply on a one-off basis.
The responses to the issue of scope are contained in Figure 2.3. Parts (a)
and (b) contain responses to what are essentially identical questions but
worded in different ways: one as a positive statement and the other as a
negative statement. The two statements do not appear contiguously in the
questionnaire. This is done to ensure that responses are meaningful and
consistent through the questionnaire and are not merely an artefact of the way
the statements are expressed. In fact the pattern of responses to the two
questions is almost identical, with over 50 per cent agreeing or strongly
agreeing that they are best characterized as focused givers. However, a sub-
stantial minority (in excess of 25 per cent) responded that they tended to give
very widely.
Finally, Figure 2.3(c) presents the results of a question as to whether
people really cared what kind of good cause they were giving to. Interest-
ingly, this question divided the sample more or less down the middle, with
just over 40 per cent agreeing or strongly agreeing and just under 40 per cent
disagreeing or strongly disagreeing.
The responses to the issue of foresight are contained in Figure 2.4. Once
again two versions of the question are used to test for sensitivity to wording.
Although the overall pattern of responses is skewed towards agreement with
these statements in both cases, that is, revealing a tendency towards sponta-
neous giving, the second statement seems to elicit a substantially higher rate
of agreement than the first, just under 70 per cent versus just under 50 per
cent.
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6
(a) ‘I tend to give to one or two favourite charities or a specific area of

charitable activity’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6
(b) ‘I don’t tend to give to a wide range of different charities’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6
(c) ‘When I give, I don’t really mind what I’m giving to as long as it’s some
kind of good cause’
Figure 2.3 Attitudes towards charitable donations: scope

strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(a) ‘I give money spontaneously when I am approached by someone’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(b) ‘I don’t plan my charitable giving carefully in advance’
Figure 2.4 Attitudes towards charitable donations: foresight
The responses to the issue of commitment are contained in Figure 2.5. In

this case both versions of the question are positive. Thus a consistent re-
sponse would require that a person who agreed with the first statement would
disagree with the second. The figures show that, on average, this is indeed the
case. Just under 60 per cent of the sample identify themselves as uncommit-
ted givers by disagreeing with the first statement and agreeing with the
second, while the converse pattern of responses identifies just under 30 per
cent of the sample as committed givers.
Once again it is interesting to consider the interrelationship between these
different types of giving habits, as shown in Table 2.8.
The results indicate that the different questions relating to scope have the
expected mutual correlations. The value is 0.20 for responses to the questions
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(a) ‘I tend to support a particular cause with regular donations over a period
of time’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(b) ‘I tend to make one-off donations with no particular long-term commit-
ment’
Figure 2.5 Attitudes towards charitable donations: commitment
addressing the range of charities to which a person gives. However, both

these questions are negatively correlated with the question indicating that
people don’t really mind what good cause they’re giving to. None of these
correlations is particularly high in absolute terms.
The different questions relating to foresight and commitment also have the
expected mutual correlations even though, in absolute terms, the values are
not particularly high. The correlation is 0.17 for responses to the questions
about giving spontaneously and not planning one’s giving, while the value is
–0.11 for responses to the questions about making one-off donations and
supporting causes regularly over time.
Table 2.8 Correlation between different approaches to monetary donations
Scope Scope Scope Foresight Foresight Commitment Commitment

(a) (b) (c) (a) (b) (a) (b)
Scope (a) 1
41
Scope (b) 0.20 1
Scope (c) –0.05 –0.11 1
Foresight (a) –0.07 –0.21 0.41 1
Foresight (b) –0.01 0.15 0.05 0.17 1
Commitment (a) 0.37 0.03 0.10 0.06 –0.18 1
Commitment (b) 0.07 –0.01 0.11 0.13 0.20 –0.11 1
There is a high correlation (0.41) between those who tend to give sponta-
neously and those who are completely indifferent about which good cause
they are supporting. There is also a high correlation (0.37) between those
who tend to give to favourite charities and those who give regularly over
time. Indeed, 62 per cent of respondents consistently agreed or disagreed
with both of these statements.
Putting together the results of the attitudinal questions on motive and those
on the process of donating uncovers some further interesting associations.
The tendency to give spontaneously is quite strongly associated both with
emotional motivations for giving (a correlation of 0.31 with the embarrass-
ment motive and 0.23 with the warm-glow motive) and with more altruistic
motives for giving (a correlation of 0.36 for the indirect use motive and 0.27
for the existence value motive). Furthermore, there is quite a strong correla-
tion (0.37) between those who tend to concentrate their giving on a few
favourite charities and those who identify themselves as being motivated to
give out of regard for the associated good cause.
By and large these results conform to prior expectations regarding the
motivations behind support of charitable activities in the light of previous
findings reported elsewhere in the literature. While no single motivation
stands out as the most important factor driving respondents’ attitudes, as
many considerations seem to play a role in individual attitudes, by and large
the more altruistic group of motivations seems to play a fundamental role in
explaining people’s attitudes towards charitable donations. As for the process
of giving, typically it is neither carefully planned nor implies a long-term
commitment by donors and it tends to be focused on a number of favourite
charities.
2.3.4 Attitudes towards Volunteering
In parallel to the investigation of the motivations behind monetary donations,

the survey also collected information on the motives behind volunteering.
Figure 2.6 summarizes the results. Once again the statements are organized
along a spectrum from selfish to altruistic motives. However, given the differ-
ent nature of volunteering, the motivational categories do not exactly match
up with those used for monetary donations.
At the selfish end of the spectrum of motivations lies personal enjoyment:
a person volunteers because the very act of volunteering is enjoyable or
personally advantageous in some way. Figure 2.6(a–c) explores three aspects
of this: volunteering as a way of occupying free time, volunteering as a way
of making friends and volunteering as a way of gaining skills. The responses
to all of these questions split the sample to a considerable extent. Just over 40
per cent agree that it is a good way of occupying free time, just over 50 per
cent agree that it is a good way of making friends, and nearly 50 per cent
agree that it is a good way of acquiring skills.
The expectation of some personal benefit may also provide the rationale for
volunteering: a person volunteers because their relatives are among the direct
beneficiary group of the services provided. This is not unlike the option value
statement posed for monetary donations. As with the option value question
discussed above, Figure 2.6(d) shows that the sample of volunteers is fairly
evenly split between those who acknowledge some kind of personal benefit and
those who do not. The former are in a slight majority at just under 50 per cent.
At the other end of the spectrum of motivations lie altruistic considera-
tions: a person volunteers because he values the ultimate good cause which
the charity works towards. Figure 2.6(e) shows that the vast majority of
people see their volunteering activity in these terms. As in the case of dona-
tions, over 80 per cent of the sample identify with this altruistic motivation.
In order to shed light on the overlap between motivations for individual
respondents, Table 2.9 reports the correlation coefficients between each pair
of attitudinal variables. In contrast to the results reported for monetary dona-
tions in Table 2.7, there are strong positive associations between all the
different motivations, indicating that many of these considerations tend to
play a role for many individuals as opposed to some individuals being driven
primarily by some motivations and others by different ones.
A final attitudinal question asked volunteers to think about what would be
the most likely alternative use of their time if they no longer had the opportu-
nity to volunteer. This makes it possible to establish what is being sacrificed
by the volunteer in order to give time to the charity and thus sheds some light
on the issue of how much the volunteer values their donation of time.
Economists have often approached the valuation of volunteer time in terms
of the net hourly wage that the volunteer could have earned had they been
working instead of volunteering (see Chapter 4). Clearly, this is only legiti-
mate to the extent that people are genuinely substituting between volunteering
and paid employment. However, the results of the attitudinal question pre-
sented in Figure 2.7 indicate that only 20 per cent of volunteers see themselves
as giving up opportunities for paid employment in order to volunteer. Hence
the traditional approach of valuing volunteer time in terms of the net hourly
wage is only valid for a small subset of the volunteering population.
Indeed, Figure 2.7 indicates that an equal proportion of just over 20 per
cent of volunteers state that they would not have anything in particular to do
if they could not volunteer. This suggests that this group of people place no
value whatsoever (or perhaps even a negative value) on the time which they
give to charities as volunteers.
By far the largest category of volunteers are giving up time that they would
otherwise devote to domestic work (about 40 per cent) or to recreation (about
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(a) ‘Volunteering is a good way of occupying my free time’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(b) ‘I find it enables me to meet people and make friends’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(c) ‘I think of volunteering as gaining valuable skills and experience’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(d) ‘I think of volunteering as helping my family and friends’
strongly agree
agree
neutral
disagree
strongly disagree
0 0.1 0.2 0.3 0.4 0.5 0.6

(e) ‘I find I enjoy contributing to a good cause’
Figure 2.6 Attitudes towards volunteering: motive
Table 2.9 Correlations between different motives for volunteering
Free Make Gain Help Good

time friends skills family cause
Free time 1
Make friends 0.19 1
Gain skills 0.34 0.31 1
Help family 0.37 0.42 0.47 1
Good cause 0.26 0.44 0.33 0.60 1
Paid Recreation
14% Other
employment 4%
20%
Nothing in Home
particular responsibilities
21% 41%
‘If for any reason you no longer had the opportunity to volunteer, what
would you be most likely to do with your time instead?’
Figure 2.7 Alternatives to volunteering
14 per cent). These uses of time could be expected to have a positive value to
respondents but one which does not necessarily bear any relationship to the
net hourly wage that they could potentially earn in employment.
2.3.5 Valuation Results
This section describes the estimation of the value of all charities and the
value of the subset of the housing and homelessness charities obtained from
both the CV and the CR versions of the general public survey.
Contingent valuation results

As described in Section 2.2, two types of elicitation questions were asked in
the CV versions of the questionnaire, for both CVALL and CVHH: first, the
dichotomous choice approach was used, whereby respondents were asked
whether they would be willing to pay a sequence of predetermined amounts
of money in taxes (the bid levels) that varied across people, to which they
only had to respond ‘yes’, ‘no’ or ‘don’t know’; and second, respondents
were presented with the payment ladder depicted in Table 2.2 and asked to
choose the amount that best described their maximum WTP. The results of
each approach will be considered in turn.
Table 2.10 describes the percentage of respondents that accepted both bids
(YY), accepted the first and rejected the second (YN), rejected the first bid
but accepted the second (NY) or rejected both bids (NN), for each of the five
bid vectors in Table 2.1. Inspection of Table 2.10 shows that, in the case of all
charities and of the housing and homelessness charities, only 8 per cent and
Table 2.10 Percentage of respondents in each response category
NN NY YN YY
All Housing All Housing All Housing All Housing

charities charities charities charities charities charities charities charities
I 31 25 6 4 13 14 50 57
II 26 32 3 14 28 20 43 34
III 32 46 14 11 31 21 23 22
IV 42 47 14 15 23 29 21 9
V 58 53 15 16 19 17 8 14
Notes: NN: rejected both bids; NY: rejected the first bid accepted the second; YN: accepted
the first bid, rejected the second; YY: accepted both bids; I–V: bid vectors as in Table 2.1.
14 per cent of respondents respectively accepted the highest bid offered

(CVALL: £25 per month, corresponding to £300 per year; CVHH: £15 per
month, corresponding to £180 per year). As expected, the percentages are
low since these were the highest proposed tax levels.
The data summarized in Table 2.10 do not provide a direct WTP value for
the services provided by charities. They only clarify whether a respondent’s
WTP is above or below certain bid levels. However, using appropriate statis-
tical models and assuming particular probabilistic distributions for the WTP,
it is possible to infer the average WTP of the sample either using only the
results from the first valuation question (the single-bounded approach) or
using both valuation questions (the double-bounded approach). The latter
approach was followed. Further details can be found in the Statistical Appen-
dix at the end of this chapter.
In the surveys of all charities and of the housing charities, 15 per cent and
18 per cent of respondents respectively were uncertain about being prepared
to pay the specified bid levels and answered ‘don’t know’ to the valuation
question. As such it was necessary to make assumptions about the most
probable direction of this type of answers. In the calculations that follow, we
chose the two uncertainty assumptions: we treated all ‘don’t know’ answers
as ‘no’ (the lower-bound approach) and also as ‘yes’ (the upper-bound ap-
proach). Thus true WTP can be said to lie between these two boundaries.
Table 2.11 displays both approaches for all charities and for housing charities
alone. Yearly bid levels were used in the estimation.
The statistical models depicted in Table 2.11 do not identify WTP directly,
but estimate the probability of accepting to pay a particular tax level as a
function of that yearly tax level. As such the coefficient on the tax level
measures its impact on the probability of acceptance. All the tax coefficients
Table 2.11 Yearly WTP estimates for different models and uncertainty assumptions
All charities Housing and homelessness charities
‘Don’t know’ = ‘No’ ‘Don’t know’ = ‘Yes’ ‘Don’t know’ = ‘No’ ‘Don’t know’ = ‘Yes’
coeff. t-stat. coeff. t-stat. coeff. t-stat. coeff. t-stat.
Constant 0.4603 5.153 0.6755 7.042 0.4300 4.789 0.7120 7.478
48
Tax level –0.9816 –16.203 –1.0687 –17.937 –1.2723 –14.124 –1.4079 –15.649
Log-L –771.19 –414.47 –372.80 –412.56

N 278 278 282 282
Yearly WTP £46.90 £63.21 £33.80 £50.57

95% CI £31.38–£62.41 £49.84–£76.58 £21.81–£45.79 £40.38–£60.77
Notes: Model specification = double-bounded probit; CI = confidence interval calculated using the Krinsky and Robb (1986) procedure.
are negative as expected: the higher the bid level presented to respondents,
the lower the probability of acceptance.
The last two rows in Table 2.11 show the mean yearly WTP for all charities
and for the housing and homelessness charities respectively and their respec-
tive confidence intervals, estimated under both uncertainty treatments (see
the Statistical Appendix to this chapter for details of the estimation procedure
used). WTP for all charitable services (over and above the current level of
donations) ranges from £47 to £63 per person per year, while WTP for the
housing and homelessness charities alone is in the interval of £34 to £51 per
person per year. As expected, the upper-bound approach of treating ‘don’t
know’ as ‘yes’ yields higher estimates than the corresponding lower-bound
treatment where uncertain answers are treated as refusals. However, due to
the imprecision associated with the estimates (which can be judged by the
magnitude of the confidence intervals around the mean), these boundary
values are actually not statistically different from each other. Another way of
putting this is that the stated confidence intervals around the mean are over-
lapping.
Finally, whatever the uncertainty assumption taken, WTP for CVHH is
smaller than that for CVALL. However, in none of the cases are the differences
between WTP amounts statistically significant. This lends support to the
insensitivity to scope hypothesis discussed initially: when valuing one subsector
on its own, people tend to overestimate its value. In our case, the WTP for the
housing and homelessness sector, estimated independently, does not seem to
be significantly different from the WTP for all charitable sectors.
Attention now turns to the results obtained from the second type of valua-
tion questions used: the payment ladder described in Table 2.2, for both the
CVALL and CVHH versions of the questionnaire. It should be noted that these
results are likely to be correlated with the estimates obtained in the dichoto-
mous choice elicitation, since the same respondents were asked both valuation
questions. The yearly WTP results from the payment ladder for all charities
and for the housing and homelessness charities alone are displayed in Table
2.12. The average maximum yearly WTP is £49 for all charities and £41 for
the housing and homelessness charities: the two values are not, however,
statistically different. As already indicated by the dichotomous choice results,
there seems to be considerable insensitivity to the scope of the charitable
sector being analysed.
Median WTP values for CVALL (£24) and CVHH (£18) are considerably
lower than their respective means, which signals the presence of outliers
(unrealistically high bids), as can be confirmed by looking at the distribution
of WTP amounts depicted in Figure 2.8. Both distributions are quite similar –
generally downward sloping as expected with some outliers at the far end of
the right-hand tail.
Table 2.12 Yearly WTP for CVALL and for CVHH from payment ladder
Housing and homelessness

All charities charities
Maximum Minimum Maximum Minimum

WTP not WTP WTP not WTP
(✔) (✘) (✔) (✘)
Mean £49.21 £66.87 £41.22 £56.94

95% confidence £40.00–£58.41 £54.81–£78.92 £33.56–£48.88 £47.23–£66.65
interval
Median £24 £36 £18 £30
Minimum £0 £1.2 £0 £1.2
Maximum £600 £900 £540 £600
N 272 266 275 274
90
80
All charities
70
Housing charities
60
50
40
30
20
10
0
0
1–25
26–50
51–75
76–100
100–125
126–150
151–175
176–200
201–400
401–600
Figure 2.8 WTP distribution in CVALL and CVHH
Comparing the WTP values directly obtained from the payment ladder
with the indirectly estimated WTP amounts from the dichotomous choice
(‘yes’ or ‘no’) procedure for all charities and for the housing charities alone
(Tables 2.11 and 2.12), it can be seen that the payment ladder values are
much closer and statistically not different from the dichotomous choice esti-
mates under the lower-bound uncertainty assumption, that is, when ‘don’t
know’ answers are treated as ‘no’. In contrast, the payment ladder values are
much lower than the upper-bound estimates from the dichotomous choice
models, which suggests that the latter may be biased upwards and not a true
reflection of people’s preferences. Further inspection of these results shows
that the upper-bound dichotomous choice estimates fall within respondents’
uncertainty range, as revealed by the payment ladder, that is, the difference
between what respondents are willing to pay for sure (the ticks) and the
amounts they are sure they would not pay (the crosses).
These results indicate that respondents seem to have been reasonably con-
sistent when answering the dichotomous choice and the payment ladder
valuation questions. The WTP estimates seem to be robust to the elicitation
method chosen when the right uncertainty assumption is made, that is, the
lower-bound approach. The ladder WTP results match the dichotomous choice
estimates when ‘don’t know’ answers are treated as ‘no’, which suggests that,
in the latter, respondents only accepted to pay a given amount when they
were almost sure they actually would do it. It is interesting to note that the
number of respondents expressing uncertainty in the payment ladder (blank
values between the last tick and the first cross) and in the dichotomous choice
questions (‘don’t know’ answers) is remarkably similar: 18 per cent in both
approaches for CVHH; and 22 per cent for ladder versus 15 per cent for
dichotomous choice for CVALL.
In the valuation of all charities questionnaires (CVALL), when they had
completed the payment ladder elicitation procedure, respondents were also
asked how many pence (out of every pound in their global WTP amount for
all charities) they thought should be given to the four charitable sectors of
interest. Table 2.13 shows the results. This procedure, whereby the total value
of a good is first estimated and then reallocated into subcomponents of that
good, is sometimes called the ‘top-down’ approach.
As expected, the top-down estimation procedure yields significantly lower
values for the subcomponent areas of the charitable sector than if these latter
values were directly estimated separately. This is apparent by looking at the
value of the housing and homelessness sector for which we have both the
WTP derived from the top-down approach (£13 per year in Table 2.13) and
the directly estimated value, either from the payment ladder (£41 per year in
Table 2.12) or from the dichotomous choice approach (£34 to £51 per year in
Table 2.11). These large differences suggest that due to the degree of insensi-
tivity to scope exhibited by respondents, the top-down approach may be
preferable for estimating the value of individual subsectors which are part of
a larger sector: first, estimate the total value of the charitable sector; then
allocate that value across subsectors. Otherwise, when faced with a subsector
on its own (as the housing and homelessness charities), respondents tend to
offer their whole available budget for all charities to the particular subsector
they are being asked to value. Arguably this phenomenon would take place
also with the other types of charities considered (social services, health and
Table 2.13 Distribution of total WTP across different areas of charitable activity
Overseas aid,
Housing and Social Health and environment Total WTP
homelessness services medical research and culture (£/year)
52
Weighting 0.24 0.26 0.37 0.13 1
Estimated mean WTP (£/year) £13.10 £18.09 £19.37 £8.21 £58.78
95% Confidence interval £10.5–£15.7 £13.3–£22.9 £15.5–£23.3 £6.4–£10.1 £47.4–£69.8
N 216 216 216 216 216
medical research and overseas aid, environment and culture), although our
survey did not test for them explicitly.
It should be noted that the total WTP for all charities drawn from the
payment ladder as described in Table 2.13 (£59 per year) is different from
that depicted in Table 2.12 (£49 per year). This is due to non-response to the
top-down approach question. Different numbers of respondents were used to
calculate the value in both cases: the top-down estimates in Table 2.13 are
based on 216 respondents vis-à-vis the 272 answers on Table 2.12.
The validity of the payment ladder WTP results was assessed using valua-
tion functions which estimate the impact of possible explanatory variables on
WTP. The valuation functions are obtained by regressing the maximum WTP
amounts (the last tick on the ladder) on a set of economic, attitudinal and
other explanatory variables. It is then possible to investigate whether these
variables affect WTP in the manner predicted by theory. For example eco-
nomic theory suggests that WTP should vary positively with income, if
charitable services are considered to be normal goods. Table 2.14 describes
the sociological, economic, behavioural and attitudinal variables thought to
influence WTP that were included in the valuation functions. The same type
of analysis was conducted on the dichotomous choice estimates with similar
results (further details can be found in EFTEC, 1997).
The results of the valuation functions for CVALL and CVHH are illustrated in
Table 2.15. The regression method used was ordinary least squares (OLS –
see the Statistical Appendix for further details). Two model specifications
were tested: in the first, personal income is included as a possible economic
explanatory variable, and in the second, income is excluded from the set of
economic explanatory variables and included instead are two proxies for
income (dummy variables reflecting full-time employment status and socio-
economic population segment AB). The reason for this dual estimation
procedure lies in the high proportion of income non-response (46 per cent),
which is not uncommon in household surveys.
Overall, the regressions with income perform better than the equations
without income but with income proxies, with an explanatory power of 58
per cent (CVALL) and 42 per cent (CVHH) vis-à-vis 33 per cent (CVALL) and 25
per cent (CVHH). These levels of explanatory power (R2) are relatively high
for cross-sectional data contingent valuation studies (where the minimum
acceptable threshold is considered to be as low as 15 per cent: Mitchell and
Carson, 1989).
Conforming to prior expectations, both income and the level of current
donations are found to be strong determinants of WTP: the richer one is and
the higher the current level of philanthropic giving, the higher the WTP for
all charities in general and for the housing charities in particular. In the
alternative specification with the income proxies the population segment AB
Table 2.14 Description of explanatory variables
Sociological variables
Sex 1 – male; 0 – female
Age Mid-point of interval
Education Age at which full-time education was completed
Economic variables
Income Adjusted mid-point of interval (individual,
monthly)
Full-time job 1 – employed full-time; 0 – otherwise
Segment AB 1 – segment AB; 0 – otherwise
Cash donations £ donated in the previous year
Attitudinal variables
Charities’ importance 3 – very important, to 0 – not important at all
Housing charities 1 – most important charity area, to 4 – least
importance important charity area
Existence values 2 – if strongly agrees that support for charities is
due to existence related motives, to (–2) – if
strongly disagrees
Indirect use values 2 – if strongly agrees that support for charities is
due to indirect use related motives, to (–2) – if
strongly disagrees
Option values 2 – if strongly agrees that support for charities is
due to option related motives, to (–2) – if strongly
disagrees
Type of donation
Diverse donations 2 – if donations are focused on a small number of
charities; (–2) – if donations are wide-ranging
Tax-efficient 1 – if donations are tax efficient; 0 – if otherwise
Charity credit card 1 – if uses a charity credit card; 0 – if otherwise
Appeals 1 – if gives in response to an appeal; 0 – if other-
wise
Merchandise 1 – if buys charity merchandise; 0 – if otherwise
Big Issue 1 – if buys The Big Issue; 0 – if otherwise
Beggars 1 – if gives directly to beggars and others; 0 –
otherwise
Other variables
Tax level Initial tax or bid level presented (in yearly terms)
Volunteering 1 – if volunteered in the previous year; 0 – if not
Diana 1 – if questionnaire was done after the death of
Diana, Princess of Wales; 0 – if before
London Sampling point: 1 – London; 0 – elsewhere
is a positive and significant determinant of giving as expected, with current

donations remaining significant only for the housing and homelessness re-
gression.
Regarding attitudinal variables, existence value motivations are found to
positively influence WTP for all charities, albeit only at the 10 per cent level.
Interestingly, in the case of housing charities, indirect use benefits seem to be
behind donations. The results suggest that those who receive indirect benefits
from the services provided by these charities, for example keeping the home-
less away from the streets and thus providing a better social atmosphere, are
more likely to pay for such services.
In CVHH, the coefficient of the variable that reflects the importance of
housing and homelessness charities in particular, vis-à-vis other areas of
charitable giving, is not statistically significant, even though it has the ex-
pected negative sign, meaning that the higher the rank obtained by this
charitable area (1, 2, 3 or 4, with 1 being the highest), the higher the probabil-
ity of accepting paying for it. Those who think charitable work is important
in general are found to have a higher WTP.
The type of donation is seen to have a significant impact on the WTP for
all charities when income is not included as a regressor but less so when
income is part of the specification. Respondents who donate to a wide variety
of charities rather than channelling their donations to particular causes (di-
verse donations) and who respond to TV, mail, telephone or newspaper appeals
are found to value charitable services more highly. In contrast, those who use
a charity credit card or buy charity merchandise tend to give less. The impact
of this type of variable on the WTP for the housing and homelessness sector
is limited. Other regressors specifically related to housing and homelessness
services such as giving directly to beggars in the street or buying The Big
Issue are not statistically significant even though they have the expected
positive signs.
Socioeconomic variables such as gender and age have a significant impact
on the WTP for all charities. The results show that men are more willing to
pay than women (which contradicts the stylized fact that women are more
generous than men, as uncovered in various other valuation studies), as are
younger people. Education has a positive though insignificant influence. None
of these variables is significant in explaining WTP for housing and homeless-
ness charities.
Interestingly, the death of Diana, Princess of Wales, does not seem to have
a significant impact on actual WTP levels, whatever the specification, al-
though the variable coefficient has a positive sign.
Finally, the tax level coefficient is positive and significant in nearly all
OLS regressions. This is an important finding as it indicates that the initial
bid level (presented in the dichotomous choice elicitation questions) posi-
Table 2.15 Valuation functions for CVALL and CVHH: payment ladder results
All charities (CVALL) Housing and homelessness charities (CVHH)
With income Without income With income Without income
Coeff. t-stat. Coeff. t-stat. Coeff. t-stat. Coeff. t-stat.
Constant –12.36 –0.24 37.50 0.92 –95.04 –1.23 –77.24 –1.11

Sex 21.03 2.00 20.80 2.32 1.33 0.13 10.62 1.37
56
Age –0.77 –2.29 –0.63 –2.06 0.03 0.07 0.21 0.65
Education 1.89 0.82 0.85 0.43 3.96 0.97 2.52 0.70
Economic variables
Income 0.02 2.32 – – 0.03 2.51 – –
Full-time job – – 18.91 2.45 – – 13.37 1.54
Segment AB – – 25.05 1.85 – – 25.01 2.18
Cash donations 0.46 2.43 0.05 0.87 0.33 3.64 0.213 2.09
Charities’ import. 3.78 0.47 3.03 0.56 12.92 1.67 12.08 2.22
Housing char. imp. –1.81 –0.39 –7.52 –1.72 –6.61 –1.09 –2.42 –0.59
Indirect use value – – – – 16.43 3.30 13.13 2.92
Option values – – – – 2.31 0.65 3.62 1.03
Existence values 10.99 1.65 9.36 1.80 – – – –
Type of donation
Diverse donations –8.05 –1.35 –16.71 –3.30 – – – –
Tax-efficient – – – – –49.84 –1.82 –18.15 –1.03
Charity credit card –24.71 –0.73 –59.60 –2.20 – – – –
Appeals 10.90 0.66 20.27 1.91 – – – –
Merchandise –23.61 –2.36 –16.53 –2.17 – – – –
Big Issue – – – – 5.72 0.35 14.74 1.14
Beggars – – – – 13.23 0.57 20.63 1.30
Other variables
Tax level 2.99 2.44 1.59 1.57 2.76 1.88 3.50 2.73
Volunteering 11.58 0.92 41.67 3.16 – – – –
London – – – – –21.32 –1.41 –6.91 –0.66
Diana 1.83 0.19 3.03 0.34 4.39 0.33 1.63 0.16
57
R2 0.58 0.33 0.42 0.25
N 113 196 111 199
tively influences the WTP outcome derived from the payment ladder. This
confirms the correlation hypothesis previously mentioned: the WTP amounts
chosen on the ladder are not independent of the preceding dichotomous
choice questions. This results from the fact that the same sample was con-
fronted with both sets of questions, rather than having a split-sample context.
This within-sample approach permits a test of internal consistency, which
was achieved by respondents: the WTP values estimated both from the pay-
ment ladder and the ‘yes’/‘no’ questions are statistically indistinguishable
(under the lower-bound uncertainty assumption). However, the significant
coefficient of the tax level in the regressions of Table 2.15 suggests that
payment ladder results may suffer some sort of anchoring bias, where the bid
levels presented in the dichotomous choice procedure affect the WTP state-
ment.
Contingent ranking results

In order to derive WTP estimates from the ranking data it is necessary to
make use of a statistical model known as the rank-ordered logit. This
statistical model estimates coefficients which indicate the sign and magni-
tude of the effect of each particular aspect of the options offered on the
ranking accorded to that option. Further details can be found in the Statisti-
cal Appendix to this chapter. Amongst the explanatory variables used in the
model are the various attributes of the options presented to respondents (see
Table 2.3). Each of these variables shows whether or not the respective
subsector is shut down under any particular option. In addition, the ‘overall’
variable represents an interaction of all four subsectors in combination, that
is, it indicates whether the whole charitable sector shuts down. The last
attribute (tax bill) represents the monetary cost associated with any particu-
lar option.
The results from this model are reported in Table 2.16. A positive coeffi-
cient on a variable indicates that the respective subsector is positively valued.
The negative coefficient on the overall variable indicates that the overall
value for all four sectors in combination is smaller than the sum of the values
of each sector when presented individually. This is consistent with the insen-
sitivity to scope phenomenon that was also apparent in the contingent valuation
version. The tax bill associated with the options has a negative coefficient,
indicating that the higher the cost of an option, the less likely people are to
select that option.
On the basis of these coefficient results it is possible to calculate the WTP
for each sector individually and for all sectors in combination (see Statistical
Appendix). The results in Table 2.16 show that the health and medical re-
search sector has the highest annual WTP for any individual charitable sector
at £149.95 on average. This is followed by social services with £99.16 on
Table 2.16 Results of contingent ranking version
Model coefficient Annual WTP

(confidence interval) t-statistic (confidence interval)
Housing and 0.1598 1.72 £35.75

homelessness (0.0669–0.2527) (–£4.21–£75.71)
Social services 0.4498 4.79 £99.16
(0.2658–0.6338) (£51.36–£146.96)
Health and medical 0.6800 6.18 £149.95
research (0.4642–0.8958) (£96.74–£203.16)
Overseas aid, environment 0.3012 2.91 £66.14
and culture (0.0985–0.5038) (£20.10–£112.18)
Overall –0.7017 –3.07 £195.52
(–1.1494– –0.2541) (£132.00–£259.04)
Tax bill –0.0046 –5.92 n/a
(–0.0061– –0.0031)
N 702
Log-likelihood –1230.81
Note: Confidence intervals estimated with the Krinsky and Robb (1986) approach.
average, overseas aid, environment and culture with £66.14 on average and
housing and homelessness with £35.75 on average. However, the confidence
intervals on all the estimated values are very wide. Consequently the WTP
for housing and homelessness charities is not significantly different from
zero. Moreover, the valuation of the housing and homelessness, social serv-
ices, and overseas aid, environment and culture subsectors are not significantly
different from each other. The value of the health and medical research
subsector is not significantly different from the social services and overseas
aid, environment and culture subsectors. Neither is it significantly different
from the estimated overall value accorded to all charities, which at £195
appears to be unrealistically high.
Debriefing questions
Table 2.17 summarizes the responses given to the common debriefing ques-
tions which were administered on each version of the questionnaire. These
responses should be regarded as indicative rather than conclusive evidence.
This is because respondents were faced with a list of precoded reasons why
they were or were not willing to pay for charitable services and were asked to
choose only one of these reasons, while, typically, WTP responses are the
end result of a number of different (maybe even contradictory) motivations,
as the attitudinal part of the survey shows. Still, they illustrate some of the
potential caveats associated with valuation surveys.
Table 2.17 Summary of debriefing questions across subsamples

(percentage of respondents aggreeing with the statement)
CVHH CVALL CR
Reasons for being willing to pay:

I think that the work done by these charities 36.2 64.5 51.0
is worth the extra contribution. (29.5–42.9) (57.6–71.4) (43.9–58.1)
I think this is a very important issue, I am 30.6 35.5 48.9
not sure if I could pay this amount but I (24.1–37.1) (28.6–42.4) (41.8–56.0)
wish I could.
I think charities in general are a very 33.2 n.a. n.a.
important sector. My answer reflects my (26.5–39.9)
concern for charities in general and not
simply for those working on housing and
homelessness.
Reasons for not being willing to pay:

I do not care that much about these charities, 5.6 7.6 5.6
and would rather spend the money on other (0.7–10.5) (2.1–13.1) (0.3–10.9)
things that are more important to me.
I could not afford the additional amount of 23.6 22.8 18.1
tax which was being asked of me. (14.8–32.4) (14.2–31.4) (9.1–27.1)
I think I already make enough voluntary 9.0 3.3 6.9
donations to charities as it is, and that other (3.1–14.9) (–0.4–7.0) (1.0–12.8)
people ought to pay more.
I think I already pay enough tax as it is. 13.5 12.0 18.1
(6.4–20.6) (5.3–18.7) (9.1–27.1)
I think people should be left to make their 44.9 50.0 44.4
own contributions if they want to rather than (34.5–55.3) (39.8–60.2) (32.8–56.0)
being forced to contribute through tax.
I did not find the idea very convincing. 3.4 4.3 6.9
(–0.3–7.1) (0.2–8.4) (1.0–12.8)
By and large, there are no significant differences between the responses

given for each version of the questionnaire, as can be seen from the fact that
the respective confidence intervals always overlap.
In the surveys, people were asked the reasons behind their WTP. This
information can be used to distinguish between those with a genuine WTP
and those who may simply be using this question to express a positive
attitude towards charities. The results indicate that a sizeable minority of at
least a third fell into this latter category. For the CVHH version of the ques-
tionnaire a further third claim to be expressing their valuation of all charities
rather than simply the housing and homelessness sector as requested in the
questionnaire (which goes some way to explain the insensitivity to scope
found). For the CVALL two-thirds of respondents confirm that they are ex-
pressing a genuine WTP, whereas for CR the proportion is substantially
lower, at about one-half.
Respondents were also asked for the reasons why they were not willing to
pay. This question attempts to distinguish between those who are genuinely
unwilling to pay and those who may simply be registering a ‘protest’ of some
kind against the questionnaire. The first three of the reasons identified (which
account for about a third of the sample) constitute a genuine unwillingness to
pay. The most common of these reasons is inability to afford the contribu-
tions. The last three of the reasons identified (which account for about
two-thirds of the sample) are more indicative of people expressing hostility
towards the scenario proposed. The most common of these reasons is a
rejection of the idea of compulsory contributions through taxation. Reassur-
ingly, only a very small proportion of respondents (3.4 per cent–6.9 per cent)
protested on the grounds that they found the questionnaire unconvincing.
2.4 DISCUSSION AND CONCLUSIONS

The discussion thus far has presented results from a wide range of methods
for assessing WTP to preserve charitable organizations in the UK. The pur-
pose of this section is to bring together the main results in a way that permits
comparisons to be made and thereby to assess which results should be used
for the purposes of valuing the sector.
Table 2.18 summarizes the results obtained for the valuation of all charities
and the housing and homelessness subsector, on the basis of four different
methods incorporated in the survey: double-bounded dichotomous choice,
payment ladder, top-down allocation and contingent ranking.
Inspection of the table reveals a high degree of consistency in the results
obtained from the various permutations of CV: the double-bounded model
and the payment ladder. These produce results in the range £47–£49 per
person per year for all charities, and £34–£41 per person per year for housing
and homelessness charities.
The results obtained from the CR are not significantly different from those
obtained from the CV in the case of the housing and homelessness charities.
However, the CR gives a value for all charities which is about four times as
large as that obtained from CV. Furthermore, the confidence intervals on the
CR estimates are substantially wider than those for the CV estimates, sug-
Table 2.18 A comparison of WTP estimates across models (£ per person

per year)
Housing and
Elicitation method All charities homelessness charities
Dichotomous choice £46.90 £33.80

(with ‘don’t know’ = ‘no’) (£31.38–£62.41) (£21.81–£45.79)
Payment ladder £49.21 £41.22
(£40.00–£58.41) (£33.56–£48.88)
Top-down allocation £58.78 £13.10
(from payment ladder) (£47.4–£69.8) (£10.5–£15.7)
Contingent ranking £195.52 £35.75
(£132.00–£259.04) (–£4.21–£75.71)
Actual donations £42.98 £5.56
(£34.90–£51.10) (£4.50–£6.60)
gesting a higher degree of uncertainty among respondents faced with the

former valuation method.
The above results suggest that the divergence between the CV and CR
results is not so much attributable to the differences between the methods as
to the fact that in the latter survey respondents were taken through a bottom-
up route to the valuation of all charities (that is to say they were presented
with a series of scenarios that focused only on the preservation of a single
subsector before they were presented with a scenario that preserved the
whole sector). This explains the very close correspondence of results with the
CV question, where respondents focused only on preserving the housing and
homelessness sector. However, it is in contrast to the CV, where respondents
were asked to focus on their WTP to preserve all sectors from the outset. This
latter approach produces the more credible and conservative estimate.
Lastly, there is a large divergence between top-down and bottom-up ap-
proaches to valuation. This is evident when the results obtained from CVHH
(about £35) are compared with the results obtained from allocating a share of
the result obtained from CVALL to the housing and homelessness subsector
(about £13). The way to interpret this divergence is that people would be
prepared to pay a lot more to prevent the closure of the housing and home-
lessness sector (if this were the only sector threatened with closure) than they
would be prepared to pay if all sectors were threatened with closure.
For policy purposes, the most conservative estimates of the value of the
charitable sector in the UK, over and above current donations, are therefore
£47 per person per year for the whole philanthropic sector and £13 per person
per year for the housing and homelessness subsector.
These results have important practical implications. Since CR and CV
valuation approaches were found to be perceived differently by respondents,
in each particular circumstance, care must be exercised to choose the most
appropriate methodology. The research findings suggest a number of relevant
guidelines for practitioners. When the policy objective is to evaluate a set of
changes or policies, then the suggested approach is direct CV on the inclusive
good. CR is likely to produce a series of single policy evaluations that, if
summed up, may seriously overestimate the value of the whole set. When the
question of interest involves evaluating a single isolated change or policy,
then both direct CV and indirect CR approaches could be used. CR has the
added advantage of being able to produce values for several of these changes
simultaneously. Finally, if the aim is to uncover the value of a good or policy
that is embedded in a more inclusive good or policy, whose other components
may also be expected to vary, then the top-down approach seems to have
considerable advantages and to produce more robust results.
Table 2.18 also includes an estimate of the survey respondents’ actual level
of donations. Inspection of the table shows that even the lowest estimates of
WTP (over and above current charitable giving) obtained from the valuation
survey are substantially greater than current levels of annual donations. This is
an important result and is suggestive of the extent of free-riding that exists with
voluntary donations to charities. Taking into account the most conservative
WTP estimates, both for all charities and for the housing and homelessness
subsector, respondents state that they are willing to pay more than double their
current level of annual donations in order to prevent the closure (about £47 on
top of the current £43 per person per year for all charities and about £13 on top
of the current £6 per person per year for housing and homelessness charities).
As most stated preference surveys, our study collected various types of
evidence, not just monetary values, as to the relative importance that people
place on each of the charitable subsectors considered. There are four ways in
which this can be measured: first, by looking at attitudinal responses: at the
beginning of the questionnaire respondents were asked to rank the four
charitable subsectors according to their relative importance; second, using
the CV responses: at the end of the CV section, respondents were asked to
allocate their total WTP to preserve all charities among the four different
charitable subsectors; third, using the CR responses: during the CR exercise,
respondents implicitly revealed the relative importance they attached to each
charitable subsector by the way in which they ranked alternatives which
preserved different areas of charitable activity; and lastly, by analysing actual
Table 2.19 Relative importance of different charitable subsectors
Current
Attitudinal CV allocation CR weighting donations
Score Rank Score Rank Score Rank Score Rank
Health and medical 1.77 1st 37.2 1st 0.68 1st 2.36 1st
research (0.03) (1.44) (0.11) (0.34)
Social services 2.29 2nd 26.4 2nd 0.45 2nd 1.14 3rd
(0.03) (1.09) (0.09) (0.13)
Housing and 2.30 3rd 23.7 3rd 0.16 4th 0.7 4th
homelessness (0.03) (0.99) (0.09) (0.10)
Overseas aid, 3.56 4th 17.4 4th 0.30 3rd 1.49 2nd
environment and (0.03) (0.84) (0.10) (0.44)
culture
donations data: during the first section of the questionnaire, information was
collected about the amounts of money that people were currently giving to
each of the different charitable subsectors.
Table 2.19 summarizes this information. A number of interesting points
emerge. Although all four methods put health and medical research charities
as the sector to which people attach the greatest importance, there is signifi-
cant disparity in the relative rankings awarded to the remaining three sectors
depending on the valuation method used. The results obtained from the CR
method provide the closest match with the pattern which shows up in actual
donations. In common with actual donations, the CR method puts the health
and medical research sector first, the housing and homelessness sector last
and only differs on the relative ranking of the social services, and overseas
aid, environment and culture charities. However, in both cases the confidence
intervals for these two intermediate sectors overlap, suggesting that it is more
accurate to think of these sectors as tying rank under both methods.
Interestingly, the CV method gives exactly the same ranking as that which
emerges from the attitudinal questions. This makes sense inasmuch as both
approaches require respondents to think explicitly about the relative impor-
tance of the sectors. This is in contrast to the other two methods, where the
rankings emerge implicitly from the decisions which people make when
faced with an opportunity to give.
With a view to estimating the total benefits provided by the charitable
sector in the UK, further analysis of the results presented in this chapter will
be undertaken in Chapter 5, incorporating information on the value attached
by users to the housing and homelessness charities (Chapter 3) and a discus-
sion of the value of volunteering (Chapter 4).
STATISTICAL APPENDIX
This appendix briefly presents an overview of the theoretical and statistical
framework used to obtain WTP welfare measures from the experimental
survey design used in this chapter. We will examine, in turn, the CV and the
CR approaches.
Contingent Valuation Approach
There are basically two types of data from CV studies: data from open-ended
or payment ladder questions and data from dichotomous choice questions.
The open-ended format directly elicits an individual’s maximum WTP
while the dichotomous choice format presents a monetary amount – the bid
level – or a sequence of amounts to the respondent and asks for a ‘yes’ or ‘no’
vote on the WTP for each bid. Dichotomous choice questions do not elicit
individual maximum WTP directly but intervals where it lies. Therefore, the
only way to obtain a mean WTP value from the data is through a statistical
model linking the money amounts offered to people’s responses.
Several variations of both these question formats were used to elicit indi-
vidual WTP for avoiding closure of the charitable sector in the UK.
Accordingly, different econometric specifications were tested to analyse the
results from the various types of questions. This section presents a summary
of the theoretical and statistical models that were used to analyse the CV
survey data.
Double-bounded dichotomous choice elicitation

Since the mid-1980s, dichotomous choice elicitation procedures have be-
come the most popular way of obtaining information about individual WTP
for environmental resources and services in CV studies. The method stems
from the seminal work of Bishop and Heberlein (1979) and was subse-
quently endorsed by NOAA’s Blue Ribbon Panel in 1993 (Arrow et al.,
1993). As noted above, the discrete choice format consists of take-it-or-
leave-it questions: respondents are asked whether or not they are willing to
pay a certain amount of money (varied across subsamples) for a change in
the good or service in question. Using the observable yes/no answers to the
payment question it is possible to infer the distribution of the underlying
(unobservable) WTP and form a statistic of interest such as the mean or
median WTP. The models that can be used to infer people’s maximum WTP
from binary choice questions are qualitative response models and survival
models.
The main advantage of using discrete response formats is the fact that it
may be easier for respondents to answer ‘yes’ or ‘no’ to a specific bid level
than to come up with a value for the change of interest. Under certain
circumstances it is also an incentive-compatible format (Carson et al., 1997).
Related disadvantages of the dichotomous choice method are as follows:
the answers are less informative than open-ended alternatives since they give
only a discrete indication of a person’s WTP; it is statistically more burden-
some to analyse qualitative answers than numerical ones and it is necessary
to specify a parametric distribution of the WTP or, equivalently, of the indi-
rect utility function to estimate the mean WTP; the choice of bids must be
done carefully in order to span the range of respondents’ true valuations; and
these formats are found systematically to result in higher mean WTP amounts
than open-ended questions, arguably because of a so-called ‘yea-saying’ bias.
This chapter makes use of the double-bounded variant of the dichotomous
choice approach (Hanemann et al., 1991), where respondents are faced with
two valuation questions. The first question asks whether respondents are
prepared to pay a fixed sum of money for some improvement (the bid level
varying across different subsamples). Subsequently, they are faced with a
follow-up payment question that depends on the response to the first bid
level: if the respondent accepted the initial bid they are asked another WTP
question with a higher bid; if the answer to the first bid level is ‘no’, then the
respondent is presented with a lower bid (with the bid amounts varying
across subsamples).
In order to estimate a monetary welfare measure from dichotomous choice
data it is necessary to employ some microeconomic model of choice. The
random utility model approach, suggested by Hanemann (1984), provides the
theoretical choice framework (an alternative approach was developed by
Cameron, 1998).
Assume that individuals have indirect utility functions of the form:
V = U (P, Y , X, C 0 ) (2.1)
where P is a vector of prices, Y is income, X a vector of individual character-

istics and C0 a vector of public goods, such as the services provided by the
charitable sector. Hanemann (1984) explains the individual yes/no answer to
the WTP question in terms of a random utility model. Since some of the
components of the utility function are not observed by the researcher, equa-
tion (2.1) can be rewritten as:
V = U (P, Y , X, C 0 ) + ε (2.2)
where again P is a vector of prices, Y is income, X a vector of individual

characteristics, C0 a vector of public goods, including charitable services, and
ε is the error term reflecting unobserved taste components.
In this new random framework, consider a possible shutdown of the chari-

table sector that reduces the vector of public goods from C0 to C1. Then, C1 <
C0 and:
U (P, Y , X, C 0 ) + ε 0 > U (P, Y , X, C1 ) + ε1 (2.3)
In a dichotomous choice context, the individual is faced with the choice of

saying ‘yes’ or ‘no’ to avoid the welfare decrease arising from the charities’
shutdown at a cost of B. The bid level B is varied across different subsamples.
The respondent accepts to pay this price if:
U (P, Y − B, X, C 0 ) + ε 0 ≥ U (P, Y , X, C1 ) + ε1 (2.4)
U (P, Y − B, X, C 0 ) − U (P, Y , X, C1 ) ≥ ε1 − ε 0 (2.5)
∆U ≥ η (2.6)
where ∆U is the utility difference U0 – U1 and η = ε1 – ε0. Hence the

probability of accepting the price is given by:
P( ‘yes’) = P( ∆, U ≥ η) = P( η ≤ ∆U) = Fη ( ∆U ) (2.7)
where Fη is the cumulative distribution function (cdf) of η, whose functional

form depends on the distribution chosen by the researcher. Typical cdfs include
the standard normal, logistic, log-normal, log-logistic and Weibull distribu-
tions. Hanemann and Kanninen (1996) provide an extensive overview.
Since we are dealing with binary choices (a ‘yes’ or ‘no’ answer to a
particular bid level) the data can be analysed in the framework of parametric
probability models. These models can be estimated by maximum likelihood
techniques (Greene, 1997). Assuming a linear-in-income utility specification,
the log-likelihood function for the double-bounded dichotomous choice model
is described by equation (2.8):
N
log L = ∑ {yyi log[ Fη ( a − bHBi )] + yni log[ Fn ( a − bIBi )
i =1
− Fη ( a − bHBi )] + nyi log[ Fη ( a − bLBi ) − Fη ( a − bIBi )]
+ nni log{1 − Fη ( a − bLBi )]} (2.8)
where yy, yn, ny and nn are dummy variables corresponding to the four
possible response pairs (yes/yes, yes/no, no/yes and no/no); LB, IB and HB
correspond to the lower, initial and high bid levels, respectively; B is the bid
level offered to the individual; a and b the regression coefficients; and Fη is
the cdf assumed by the researcher.
Once the unknown parameters in (2.8) have been estimated, the mean
WTP from dichotomous choice data can be determined. Hanemann (1984)
and Hanemann and Kaninnen (1996) show that, for the linear utility model
referred to above, and on the assumption of a zero mean symmetrically
distributed error term, mean WTP can be calculated as a ratio of coefficients
as given by expression (2.9). This welfare measure corresponds to a Hicksian
equivalent variation.
WTP = − a / b (2.9)
Although the asymptotic distribution of the maximum likelihood estimator

for the parameters a and b is known, the asymptotic distribution of the maxi-
mum likelihood estimator of the welfare measure is not, since mean WTP is a
non-linear function of the parameter vector. One way of obtaining confidence
intervals for this measure is by means of the procedure developed by Krinsky
and Robb (1986). This technique simulates the asymptotic distribution of the
coefficients by taking repeated random draws from the multivariate normal
distribution defined by the coefficient estimates and their associated covariance
matrix. These are used to generate an empirical distribution for the welfare
measure, and the associated confidence intervals can then be computed.
The previous analysis can be readily generalized in the presence of more
explanatory covariates. However, the statistics that are usually of interest to
the researcher, the population (unconditional) mean or median, can be as
easily estimated by the marginal methods described above as by a conditional
approach that first estimates conditional mean WTP as a function of covariates
and then finds its average with respect to an estimate of the density of the
covariates (McFadden, 1994).
Payment ladder elicitation

The charities survey also included payment ladder WTP questions as follow-
ups to the dichotomous choice elicitation mechanism. These questions
consisted of presenting respondents with a payment card and asking them to
identify with a tick the amounts they would be willing to pay to avoid the
closure of the charitable sector and crossing the amounts they were sure they
would not pay.
The main advantage of these data is that they are statistically easy to
manipulate, providing more information than dichotomous choice questions,
that is, exact WTP values and not WTP intervals, so that very few assump-
tions are needed to estimate the mean WTP.
A possible disadvantage of payment cards is that respondents may find it

more difficult to come up with an answer. Assigning meaningful monetary
values to sometimes complex environmental changes without some kind of
external assistance can be a very difficult task to complete in the short time
the CV interview lasts.
A common way of explaining estimated WTP values from payment card
data is by modelling them as a function of possible explanatory factors.
Equation (2.10) illustrates what has been called in the contingent valuation
literature ‘valuation function’ or ‘bid curves’:
WTPi = f (X i ) (2.10)
where WTPi is the reported WTP for individual i and Xi a vector of explana-
tory variables thought to influence individual valuation. These regressors
may be socioeconomic characteristics of interest (such as education and
income) or variables reflecting general attitudes towards the charitable sector,
for example.
The most common specification of the valuation function assumes a linear
relationship between the regressors:
WTPi = β‘ X i + ε I
ε i ~ N (0, σ 2 ) (2.11)
where β is a vector of unknown parameters reflecting the impact of changes

in a given explanatory variable on WTP estimates and εi is a random error
term reflecting factors affecting utility that the researcher is unable to ob-
serve. εi is assumed to be normally distributed with zero mean and constant
variance.
Valuation functions using WTP estimates resulting from open-ended or
payment ladder questions may be modelled by a simple classical ordinary
least squares (OLS) regression. The results from the payment ladder elicitation
procedure reported in this chapter were modelled using this procedure. Note
that these data can also be modelled using interval data procedures as maxi-
mum WTP can be interpreted to be an unobserved amount lying in between
the highest amount ticked in the payment ladder and the next amount up.
Contingent Ranking Approach
The CR method (Beggs et al., 1981) is part of a number of stated preference

techniques originally designed by marketing practitioners to isolate the value of
individual product characteristics or attributes typically supplied in combina-
tion with one another. These techniques, also survey-based, provide a natural
way of analysing multidimensionality of goods and services and situations
where trade-offs between product attributes are of particular interest.
Stated preference techniques such as contingent ranking are gaining in-
creased popularity among valuation practitioners (Laureau and Rae, 1985;
Adamowicz et al., 1994; Johnson and Desvousges, 1997; Foster and Mourato,
2000; Atkinson et al., 2000). As noted by Hanley et al. (1998), these methods
share with the dichotomous choice approach a common theoretical frame-
work in the random utility model as well as a common basis of empirical
analysis in limited dependent variable econometrics. Their presumed advan-
tage lies in their ability to identify part-worths for different components of
the change in question. Other benefits which have been claimed for these
methods are the avoidance of the anchoring problem associated with dichoto-
mous choice (Adamowicz, 1995) and of the part–whole bias problem which
arises in CV more generally (Hanley et al., 1998). On the other hand, these
techniques tend to impose a considerable cognitive burden upon respondents,
increasing the likelihood of unreliable and even inconsistent choices.
In a CR experiment, respondents are asked to rank a number of alterna-
tives, each one consisting of a combination of attributes and prices, set at
varying levels. From the ordinal rankings, the monetary welfare change asso-
ciated with each attribute can be indirectly calculated. In the exercise presented
in this chapter, respondents were asked to rank according to their preferences
three hypothetical options implying closure of one or more charitable
subsectors, each different option implying a different tax burden. The at-
tributes of the choice were therefore the four charitable subsectors of interest
(described in the chapter) and the tax level. Each attribute was presented at
different levels: the charitable sectors could either be ‘open’ or ‘closed’,
while seven different tax levels were considered.
As in the dichotomous choice approach, the random utility model provides
the economic theory framework for analysing the data from the CR exercise.
According to this framework, respondents will select the option that maxi-
mizes their utility or satisfaction. Since the researcher does not observe all
the determinants of individual choice, the utility function for each respondent
i can be decomposed into two parts: a deterministic element, which is a linear
index of the attributes (X) of the j different alternative options in the choice
set; and a stochastic element (ε) which represents unobservable influences on
individual choice. This specification is shown in equation (2.12).
Vij = bXij + ε ij (2.12)
Under the assumption of an independently and identically distributed ran-

dom error (εij) with a Weibull distribution, Beggs et al. (1981) developed a
rank-order logit model capable of using all the information contained in a

survey where alternatives are fully ranked by respondents. It can be shown
that the probability of any particular ranking of options for the charitable
sector being made by individual i can be expressed as:
 
2  
exp(bXij ) 
Pi (Vi1 > Vi 2 > Vi 3 ) = ∏  3 (2.13)
 
 ∑ exp(bXik ) 
j =1
 k = j 
The parameters of the utility function can be estimated by maximizing the

log-likelihood function given in equation (2.14):
 
N  2 
exp(bXij ) 
log L = ∑ log∏  3 (2.14)
 
 ∑ exp(bXik ) 
i =1 j =1
 k = j 
After the parameters in (2.12) have been estimated by maximizing the log-
likelihood function given in (2.14), the welfare measures of interest, that is,
the WTP to avoid the closure of one or more sectors of charitable activity, can
be readily calculated as the marginal rate of substitution between the relevant
charitable subsector and the tax level (a ratio of model coefficients), as
indicated in equation (2.15):
∂U j
∂Charitable sectorij
WTP = − (2.15)
∂U j
∂Taxij
Confidence intervals for the welfare measure can be calculated using the
Krinsky and Robb (1986) procedure described above.
3. The benefits of charities to users: the
homeless
3.1 INTRODUCTION
According to the broad definition of homelessness, a homeless person is any-
one living in precarious, insecure or short-term accommodation, that is, in
hostels, hotels, bed and breakfast (B&B), squatting, sleeping rough and hidden
homelessness (those who sleep around friends and family). Although accurate
statistical information is difficult to obtain, it is estimated that there were
around 140 000 single homeless people in the UK in 1994 (personal communi-
cation, Shelter, 1997). Of those, around 270 were sleeping rough every night.
Hostels are a necessary first step in the resettlement process of most
homeless people. In London alone, there are about 26 000 hostel bed spaces
in over 600 buildings (Resource Information Service, 1996). A common view
of hostels is that they are large buildings, offering poor accommodation in
dormitories, with regimes dominated by strict rules and regulations. How-
ever, while there are still some hostels that conform to this Victorian image,
typically the reality is quite different.
Today nearly 20 different types of hostels exist (Resource Information
Service, 1996). The term ‘housing project’ is increasingly being used to
describe many of the more recently established schemes run by the charitable
sector. Apart from accommodation and food, hostels or housing projects also
offer a range of support and counselling services that are needed by a large
proportion of homeless people for problems related to drink, mental health,
drugs or physical health, among others.
This chapter summarizes a study which aimed to place an economic value
on the services provided by hostels, that is, the housing and homelessness
charities, to their direct beneficiaries, that is, the homeless people who use
hostel services. This study was carried out as part of a wider project on
valuing the output of the charitable sector in the UK. A parallel study was
undertaken to estimate the value for society at large, that is, indirect users and
non-users of the housing and homelessness subsector of charitable activity.
This is discussed in Chapter 2. Together, the two studies permit a complete
evaluation of the housing and homelessness charities, taking into account all
stakeholders involved.
72
Benefits of charities to users 73
The price of hostel accommodation does not reflect the value of the
services provided to users, as in the case of other private goods, first of all,
because the current price of hostel accommodation is subsidized and thus
does not reflect an equilibrium price between supply and demand, and
second, because some of the support services provided by hostels, over and
above food and accommodation (counselling, medical help, financial ad-
vice and so on) are probably not easily found or affordable by the homeless,
outside the hostel circuit. Hence, for the target group of hostel residents,
the benefits are expected to exceed the price currently paid for hostel
services.
In previous work, the value of this sector has been assessed by looking at
the cost side, that is, the income received or the expenditures made by
charitable organizations (Jurgen, 1988; Jencks, 1994; O’Flaherty, 1996). How-
ever, the income received by these charities through public donations of time
and money and government grants is not the correct value of the services they
provide. This is because: (i) given the voluntary nature of charitable giving,
people tend to free-ride and to take for granted that donations by others will
make up for their own lack of generosity and hence individual giving is
suboptimal; (ii) government grants are not calculated in a manner that would
compensate for the existence of free-riding behaviour. They are determined
by political interests and lobbying forces and are not a result of an efficient
choice.
To overcome these problems, the approach introduced in Chapters 1 and 2
was used: in order to estimate the value of the housing and homelessness
charities’ output we looked directly at the benefits they provide. This tech-
nique is known as contingent valuation (Mitchell and Carson, 1989). This
chapter reports the results of a contingent valuation (CV) survey of hostel
residents designed to evaluate the user benefits of hostels. To our knowledge,
it is the first time ever that a minority of the population, living in precarious
conditions and with a range of special support needs, has been administered
an economic valuation survey.
In this framework, to make respondents consider how much the sector was
worth to them, the hypothetical scenario presented in the questionnaire was
the following: due to a financial crisis all the hostels in the country were
facing the prospect of shutting down for a whole year, leaving many people
with no alternative but to sleep rough on the streets. A sample of homeless
people were then asked how much compensation they would need to com-
pensate them for the loss of hostel services over and above any state benefits
they may already receive.
This willingness to accept compensation (WTA) approach has also been
applied in the environmental economics literature (for example Adamowicz
et al., 1993; Shyamsundar and Kramer, 1996; Smith et al., 1997). While, in
theory, it should produce comparable results to the WTP approach (Randall

and Stoll, 1980; Willig, 1976), in practice it has tended to give substantially
larger values (Coursey et al., 1987; Cummings et al., 1986). Explanations for
this phenomenon include income effects (Hanemann, 1991), loss aversion
(Kahneman and Tversky, 1979) and strategic behaviour induced by the ab-
sence of a budget constraint. For all of these reasons, particular care must be
taken in using a compensation-based measure.
The chapter is organized as follows. Section 3.2 describes the survey
design and Section 3.3 presents the survey results. Conclusions are presented
in Section 3.4.
3.2 SURVEY DESIGN

The main objective of the CV survey was to elicit direct user benefits from
housing and homelessness charities. Hostels dominate this subsector of chari-
table activity and its users are homeless people. Hence the CV interviews
took place in hostels or associated day centres in London with a random
sample of homeless people.
The survey collected information about the reasons that led respondents to
become homeless, their current use of hostel accommodation, the quality of
the services provided by hostels and possible substitutes for hostel accommo-
dation. Respondents were also asked about their current level of income and
to estimate how much of that income was spent on accommodation and other
expenditure categories.
The valuation sections of the survey aimed to uncover, in money terms, the
benefits received by the direct recipients of hostel services. For that purpose,
as described above, respondents were faced with a hypothetical scenario –
where all hostels would close down for a year – that was intended to make
them consider carefully how much accommodation and other hostel services
were worth to them. The wording was as follows:
Hostels provide you with accommodation and a number of other support services.
We would like to find out how much you value all of this. We will use the following
question to try and find out how much the accommodation and services are worth
to you.
Imagine that, for some reason, all the hostels in the country had to shut down
for a whole year and you had to find somewhere to stay. (Don’t worry, this is
definitely not going to happen! But we would like you to think about what your life
would be like if it did happen.)
Suppose in order to compensate you for not being able to use the hostels
anymore, you were given an additional cash payment each week, over and above
what you said you receive at the moment.
The shutdown had to be on a national scale so as to eliminate the possibility

of migrating to find alternative hostel accommodation elsewhere. It also had to
last for a considerable period of time to oblige respondents to think about a
formal alternative as opposed to just sleeping rough for a few weeks. However,
arguably, the very dramatic nature of the shutdown described in the question-
naire could have undermined its credibility or caused undue anxiety to the
hostel residents. For these reasons, the presentation encouraged respondents to
regard the scenario as a thought experiment rather than as something that might
actually happen. The WTA question was worded as follows:
How much money would you have to receive each week during the whole year to
give you the same quality of life as you have now (not better or worse, but just the
same)?
The estimated WTA welfare measure is a compensating variation; further

details of the theoretical framework behind welfare measures can be found in
the Statistical Appendix to Chapter 2. After extensive pre-testing it was found
that respondents seemed to be most comfortable with a simple open-ended
elicitation framework, where they were directly asked for their required com-
pensation amount.
Open-ended WTA procedures may be problematic and result in respond-
ents asking for unrealistically high amounts. This possibility was taken into
account in the survey design and, to try to minimize the occurrence of
overstated compensatory amounts, respondents were explicitly asked for sub-
sidies that maintained (and not improved or reduced) their quality of life. To
confirm that this was the case, two screening mechanisms were adopted.
First, interviewers were instructed to repeat the valuation question to every
respondent who asked for more than £500 per week, stressing that they
should think of an amount that would afford them the same quality of life
they have now, not a better one. Second, the sampled homeless were asked
how they would spend the compensatory amount stated (accommodation,
food and so on), were they to actually receive it. Then they were asked the
following debriefing question:
Taking into account what you have just said about how you would spend the
money, do you think that your quality of life would be the same, better or worse
than what you have at the moment?
If the answer was better or worse, the WTA question was repeated and
respondents were urged to state a compensation amount that would leave
them as well off as they were at the current moment. The valuation section
concluded with a follow-up question for respondents who did not answer the
WTA question, to try to uncover the motives behind the refusal.
The final part of the questionnaire collected basic socioeconomic informa-

tion about the respondents, including gender, age, marital status, educational
attainment, employment status and nationality.
All interviews were conducted in person, on site at the housing establish-
ments, and took about 20 minutes to complete. Due to the difficult nature of
the work, the fieldwork was assigned to a firm specialized in conducting
surveys of homeless people. The interviewers were extensively involved in
the development of the questionnaire, and carefully briefed on the nature of
the research methodology. After the questionnaire had been piloted in the
field, the main survey began in June 1997 and ran over a period of six
months.
The sample was designed to cover the full range of accommodation op-
tions available to homeless people, which are extremely diverse. Table 3.1
identifies some of the main kinds of institutions and the types of services that
they offer. From this range of alternatives, ten establishments were initially
selected from the London Hostels Directory (Resource Information Service,
1996). Within each hostel, interviewers were instructed to select at random
the specified quota from the register of current residents.
In practice, it proved impossible to implement this sampling strategy. A
number of the hostels in the original sample were unwilling to collaborate
with the research so that alternative institutions had to be identified at short
notice. Furthermore, some of the selected residents in self-contained accom-
modation refused to open their doors to the interviewers so that alternative
respondents had to be found. As a result of these difficulties, out of the 200
interviews originally scheduled, only 150 were completed. Of these, 119
provided usable responses.
In spite of the logistical problems and the lack of cooperation from hostels,
the interviews that took place seemed to work well. A substantial majority of
respondents (58 per cent) stated that they liked the questionnaire and found it
interesting. Only 5 per cent of the sample considered the proposed idea to be
unrealistic and fewer than 10 per cent found the questionnaire difficult to
understand. These results indicate that the initial designing and pre-test phases
were successful in producing an effective survey instrument that was credible
and understandable.
The characteristics of the participating 14 hostels are summarized in Table
3.2. These establishments account for 911 out of a total of some 26 200 beds
in London. Although a wide coverage of the different types of institutions
was accomplished, the final distribution cannot be regarded as fully repre-
sentative. In particular, there is an under-representation of supportive projects
and an over-representation of emergency night shelters. This problem was
taken into account when presenting the valuation results.
Table 3.1 Definition of services provided by different types of hostels
Emergency night shelters Provide direct access to short-term accommo-

dation, on a nightly basis, for people with no
alternative. Only basic accommodation and
support are available. Patrons are expected to be
out during the day.
Low-support hostels Offer accommodation to people in housing
need, but with only very limited support. Many
of these schemes are quite large and resettle-
ment advice is not necessarily a main focus of
their work. In many cases, the projects are
staffed only by resident wardens.
Semi-supportive projects Aimed at those who want to live fairly inde-
pendently, with some practical or personal
support available. The main emphasis is usually
on finding permanent housing and preparing
residents for independent living. Most do not
offer 24-hour staff cover.
Housing schemes Organizations running a number of flats, bedsits
or shared houses offering good-quality accom-
modation and little more support than help with
benefit claims. Single-room, self-catering
accommodation. Most accommodation is either
intended as permanent or otherwise as a
precursor to permanent accommodation.
Supportive projects Offer medium- to long-term accommodation
and place considerable emphasis on counsel-
ling, education and training with the aim of
preparing residents for independent living.
Most offer individual support (through a key
worker system) and 24-hour staff cover.
Traditional hostels Large, long-established hostels, which are
traditionally used by long-term homeless
people. In some cases, accommodation is still
in large domitories.
Specific needs projects Cater for specific individual situations such as
ex-offender, drug addict, single mother. The
type of accommodation and the level of support
vary considerably.
Table 3.2 List and characteristics of selected hostels
Hostel Sex Age Cost Support N Total
Emergency night shelters

Centrepoint Shelter (Berwick Street) M/F 18 £20 D/F/H/O 15 27
Lord Clyde Nightshelter M/F 19 0 B/F/H/R 10 27
St Mungo – Cedars Road M 35 £6 A/B/D/E/F/H/R 6 110
Westminster Cathedral Nightshelter M/F 42 0 F/H/R 8 50
(Passage Day Centre)
Low-support hostels
Look Ahead – Aldgate Hostel M/F 25 £20 B/E/F/H/O/R 30 163
Semi-supportive projects
Opendoor – De Laune Street F 29 £7 A/B/D/E/H/R 5 12
Providence Row – Daniel Gilbert House M/F _ £18 E/F/H/O/R 12 82
78
Housing schemes
Single Homeless Project
(Camden – Kings Cross Road) M/F _ £54 B/O/R 1 74
Single Homeless Project (Westminster) M/F _ £5–£55 B/O/R 9 31
Supportive projects
Thames Reach – Stamford St Hostel M/F/C 35 £6 B/F/H/O/R 4 22
Traditional hostels
English Churches – Queen Mary’s Hostel F 29 £10–£25 B/F/H/O/R 16 60
Salvation Army – Booth House M 57 £24 B/F/H/R 18 166
Salvation Army – Hopetown F 42 £10 B/F/H/R 10 54
Ex-offenders projects
Opendoor – Warwick Road M 30 £16 D/F/H/O/R 6 33
Notes: M=male; F=female; C=couples; Age: average age of hostel residents; Cost: approximate average cost required from residents in receipt of
benefits, in £ per week; Support: A=alcohol; B=benefits; D=drugs; F=food; E=employment; H=health; O=referral or liaison with other agencies for
additional support; R=resettlement; N: number of interviews conducted; Total: total number of bed spaces available.
3.3 THE SURVEY RESULTS
3.3.1 Socioeconomic Characteristics
Table 3.3 provides a socioeconomic profile of the hostel residents interviewed.

Both male and females were interviewed. The age range covered was wide,
Table 3.3 Summary statistics of selected socioeconomic variables
Total number of individuals 150
Demographic variables
Males (%) 65
Age (mean in years) 36
Less than 18 years old (%) 6
18–25 years old (%) 23
26–39 years old (%) 33
40–59 years old (%) 28
60 years old or above (%) 10
Single (%) 69
Divorced/separated (%) 21
UK citizen (%) 81
Education
Primary (%) 6
Secondary without O levels/GCSEs (%) 41
O levels/GCSEs (%) 31
University (%) 14
Age completed education (mean in years) 17
Employment
Looking for work (%) 44
Sick/disabled (%) 19
Student (%) 11
Economic variables
Income non-response (%) 5
Weekly income: mean (s.e.) £59.6 (£3.3)
median £48.2
range £19.5–£300
Expenditure non-response (%) 9
Weekly expenditure: mean (s.e.) £54.6 (£3.1)
median £47
range £9–£300
Note: s.e.: standard error

varying from 16 to 88 years, with an average of 36. As expected, since few

hostels cater for families, most respondents lived alone, being either single or
divorced/separated (90 per cent). Nearly 20 per cent were foreigners, a percent-
age much higher than in the overall population. This substantiates the idea that
many emigrants arrive in the UK in precarious conditions. Nearly 50 per cent
of the sample did not complete their O levels/GCSEs but a surprising 14 per
cent claimed to have university frequency. This could be explained by the fact
that 11 per cent of the sample were students. The age at which full-time
education was completed broadly ranged from 5 to 36, with an average of 17
years of age. However, this variable does not necessarily give the right picture
in terms of educational attainment, as respondents may either have taken longer
to complete lower levels of education or have had periods of interruption in
their studies – the fact that 12 per cent reported having finished their education
between the ages of 21 and 36 suggests that this is the case.
It is interesting to note that only 5 per cent of respondents refused to reveal
their income. This is substantially lower than typical income non-response
rates for general population surveys, which are often as high as 25 per cent to
50 per cent. The average weekly income was around £60, consistent with the
magnitude of the state benefits that constitute the major source of income of
hostel users. As illustrated in Figure 3.1, only 6 per cent of the sample were
receiving income from employment, with the remainder being almost solely
reliant on government benefits (income support and job seeker’s allowance
alone constitute a source of income for 70 per cent of the sample).
Respondents were also asked to break down their weekly expenditure into
a number of categories: the sum of all expenditures provided an alternative
Disability Other benefits

13% Begging
allowance 3%
11% Wages
6%
Income support Job seeker’s

40% allowance
30%
Note: multiple sources of income are possible.
Figure 3.1 Sources of income

Personal items Recreation

Transport 9% 4% Others
8% 8%
Alcohol/drugs/
tobacco
20%
Hostel
30%
Food
21%
Figure 3.2 Current expenditure breakdown
measure of income. Some respondents had difficulty in recalling specific

purchases and so the non-response rate is slightly higher than for income, at 9
per cent (Table 3.3). However, reassuringly, the average reported total ex-
penditure per week (£55) is not statistically different from the reported income
(£60). Figure 3.2 presents an overview of the expenditure allocation pattern.
The largest proportion of respondent’s income (30 per cent) is allocated to
70 Income
Expenditure
60
Percentage of sample
50
40
30
20
10
0
<£25
£26–50
£51–75
£76–100
£101–125
£126–150
£151–175
£176–200
£201–250
£251–300
£301–350
£351–400
>£400
Figure 3.3 Percentage frequency distribution of current weekly income and

expenditure
hostel accommodation and services (which, in many cases, include some

meals). Additional expenditures for food and for alcohol, drugs and/or
tobacco account for 21 per cent and 20 per cent of the available income,
respectively.
Figure 3.3 depicts the percentage distribution of current weekly income
and expenditure levels. In both instances, there is a high concentration of
cases in the £26–£50 interval. Both distributions match up quite closely.
A number of introductory questions aimed to reveal respondents’ housing
history. The responses are summarized in Table 3.4. Around 37 per cent of
respondents said that the house of their parents (or other relatives) had pro-
vided the last place they thought of as a home, while 5 per cent were formerly
Table 3.4 Housing history
Which was the last place you thought of as home?

Parents’/foster parents’ home 29
Private rented accommodation 17
Hostel 13
Council or housing association tenancy 12
Relatives’ home 8
Own house/flat/bedsit 6
Friends’ home 5
B&B 4
Sleeping rough 3
Never had a home 3
Special institution 1
What was the main reason that you left home?
Unemployment/money/accommodation problems 16
Evicted/asked to leave/disputes with other residents 14
Problems with parents 12
Relationship problems with partner 11
Wanted to leave the family home/to look for work 11
Situation in another country 9
Health reasons 4
Family death 3
Domestic violence/abuse 2
Depression/loneliness 1
Drink/drug problems 1
Left institution 1
living with friends (hidden homeless). For a further 29 per cent of the sample,
their last home had been accommodation rented either from the private sector
or from a local council or housing association, while 17 per cent considered a
hostel or a B&B to be the last place thought of as home – which is suggestive
of long-term homelessness. Only 6 per cent of respondents had a history of
home ownership. For 3 per cent of the sample the street was considered to be
their home, while a further 3 per cent declared never to have had a home.
The largest single cause of homelessness appeared to be the breakdown of
relationships with parents (12 per cent), partners (11 per cent) or neighbours
(14 per cent), leading to voluntary departure in some cases or even eviction in
others. The next most important cause was work-related, with 16 per cent
abandoning homes as a result of financial hardship ensuing from unemploy-
ment and a further 11 per cent leaving home in search of a job. As many as 9
per cent of those interviewed were refugees fleeing the situation in another
country.
Figure 3.4 presents the frequency distribution for the duration of homeless-
ness up until the time of the survey and the expected remaining period before
a more permanent home is found. The figure indicates that about half of those
interviewed had been homeless for less than a year, while two-thirds ex-
pected to have found longer-term accommodation within less than a year (in
interpreting these results it is important to note that just over half of the
respondents were unable to answer the question about the expected delay
before more permanent accommodation would be found). The median ex-
50 Time since lost old home

45 Time until new home expected
40 Expected total duration of homelessness

35
30
25
20
15
10
5
0
<6 mths
7–12 mths
13–24 mths
25–36 mths
37–48 mths
49–60 mths
61–120 mths
>120 mths
Figure 3.4 Duration of homelessness

pected overall duration of the spell of homelessness was two years, even
though there were some homeless people in the sample with average duration
of homelessness of almost five years. Seventy-five per cent of the sample
expected to be resettled within a four-year period. There was, however, a
significant group, representing just under 20 per cent of the sample, who
expected to be homeless for more than ten years; that is, for all practical
purposes, indefinitely.
3.3.2 Use of Hostel Services, Attitudes and Preferences
The CV survey included a number of questions aimed at assessing the level

of use of hostel facilities, the degree of satisfaction with these services and
the preferred short-term and long-term accommodation alternatives. This
section presents the results from these questions.
Figures 3.5 and 3.6 show the level of use of general hostel facilities (food,
laundry and washing services) and of counselling and support services for
special needs (finding a home, money/benefits, training/finding work, general
health/mental health problems, drug/alcohol problems).
As depicted in Figure 3.5, the vast majority of respondents were making
use of the basic services provided, such as meals (87 per cent), washing (99
per cent) and laundry (91 per cent). During the pre-test stages of the ques-
tionnaire, respondents repeatedly told interviewers that, although they were
in precarious accommodation or sleeping rough, they still endeavoured to
maintain their standards of cleanliness. This seemed to be an important issue
for many people.
Figure 3.6 shows that as many as 74 per cent of respondents used one or
other of the counselling facilities available in hostels, which range from
99%
100 91% 87%
90
80
70
60
50
40
30
20
10
0
Washing Laundry Food
Figure 3.5 Percentage of respondents using general hostel facilities

100
90
74%
80
70
60 51% 51%
50
40
26% 25%
30
17%
20 12%
10
0
Mental health problems

Money/benefits
Training/finding work
Health problems
Drug/alcohol problems
Total
Finding a home
Figure 3.6 Percentage of respondents using counselling and support

services
support for health problems and substance abuse to advice on seeking em-
ployment and claiming benefits. As expected, the most popular type of
assistance was finding long-term accommodation and claiming benefits, each
used by 51 per cent of respondents. A quarter of the sample had support in
finding a job and another quarter in dealing with general health problems; a
further 17 per cent and 12 per cent sought counselling regarding drug or
alcohol addiction and mental health problems, respectively.
In general, the level of satisfaction with these services was relatively high.
Figure 3.7 indicates that, in what concerns general facilities, only a fifth of
the sample or less considered any of the services provided to be poor. Satis-
faction was higher for issues such as opening hours and security and safety:
78 per cent and 76 per cent of the sample, respectively, considered them to be
good. Service quality was considered relatively less good for meals and
washing facilities, with only 34 per cent and 49 per cent considering them to
be good, respectively.
The most surprising results relate to views held on house rules and access
hours. Almost all of the hostels where the interviews took place did not allow
alcohol on their premises, and four of the selected hostels had strict opening
and access hours, requiring residents to leave their bedrooms by 7 a.m., 8
a.m. or 9 a.m., and to return by 8 p.m., 8.30 p.m. or midnight, depending on
Staff attitudes
Opening hours
Hostel rules
Privacy/space
Security/safety
Laundry
Cleanliness
Food
Washing
0 20 40 60 80 100
Good Average Poor
Figure 3.7 Satisfaction with the quality of general hostel facilities and
services
the case. Despite such restrictions, Figure 3.7 shows that staff attitudes,
opening hours and hostel rules are considered to be good by at least 60 per
cent of the sample.
As for those respondents who made use of counselling and support serv-
ices, Figure 3.8 shows that over 60 per cent rated these services as being
good, with the notable exception of help in finding a home, which was
considered good by only 47 per cent of the sample. Furthermore, help in
finding permanent accommodation and mental health counselling were thought
to be of bad quality by about 30 per cent of the sample.
Overall, these results indicate that hostel residents not only make use of the
range of services that are provided by hostels in addition to bed spaces, but also
consider these services to be generally of good quality. This is consistent with
earlier literature which has found that 56 per cent of hostel residents were
satisfied with the facilities provided (Randall and Brown, 1996). The implica-
tion is that support services seem to increase the well-being of their users and
that as such their value should be significantly positive. It also suggests that
Training/finding work
Money/benefits
Finding a home
Mental health problems
Drug/alcohol problems
Mental health
0 20 40 60 80 100
Good Average Poor
Figure 3.8 Satisfaction with the quality of hostel counselling and support
services
hostel users might prefer hostel accommodation vis-à-vis private sector equiva-
lents (such as bed and breakfast accommodation), where support services are
not available, on the assumption that they could be admitted into private
accommodation and pay for it through their housing benefit.
As regards accommodation preferences, the survey asked respondents to
think about where they would be most likely to go in the short term (if they
could not use the hostels for a couple of weeks), in the medium term (if
hostels were shut down for an entire year), and in the long term (when their
preferred housing choice might become available). The answers to these
questions are illustrated in Figure 3.9 and reveal a striking variation in prefer-
ences across different time horizons as well as a remarkably high degree of
unanimity among respondents.
When faced with a sudden short-term shortage of hostel accommodation
and no additional money, as many as 60 per cent of respondents would
choose to sleep rough rather than seek paid accommodation or stay with
family and friends. This finding indicates that private sector accommodation
is not a viable alternative for many homeless people, probably due to inad-
equate financial resources and/or difficulties in claiming benefit. This
explanation is reinforced by the fact that, for a further 32 per cent of respond-
ents, staying with family or friends or in a squat seems to be the preferred
solution to a short-term shortage of hostel accommodation.
Status quo
Home ownership
Private rental
Institution
Council
B&B
Family
Friends
Squat
Sleeping rough
0 20 40 60 80
Long term Medium term Short term
Figure 3.9 Frequency distribution of housing preferences across different

time horizons
On the other hand, when presented with a longer period of exclusion from
hostel facilities and provided with the compensation payment described in
the survey, some 55 per cent of respondents would opt to live in private
rented accommodation. The next most popular alternative in this situation
would be a B&B, which was selected by 16 per cent of the sample. This
suggests that in the long term the informal arrangements selected for the
short term would not be adequate for most residents and the financial com-
pensation paid brings the private sector alternatives within reach. Nevertheless,
one-fifth of the sample could not indicate an alternative form of accommoda-
tion. For many, a medium-term alternative to hostel accommodation may
simply not exist in the private sector or in council estates.
However, in the longer term the vast majority of the hostel residents (75
per cent) would prefer to obtain their permanent accommodation from the
local council or housing association. The fact that this was not selected in the
medium-term scenario suggests either that such accommodation is not avail-
able within that particular time horizon or possibly that respondents would
only be able to afford the private sector options with the compensation
payment defined in the hypothetical scenario.
Overall, these results show that respondents were realistic in assessing
their future resettlement prospects over different time periods.
3.3.3 Valuation Results
The first interesting finding that arose from the valuation section of the hostel
users’ survey was the fact that the majority of respondents did not consider
the idea of hostel closure followed by compensation, as presented in the CV
hypothetical scenario, to be an attractive one. As illustrated in Figure 3.10,
nearly 25 per cent of the sample disliked the idea, while a further 40 per cent
actually claimed to have hated it. On the other hand, the figure also shows
that about a quarter of respondents did like the idea.
Would hate it
Would not like it
Don’t mind
Would like it
Would very much like it
0 10 20 30 40 50
Figure 3.10 Answers to the question ‘Would you like this idea to actually
happen?’
Table 3.5 presents an analysis of the correlation between disliking the

proposed scenario and considering the services provided by hostels to be of
good quality. This relationship is expected to be positive, as the more an
individual appreciates the services provided by hostels, the more likely he/
she is to dislike giving them up. The results indicate that, as expected, all the
correlation coefficients are positive although generally not very high in abso-
lute terms. The highest correlations are obtained for those who appreciate the
quality of health-related support (0.43), the cleanliness of the premises (0.34),
the food (0.29) and hostel rules (0.29).
As described above, the WTA valuation question was open-ended, asking
hostel residents to state their required compensation to forego hostel accom-
Table 3.5 Correlation between attitudes towards the proposed scenario

and attitudes towards hostel services
Appreciating the quality of hostel services Disliking the proposed scenario
General facilities
Food 0.29
Washing facilities 0.18
Laundry 0.15
Cleanliness 0.34
Privacy/space 0.20
Security/safety 0.09
Staff attitudes 0.14
Opening hours 0.08
Hostel rules 0.29
Counselling services
General health 0.43
Mental health 0.23
Drugs and alcohol 0.07
Finding a home 0.22
Money/benefits 0.24
Training/finding work 0.20
modation for one year and still be as well off as before. Nearly 80 per cent of
respondents were able to provide answers to the hypothetical valuation ques-
tion.
Table 3.6 presents the compensatory amounts required by hostel users to
give up hostel accommodation and still maintain their initial welfare, while
Figure 3.11 depicts the distribution of these WTA amounts.
The overall (unweighted) mean WTA was found to be £157 per week (over
and above current income), ranging between £7 and £1600, with a median of
£125. These values are reasonable overall and indicate that, with few excep-
tions, respondents took the exercise seriously and did not react strategically
to the valuation question by asking for extremely high compensations. Figure
3.11 also shows that the compensation amounts are concentrated on the left-
hand side of the distribution with very few outliers (unrealistically high bids),
as the small difference between the mean and median WTA already indi-
cated.
Table 3.6 also shows that the required compensation is higher for hostels
providing more support and counselling services. For example, residents of
low-support projects asked for an average compensation of £127 per week
Table 3.6 WTA compensation to give up hostel services (£ per person per week)
Confidence
Hostel type Mean interval Median Minimum Maximum N
Emergency night shelters £140.8 £102.3–£179.3 £107.5 £20 £500 34

Low-support projects £127.4 £106.7–£148.1 £120.6 £20 £230 26
91
Semi-supportive projects £162.1 £92.1–£232.1 £140.0 £50 £500 12
Housing schemes £147.9 £101–£194.8 £140.0 £50 £250 7
Supportive projects £231.0 £37–£425 £231.0 £132 £330 2
Traditional hostels £210.6 £106.2–£315 £150.0 £7 £1600 29
Ex-offenders projects £117.0 £86.9–£147.1 £111.0 £70 £170 6
Total WTA £157.4 £128–£186.8 £125.0 £7 £1600 119
25
20
15
10
0
<£25
£26–50
£51–75
£76–100
£101–125
£126–150
£151–175
£176–200
£201–250
£251–300
£301–350
£351–400
>£400
Weekly WTA compensation per person (in £)
Figure 3.11 Distribution of WTA compensation amounts (in £ per person

per week)
while users of traditional hostels and supportive projects required compen-

sations above £200 per week. Aggregating hostels into two broad types,
according to the level of support provided, gives the following average
result: low support (night shelters, low support, housing schemes, ex-
offenders): £133 per person per week; and medium and high support
(semi-supportive, traditional and supportive projects): £201 per person per
week. The difference of £68 could be viewed as a rough lower-limit esti-
mate of the value of the services provided by hostels over and above the
provision of a bed space. This is consistent with the findings from the
attitudinal part of the questionnaire reported above that suggested that
respondents positively valued the range of services provided by hostels.
However, due to the small number of interviews conducted in some of the
hostels, the standard error of these estimates is quite high and the differ-
ences are not statistically significant.
A regression analysis was performed on the WTA results to assess their
theoretical validity, that is, to see whether a number of relevant explanatory
variables manage to explain a significant proportion of the total variation in
compensatory estimates in ways that are compatible with economic theory.
Table 3.7 presents an overview of the variables that were selected as possible
explanatory factors. These include sociological, economic and attitudinal
variables and a number of dummy variables representing the type of hostel
where the interviews took place.
Table 3.7 Description of explanatory variables
Sex 1 – male; 0 – female
Age Average age in years
Primary education 1 – if primary is highest level of
education attained; 0 – if otherwise
Age education was completed Age at which full-time education was
completed
Economic variables
Expenditure per person Weekly average (in £)
Disliked the proposed idea 1 – if disliked the idea; 0 – if otherwise
Disliked the questionnaire 1 – if disliked the questionnaire; 0 – if
otherwise
Likes hostel rules 1 – if likes hostel rules; 0 – if otherwise
Likes hostel counselling and support 1 – if likes support services; 0 – if
services otherwise
Type of hostel (baseline: emergency
nightshelters)
Low support hostels 1 – if low support hostel; 0 – if otherwise
Semi-supportive projects 1 – if semi-supportive project; 0 – if
otherwise
Housing schemes 1 – if housing scheme; 0 – if otherwise
Supportive projects 1 – if supportive project; 0 – if otherwise
Traditional hostels 1 – if traditional hostel; 0 – if otherwise
Ex-offenders projects 1 – if ex-offenders project; 0 – if
otherwise
Other variables
Looking for a job 1 – if looking for a job; 0 – if otherwise
Would seek private rented 1 – if seeking rented accommodation;
accommodation after hostel closure 0 – if otherwise
The results from estimating the valuation function obtained by regressing

the WTA amounts against the variables in Table 3.7 are reported in Table 3.8.
The statistical method used was ordinary least squares (OLS) with correction
for heteroskedasticity (more details about this approach can be found in the
Statistical Appendix to Chapter 2). The best fit of the model was obtained by
using the logarithm of WTA as the dependent variable. Overall, in spite of the
small sample size, the regression manages to explain 36 per cent of the total
variation, which is a good result for the type of cross-sectional data being
analysed.
As expected, weekly expenditure was found to be a strong positive deter-
minant of the desired compensation amounts: the higher the expenditure
level, the higher the compensation required. Expenditure is a good proxy for
Table 3.8 Valuation function: WTA compensation for hostel closure
Coefficient t-statistic
Constant 3.26 7.74

Sex 0.05 0.35
Age 0.003 0.45
Primary education –0.54 –2.13
Age education was completed –0.03 –1.82
Economic variables
Expenditure 0.01 5.96
Disliked the proposed idea 0.12 2.40
Disliked the questionnaire 0.40 2.33
Likes hostel rules 0.20 1.68
Likes hostel counselling and support services 0.03 0.92
Type of hostel (baseline: emergency night shelters)
Low-support hostels 0.32 1.43
Semi-supportive projects 0.18 0.88
Housing schemes 0.10 0.35
Supportive projects 0.69 2.30
Traditional hostels 0.47 1.98
Ex-offenders projects 0.31 1.74
Other variables
Looking for a job 0.53 3.19
Would seek private rented accommodation after 0.38 2.62
hostel closure
R2 0.36
N 93
Note: Estimation method: OLS with correction for heteroscedasticity. Dependent variable:
log(WTA).
income in this model and actually performs better, in statistical terms, than
income itself (whose coefficient was also positive and significant in an alter-
native specification of the valuation function not reported here). Thus higher
levels of income/expenditure are indicative of a higher quality of life and,
arguably, of higher opportunity costs from losing hostel accommodation.
This argument is intuitive but is also difficult to verify. Weak supporting
evidence can be found by looking at the low but positive correlation between
expenditure and the modern housing projects included in the sample (0.12)
and the small but negative correlation found with the more traditional hostel
types (–0.01).
Regarding sociological factors, those respondents who only have primary
education are willing to accept less compensation than others with further
educational degrees. Again this may be indicative of a lower quality of life
that is reflected in lower WTA amounts. Sex and age do not have a statisti-
cally significant effect on WTA, and the age at which education is completed
is only significant at the 10 per cent level.
A number of dummy variables were included to represent the various
different types of hostels where the interviews took place. Emergency night
shelters were taken as the baseline: this means, for instance, that the coeffi-
cient of the supportive projects variable (0.6899) indicates that hostel residents
in supportive projects required compensation amounts nearly £0.7 higher
than the logarithm of the WTA amount of those using emergency night
shelters. The prior expectation is that, the larger the level of support and
counselling services offered, the higher the opportunity cost of hostel clo-
sure. The results reported in Table 3.8 confirm this expectation: all included
hostel categories have positive coefficients, reflecting the fact that emergency
hostels typically provide the lowest level of support services and thus are less
valued. Supportive and traditional hostels have the highest coefficients, both
significant at the 5 per cent level.
Finally, regression results also show that respondents actively looking for a
job seek higher compensations, as do those who would opt for renting private
accommodation after the prospective hostel closure. Once more, this suggests
that these individuals have higher opportunity costs and, possibly, a higher
standard of living and hence require higher compensations for hostel closure,
in order to be as well off as initially.
By and large, these results conform to initial expectations regarding the
influence of the included explanatory variables on compensation amounts
and strike a positive note on the theoretical consistency of the survey results.
Debriefing questions
A number of debriefing questions were included in the questionnaire aimed
at exploring the consistency of the reported compensation amounts and the
reasons behind not answering the valuation question.
Respondents who stated a positive WTA amount were asked to allocate the
new hypothetical income they would have under the proposed scenario (cur-
rent income plus compensation) among the same categories used in the
question about current expenditure composition, described in Figure 3.2. The
prospective expenditure breakdown question is useful to check whether re-
spondents would roughly maintain their pattern of consumption, as would be
expected since the new income should only afford them a standard of living
similar to their current one, or if instead they anticipate large changes in the
composition of their expenditures which could be indicative of upward-
biased answers in the valuation question.
Recreation Others
7% 19%
Personal
items Alcohol/drugs/
9% tobacco
18%
Transport
11%
Food
Accommodation 17%
19%
Figure 3.12 Predicted expenditure breakdown after the proposed scenario
The results are depicted in Figure 3.12. Although with a lower importance
in absolute terms than before, accommodation costs still constitute the largest
proportion (19 per cent) of total expenditure, followed by alcohol/drugs/
tobacco (18 per cent) and food (17 per cent). This pattern is similar to the one
depicted in Figure 3.2 but for the fact that the difference between accommo-
dation costs and food and alcohol/drugs/tobacco is now much narrower: 1–2
percentage points versus 9–10 percentage points. Contributing to this smaller
difference is the fact that, in most cases, hostels provide some meals while
private accommodation does not, and hence the relative proportion of the
budget spent on food would be expected to increase. The relative importance
of transport, recreation and personal items under the hypothetical scenario is
similar to the status quo. The major difference between the post-hostel clo-
sure and the current situation seems to lie in the residual category denominated
other expenditures. Initially this corresponded to some 8 per cent of the
budget while now 19 per cent fall in this category. This increase could be
indicative of the fact that hostel users would need to make up for the lack of
hostel general support services by incurring a number of other costs, for
example laundry, counselling, rehabilitation programmes and so on. But since
no further information was recorded, this argument is merely speculative.
To make sure that respondents were stating compensation amounts that
would enable them to achieve their current quality of life after the proposed
change, the questionnaire included a debriefing question about whether they
felt their quality of life would improve, worsen or stay the same after the
compensation had taken place. Respondents who expected their quality of
life to change were given an opportunity to revise their stated compensations
in order to arrive at an amount that would leave them as well off as they
currently were. Only 11 per cent of the sample indicated that their welfare
would change after the compensation; and when asked to revise their WTA
amounts accordingly, only a third were able to do so. As such, there may be a
small bias on the estimated compensation suggests amounts by the small
minority of respondents who failed to revise their answers. But the small
proportion of the sample that changed its initially stated compensation that
the careful questionnaire wording may have helped in inducing truthful re-
sponses for most of the sample.
Finally, those who failed to provide an answer to the WTA question were
asked for the reasons behind their lack of response. In particular they were
asked to agree or disagree with a number of related statements. Table 3.9
displays the results.
Table 3.9 Reasons for not answering the valuation question
Agree Disagree
(%) (%) Number
I don’t like the idea of not being able to use 53.9 46.2 26
hostels any more
Even with the money, I wouldn’t be accepted 0.0 100.0 24
anywhere else
I couldn’t find the services hostels offer 64.3 35.7 28
anywhere else
I find it very difficult to manage my own money 12.5 87.5 24
I would end up spending the extra money on 0.0 100.0 24
other things
I don’t think I would be given the amount of 4.2 95.8 24
money I want
I did not find the idea very easy to understand 48.2 51.9 27
or imagine
A total of 31 individuals did not answer the valuation question (21 per cent
of the total sample). Sixty-four per cent said they could not find the type of
services provided by hostels elsewhere. It is interesting to note that these

individuals chose not to answer the valuation question rather than to bid for
very high compensatory amounts that would afford them the sort of special
support they might need. A general dislike of the proposed idea (54 per cent)
was the second most important reason that led to non-response, while 48 per
cent found the hostel closure idea difficult to understand. Lastly, 13 per cent
of hostel residents noted that they had difficulties managing their money, and
would probably end up sleeping in the streets as a result. For this subgroup,
the alternative scenario would probably never enable them to have a similar
quality of life as the current one.
None of the 31 non-response cases indicated that they didn’t answer be-
cause they didn’t value hostel services. Their motives reflected some type of
protest against the proposed idea, an inability to understand the scenario or a
feeling that hostels were just irreplaceable, which was translated into a non-
response instead of a high compensation.
By and large, these findings indicate that respondents took the valuation
exercise seriously, providing answers that were coherent and accorded to
previous expectations.
3.4 DISCUSSION AND CONCLUSIONS

Much has been written about the difficulties of applying economic valuation
methods based on hypothetical markets in the context of developing coun-
tries, where the population of interest may not know how to read or write,
where money may not be a common means of exchange and where markets
may not be well developed. However, little research, if any, has been done on
applying such techniques to minority groups in developed countries, living
outside the prevailing culture and where many of the difficulties in conduct-
ing valuation studies encountered in apparently different contexts may strike
a parallel.
This chapter reported on the results of a novel application of economic
valuation techniques to homeless people, with a view to finding out how
much the services provided by charitable hostels was worth to them. In order
to do that, a carefully designed CV questionnaire was administered to 150
hostel residents. The scenario proposed described a hypothetical hostel clo-
sure during one year accompanied by monetary compensation to users:
respondents were asked for the minimum compensation required, in such a
scenario, that would afford them the same quality of life as they enjoyed at
present.
A number of conclusions can be drawn from the survey. CV was success-
fully administered to a minority group (that of single homeless people) living
an unconventional lifestyle, without permanent accommodation, most with-

out a job and with special support needs. Nevertheless, the majority of
respondents not only answered the valuation question, but also described the
whole questionnaire in emphatic terms.
The valuation question was framed as a WTA compensation to forego the
use of hostels. This format, as opposed to the more common WTP format,
can give rise to strategic behaviour in the form of overbidding. However, in
the present survey, careful design and pre-test of the questionnaire, and
specifically of the wording of the scenario and the structure of the valuation
section, ensured that the estimated WTA distribution was not skewed towards
unrealistically high answers.
The overall unweighted average WTA was found to be £157 per person per
week. As expected, the average WTA varied across different types of hostels:
hostels providing higher levels of support yielded higher-value estimates than
those offering only basic services. This indicates that the range of additional
support and counselling services for special needs offered by these charitable
institutions has a positive value to its users. Moreover, considering the price
of rentals in the London area, respondents could probably find accommoda-
tion for less than £157 per person per week. The positive difference would be
an estimate of the value of hostel support services. The majority of respond-
ents who refused to answer the valuation question did so because they thought
they could not find the type of services offered by hostels in any other place.
The significantly positive monetary values uncovered in the valuation sec-
tion were supported by the results of the attitudinal questions included in the
questionnaire. Nearly all respondents made use of hostel support, counselling
and general services, over and above the simple bed space provided. And,
most importantly, the large majority classified these additional services as
being good, even being satisfied with unlikely items such as hostel rules and
opening hours. As a result, when confronted with the closure scenario, ac-
companied by monetary compensation, more than 60 per cent of respondents
openly expressed their dislike of such an idea.
The estimated compensation amounts also performed well in terms of
standard tests of coherence and validity. Main determinants of the requested
compensation amounts were the type of hostel, the attitudes towards the
proposed scenario and the questionnaire in general, and several indicators of
the quality of life of hostel residents such as the current level of expenditure.
After receiving the hypothetical compensation, respondents did not predict
major changes in their pattern of consumption, leading us to believe that their
stated WTA amounts would actually be just enough to afford them their
present lifestyle.
The rationale for the homeless survey was to provide an input into the
estimation of the total value of the housing and homelessness charitable
sector, by uncovering the value of the services provided by hostels to its

direct beneficiaries. Therefore, in Chapter 5, the results presented in this
chapter will be further analysed, together with the findings from the survey of
the general public reported in Chapter 2, and value of volunteering reported
in Chapter 4.
4. The benefits of volunteering
4.1 INTRODUCTION
A question of enduring interest in voluntary sector research is the appropriate
value to be accorded to volunteer time. The methodological difficulties asso-
ciated with such a valuation exercise have been well documented, and it is
perhaps for this reason that there have been comparatively few attempts
actually to perform it.
An important starting-point is the exercise conducted by the Volunteer
Centre UK (1995), which took as a proxy for the value of volunteer time
the national average for the gross hourly wage rate (£7.83 in 1993). Apply-
ing this figure to estimates of the total annual number of hours volunteered
in the UK yielded a total value for volunteering of £24.5 billion in 1993.
However, Knapp (1990) is critical of any approach based on ‘a blanket
foregone wage for all volunteers’, arguing that ‘some are retired or unem-
ployed, and others could be giving up a variety of forms of paid employment
in order to participate’. Nevertheless, the authors of the 1995 study by the
Volunteer Centre UK readily acknowledge that a simple approach of this
nature is only ‘a taste of what could be achieved in a more detailed analy-
sis’.
The purpose of this chapter is to refine and extend the valuation approach
applied by the Volunteer Centre UK (1995), by responding – at least in part –
to some of the criticisms set out by Knapp (1990).
One of the reasons why valuing volunteer time is such a complex issue is
that there are three very different perspectives that can be taken on the
problem, each with a particular validity of its own.
First, one could take the perspective of the volunteer and ask to what extent
they value the time that is being given up to volunteering. In other words,
what is the opportunity cost of volunteer time to the volunteer? This perspec-
tive is a particularly important one in terms of the overall enterprise of this
book, which is to measure the net social value of the charitable sector. Since
volunteers could be expected to give time up until the point where the
marginal benefit they experience from volunteering is equal to the marginal
cost, knowing the opportunity cost of volunteer time provides a way of
estimating the benefits that volunteers receive from this activity.
101
Second, one could take the perspective of the charity manager and ask to
what extent volunteering enables charities to economize on employment-
related costs. This perspective is also important as regards the overall objective
of measuring the net social value of charities, since it permits an estimate of
the resources that charities absorb in order to provide their services. The cost
of replacing volunteers with paid employees is therefore an important com-
ponent of the full social cost of the charitable sector, and one that has tended
to be overlooked in traditional measures of charitable expenditure.
Third, one could take the perspective of the charity’s beneficiaries and ask
to what extent volunteering has a positive impact on their well-being. This
perspective looks at the productivity of volunteers in terms of the output they
produce.
Each of these perspectives suggests a different methodological approach to
the valuation of volunteer time, which will be developed in greater detail
below.
To the extent that is possible, the approaches developed will be illustrated
using data on volunteering from the Individual Giving Survey (IGS) (Half-
penny et al., 1992, 1993, 1994). The IGS was an annual repeated cross-section
survey of philanthropic behaviour in the UK funded by the Charities Aid
Foundation, which ran from 1987 to 1993. Each year the survey covered
about 1000 individuals who were selected so as to be representative of the
UK population. A structured interview conducted in the respondent’s home
was used to collect information about philanthropic activities during the
month before the survey. In particular, respondents were asked to report the
number of hours per month that they devoted to a wide range of volunteering
activities.
4.2 AN OPPORTUNITY COST APPROACH

One way of thinking about volunteering is as a ‘donation in kind’. Whereas
some people make financial contributions to charity, others may prefer to
contribute their time. When people choose to volunteer, they are implicitly
foregoing the opportunity to do something else, whether that be paid employ-
ment, working in the home, or taking leisure. Thus the cost of giving time to
charity from the perspective of the volunteer is effectively the same as the
value of time that could have been spent in the alternative activity.
A key question in valuing volunteer time in accordance with this approach
is whether people work under flexible employment contracts which allow
them freely to substitute time and money at the margin (for example, the self-
employed or those able to work paid overtime at will). Where this is the case,
people are at liberty to allocate their time between alternative uses with the
Benefits of volunteering 103
result that the value of a marginal hour assigned to any activity should be the
same across uses, and in all cases equal to the net hourly wage from employ-
ment.
In these circumstances, the hourly value of time allocated to volunteering
(Vv) will be equivalent to the take-home (or net) hourly wage rate of the
volunteer (WN) (Vv = WN). The reason is that volunteers will be able to adjust
the number of hours they devote to charitable work up to the point where the
benefits of a marginal hour of volunteering will be equal to the benefits of a
marginal hour in employment. In this calculus, the benefits of volunteering
will include not only the altruistic satisfaction of contributing to a good
cause, but also any private benefits that the volunteer gains from the experi-
ence; such as social contacts or acquisition of skills.
As an illustration of this approach, a net hourly wage can be estimated for
volunteers recorded in the IGS (Halfpenny et al., 1992, 1993, 1994). Al-
though the surveys do not collect data on wage rates, they do contain
information on many of the individual characteristics (such as sex, age and
education) which are important in determining wages. The presence of this
information makes it possible to impute a net hourly wage to respondents in
the IGS survey. In order to do so, it is necessary to have access to a comple-
mentary dataset containing information on the net hourly wage and on the
same individual characteristics recorded in the original datasest. In this case,
the Family Expenditure Survey (FES) was identified as a suitable comple-
ment to the IGS (Arellano and Meghir, 1992).
Using the FES data, a regression model was estimated capable of predict-
ing the net hourly wage on the basis of observable characteristics common to
both datasets. One of the difficulties with estimating such an equation is that
net wages are only observed for those respondents currently in employment.
To get round this problem, a Heckman model is used. This model takes into
account the fact that individuals select whether or not to join the labour
market on the basis of their wage prospects. Thus the wages observed among
those who work are likely to be greater than the wages that would have been
earned by those who do not work had they chosen to do so. Accordingly,
there are two components to the model: a wage equation which captures the
relationship between net wages and respondent characteristics for those who
work, and a participation equation which explains how observable character-
istics influence the decision whether or not to enter the labour market
(Heckman, 1979).
Table 4.1 reports the coefficients and t-statistics for the two separate mod-
els estimated using male and female respondents in the FES. The participation
equation indicates that being married has a major positive impact on labour
force participation for men while, for women, the presence of pre-school-
aged children in the household substantially reduces the likelihood of observing
Table 4.1 Summary of Heckman selectivity models used to fit net hourly
wage
Males Females
Coefficient t-statistic Coefficient t-statistic
Wage equation
Age –0.023 –2.382 –0.068 –6.139
Age squared –0.002 –4.319 0.003 7.266
Age × education 0.003 6.205 0.003 5.782
Education –0.951 –3.701 –0.726 –2.664
Education squared 0.028 3.716 0.021 2.645
Greater London 0.199 8.075 0.272 10.517
Southeast 0.118 6.207 0.070 3.391
North of England 0.039 2.172 0.034 1.790
Scotland or Wales 0.055 2.513 0.051 2.205
Constant 8.858 4.024 7.716 3.296
Participation equation
Age 0.190 15.086 0.208 15.898
Age squared –0.002 –41.442 –0.020 –39.551
Age × education –0.003 –3.989 –0.004 –6.018
Education 2.325 5.849 1.034 2.841
Education squared –0.062 –5.265 –0.022 –2.028
Greater London –0.078 –2.049 –0.057 –1.607
Southeast 0.090 2.937 0.060 2.110
North of England –0.121 –4.439 –0.035 –1.378
Scotland or Wales –0.145 –4.381 –0.099 –3.186
Single-parent status 0.020 0.496 –0.469 –14.453
Living with partner 0.399 16.892 –0.038 –1.665
Pre-school children –0.075 –2.932 –0.425 –20.061
Constant –23.206 –6.910 –12.902 –4.146
Correlation –0.854 –0.865

Variance 0.7737 0.7838
Log-likelihood –19 809.78 –19 120.78
Observations 18 737 20 634
labour force participation. For both genders, net wage levels show a strong
positive correlation with age and education. There is also some evidence of
regional wage differentials; particularly for the Greater London area. Almost
all variables show a very high degree of statistical significance, in excess of

the 99 per cent confidence level.
In order to test the performance of the model, predictions were calculated
for the FES data and then compared with the original values of the net hourly
wage. It was found that, on average, the fitted wage variable tends to
underpredict the true variable by £0.87 per hour for males and £0.77 per hour
for females. As might be expected, the range of variation exhibited by the
fitted wage is considerably narrower than that found in the original wage
variable, illustrating the difficulty of modelling outliers in the data. Overall,
this performance is thought to be adequate for the present purposes, subject
to the caveat of underprediction.
By applying this model to the IGS data it is possible to impute the value of
the net wage variable to the initial dataset. The resulting hourly values of
volunteer time for males and females during the early 1990s are reported in
Table 4.2. The approach produces estimates of over £4 per hour for females
and under £6 per hour for males. As noted above, these imputed values are
probably somewhat below the true net hourly wages for the IGS sample.
Table 4.2 Estimated value of volunteer time using the opportunity cost
approach* (£1993/hour)
1991 1992 1993
Males 5.80 5.89 5.51

Females 4.13 4.40 4.32
Note: * Calculated for that subsample of the IGS for which it was feasible to impute the
opportunity cost.
A possible application of this particular measure of the value of volunteers’

time would be in the calculation of matching grants. The level of such grants
is usually set with reference to the total volume of monetary donations from
the general public, with the result that charities receiving relatively high
levels of volunteer support and relatively low levels of financial support lose
out in the calculation of such grants. This situation could be rectified by using
an opportunity-cost-based valuation of volunteer time. The reason why the
approach is particularly appropriate to this context is that it captures the cost
to the volunteer of contributing their time, and thus the size of the monetary
donation that would be equivalent to the number of hours volunteered.
Clearly, the main limitation with the approach described is that the model
of fully flexible allocation of time between paid employment and other activi-
ties is unlikely to hold in many cases, for example, if people are employed on
fixed-time contracts with unpaid overtime, or if they are retired, unemployed

or otherwise engaged. In these cases, the principle of valuing volunteer time
at the opportunity cost of the next best alternative still holds. However, the
measurement of this opportunity cost becomes much less straightforward
than simply computing the net hourly wage. Instead, a more complex survey-
based approach, requiring respondents to trade off money and alternative
uses of time, would probably be required to ascertain the appropriate value.
4.3 A REPLACEMENT COST APPROACH

By relying on volunteers to assist in the organization and running of their
programmes, charities may be able significantly to reduce the payroll costs
associated with their operations. Thus an obvious way of valuing the contri-
bution of volunteers from the perspective of the charity is to estimate the cost
of replacing volunteers with paid staff.
The relevant measure for the hourly value of volunteer time to the charity
(VC) is no longer the net wage of the volunteer, but rather the full labour costs
that would be associated with a substitute employee. These are equivalent to
the gross hourly wage (WG), multiplied by a mark-up that reflects the margin
of non-wage labour costs (λ). From this must be deducted the hourly finan-
cial costs which charities may incur in the course of managing and deploying
volunteers but which would not arise in the case of paid employees (κ).
VC = WG (1 + λ ) − κ
It is worth noting that the volunteer’s own wage in employment is not

necessarily a good guide to the wage required to secure a substitute em-
ployee, in the sense that the activities performed by a volunteer may be very
different from what they would do in paid employment. For example, if a
merchant banker volunteers at a soup kitchen over the weekend, the wage
earned by the merchant banker in their usual employment will be very differ-
ent to the wage that the charity would have to pay to get an employee to do
kitchen work.
Therefore, in order to apply this approach, it is necessary to have informa-
tion on the exact type of activity undertaken by the volunteer. Fortunately, the
IGS is helpful in breaking down volunteer hours into a number of activity-
based categories. By matching up these activities against occupational
categories cited in the national New Earnings Survey (NES) (Department of
Employment, 1991, 1992, 1993) it becomes possible to estimate the hourly
wage corresponding to each activity. This process is conducted in Table 4.3,
which shows considerable variation in the resulting hourly wage rate both
Table 4.3 IGS volunteer categories against equivalent NES occupational categories with corresponding average male and
female hourly wage rates for 1993 (£1993/hour)
IGS volunteer category Equivalent NES occupational category Males Females
Raising money Managers and administrators (marketing and sales) 14.40 10.19
Serving on a committee Managers and administrators (officials of charities) 13.09 10.01
Organizing events Managers and administrators 8.76 6.26
107
Visiting people Professional occupations (social workers) 9.23 8.06
Providing information Associate professional occupations (information officers) 8.91 7.92
Doing secretarial work Clerical and secretarial occupations 6.55 5.86
Providing transport Plant and machine operatives (road transport) 5.26 4.57
Providing other services Personal service occupations 6.87 4.93
Representing a charity Managers and administrators (officials of charities) 13.09 10.01
Making goods for sale Craft and related occupations 5.63 4.20
Helping in a church Professional occupations (clergy) 9.23 8.06
between the sexes and across the different types of activities undertaken by
volunteers.
Not only is it necessary to identify the appropriate hourly wage, but this
must be adjusted to take into account the additional payroll costs – such as
National Insurance contributions – that would be incurred by the charity in
employing such an individual. The national Labour Cost Survey (Janes and
Roberts, 1990) is designed to track the proportion of total labour costs ac-
counted for by wage and non-wage elements. For the years 1991–93, the
survey records that wage costs accounted for just over 80 per cent of full
labour costs. The resulting hourly values of volunteer time for males and
females during the early 1990s are reported in Table 4.4. The approach
produces estimates of over £8 per hour for females and around £12 per hour
for males.
Table 4.4 Estimated value of volunteer time using the replacement cost
approach (£1993/hour)
1991 1992 1993
Males 11.13 11.64 12.23

Females 7.35 8.64 9.10
A potential application for this particular measure of volunteer time would

be to correct charity accounts for differences in the use of volunteers, thereby
permitting more meaningful efficiency comparisons between charitable insti-
tutions. For example, take the case of two charities of a similar size providing
the same type of service. Imagine that one of the charities is supported
primarily through financial contributions, while the other is supported prima-
rily through volunteering. Then the expenditure accounts of the two charities
would look very different. Indeed, a superficial comparison between the two
might lead to the conclusion that the volunteer-intensive charity was the more
efficient, in the sense of incurring lower financial costs. One way of correct-
ing for this distortion would be to adjust the expenditure figures of the
volunteer-intensive charity by adding a shadow expenditure figure that re-
flected the total value of contributions of volunteer time according to the
replacement cost method. The replacement cost approach is particularly ap-
propriate in this context because it reflects the amount that charities would
actually have to spend if volunteers were replaced with paid employees.
Like the previous approach, this methodology also has a number of limita-
tions. In particular, the procedure described relies upon two assumptions that
are not necessarily supported in reality.
The first questionable assumption is that, if volunteers were no longer

available, charities would choose to replace them with paid employees who
would do exactly the same type and amount of work. For example, it may be
that some of the types of work done by volunteers would not be worth doing
if charities had to pay for them at full market rates. Alternatively, it may be
that the provision of opportunities for volunteers is one of the objectives of
the charity and thus that it would not make sense to replace volunteers with
employees.
The second questionable assumption is that the quality of work done by
volunteers is no different to the quality of work done by paid employees. It
seems likely that the relative quality of work performed by these two groups
could differ in either direction. On the one hand, in some areas of activity,
paid employees may perform their work with a higher degree of professional-
ism and therefore attain a higher level of productivity than volunteers. On the
other hand, in other areas of activity (for example, visiting prisoners or
befriending the desperate or the destitute), the very absence of a formal
employment contract may make the interventions of a volunteer more valu-
able in the eyes of the beneficiary than those of a paid employee.
4.4 AN OUTPUT-BASED APPROACH

The third perspective on the value of volunteer time is perhaps the most
important, and at the same time the most problematic to quantify. This
approach begins with the assumption that the main purpose of volunteering is
to provide a service to the charity’s intended beneficiaries. From this perspec-
tive, the hourly value of volunteering is essentially the value of the
corresponding output to the charity beneficiary (VB). This in turn can be
broken down into two components: the number of physical units of output
(O) produced by the volunteer during the course of an hour (H), and the price
(or monetary value to the beneficiary) of each physical unit of output pro-
duced (PB).
VB =   × PB
O
 H
Quantifying the productivity of volunteer effort involves identifying the

actual physical output produced per hour, whether that be cooking a certain
number of meals for the elderly, or restoring a certain number of metres of
footpath. Less straightforward is valuing each unit of this output, in the
sense of estimating how much the service is worth to the beneficiary group.
The studies presented in Chapters 2 and 3 of this volume illustrate a novel
attempt to do precisely this. However, there is little if any wider literature

on this subject. Extending this third approach to the valuation of volunteer
time therefore depends on further progress in the valuation of charitable
output.
4.5 COMPARING APPROACHES

Since it is not possible to produce valuations based on the output approach,
the comparisons will be limited to the two initial methodologies presented:
the opportunity and replacement cost approaches. As a point of reference, the
methodology adopted by the Volunteer Centre UK will also be considered.
Comparing the results reported in Tables 4.2 and 4.4 indicates that the
hourly values resulting from the replacement cost approach (at between £8
and £12 per hour) are approximately double the size of those resulting from
the opportunity cost approach (at around £4 to £6 per hour). These can be
compared to the corresponding values that arise from taking the national
average for the gross hourly wage as suggested by the Volunteer Centre UK.
This latter approach gives values of around £6 to £8 per hour, which lie
exactly in between the values arising from the other two methodologies.
It is pertinent to ask whether the differences that are in evidence between
the three methodologies are merely attributable to the fact that one estimates
a net wage, one a gross wage and the other a full labour cost. To investigate
this issue, all three measures are converted into an hourly gross wage equiva-
lent to see whether there are any underlying differences in the value of time
associated with each of the approaches.
The gross hourly wage equivalents for all three methodologies are pre-
sented in Figures 4.1 and 4.2, with the corresponding values for tax and
non-wage costs reported alongside. The figures show that substantial differ-
ences in the hourly measure remain between the three methods, even after
adjusting for tax and non-wage labour costs. Although the spread of the three
estimates narrows considerably as a result of the adjustment, the initial rank-
ing is preserved, with the opportunity cost of time continuing to give the
lowest value, the replacement cost the highest value, and the Volunteer Cen-
tre methodology giving a value somewhere in between.
The differences between the estimates are, of themselves, quite informa-
tive. The first point to note is that the replacement cost method still gives a
gross hourly wage equivalent that is substantially higher than that obtained
from the Volunteer Centre approach. This may well be a reflection of the fact
that the charitable sector requires relatively highly skilled labour inputs,
which would command a wage above the national average level if recruited
from the marketplace.
Replacement
cost
Non-wage costs
Volunteer Tax costs
Centre Wage costs
Opportunity
cost
0 2 4 6 8 10
Figure 4.1 Breakdown of differences between valuation methodologies for

males in 1993 (£1993/hour)
Replacement
cost
Non-wage costs
Volunteer Tax costs
Centre Wage costs
Opportunity
cost
0 2 4 6 8 10
Figure 4.2 Breakdown of differences between valuation methodologies for

females in 1993 (£1993/hour)
The second point to note is that the opportunity cost method gives a gross
hourly wage equivalent that is substantially lower than that for the replace-
ment cost approach. There are two possible explanations for this. One is that,
on average, people who volunteer have personal characteristics that may
prevent them from commanding an employment wage comparable to the

national average. The other explanation could be that the difference is simply
due to the fact that the imputation process described above underestimates
the true value of the net hourly wage. It is difficult to distinguish between
these two hypotheses, although the differences between equivalent gross
hourly wages observed in the two methodologies are quite close to the size of
the underprediction resulting from the use of the imputation model (of the
order of £1).
4.6 FACILITATING FUTURE VALUATION

Clearly, these estimates could be further improved if additional data were
collected. An important feature of the methodological framework developed
above is that it points quite specifically to the kind of information that would
need to be collected in future surveys of volunteering in order to permit a
more sophisticated valuation exercise.
In order to develop the opportunity cost approach, it would be necessary
for volunteer surveys to pay greater attention to documenting alternative uses
of the volunteer’s time. Key questions are what the volunteer would be doing
if not volunteering, and how highly the volunteer would value time spent in
this alternative activity. As part of this process, it is important to collect
reliable information on the net hourly wage from employment. Of equal
value are details of the volunteer’s employment contract which would help to
establish the extent to which time in volunteering could be substituted at the
margin for time in employment, and at what rates.
As regards implementation of the replacement cost approach, the key
question is to identify the type of services that the volunteer performs for
the charity and to consider how these might be replaced by paid employees
and at what cost. In undertaking these calculations, it is important to take
into account the costs that the charity already incurs in deploying the
volunteer.
Finally, the application of the output-based approach would require sub-
stantial empirical advances in the economic analysis of the charitable sector.
On the one hand, it would be necessary to improve understanding of the
charitable production function, in other words the relationship between in-
puts of volunteer time and the output of philanthropic services. On the other
hand, further research is needed on the valuation which charity beneficiaries
place on the services delivered. A range of methodological alternatives for
doing this are presented in Chapter 1 and one particular methodological
approach is illustrated in Chapters 2 and 3. There is clearly considerable
scope for further research on both of these questions.
The valuation figures presented in this chapter are therefore in no sense

definitive. Their primary purpose has been to illustrate how it is possible to
refine previous estimates of the value of volunteer time even on the basis of
existing information, but more importantly to provide guidance for future
valuation attempts.
5. The aggregate benefits of the charitable
sector: summary
5.1 INTRODUCTION
A methodology for valuing the output of the charitable sector was developed
in Chapter 1. The methodology requires an estimation of the net social value
of the charitable sector, defined as the difference between total social benefits
and total social costs.
The main challenge of implementing this approach is to put monetary
values on the benefits that charities generate for their immediate target groups,
as well as for society at large. This task was undertaken in Chapter 2, where
members of the general public were asked to state what additional contribu-
tions they would be willing to make over and above their existing donations
in order to avoid losing the services provided by housing charities, among
others, and in Chapter 3 for the beneficiaries of hostels for the homeless in
London, who were asked to state their willingness to accept compensation for
the loss of these services. Issues relating to the valuation of volunteer time
were discussed in Chapter 4.
Now that all of these missing components are in place, it is possible to
proceed with the overall valuation exercise outlined in Chapter 1.
5.2 ACCOUNTING FRAMEWORK
5.2.1 Total Social Value
As noted in Chapter 1, the total social value of charities is the sum of direct
benefits to users, benefits to volunteers and indirect benefits to the general
public.
Total social value (TSV) =

direct user benefits + benefits to the general public + benefits to volunteers
Given that charity services are typically provided at zero or highly subsi-
dized cost, the fees paid by users do not fully capture the benefits that this
114
Aggregate benefits of the charitable sector 115
group receives. To capture the full benefits to charity users it is necessary to

have some measure of WTP (or WTA compensation) over and above any
existing fee payments. As a result of the empirical work reported in Chapter
3, this measure is available for charities in the housing and homelessness
sector, though not at present for other types of charities. Specifically for this
case, it was found that WTA compensation for the loss of hostel services was
£8164 per beneficiary per year.
Turning to the general public, it was noted above that actual donations
comprise cash gifts, profits from the sale of goods and investment income on
accumulated past donations. The survey reported in Chapter 2 found that
actual donations totalled £8 per year for housing and homelessness charities
and £76 per year for charities overall. Due to the free-rider problem, actual
donations are likely to underestimate the benefits that accrue. Once again, it
is necessary to have some measure of WTP for charity services over and
above existing donations. The empirical work reported in Chapter 2 provides
an estimate of this value for all charity sectors. The most conservative meas-
ure of incremental WTP for housing and homelessness surveys derived from
this survey was £13 per person per year, while for the charitable sector as a
whole, the most conservative estimate lies in the range £47 to £58 per person
per year.
The benefits to volunteers can be estimated in terms of the opportunity cost
of volunteer time, since, to the extent that volunteers behave rationally, they
will give time up until the point where the marginal benefit of doing so is
equal to the marginal cost. The survey of the general public reported in
Chapter 2 also collected information on the average number of hours volun-
teered to the various different types of charities considered. However, owing
to the absence of detailed information on the net hourly wage in employment,
it was not possible to estimate the true opportunity cost of time. Hence an
indicative value of £5 per hour was used. Applying this value to the average
number of hours volunteered yields a total value of time volunteered of £7
per person per year for housing and homelessness charities and £89 per
person per year for the charitable sector as a whole.
The above can be summarized by the following formula, where AF repre-
sents the actual user fees paid to charities, IWTPdir is the incremental WTP
(or WTA) of the direct beneficiaries, AD are the actual donations made to
charities, IWTPgp is the incremental WTP of the general public, PRODPROF
are the profits from the sale of goods, I is the interest on income invested by
charities, and Bvol is the benefits to volunteers.
Total social value (TSV) =

AF + IWTPdir + AD + IWTPgp + PRODPROF + I + Bvol
5.2.2 Total Social Costs
As indicated in Chapter 1, the costs of providing charitable services can be

measured by the income they receive. This is summarized in the following
formula, where AF are fees paid by users, AD are actual donations,
PRODPROF are the profits from the sale of products, I is interest income
from investment, Cvol are the costs of volunteering and Cgovt represents the
cost of government grants.
Most of these items can be derived directly from published charity ac-
counts. A key exception is the cost of volunteer time, which according to the
methodology developed in Chapter 4 should be valued in accordance with its
replacement cost. Since it was not possible accurately to measure the replace-
ment cost of volunteer time on the basis of the survey of the general public
reported in Chapter 2, an indicative value of £5 per hour is once again used.
Total social costs (TSC) = AF + AD + PRODPROF + I + Cvol + Cgovt
5.2.3 Net Social Value
The net social value is the difference between total social value and total
social costs that measures the net social value of the charities, the social
‘valued added’, stated in equation form below. As can be seen, all the expres-
sions for fees and donations drop out since they appear in both equations,
leaving the following reduced expression for net social value. Net social
value is essentially the incremental WTP of donors and direct beneficiaries
over and above the value of government grants, plus the net benefits to
volunteers. Clearly, net social value could be positive or negative, depending
on the relative value of the terms.
Net social value (NSV) = IWTPdir + IWTPgp + Bvol – Cgovt – Cvol
5.3 ESTIMATING THE NET SOCIAL VALUE

Applying the above accounting framework to the values estimated in the pre-
ceding chapters, the net social value of the charitable sector can be calculated.
For charities in the housing and homelessness sector it is possible to
calculate the full net social value including the benefits to direct users. This
calculation is reported in Table 5.1. For the charitable sector as a whole, the
benefits to direct users are not available and thus can only be estimated.
Consequently, the calculations presented in Table 5.2 are only preliminary,
and as such the results can only be regarded as tentative and illustrative.
Table 5.1 The net social value of housing and homelessness charities,
1997
Element Unit value Relevant population Total value
Total social value £2.56bn

AF £884 0.14m (target group) £0.12bn*
IWTPdir £8164 0.14m (target group) £1.14bn*
AD £8 46.2m (adult population) £0.37bn*
IWTPgp £13 46.2m (adult population) £0.60bn*
PRODPROF n.a. n.a.
I n.a. n.a.
Bvol £7 46.2m (adult population) £0.32bn*
Total social costs £0.84bn
AF £0.12bn*
AD £0.37bn*
PRODPROF n.a.
I n.a.
Cvol £0.32bn*
Cgovt £0.03bn†
Net social value £1.72bn
Sources: figures marked * are from surveys undertaken as part of the research in Chapters 2–
4. Figures marked † are based on the ONS 1994–95 survey as reported in Pharoah (1997).
In the case of housing and homelessness charities, the results reported in

Table 5.1 derive directly from the unit values reported above. The results
show that by far the largest category of benefits is the incremental WTP of the
direct beneficiaries; this reflects the high cost of housing and thus the high
WTA compensation of the target group. It is interesting to note that the actual
fee revenue is less than 10 per cent of the incremental WTP. In the case of the
general public, incremental WTP is somewhat less than half the value of
actual donations. The overall benefits to the general public at £0.97bn are
considerably smaller than the overall benefits to the target group of £1.26bn.
In the case of the charitable sector as a whole, the absence of information
about the incremental WTP of beneficiaries meant that additional assump-
tions had to be made. In order to get around this problem, it was assumed that
the ratio between the incremental WTP of the target group and that of the
general population would be the same as that found for the housing and
homelessness sector. On this basis, an estimate of the incremental WTP of
direct users was obtained by applying a factor of 1.3 to the incremental WTP
of the general population.
Table 5.2 The net social value of all charities, 1997
Element Unit value Relevant population Total value
Total social value £20.47bn–£22.02bn

AF £3.94bn
IWTPdir £4.34bn–£5.36bn
AD £76 46.2m (adult population) £3.52bn*
IWTPgp £47–£58 46.2m (adult population) £2.17bn–£2.68bn*
PRODPROF n.a. n.a.
I £2.39bn†
Bvol £89 46.2m (adult population) £4.11bn*
Total social costs £15.63bn
AF £3.94bn
AD £3.52bn*
PRODPROF n.a.
I £2.39bn†
Cvol £4.11bn*
Cgovt £1.67bn†
Net social value £4.84bn–£6.39bn
Sources: figures marked * are from surveys presented in Chapters 2–4. Figures marked † are
based on the ONS 1994–95 survey as reported in Pharoah (1997).
5.4 INTERPRETATION AND POLICY IMPLICATIONS

Although incomplete, the results in Tables 5.1 and 5.2 show that charities do
indeed add value to the resources that they absorb. The exact level of value
added must await further research, particularly of the direct users of charities,
since only homeless direct beneficiaries were surveyed in this research. There
is also a need for further work on the value of volunteer time.
Notwithstanding these limitations, the results reported in Tables 5.1 and
5.2 tentatively suggest that the charitable sector as a whole adds value of
around 35 per cent to total social costs (£21 billion/£16 billion), while the
housing and homelessness charities add value by some 200 per cent (£2.6
billion/£0.8 billion).
These figures can be thought of as demonstrating economic value. The
issue for charities is to capture as much as possible of this full economic
value, that is, the WTP over the fees charged to direct users, and the dona-
tions contributed by the general public. Table 5.2 suggests that the general
public alone have an incremental WTP for the services of the charitable
sector in the order of £2.5 billion per year, a sum that is not captured by
charities.
It is interesting to compare the incremental WTP of the general public with

the overall value of government grants to the charitable sector. Government
grants represent one way of overcoming the free-rider problem by taking
collective action to increase the flow of resources to charities. The results
reported in Table 5.2 indicate that the current value of government grants at
£1.7 billion falls short of incremental WTP, which stands at £2.5 billion.
One possible conclusion from this comparison is that there is a case for
increasing the size of the government grant to the charitable sector. However,
government grants are not the only way of allocating additional resources to
charities. Alternative methods include sharpening fiscal incentives for giving,
or increasing charity fundraising efforts. Part II of this book turns to the
problem of capturing economic value and examines the relative efficacy of
grants, fiscal incentives and fundraising approaches.
PART II
Capturing the Economic Value of Charities

6. Providing fiscal incentives for giving
6.1 INTRODUCTION
In many countries, including the United Kingdom, charitable donations re-
ceive special tax-exempt status. The rationale for this policy is precisely to
stimulate charitable giving by reducing the ‘price’ of making a donation,
thereby counteracting the free-rider tendency and helping charities to capture
their full economic value. Since the government contributes µ per cent of
every £1 given to charity, where µ is the marginal rate of tax rate, it only costs
£(1 – µ) to give £1 to charity. Clearly, the effectiveness of this policy depends
on the ‘price’ elasticity of donors.
A major focus of the US literature on giving and volunteering has been the
estimation of price and income elasticities, particularly in the context of
evaluating the efficacy of fiscal inducements towards philanthropic behav-
iour. Econometric studies have in general found that in the USA monetary
donations are price-elastic with respect to tax incentives, even though in the
case of volunteering the results are not so clear.
At a time when the UK government has been reviewing its own – very
different – tax treatment of the charitable sector (Banks and Tanner, 1998),
the econometric literature gives very little guidance on the price-responsive-
ness of British philanthropists. The few British studies that exist either exclude
the issue of price-responsiveness altogether (Banks and Tanner, 1997; Pharoah
and Tanner, 1997) or are subject to data limitations which make it difficult to
interpret the price effects obtained (Jones and Posnett, 1991a,b). Yet such
estimates are critical in informing fiscal policy towards the charitable sector.
Barrett (1991) shows how elasticity estimates can be used to test for the
neutrality and treasury efficiency of fiscal policy.
A priori, an increase in the marginal tax rate has an ambiguous effect on
charitable donations since it will both reduce the price of making tax-efficient
donations (thereby stimulating philanthropic gifts) and reduce the after-tax
income of donors (thereby dampening their generosity). It can be shown that
as long as the income and tax-price elasticities are equal in absolute magni-
tude the tax change will be ‘neutral’ in its impact on giving behaviour. If the
tax-price elasticity exceeds the income elasticity in absolute terms, the tax
rise will have a positive net effect on donations, and vice versa.
123
124 Capturing the economic value of charities
Another important concern in the design of fiscal policy is whether the

extra donations induced by a tax incentive are greater or less than the result-
ing loss in government revenues. Barrett (1991) shows that if donations are
tax-price-elastic, the induced flow of resources into the charitable sector
more than offsets the reduction in tax revenues. This situation is described as
treasury efficiency.
Finally, if donations are found to be inelastic with respect to income, this
would suggest that subsidizing donations may be a regressive way of achiev-
ing an increase in the aggregate supply of public goods.
This chapter reports estimates of price and income elasticities for both
monetary giving and volunteering in the UK, based on a new and hitherto
unexploited source of information: the Individual Giving Survey (Halfpenny
et al., 1992, 1993, 1994). The dataset is attractive in a number of ways. First,
it is made up of repeated cross-sections of data across a period of substantial
variation in the UK tax system, thereby incorporating variation in the tax-
price variable which is not merely cross-sectional. Second, it contains very
detailed information on gifts of time and money. Third, it records respond-
ents’ perceptions of the price of giving as well as the objective value of the
price itself, thereby making it possible to take into account the problem of
imperfect information about prices.
The results indicate that British philanthropists show significant responses
to fiscal incentives, even though price elasticities are somewhat smaller than
those found in the USA. Since tax-price elasticities are generally found to
exceed income elasticities, an increase in the marginal rate of income tax can
be expected to bring about a net increase in tax-efficient donations. However,
the fact that donations are inelastic with respect to the tax price indicates that
the induced increase in donations will not be large enough to offset the loss
of revenue to the Treasury.
The empirical analysis does not limit itself to considering conventional tax-
price elasticities, but also examines the extent to which contributions are affected
by two additional concepts of price. The first is the efficiency price or the rate
at which charities absorb donations on administrative expenditures. The second
is the price of time, that is the productivity differential between time in employ-
ment and time in volunteering. These wider price effects are also found to be
significant in determining philanthropic behaviour.
The remainder of the chapter is organized as follows. Section 6.2 provides
an overview of the literature on the price-responsiveness of giving and volun-
teering behaviour both in the UK and North America. Section 6.3 introduces
the dataset on which the analysis is based, while Section 6.4 explains how it
was used to construct the relevant price variables. Section 6.5 reports the
empirical results and calculates the corresponding price and income elasticities.
Section 6.6 concludes.
Providing fiscal incentives for giving 125
A more extensive discussion of the material can be found in Foster (1999c).
6.2 LITERATURE OVERVIEW
6.2.1 Giving
The existing literature on price and income elasticities of charitable giving is

summarized in Table 6.1.
The issue of charitable donations began to be studied empirically in the
USA from the late 1960s onwards (Taussig, 1967; Schwartz, 1970; Feldstein
and Clotfelter, 1976; Feldstein and Taylor, 1976; Boskin and Feldstein, 1977).
A primary objective of the US literature has been to gauge the effect of
taxation policy on philanthropic behaviour. Under US tax rules, summarized
in Table 6.2, ‘itemized’ charitable donations listed on the annual tax return
are deducted from overall income before the assessment of tax liability. Thus
donating a dollar to a charitable cause only costs the donor (1 – µ) dollars of
income, where µ is the marginal income tax band.
The earliest US studies set about estimating the elasticity of donations to
this tax price using time-series data from tax returns. However, it was soon
noted that use of such data carried a number of significant drawbacks (Feldstein
and Clotfelter, 1976). Particular problems were the partial non-identifiability
of donations by low- and middle-income households (who may choose not to
itemize their deductions), the possible over-reporting bias of donations by
itemizers (in an attempt to minimize tax liability), and the limited availability
of explanatory variables (particularly measures of permanent income and
wealth). These problems were avoided by altering the basis of empirical work
from tax-returns data to survey-based data (Clotfelter, 1985).
Although early studies based on tax returns found relatively low tax-price
elasticities (for example Taussig, 1967; Schwartz, 1970), a consensus was
reached in later work that donations were tax-price-elastic, with most esti-
mates lying in the range –1.0 to –1.5 in absolute value. Income elasticities, on
the other hand, were invariably found to be below unity, with a typical value
of 0.8. These estimates were derived for the most part from simple ordinary
least squares (OLS) models of donations, often using logarithmic specifica-
tions.
Two important innovations to be introduced during the 1980s were the use
of panel data (Clotfelter, 1980) and the correction of possible endogeneity
problems (Reece and Zieschang, 1985). Ironically, both of these refinements
had the effect of depressing the resulting estimates of tax-price elasticity
below unity and back towards the much lower levels found in the earliest
studies (Taussig, 1967; Schwartz, 1970).
Table 6.1 Summary of literature on price and income elasticities of giving
Data Data Model Functional Tax-price Income

Authors Year Country type source Observations specification form elasticity elasticity
Taussig 1967 USA CS Tax returns Individuals OLS Log-linear 0.2–0.4 1.3–3.1
Schwartz 1970 USA TS Tax returns Groups OLS Log-linear 0.4 –0.8 0.2–0.8
Feldstein and Clotfelter 1976 USA CS Survey Individuals OLS Log-log 1.6 0.8
Feldstein and Taylor 1965 USA CS Tax returns Individuals OLS Log-log 1.0–1.5 0.8
Boskin and Feldstein 1977 USA CS Survey Individuals OLS Log-log 2.5 0.7
Abrams and Schmitz 1978 USA TS Tax returns Groups OLS Log-log 1.0–1.1 0.8–0.9
Reece 1979 USA CS Survey Individuals Tobit Linear 1.0 –1.4 0.6–1.4
126
Clotfelter 1980 USA Panel Tax returns Individuals Fixed effects Log-log 0.6–1.1 0.4–0.7
Reece and Zieschang 1985 USA CS Survey Individuals Tobit Linear 0.1–0.2 1.3–1.4
Schiff 1985 USA CS Survey Individuals OLS Linear 2.8 0.8
Kingma 1989 USA CS Survey Individuals Tobit Linear 0.4 1.0
Barrett 1991 USA Panel Tax returns Individuals Random effects Log-log 1.1 0.2
Brown and Lankford 1992 USA CS Survey Individuals Tobit Log-linear 1.6–1.8 0.4
Jones and Posnett 1991a UK CS Survey Individuals Selectivity Log-linear n.a. 0.4–0.6
Jones and Posnett 1991b UK CS Tax returns Individuals Selectivity Log-log n.a. 0.5–1.7
Banks and Tanner 1997 UK TS Survey Individuals Selectivity Log-log n.a. 0.1–1.2
Notes: regarding data type, CS stands for cross-sectional and TS for time-series. Regarding model specification, OLS stands for ordinary least squares.
Regarding tax-price elasticities, reported figures are absolute values; all estimates have the expected negative sign.
Table 6.2 Contrast between tax treatment of charitable donations in the USA and the UK
USA UK
Federal income tax liability is assessed by subtracting allowable deductions Charitable donations are only eligible for tax relief if they are
from gross income. Allowable deductions include both business expenses and channelled through one of the four following schemes:
charitable donations. There are, however, certain restrictions regarding:
● The Deed of Covenant scheme, which allows charities to
● the types of recipient organizations whose contributions qualify for reclaim tax paid on a regular annual gift committed over a
deductability; period of at least four years. There are no upper or lower
● the proportion of income given which can qualify for deductibility. limits on the value of sums that can be covenanted. The
payments must be accompanied by a written Deed of Cov-
US taxpayers can be divided into two categories according to how they approach
enant between the donor and the charity.
the calculation of their deductions:
● The Payroll Giving scheme (in operation since 1987), which
● Non-itemizers are those who elect to stake a standard deduction against is operated by some employers and available only to their
their gross income, rather than separately itemizing their deductible ex- employees. The scheme allows people to have donations
127
penditures. This innovation was introduced in 1944 in an attempt to simplify deducted directly from their pay before tax liability is as-
income tax assessment. Effectively, non-itemization is an attractive option sessed. There is an upper limit on the value of the donations
to all taxpayers whose deductible expenses sum to no more than the value (which was £1200 per annum in 1996/97).
of the ‘standard deduction’. It is thus more prevalent among low- and ● The Gift Aid scheme (in operation from October 1990),
middle-income households. An important implication of this choice is that which allows charities to reclaim tax paid on lump-sum
any charitable donations made by such households will not be subject to gifts in excess of a value threshold (which was £250 in
any tax incentive, facing instead a marginal price of unity. 1996/97) with no upper limit. A Gift Aid form must be
● Itemizers, are those who elect to assess their allowable deductions by made out to the charity as an accompaniment to the net gift.
individually ‘itemizing’ on their tax returns all deductible expenditures ● The Millennium Gift Aid scheme (in temporary operation
incurred during the past tax year, rather than using the ‘standard deduc- between July 1998 and December 2000), which allows chari-
tion’. Effectively, ‘itemization’ is an attractive option to all taxpayers ties to reclaim tax paid on lump-sum gifts in excess of a
whose deductible expenses sum to more than the value of the ‘standard value threshold of £100. The scheme only applies to chari-
deduction’. It is thus more prevalent among high-income households. An ties supporting education, health and poverty relief
important implication of this choice is that any charitable donations made programmes in 80 specific developing countries.
by such households will not be subject to a tax incentive, facing a price
equal to one minus the corresponding marginal income tax rate.
In the UK, the empirical analysis of charitable donations did not begin
until the 1990s (Jones and Posnett, 1991a; Banks and Tanner, 1997, 1998;
Pharoah and Tanner, 1997). These studies made use of data contained in the
Family Expenditure Survey, which requires individuals to keep diary records
of casual donations over the period of a fortnight, and also collects informa-
tion on regular donations made by bank standing orders or payroll deductions.
A further study by Jones and Posnett (1991b) makes use of Inland Revenue
data on giving by covenant.
Unlike some of the earlier US work, these UK studies give explicit consid-
eration to the problem of modelling zero donations. Banks and Tanner (1997)
identify three possible interpretations of zero donations, namely: a genuine
preference for not making donations; infrequency of donations (leading to
none being recorded during the survey period) or recording error (as a result
of respondents failing to recall donations that were actually made). In the
early US studies, this problem was dealt with by discarding zero observations
or adding a small positive quantity to permit their inclusion within a logarith-
mic specification.
Both of the UK studies considered here, as well as some of the later US
studies, use limited dependent variable specifications to permit a more so-
phisticated incorporation of the zeros. Banks and Tanner (1997) distinguish
between the Tobit specification and the selectivity model. The former models
participation and expenditure as a single process and thus implicitly assumes
that all zeros represent a genuine preference for not making donations. The
latter allows participation and expenditure to be governed by separate pro-
cesses, thereby providing greater flexibility in modelling zero donations.
Both of the UK studies indicate a preference for the selectivity specification.
The estimation of tax-price elasticities has not been such a prominent
issue in the UK literature as in the US literature. This is largely a reflection
of the differences in the tax treatment of charitable contributions between
the two countries, as indicated in Table 6.2. Whereas in the USA ‘itemizers’
are at liberty to claim tax relief on all charitable gifts, in the UK tax relief is
confined to a number of special modes of giving. These schemes are quite
restrictive in the sense that they require either a long-term commitment to
making regular contributions (covenants and payroll giving), or a relatively
large lump-sum contribution (Gift Aid and Millennium Gift Aid). Perhaps
as a consequence of this, tax-efficient giving remains a relatively small
proportion of individual giving in the UK. For example, in 1992/3 tax-
efficient giving amounted to £1200m (Inland Revenue, 1998) or about 24
per cent of the estimated £5000m total value of charitable giving for the
corresponding year (Halfpenny et al., 1994). All remaining donations can
be described as non-tax-efficient in the sense that they do not attract any tax
advantage.
None of the UK studies summarized in Table 6.1 presents estimates for

price elasticities. Jones and Posnett (1991a,b) do in fact estimate significant
price effects in both of their papers, but do not report price elasticies. In part,
this reflects the authors’ concerns that the estimated price effects may be
spurious due to limitations with the data. In the first study (Jones and Posnett,
1991a), which is based on the Family Expenditure Survey, the authors are
surprised to find a significant price effect given relatively low participation in
tax-efficient giving and consequently conjecture that the result may simply
reflect collinearity beween tax-price and income variables. In the second
study (Jones and Posnett, 1991b), which uses tax return data on giving by
covenant, the authors claim that the significance of the tax-price variable is
largely an artefact of the Inland Revenue system. The reason that they give
for this is that only higher-rate taxpayers have any real incentive to report
covenants to the Inland Revenue.
As regards income elasticities, the estimates presented span the range of
those which have been obtained for the USA. On the basis of their selectivity
model Banks and Tanner (1997) find that while the income elasticity of
participation is very low (0.1), the income elasticity of gifts among donors is
relatively high (1.2).
6.2.2 Volunteering
The existing literature on price and income elasticities of volunteering is

summarized in Table 6.3.
Largely owing to a paucity of suitable data, the empirical analysis of
volunteering behaviour does not begin until considerably later than the re-
search into monetary contributions. In the case of volunteering, the appropriate
price variable is the opportunity cost of the time given to charity. A widely
used measure for this is the net hourly wage rate, although strictly speaking
this is only valid for those who are able freely to substitute time between
work and leisure activities at the margin. Given the lumpy and restrictive
nature of employment contracts, this is unlikely to be true except in a minor-
ity of cases.
To the extent that there is substitutability or complementarity between
giving and volunteering, the tax price of giving may become relevant as a
cross-price variable. The earliest study by Dye (1980) confines its attention to
estimating this cross-price effect which is found to be negative and signifi-
cant and thus provides evidence of complementarity between gifts of time
and money.
A much more comprehensive treatment of the subject is provided in the
seminal paper by Menchik and Weisbrod (1987). These authors estimate a
cross-tax-price elasticity of –1.3, which is found to be substantially higher
Table 6.3 Summary of literature on price and income elasticities of volunteering
Tax
Data Data Model Functional Wage-price cross-price Income
Study Year Country type source Observations specification form elasticity elasticity elasticity
Dye 1980 USA CS Survey Individual Probit and Log-log n.a. 0.6 Insignificant
OLS
Menchik 1987 USA CS Survey Individual Tobit Linear 0.4 1.3 0.7
and
Weisbrod
Brown 1992 USA CS Survey Individual Tobit Log-linear n.a. 1.1–2.1 n.a.
130
and
Lankford
Freeman 1997 USA CS Survey Individual Probit and Log-log Insignificant n.a. n.a.
OLS
Day and 1996 Canada CS Survey Individual Selectivity Log-linear n.a. n.a. Insignificant
Devlin
Knapp 1995 UK CS Survey Individual Selectivity Log-linear Insignificant 0.1 n.a.

et al.
Notes: regarding data type, CS stands for cross-sectional. Regarding model specification, OLS stands for ordinary least squares. Regarding tax-price
elasticities, reported figures are absolute values; all estimates have the expected negative sign.
than the own-wage-price elasticity of –0.4. The negative sign of the cross-
tax-price elasticity once again suggests complementarity between giving and
volunteering behaviour. They also estimate an income elasticity of 0.7 for
volunteering, which lies in the range of results obtained for monetary gifts.
Subsequent studies have failed to produce such clear-cut results. For exam-
ple, Freeman (1997) finds positive and significant coefficients on income and
the net hourly wage in a model of volunteer participation. Freeman argues
that one explanation for this result is that hourly wage acts as a proxy for the
productivity of volunteer time rather than capturing the opportunity cost of
hours volunteered. He hypothesizes that the correct price variable for volun-
teering is not so much the net hourly wage as the ratio between the net hourly
wage and the productivity of time in volunteering, since the lower this ratio,
the more attractive it becomes to give time rather than money. Freeman
constructs this variable on the basis of a subjective respondent assessment of
the relative productivity of time in work and volunteering and in a second set
of models finds a statistically significant relationship between hours volun-
teered and this revised price variable.
There has only been one UK econometric study of volunteering behaviour
to date (Knapp et al., 1995), making use of data collected by Volunteer Centre
UK/Social and Community Planning Research survey (1991). Based on a
selectivity model, the study finds that both income and the cross-tax-price of
giving positively affect the probability of participation but do not influence
the number of hours volunteered. The finding of a positive cross-tax-price
effect indicates that giving and volunteering are substitutes, and is thus at
odds with the findings of the US literature. As in the Jones and Posnett paper
(1991a), the authors of this study express some surprise at uncovering any
statistically significant cross-tax-price coefficient given the limited extent of
tax-efficient giving in the UK. Finally, Knapp et al. (1995) do not find any
statistically significant effect from the own-wage price of volunteering. They
suggest that this may be due to the fact that this variable was not reported in
their dataset but rather had to be proxied by average wage rates for particular
occupational groups.
Summary remarks
Thus, overall, the analysis of philanthropic behaviour has reached a much
higher level of development in the case of monetary giving than in the case of
volunteering. Moreover, the associated empirical literature is far more exten-
sive in the USA than in the UK. While there is a considerable body of
evidence which supports the finding that US donations are tax-price-elastic,
some of the more recent and sophisticated studies reach the opposite conclu-
sion (Clotfelter, 1980; Reece and Zieschang, 1985). By contrast, the few
British studies that exist either exclude the issue of price-responsiveness
altogether (Banks and Tanner, 1997; Pharoah and Tanner, 1997) or are sub-
ject to data limitations which make it difficult to interpret the price effects
obtained (Jones and Posnett, 1991a,b).
6.3 THE DATASET

The empirical analysis presented in this chapter is based on data collected in
the Individual Giving Survey, during three repeated annual cross-sections
covering the years 1991–93 and comprising just over 1000 observations each
(Halfpenny et al., 1994). Each of these annual surveys was based on a
representative cross-section of the British population (excluding Northern
Ireland), achieved by quota sampling on demographic characteristics such as
sex, age and size of household. Interviews were conducted face to face in the
respondents’ homes and spread evenly throughout the year to avoid seasonal
biases.
It is important to stress the value of being able to rely on a repeated cross-
section of data as opposed to a single cross-section, as has been the case with
much of the philanthropic literature (recall Tables 6.1 and 6.3). Of particular
importance for the identification of price effects is the fact that there were
numerous modifications to the tax system during the period 1990 to 1993.
The introduction of independent taxation occurs just a few months before the
beginning of the study period, while the creation of the additional 20 per cent
income tax band took place right in the middle of the study period, in 1992–
93. There were also numerous modifications to the real value of tax allowances,
most notably the fall in the value of the married couples’ allowance by more
than 10 per cent in real terms. All of this means that the tax-price effects are
not merely being identified from differences in behaviour between individu-
als at very different income levels, but also from variations in the tax price
faced by individuals at similar points in the income distribution.
During the interviews, respondents are required to recall philanthropic
behaviour during the last month (or year in the case of tax-efficient methods
of donation). This can be contrasted with the Family Expenditure Survey
(FES) – used in earlier UK empirical studies – which requires respondents to
keep a diary of casual cash donations over a two-week period.
The various differences that exist between these two survey methodologies
suggest that the corresponding measures of donations are likely to diverge.
This is so for a number of reasons. First, the IGS definition of charitable
donations is broader than that adopted in the FES, in that it includes pur-
chases of goods and services from charities in addition to pure donations.
This will tend to inflate the estimate of donations resulting from the IGS.
Second, the method of collecting donations differs. In the FES respondents
are asked to keep a diary of their donations, while in the IGS respondents are
asked to recall their donations during a household interview. To the extent
that respondents have trouble in remembering what contributions they have
made, this may create a downward bias in the IGS data. Finally, the FES
collects information about donations over a two-week period while the IGS
uses a monthly period. It is possible that the use of a longer recall period may
lead to ‘telescoping’ effects in the IGS, whereby respondents erroneously
ascribe donations from the more distant past to the survey period. This would
tend to inflate the donations measure obtained from the IGS and would
consequently offset the earlier problem associated with imperfect recall.
Comparing summary statistics from the FES and IGS for 1993, it becomes
possible to gauge the relative magnitude of these effects. The IGS records
participation rates that are very much higher than the FES (77.5 per cent
versus 29.1 per cent), which is probably a reflection of the longer survey
period and also the wider definition of charitable gifts. However, in terms of
the average weekly donations recorded, the FES gives slightly higher values
than the IGS (£4.11 per week versus £3.46 per week), although in terms of
the median this ranking is reversed (£1.50 per week in the IGS and £1.23 per
week in the FES).
6.4 ANALYTICAL FRAMEWORK

This section introduces the three different concepts of price that are deemed
relevant to philanthropic behaviour and explains how the data contained in
the IGS were used to construct the corresponding variables. As noted in the
introduction, the concept of price considered here includes but is not limited
to the tax price, which has formed the central focus of the earlier literature. In
addition, the analysis will examine to what extent gifts of time and money are
affected by the efficiency price and the time price. The efficiency price refers
to the rate at which charities absorb donations on ‘unproductive’ administra-
tive expenditures, whereas the time price is related to the productivity
differential between time in employment and time in volunteering.
6.4.1 Tax Price
When donors give via one of the government’s tax-efficient schemes, de-
scribed in Table 6.2, they effectively obtain a reduction in the price of giving.
This is because the government waives the tax that would normally be levied
on the donation. Thus, in order to give £1 to a charity, the donor will only
need to sacrifice (1 – µ) of his after-tax income, where µ is the marginal
income tax rate. This tax price of giving is defined mathematically as below.
Since the marginal income tax rate is always greater than or equal to zero, the
tax price will always be less than or equal to one:
PXi = (1 − µ i ) (6.1)
In order to ascertain the tax price faced by each survey respondent it is

necessary to have data on their marginal income tax rates (µ). Unfortunately,
the Individual Giving Survey does not contain this information. However, the
survey does include many of the underlying variables that determine an
individual’s income tax position, such as personal income, household in-
come, age, marital status, and single-parent status. On the basis of these
variables, and guided by the tax rules for the relevant period, it is possible to
create a simple algorithm which allocates respondents to a particular mar-
ginal income tax band.
The algorithm generates the real objective tax price faced by each respond-
ent. However, it is legitimate to question whether people actually perceive the
tax price in this way. One of the questions included in the survey asks
respondents whether they are aware of the opportunity to give through one of
the government’s tax-efficient schemes. Interestingly, only 70 per cent re-
sponded that they knew about the schemes. The significant minority did not
effectively face a tax price of unity irrespective of their marginal income tax
band. In order to ensure a closer reflection of reality, the tax price variable
was adjusted accordingly. Consequently, the analysis is based on this subjec-
tive tax-price variable, as opposed to the objective one.
50
40
30
20
10
0
0.60 0.75 0.80 1.00
Figure 6.1 Percentage frequency distribution of tax price (£)
Figure 6.1 presents the percentage frequency distribution of the subjective

tax-price variable which indicates that about half of the sample face a tax
price of unity and thus do not perceive any fiscal incentives to give. The next
largest group, representing 40 per cent of the total, face a tax price of 0.75,
corresponding to the standard income tax band. Fewer than 5 per cent of the
sample face the lowest tax price of 0.6, corresponding to the higher-rate
income tax band.
6.4.2 Efficiency Price
Not all of the money that donors give to charity will end up being spent on
the good cause. The reason is that charities are not perfectly efficient at
converting gifts into philanthropic services (Rose-Ackerman, 1986; Steinberg,
1986; Weisbrod and Dominguez, 1986). Inevitably, some fraction of the
receipts (α) will be absorbed in administrative expenditures needed to run the
organization. Hence a £1 donation will increase philanthropic expenditure by
only (1 – α). This efficiency price of giving is defined mathematically as
follows:
 1 
PEi =   (6.2)
 1 − αi 
The IGS does not contain information on the precise charity to which a
particular respondent contributes, thus it is not possible to ascertain the
objective value of α, which would (otherwise) be retrievable from charity
accounts. However, the IGS does collect data on the subjective value of (1 –
α), by explicitly asking respondents: ‘For every £1 you give to charity, how
much do you think gets to the needy cause?’ This subjective value is – in fact
– more helpful for the construction of the corresponding price variable than
the objective value would have been. This is because it seems likely that
many donors are uninformed as to the true value of α; thus their behaviour
will in all probability be guided more by their perceptions of this price
variable than by what it actually is.
Figure 6.2 illustrates the percentage frequency distribution of the efficiency-
price variable. The modal value of the distribution lies in the 1.5 to 2.0 range,
indicating that more than a third of respondents believe that administrative
expenditures absorb between 25 per cent to 50 per cent of each pound
donated. It is striking that as many as 15 per cent of respondents believe that
the efficiency price is greater than 4, or in other words that administration
absorbs more than 75 per cent of each pound donated.
These subjective assessments indicate a considerable degree of scepticism
about the efficiency of voluntary organisations. However, the reality is more
50
40
30
20
10
0
1–1.5 1.5–2 2–4 >4
Figure 6.2 Percentage frequency distribution of efficiency price (£)
favourable than respondents seem to imagine. During the period of the early
1990s covered by the survey, the top 500 of the UK’s fundraising charities were
allocating on average no more than 15 per cent to expenditures such as adminis-
tration and fundraising which do not contribute directly to the good cause.
6.4.3 Time Price
The concept of price for volunteering is somewhat more complex, although

many of the foundations have already been laid in Chapter 4. A person who
volunteers self-evidently sacrifices, say, an hour of their own time in order to
give that time to charity. The value that the volunteer places on their own
time will, in all likelihood, be very different from the value that the charity
places on the time they have volunteered. For reasons already discussed at
length in Chapter 4, the value that the volunteer places on their own time is
often approximated as their net hourly wage in employment (ϖ), while the
value that the charity places on the volunteer’s time will be related to the full
hourly (gross wage plus non-wage labour) costs of employing somebody else
to perform the same task as the volunteer (υ). The ratio of the value of the
time sacrificed by the volunteer to the value of the time gained by the charity
effectively defines the relevant price of time. This is stated mathematically as
follows:
ϖ 
PTi =  i  (6.3)
 υi 
Any differences between ϖ and υ are attributable to one of two sources.

First, there is the fact that υ includes taxation and non-wage labour costs
whereas ϖ does not. Second, it is conceivable that time spent in volunteering
may be more or less productive than time spent in employment, which would
be reflected in differences between the corresponding wages. For example, if
a merchant banker volunteers at a soup kitchen over the weekend, the wage
earned by the merchant banker in their usual employment will be very differ-
ent to the wage that the charity would have to pay to get an employee to do
kitchen work.
In order to quantify the time-price variable defined above, it is necessary to
have a measure of the net hourly wage rate of each respondent, together with
the full hourly labour cost to the charity of replacing the volunteer with a paid
employee. While neither of these two variables is recorded directly in the
IGS, the survey does contain adequate information from which to construct
estimates of the two corresponding wages for each respondent. A detailed
description of how this was done can be found in Chapter 4.
Figure 6.3 summarizes the percentage frequency distribution of the price
of time for those who volunteer. The first point to note is that the estimates
are extremely low in absolute value. This may in part reflect the underprediction
of the net hourly wage noted in Chapter 4 above. However, even if the net
hourly wage rate is multiplied by a factor of three (which is undoubtedly a
gross overestimate of the degree of underprediction), the average price of
time among volunteers remains low in absolute value, rising from £0.16 per
hour to £0.62 per hour. Moreover, the proportion of volunteers who face a
50
40
30
20
10
0
<0.05
0.05–0.10
0.11–0.15
0.16–0.20
>0.20
Figure 6.3 Percentage frequency distribution of time-value price (£)

time price in excess of unity rises only slightly, from 2.4 per cent to a mere
12.4 per cent.
Whatever the true absolute level of the time price of volunteering, Figure
6.3 remains informative in illustrating the relative level of this price among
volunteers. In particular, there is a noticeable skew in the distribution of the
price of time towards the lowest values, which provides some evidence that
participation in volunteering may be price-sensitive.
6.4.4 Analysis of Price Variables
Before proceeding to estimate a full econometric model of philanthropic

contributions, it is helpful to analyse some of the broad patterns discernible
in the data.
To this end, Table 6.4 presents the percentage participation rates for giving
and volunteering at different levels of the three price variables developed
above. What is immediately striking about Table 6.4 is the presence of a clear
negative monotonic relationship between price levels and participation rates.
This relationship, which applies to all three of the price variables for both
Table 6.4 Monthly philanthropic participation rates against different price

variables (%)
Giving Volunteering
Efficiency price
1.0–1.5 84.12 35.30
1.5–2.0 81.40 24.94
2.0–4.0 78.67 24.61
>4.0 70.96 21.84
Tax price
0.60 93.02 50.00
0.75 88.36 34.40
0.80 86.84 28.95
1.00 73.55 20.12
Time price
<1 92.12 –
>1 73.27 –
Overall 77.66 26.43
Notes: since the time-price variable is observed only for volunteers, the participation rate for
volunteering against the time price cannot be calculated.
giving and volunteering, provides prima facie evidence of price-responsive-

ness in philanthropic behaviour. It is also interesting that participation in
volunteering appears to be associated with the tax price, while participation
in giving appears to be associated with the price of time. This is suggestive of
a cross-price effect whose negative sign would seem to indicate comple-
mentarity between giving and volunteering (Dye, 1980; Brown and Lankford,
1992).
Table 6.5 presents a parallel set of results, showing how the size of the
average monthly philanthropic contribution varies with the price level. There
continues to be some evidence of a negative association between the two,
even though this looks somewhat weaker than the evidence presented for
participation rates.
Table 6.5 Mean monthly donations against different price variables
Giving (£) Volunteering (hours)
Efficiency price
1.0–1.5 17.15 16.48
1.5–2.0 12.96 12.50
2.0–4.0 11.45 16.86
>4.0 12.46 17.99
Tax price
0.60 24.51 15.21
0.75 14.67 16.63
0.80 11.59 16.77
1.00 9.42 12.56
Time price
<1 21.65 16.21
>1 10.34 13.51
Overall 13.65 15.94
However, given the correlation between the tax price and the household
income level, it is not possible to say to what extent these results reflect a
pure price effect or simply an indirect income effect. Thus a firm conclusion
on price-responsiveness must await the results of multiple regression that
controls for factors such as income.
Another way of analysing the data is to examine how the level of the price
variable differs between those who make philanthropic contributions and
those who do not. The data indicate that the tax price is somewhat lower for
those who make monetary donations as opposed to those who do not, at 0.86
versus 0.92. The contrast is much more striking in the case of the efficiency
price, which takes an average value of 3.46 for contributors and 5.59 for non-
contributors.
In the case of volunteering, the data for the time price are only available
for those who contribute, so that the same comparison between contributors
and non-contributors cannot be made. However, it is notable that the mean
time price faced by volunteers at 0.50 is substantially lower than the prices of
monetary giving. This finding raises questions as to why donors do not have
greater recourse to this relatively low-cost mode of philanthropy.
One possible explanation is that volunteers face restrictions on the avail-
ability of time. The data provide only mixed support for this hypothesis.
Participation in volunteering is indeed higher among part-time workers than
full-time workers, particularly for women, where the figures are 37.5 per cent
versus 27.4 per cent. However, participation is actually lower among those
who are not in employment than among those who are, both for men and
women. As regards the hours volunteered, there is evidence that men in full-
time employment contribute fewer hours (less than 15 hours per month) than
men who are not in full-time employment (who volunteer more than 20 hours
per month). However, there is no such effect for women.
6.5 EMPIRICAL RESULTS
6.5.1 Coefficient Estimates
This section presents the results of estimating giving and volunteering equa-
tions based on the IGS dataset and incorporating the various price variables
defined above. The objective of these equations is to estimate the extent to
which donations of time and money are sensitive to the associated prices. The
regression makes it possible to control for other (non-price) influences on
philanthropic behaviour, so that the price effect can be isolated with greater
precision.
Following the earlier UK literature, two separate Heckman selectivity models
are estimated for giving and volunteering (Jones and Posnett, 1991a,b; Banks
and Tanner, 1997; Pharoah and Tanner, 1997). The Heckman specification
has the strength of incorporating separate models for the participation deci-
sion and the decision of how much time or money to contribute. This allows
for the fact that separate processes may be governing these two aspects of the
philanthropic decision (Greene, 1993).
Thus the demand for charitable contributions of time and money is speci-
fied as a function of prices and income and conditioned on a range of
socioeconomic characteristics such as age, birth year, education, employment

status and region. The equations include both own-price and cross-price
terms. However, the time price could only be included in the hours equation
for volunteering given that it is only observed in the case of those who
actually volunteered. The reason for including age and birth year separately
is to be able to distinguish between pure age effects and cohort effects. This
is one of the advantages of having repeated cross-sections of data over a
series of years.
Following an extensive specification search, the results presented here are
based on a double log functional form (Foster, 1999c). That is to say that the
dependent variable is the logarithm of monthly donations or monthly hours
volunteered respectively, while the key economic variables, price and in-
come, are also entered logarithmically. Table 6.6 presents the results of the
Heckman models for giving and volunteering.
The giving model reported in the left-hand side of the table is identified by a
variable that captures the importance of religion to the respondent. To wit,
respondents were presented with the statement ‘Religion is very important in
my life’ and were asked to what extent they agreed or disagreed with it. The
answers are recorded on a five-point Likert scale, ranging from –2 (strongly
disagree) to +2 (strongly agree). Religious views are thought to be a suitable
instrument for the identification of the selectivity term. This is because it seems
plausible that religious beliefs may be correlated with altruistic preferences and
hence could explain participation in philanthropic giving (Smith et al., 1995).
Both the efficiency and tax-price variables are found to have a significant
impact on participation in spite of the fact that only about 10 per cent of the
sample give through one of the government’s tax-efficient schemes. Partici-
pation is also positively associated with higher income. Turning to the
socioeconomic factors, giving was found to be significantly more likely
among the highly educated and among women. There is a negative effect
attributable to age, combined with a negative cohort effect. This suggests
that, within any particular generation, people give less money as they grow
older. However, looking across generations, older generations tend to be
more generous than younger generations. Finally, the variable capturing reli-
gious beliefs is statistically significant and positively associated with
participation.
Turning to the contributions equation, both income and the efficiency price
continue to be statistically significant, with the expected signs. The positive
coefficient on birth year indicates that, although younger generations are less
inclined to give to charity, the gifts that they do make tend to be larger than
those of older generations.
The results of the volunteering model are reported on the right-hand side
of Table 6.6. The selectivity term for this model is identified by means of a
Table 6.6 Heckman selectivity models for giving and volunteering
Giving Volunteering
Participation Contribution Participation Contribution
Efficiency price –0.645 –2.878 –1.076 –3.572 –1.204 –5.243 –0.022 –0.048
Tax price –1.087 –3.527 –0.366 –0.941 –0.922 –3.131 –0.758 –1.522
Hourly wage –0.007 –0.150
Replacement wage –0.543 –4.869
Household income 0.170 2.531 0.193 2.201 0.247 3.574 –0.204 –1.691
Years of education 0.096 2.279 0.122 2.335 0.103 2.608 0.088 1.431
Sex –0.286 –3.123 0.037 0.307 –0.107 –1.190 0.084 0.573
142
Age –0.012 –4.242 0.003 0.907 0.006 2.254 0.010 2.206
Birth year –0.419 –10.166 0.242 3.943 –0.162 –4.158 0.071 0.980
Full-time employed 0.061 0.547 –0.325 –2.305 –0.136 –1.210 –0.443 –2.652
Part-time employed 0.131 0.975 –0.086 –0.523 0.170 1.349 –0.647 –3.362
North of England 0.052 0.525 –0.319 –2.534 –0.071 –0.728 –0.054 –0.366
Scotland or Wales 0.099 0.869 –0.259 –1.802 –0.193 –1.691 –0.065 –0.360
Southeast of England –0.133 –1.204 –0.163 –1.149 –0.158 –1.437 0.341 2.068
School-aged children 0.078 3.199 0.147 3.395
Constant 35.623 9.576 –19.192 –3.618 9.263 2.653 –5.354 –0.895
Correlation –0.76 –0.19

Variance 1.52 1.10
Log-likelihood –2117.11 –1291.75
Observations 1254 1307
dummy variable capturing the presence of school-aged children in the house-

hold. The reasoning behind this is that many voluntary activities are organized
in and around schools, suggesting that parents of school-aged children are
more likely to become involved.
In the participation equation, the price and income variables are statisti-
cally significant, with the expected signs. Other variables found to be positively
associated with volunteering are age, education and the presence of school-
aged children in the household. The greater participation of older people may
reflect the fact that many of them are retired and thus have more time
available. Birth year is found to be negative and significant, indicating that
younger generations are less inclined to participate in volunteering. It is
perhaps surprising that neither full-employment status nor gender is found to
have a significant effect on participation in volunteering.
In the contribution equation, the time-price variable has been disaggregated
into its two constituent components – the net hourly wage and the (inverse of)
the replacement wage – so that the corresponding impacts can be separately
identified. The only economic variable which proves to be significant in
explaining the hours supplied by volunteers is the inverse of the replacement
wage, indicating that the volunteer labour supply responds positively to the
opportunity to participate in relatively high-productivity tasks. As might be
expected, employment status has a significant negative effect on the number
of hours volunteered, reflecting the fact that those who work have fewer
hours available for other activities. The results show that older people, as well
as being more likely to participate in volunteering, also tend to supply a
larger number of volunteer hours.
In order to gauge the impact of using the subjective tax-price variable
advocated here as opposed to the objective tax-price variable used in the
earlier literature, the above models were re-estimated using an objective tax
price. In all cases, the use of the objective tax price substantially reduced the
statistical significance of the variable but did not otherwise materially affect
the pattern of results obtained.
6.5.2 Elasticities
The price and income elasticities implicit in the model coefficients reported
above are calculated and presented in Table 6.7. Separate elasticities are
calculated for the participation decision and the level of contributions. Finally
an overall elasticity estimate takes into account the combined effect of price
and income variables on participation and contribution.
In general terms, philanthropic behaviour is inelastic with respect to in-
come and price, since just about all of the overall elasticities are below one in
absolute value.
Table 6.7 Price and income elasticities for different types of philanthropic
activity
Time price
Efficiency Tax Hourly Replacement

price price wage wage Income
Giving
Participation –0.101 –0.557 0.087
Contribution –0.342 –0.366 0.193
Overall –0.459 –0.969 0.286
Volunteering
Participation –0.452 –1.135 0.303
Contribution –0.007 –0.758 –0.007 –0.543 –0.204
Overall –0.067 –0.904 –0.007 –0.543 –0.165
The own-tax-price effect for giving as well as the cross-tax-price effect for
volunteering both lie in the interval –0.9 to –1.0, indicating that philanthropic
behaviour is inelastic with respect to the tax price. The negative sign on the
cross-tax-price of volunteering provides evidence of complementarity be-
tween these two types of philanthropic activity.
The efficiency-price elasticity for giving has a substantial effect on behav-
iour, with an overall value of around –0.46. However, this is approximately
half the size of the corresponding tax-price elasticity. In the case of volun-
teering, the efficiency price has a substantive effect on participation, with a
value of around –0.45 but the overall impact is negligible.
As regards the time price, the impact of the net hourly wage is shown to be
negligible. However, the replacement wage does have a substantial effect
upon the contribution of volunteer hours, given an elasticity of –0.54. It
should be noted that the time-price effects are in all likelihood underesti-
mated. This is partly because the method used to impute the hourly wage was
such as to dampen the degree of variation in this variable. Furthermore, the
fact that the time price could not be observed for non-volunteers means that it
is not possible to take into account the impact of this price variable on the
decision to participate in volunteering.
The income elasticities are small. For giving, the overall income elasticity
lies just below 0.30, while for volunteering, the overall income elasticity is
close to zero, given that the positive elasticity for participation is more than
offset by the negative elasticity for contribution. Thus, in general, income
elasticities are substantially lower than price elasticities.
These results can be related back to the concepts of treasury efficiency and
neutrality developed by Barrett (1991) and defined in the introduction to this
chapter. The finding that the tax-price elasticity of giving generally exceeds
the income elasticity indicates that an increase in the marginal tax rate will be
non-neutral, generating a positive net impact on donations. The reason is that
the positive incentive arising from a lower tax price is large enough to
outweigh the reduction in giving that arises from a reduction in net household
income. However, this same increase in the marginal tax rate will not be
treasury-efficient. This is because donations are inelastic with respect to the
tax price, so that the additional donations resulting from the lower tax price
will not be large enough to offset the resulting loss of revenue to the Treasury.
Consequently the overall supply of public goods (those provided by charities
plus those provided by the state) will fall.
Finally, the results obtained raise an important concern about equity. Given
that the tax system is broadly progressive, whereas the income elasticity of
voluntary contributions is well below unity, a switch in government policy
away from direct grants to voluntary organizations and towards enhanced fiscal
incentives for giving will tend to be regressive in its distributional impact.
6.6 CONCLUSIONS
This chapter has identified three concepts of price which are relevant in
studying the economic aspects of philanthropic behaviour: the efficiency
price, the tax price and the time price. Furthermore, it was argued that these
price variables should ideally be measured in subjective rather than objective
terms, taking into account the effect of donor perceptions. The data indicated
that the perceived efficiency price of contributions is comparatively high,
while the time price among those who undertake volunteering is compara-
tively low.
The resulting estimates of price and income elasticities have important
implications for fiscal policy towards the sector. On the one hand, the finding
that tax-price elasticities are in general higher than income elasticities pro-
vides comfort to charities that fiscal incentives do indeed have a net positive
effect on the overall volume of contributions. On the other hand, the fact that
tax-price elasticities were found to be somewhat below unity indicates that
fiscal incentives for philanthropic giving come at the expense of reducing the
overall supply of public goods produced in aggregate by the public and
charitable sectors. This result clearly reduces the attractiveness of such fiscal
incentives from the government’s point of view and suggests that direct
grants may be a more efficient means of lending public sector support to
voluntary organizations.
The study has also revealed that other concepts of price, which fall more
directly within the influence of charity fundraisers, also have an important
impact on philanthropic behaviour. In particular, both giving and volunteer-
ing are significantly affected by the level of charity efficiency, suggesting that
improvements in this parameter will yield dividends in terms of increased
donations.
Given that donors were found to underestimate substantially the efficiency
with which charities convert monetary contributions into ‘good works’, it
probably makes most sense for charities to begin by bringing public percep-
tions of efficiency closer to reality rather than by improving the actual
underlying level of efficiency. Furthermore, the supply of volunteer labour is
positively related to the productivity of the activities undertaken by volun-
teers. This suggests that it may be possible to increase the number of volunteer
hours by paying more careful attention to the nature of volunteering opportu-
nities and in particular by ensuring that these include a greater number of
relatively high-value activities.
An important limitation of the study summarized in this chapter was the
absence of data that might provide a more direct estimate of the price of time.
As a result of this, the time price had to be inferred using procedures that
were not always entirely reliable. In addition, these procedures could only be
applied to those who were actually volunteering, thereby precluding any
estimates of the impact of the time price on the participation decision or of
the cross-price effects on giving. Future surveys of volunteer behaviour should
pay more careful attention to this issue.
7. Choosing fundraising methods
7.1 INTRODUCTION
The previous chapter showed that fiscal incentives in the UK do not have
such a powerful effect on the disposition to give. Probably more effective
than government efforts to promote charitable giving are those that charities
make themselves in the form of fundraising. This chapter estimates the re-
turns to these fundraising efforts, and explores the relative efficacy of
alternative fundraising techniques.
The charity fundraiser is a ubiquitous figure in the real world of philan-
thropic finance. Indeed, everyday experience suggests that most philanthropic
gifts are made in response to some sort of request by a charity fundraiser. A
number of voluntary sector statistics indicate that charitable organizations
have developed a substantial fundraising apparatus. To give an idea of scale,
among the top 500 fundraising charities in the UK, fundraising accounted for
8.8 per cent of total expenditure in 1996/97, amounting to nearly £370
million for the year (Pharoah and Smerdon, 1998). Indeed, between profes-
sional charity fundraisers and volunteers, fundraising is estimated to occupy
the equivalent of over 150 000 full-time employees.
Notwithstanding these fairly well-known features of voluntary sector or-
ganization, the charity fundraiser has featured surprisingly little in economic
models of philanthropic behaviour, most of which assume that giving is a
spontaneous decision of the utility-maximizing consumer. Although there
have been some notable recent attempts to remedy this omission (Andreoni,
1998), very little attention has been devoted to the question of how the efforts
of the charity fundraiser affect the potential donor’s disposition to make a
charitable contribution.
The purpose of this chapter is to incorporate the charity fundraiser into the
conventional model of philanthropic behaviour, and to conduct empirical
tests of the resulting theoretical predictions exploiting a rather unique data
source. The results provide some considerable support for the view that the
role of the fundraiser is to enhance the net private benefit – or warm glow –
associated with philanthropic giving, both by reducing transaction costs and
increasing social prestige. They also suggest that some fundraising method-
ologies are more successful than others in this respect.
147
The remainder of this chapter is organized as follows. Section 7.2 puts the
present research in the context of the existing literature on the motivation for
philanthropic giving. Section 7.3 develops a conceptual framework which
incorporates the charity fundraiser into the conventional models of charitable
giving and uses this framework to derive a number of testable hypotheses.
Section 7.4 introduces the dataset and conducts some preliminary hypothesis
tests based on simple descriptive statistics. Section 7.5 develops a statistical
model of charitable giving which controls for the intensity of fundraising
activity and the socioeconomic characteristics of the targeted population. The
predictions of this model are used to repeat the earlier hypothesis tests in a
more reliable manner. The main conclusions of the analysis are drawn out in
Section 7.6.
A more extensive discussion of the material can be found in Foster
(1999b).
7.2 MOTIVATION
According to traditional theory, philanthropic giving is purely an expression
of donors’ altruistic preferences for the supply of the associated public good
(Samuelson, 1954; Olson, 1965). As noted by Andreoni (1988, 1989, 1990),
the theory of pure altruism predicts that in large economies the free-riding
problem becomes so acute as to drive charitable donations virtually to zero.
This is because the incentive to free-ride increases as the effect of each
individual’s contribution on the overall supply of the public good becomes
vanishingly small. Such a conclusion is clearly contradicted by the existence
of a substantial voluntary sector.
Much of the recent theoretical literature on charitable donations has in
large part been motivated by the need to reconcile the traditional theory with
observed reality (Sugden, 1982; Steinberg, 1987). To this end, a number of
extensions have been made to the traditional model, and while there remain
subtle but important differences between them, what they all seem to have in
common is the incorporation of some sort of private benefit which is gener-
ated as a by-product of philanthropic giving.
Thus Sugden (1984) hypothesizes that individuals hold moral principles
which require them to contribute their fair share of effort to the financing of
public goods; that is to say, they believe it is morally wrong to free-ride as
long as other people are giving. Consequently, utility functions depend not
only on the aggregate supply of the public good, but also on the effort that
each individual contributes in relation to the effort of his peers. In a recent
empirical study Andreoni and Scholz (1998) find some limited, though not
overwhelming, evidence in support of this view.
Choosing fundraising methods 149
Andreoni (1989) shows how the difficulties associated with the traditional
model can be resolved by taking into account the fact that donors experience
a warm glow from the act of contributing that is somehow related to the
magnitude of the donation. He describes the resulting model as one of impure
altruism, to reflect the fact that people care both about the supply of the
public good and about the size of their own personal contribution to it. The
exact nature of the warm glow is not specified in the paper, but the concept is
susceptible to a wide range of interpretations. Perhaps the simplest example
of a warm glow is the case where the charitable donation is tied to the
purchase of a private good, such as charity Christmas cards or a ticket to a
charity concert. However, the warm glow could equally be taken to be refer-
ring to a sense of moral satisfaction which attaches to the philanthropic act,
and to that extent bears some relationship to Sugden’s theory.
Finally, a more recent strand of the literature has postulated that particu-
larly those donors who make relatively large-scale donations reap private
benefits in the form of social prestige (Glazer and Konrad, 1996; Harbaugh,
1998). This theory is based on the observation that charities often take special
care to publicize the identity of donors. Thus the authors argue that charitable
giving provides a signal of wealth that could be considered as a substitute for
more conventional forms of conspicuous consumption. Once again, there are
parallels with the warm glow, except that in this case the amount of prestige
attached to a particular donation will depend not only on the size of the
donation itself but also on the methods available to the recipient charity for
generating prestige.
All of these theories implicitly assume that the decision to give is taken
spontaneously by the utility-maximizing consumer. This overlooks a particu-
larly striking and prevalent feature of charitable giving as it occurs in practice,
namely the fact that philanthropic donations are invariably made in response
to some form of direct elicitation from a charity fundraiser, for example,
being targeted in a door-to-door collection or through a charity mail shot.
Indeed, the voluntary sector has developed a substantial fundraising ap-
paratus for undertaking such activities. As noted above, among the top 500
fundraising charities in the UK, fundraising accounted for 8.8 per cent of
total expenditure in 1996/97, amounting to nearly £370 million for the year
(Pharoah and Smerdon, 1998). Moreover, the UK Institute of Charity
Fundraising Managers, which covers the majority of paid professional
fundraisers active in the country, boasts a current membership of some
3000 people. However given that the bulk of fundraising effort is provided
by volunteers, these figures substantially understate the true scale of
fundraising activity. For example, during the early 1990s, fundraising ab-
sorbed some 15.5 per cent of time volunteered to charitable organizations
(Halfpenny et al., 1994). This is approximately equivalent to a further
150 000 full-time employees, or about 50 times the number of professional

fundraisers.
A recent paper by Andreoni (1998) goes some way towards acknowledging
the importance of fundraising activity by developing a model of the role of
charity fundraisers in securing seed money for large-scale capital campaigns.
The model explains the existence of fundraising activities as the solution to a
potential coordination failure that can arise when there are increasing returns
to scale in the supply of the public good. If the public good can only be
provided when aggregate donations exceed a certain high-level threshold,
donors will be unwilling to give at all unless they can be assured that comple-
mentary funds will be forthcoming from other sources. Charity fundraisers
are able to provide this assurance by initially securing a small number of
large gifts from strategic donors. This seed money subsequently serves to
elicit a large number of small donations from the general public.
However, Andreoni’s paper (1998) does not consider how the interaction
between the fundraiser and his target influences the latter’s decision to make
a contribution. This effect was found to be particularly important in Freeman’s
(1997) study of volunteering in the USA. Freeman shows that the single most
important factor determining volunteer status is whether or not a person has
been asked to volunteer. Specifically, his results are that 89 per cent of those
asked to volunteer during the course of the year did so, whereas only 29 per
cent of those who were not asked spontaneously chose to participate in
voluntary activities.
When a dummy variable indicating whether or not a person was asked to
volunteer is incorporated into a probit equation for participation in volunteer-
ing, the resulting effect is found to be much larger than that obtained from
conventional socioeconomic and demographic factors, and indeed reduces
the coefficients attached to these other variables. On the basis of these results,
Freeman (1997) hypothesizes that volunteering is a conscience good and
concludes that ‘people have a latent demand for such a good, which a request
brings to the fore, even if they would prefer to free-ride on the provision of
the good’.
Building on these findings, the purpose of this chapter is to develop a
simple conceptual model of philanthropic contributions which acknowledges
the central role of the charity fundraiser. This model generates a number of
hypotheses, which are subsequently subjected to empirical testing.
7.3 CONCEPTUAL FRAMEWORK

The standard model of impure altruism developed by Andreoni (1989, 1990)
postulates that philanthropists derive utility from their donations in two dis-
tinct ways. They benefit indirectly, in so far as their gift contributes to

expanding the supply of the philanthropic good. Although in situations where
the individual’s gift makes only a negligible contribution to the overall sup-
ply of the public good, as is the case for the vast majority of donors who
make small contributions to large charities, these indirect benefits could be
expected to be vanishingly small. Donors also benefit directly from their
charitable gifts inasmuch as these generate warm glows. These warm glows
are typically modelled as a function of the size of the donation that is made.
In the present chapter, this conventional framework will be extended by
postulating that the magnitude of the warm glow depends not only on the size
of the donor’s charitable gift, but also on the amount of fundraising effort
directed towards the individual. Whereas the size of the charitable gift is
clearly a choice variable for the individual, the exposure to fundraising effort
is determined exogenously by the charity fundraiser. Warm glows thus arise
out of the interaction between the actions of the charity fundraiser and the
reactions of the potential donor. That is not to say that warm glows would not
exist at all in the absence of interventions by fundraisers, but rather that the
initiatives of fundraisers serve to amplify the net private benefits of giving.
There are at least two different mechanisms that explain how the actions of
a charity fundraiser might be expected to magnify the warm glow experi-
enced by the donor. The first of these is that charitable fundraisers may
reduce the transaction costs of making a donation. They do this both by
passing on information about charitable causes and by providing a simple
procedure for giving, such as handing over a few coins, placing a cheque in
an envelope or ringing a toll-free number. By reducing the private costs
associated with making a donation, fundraising efforts serve to increase the
net private benefit of giving.
The second way in which fundraising efforts may enhance the warm glow
of giving is that they often create situations where the donor’s philanthropic
actions can be observed by others and thus attract some degree of social
recognition or prestige. This prestige mechanism could also function in re-
verse. For example, while there may not be a great deal of prestige associated
with placing a few coins in a plate, a person who failed to do so might well
meet with social disapproval. As noted by Becker (1974), ‘apparent “charita-
ble behaviour” can also be motivated by a desire to avoid the scorn of others’.
Thus the encounter with a charity fundraiser may elicit in the potential donor
a sense of shame, embarrassment or remorse about not making a charitable
contribution. In this case, the act of giving provides relief from the moral or
social discomfort generated by the charity fundraiser.
The level of fundraising effort targeted at any given individual can be
measured along two dimensions. The first is the number of times the indi-
vidual is approached by a charity fundraiser, and the second is the quality of
the fundraising methodology used. High-quality methods are defined as those

that, other things being equal, have the greatest capacity to enhance the warm
glow of giving. For example, it appears likely that any fundraising method
entailing direct face-to-face contact with the potential donor will be of higher
quality in this sense than a fundraising method entailing only remote contact
with the potential donor, such as via mail appeals, or newspaper or television
advertisements.
There are two reasons for thinking that this might be the case. The first is
that direct fundraising methods are likely to represent lower transaction costs
for donors than remote methods. This is because the former tend to involve
simply handing over cash, whereas the latter tend to require some sort of
formal financial transaction such as writing a cheque, which involves addi-
tional time and effort on the part of the donor.
The second reason for thinking that direct fundraising methods may be
more effective at generating warm glows is that they are more likely to be
witnessed by third parties, and thus more effective in generating social
recognition. At the very least, the charity fundraiser will be there to observe
whether or not the person gives, and in some cases also how much they
give. Furthermore, in many cases, direct fundraising methods exploit social
situations such as the workplace, pub or church. Thus the philanthropic
behaviour of potential donors is observed by a wider peer group, whether it
be fellow colleagues, drinkers or worshippers. In contrast to this, remote
giving tends to be more of a private decision that may not be observed by
anyone else.
The conclusion of all this is that a person who gives by a direct fundraising
method will experience a larger warm glow, and thus a higher utility level
than a person who gives by means of a remote fundraising method. This
implies that, other things being equal, those approached by direct methods
will be more inclined to give than those approached by remote methods.
As a basis for testing these conjectures, two null hypotheses are estab-
lished. According to these, the choice of fundraising method, whether direct
or remote, will have no significant effect either on the probability of obtain-
ing a charitable gift or on the size of the gift obtained.
1. Participation null hypothesis (H01): direct fundraising methods are no

more successful than remote fundraising methods in eliciting charitable
contributions.
Alternative participation hypothesis (HA1): direct fundraising meth-
ods are more successful than remote fundraising methods in eliciting
charitable contributions.
2. Size of the gift null hypothesis (H02): there is no difference in the size of
the donations generated by direct and remote fundraising methods.
Alternative size of the gift hypothesis (HA2): donations generated by

direct fundraising methods are larger than those generated by remote
fundraising methods.
Failure to reject both of these null hypotheses would contradict the overall
theory that the choice of fundraising method affects the magnitude of the
warm glow obtained. Rejection of the first hypothesis, without being able to
reject the second hypothesis, would lead to the conclusion that the choice of
fundraising method does matter, but that it matters only in determining the
probability of participation. Finally, rejection of both null hypotheses in
favour of the corresponding alternatives would indicate that the choice of
fundraising method influences both the probability of participation and the
size of the resulting gift.
7.4 RESULTS OF SIMPLE HYPOTHESIS TESTS

The hypotheses advanced in the previous section are tested making use of data
on charitable donations contained in the Individual Giving Survey (IGS) for the
period 1990–93 (Halfpenny et al., 1992, 1993, 1994). The IGS was an annual
repeated cross-section survey of philanthropic behaviour in the UK funded by
the Charities Aid Foundation up to 1993. It covered about 1000 individuals
each year, chosen to be representative of the UK population. A structured
interview conducted in the respondent’s home was used to collect information
about philanthropic activities during the month before the survey.
The IGS has a number of unique features that make it particularly well
suited to testing the hypotheses stated above, and more generally to examin-
ing the role of fundraising in generating charitable gifts. First, the IGS
distinguishes between 11 different fundraising methods commonly used for
eliciting philanthropic contributions. These are street collections, door-to-
door collections, church collections, sponsoring schemes, shop-counter
collections, pub collections, work collections, advertisement appeals, tele-
vision appeals, mail appeals and telephone appeals. Second, for each of these
fundraising methods, the IGS establishes how many times the individual was
approached, how many times they actually gave, and the total value of dona-
tions. Third, the IGS distinguishes between a number of different methods of
obtaining charitable donations by means of selling conventional goods and
services, such as raffles, charity events, charity shops, jumble sales and
charity goods catalogues. However, attention here is confined to fundraising
methods that aim to elicit gifts for which there is no tangible private return.
One limitation is that the survey does not record how much was donated on
each giving occasion, but only the total donations for each fundraising method.
This creates a problem when it comes to comparing methods that are used
with differing intensities, in that only average gifts can be compared. To the
extent that the size of the marginal contribution is likely to decline with each
successive gift, the fact that one method elicits a higher average gift than
another may simply be attributable to its lower frequency of use rather than to
any intrinsic characteristic of that method.
A second limitation of the dataset is the possibility that respondents may
have had the incentive to overstate the extent of their donations in order to
appear charitable to the interviewer. This would have the effect of biasing the
survey data upon which this analysis is based. It is not possible to prove
whether or not this effect is actually present in the data. However, even if it
were, the resulting bias should only affect the absolute levels of the variables
rather than their relative values across different fundraising methods, and the
tests reported in this chapter are concerned solely with relativities.
Table 7.1 provides an initial overview of the relationship between fundraising
approaches and giving behaviour. The first point to note is the wide reach of
fundraising activity in that 74.1 per cent of respondents had been approached
by a charity fundraiser during the preceding month. Furthermore, fundraising
approaches seem to enjoy a very high success rate: among those who were
approached at least once, as many as 91.2 per cent gave money at least once.
On the other hand, spontaneous unsolicited giving is very much a minority
activity; only 1.9 per cent of those who were not approached by a charity
fundraiser during the preceding month had chosen to make a donation.
Table 7.1 Cross-tabulation of fundraising approaches and giving

behaviour (% of the sample)
Gave at least once?
No Yes
Approached at least once? No 25.5 0.5

Yes 6.5 67.6
Table 7.2 provides a more detailed breakdown of donating behaviour

according to which of the 11 fundraising methods was used. The table
presents a number of different summary statistics for each method. The first
column refers to the response that respondents made to an attitudinal ques-
tion enquiring how likely they were to give when approached in each of
these different ways. The results are coded on a Likert scale ranging be-
tween +2 (signifying ‘very likely to give’) and –2 (signifying ‘very unlikely
Table 7.2 Summary statistics for different fundraising methods
Attitude towards Number of Probability of

fundraising approaches making a
method (#) per person (#) gift (%) Donation (£) Revenue (£)
Mean Std. err. Mean Std. err. Mean Std. err. Mean Std. err. Mean Std. err.
Direct methods
Street collection 0.54 0.02 0.87 0.04 84.32 1.12 0.67 0.03 0.51 0.02
Door-to-door collection 0.68 0.02 0.63 0.02 85.57 1.01 1.76 0.65 1.36 0.54
Church collection 0.27 0.03 0.50 0.03 95.41 0.92 3.27 0.54 2.62 0.36
Sponsor scheme 1.17 0.02 0.41 0.02 93.73 0.83 2.41 0.11 2.28 0.11
155
Shop-counter collection –0.14 0.02 0.34 0.02 77.78 2.12 0.43 0.03 0.33 0.03
Pub collection –0.29 0.03 0.25 0.02 92.22 1.47 0.91 0.09 0.80 0.07
Work collection 0.22 0.03 0.10 0.01 91.04 2.02 3.82 1.42 3.17 1.17
Remote methods
Appeal by advertisement –0.61 0.02 0.14 0.02 24.84 3.50 9.96 2.97 2.18 0.80
Television appeal –0.06 0.03 0.11 0.01 48.39 3.18 7.57 0.95 3.17 0.48
Mail appeal –1.05 0.02 0.21 0.02 27.21 2.60 9.16 1.15 1.97 0.33
Telephone appeal –1.47 0.01 0.03 0.01 23.26 6.52 16.28 7.69 2.17 1.05
Overall
Direct methods 0.38 0.01 3.05 0.08 91.55 0.43 1.86 0.13 1.46 0.08
Remote methods –0.80 0.01 0.48 0.03 37.54 2.00 8.85 0.80 2.69 0.31
All methods –0.06 0.01 3.58 0.09 91.18 0.60 1.94 0.13 1.40 0.08
to give’). The second column reports the average number of times a re-
spondent was approached by each method, while the third column gives the
probability that those who were approached at least once by each method
went on to make a donation. The average size of the donation is reported in
the fourth column, and the average revenue per approach in the fifth and
final column.
The final row of the table, which reports the summary statistics for all
fundraising methods, confirms the earlier impression that fundraising is a
ubiquitous phenomenon. On average, the survey respondents were approached
by a fundraiser 3.6 times during the course of the month, or about once every
eight and a half days. About 91.2 per cent of those approached made a
donation and the average value of such donations was £1.94. Overall this
implies an average return to each fundraising approach of £1.40.
Summary statistics for direct and remote fundraising methods provided in
Table 7.2 testify to some striking differences in these two different fundraising
methods. The direct methods are characterized by relatively high probabili-
ties of making a gift (in excess of 90 per cent) and relatively low average cash
donations (of less than £2.00). By contrast, the remote methods are character-
ized by relatively low probabilities of making a gift (less than 40 per cent)
and relatively large average cash donations (in excess of £8.00). Interestingly,
the lower disposition to make a donation when approached by a remote
method is reflected in the lower scores that these methods received on the
attitudinal question, where respondents were asked subjectively to evaluate
their likelihood of making a donation. Thus direct methods score an average
of 0.38 while remote methods score only –0.80.
In order to test whether these differences are statistically significant, and
the extent to which they conform to the conjectures stated above, a series of
more formal hypothesis tests is conducted.
The first hypothesis stated above concerns possible variations in the prob-
ability of obtaining a gift using different fundraising methods. This is tested
by creating a variable (∆P), which is defined as the difference between the
probability (P) of obtaining a gift from the direct (D) versus the remote (R)
methods. The null hypothesis, which is stated formally as (H01), stipulates
that this difference should be equal to zero, while the alternative (HA1) holds
that this difference should be strictly positive. Using the difference variable, a
simple t-test results in the rejection of the null hypothesis in favour of the
alternative. This indicates that there is a significantly higher probability of
success (the average difference is 41.9 per cent) when direct fundraising
methods are used as opposed to remote methods.
(H01) ∆ P = PD − PR = 0 (HA1) ∆ P = PD − PR > 0 (7.1)

The second hypothesis stated above concerns possible variations in the

average donations received from different fundraising methods. Once again,
this is tested by creating a difference variable (∆C) for the average donations
(C/N) resulting from the two types of fundraising methods. In this case, the
null hypothesis (H02) requires that this difference be equal to zero, while the
alternative hypothesis (HA2) holds that the difference is strictly positive.
Once again, a t-test procedure based on the difference variable results in
rejection of the null hypothesis. There are significant differences in the size
of the average gift obtained from direct versus remote fundraising methods.
However, curiously this difference does not run in the direction implied by
the alternative hypothesis (HA2). In fact, rather, the average cash gifts ob-
tained from direct fundraising methods are significantly smaller than those
obtained from remote fundraising methods. The average difference is –£6.76.
CD CR CD CR
(H02) ∆ C = − =0 (HA2) ∆ C = − >0 (7.2)
ND NR ND NR
Further confirmation of this result comes from comparing the frequency

distributions of the average donations obtained by the two methods. These
are illustrated graphically in Figure 7.1. From the figure it is clear that the
distribution of average gifts generated by direct fundraising is heavily skewed
towards smaller values, with the modal point occurring in the size category of
45
40
35
30
25
20
15
10
5
0
<£1 £1–2 £2–3 £3–4 £4–5 £5–10 >£10
Direct Remote
Figure 7.1 Percentage frequency distribution of average donations for

different fundraising methods (£)
gifts worth less than a pound. By contrast, the distribution of average gifts
from remote fundraising methods is skewed towards larger donations. The
overall distribution is bimodal with a local mode occurring in the size band
£2 to £3 and the overall mode occurring in the size band £5 to £10. Using the
Kolmogorov–Smirnov procedure (Daniel, 1990), it can be shown that the
differences between the two distributions are in fact statistically significant.
Thus, overall, the results of the hypotheses tests provide strong support for
the general view that the choice of fundraising method materially affects the
disposition to give. While the higher success rate of direct fundraising meth-
ods is in line with prior conjectures, the fact that remote methods tend to
generate larger gifts provides something of a puzzle. One possible explana-
tion is that the larger transaction costs associated with making a remote gift
via a formal financial transaction serve to discourage smaller donations.
People rarely write (and institutions rarely accept) a cheque for a sum as
small as 50 pence, whereas they often make cash purchases of this size. The
frequent use of cash in direct giving may therefore account for the prevalence
of smaller gifts by this means.
7.5 CONTROLLED HYPOTHESIS TESTS
7.5.1 Potential Biases in Previous Results
The results presented in the preceding section are potentially misleading for a
number of reasons. In particular, on the basis of these simple tests, it is not
possible to say whether the differences in performance observed between the
two types of fundraising methods are attributable to the methods themselves.
They may instead be due to some other difference, which happens to be
correlated with the use of one or other of the two methods.
One possible reason why the earlier results may not be entirely trustwor-
thy is that the two fundraising methods are used with differing degrees of
intensity. For example, if people tend to be targeted more frequently by
direct fundraising methods than by remote fundraising methods, it is not
surprising that the average donation reported would be lower for direct
methods than remote methods. Table 7.2 confirms that direct methods are
indeed more intensively used than remote methods, with an average of 3.05
approaches per month for the former and 0.48 approaches per month for the
latter.
A second plausible explanation is that the direct methods have a higher
success rate not because of their superior ability to generate warm glows but
because their use tends to be concentrated in segments of the population that
have a higher pre-existing disposition to give. For example, it may be that
direct methods are more likely to be used on relatively high-income individu-

als and that high-income individuals are more likely to make charitable
donations. In order to test for this effect a number of ordinary least squares
(OLS) regression models were estimated relating the logarithm of the number
of fundraising approaches to a range of exogenous factors including socio-
economic characteristics of the targeted respondents, regional and population
density factors, and seasonal dummies.
The results are reported in Table 7.3, which presents separate models for
direct and remote fundraising methods. The overall explanatory power of
these models is comparatively low, with adjusted R2 statistics ranging be-
tween 6 per cent and 7 per cent. However, the statistics for the F-test, which
lie in the range 8 to 12, indicate that the regressions are statistically signifi-
cant overall. Moreover, the presence of a number of statistically significant
Table 7.3 Ordinary least squares regression models for the number of
fundraising approaches
Direct methods Remote methods
Coeff. t-stat. Coeff. t-stat.
Log (household income) 0.119 3.878 0.075 4.047

Years of education 0.121 5.694 0.077 6.056
Sex –0.024 –0.528 0.007 0.026
Age 0.019 0.915 –0.020 –1.592
Birth year –0.022 –1.077 0.022 1.742
Full-time employed –0.095 –1.669 –0.020 –0.584
Part-time employed 0.039 0.578 –0.013 –0.325
North of England –0.107 –2.047 –0.085 –2.763
Scotland or Wales 0.067 1.095 –0.108 –3.081
Southeast of England –0.180 –3.056 –0.020 –0.589
Population density –0.051 –2.838
Winter season 0.089 2.525
Autumn season 0.051 1.427
Spring season 0.031 0.877
Constant –3.672 –1.946 0.211 0.180
Correlation 0.267 0.257

F-statistic 11.20 8.48
Adjusted R2 0.07 0.06
Observations 1615 1578
coefficients indicates that the targeting process used by charity fundraisers is

far from random.
A comparison of the separate models estimated for direct and remote
fundraising methods reveals some substantial differences in the magnitude of
coefficients and in the pattern of significance of the explanatory variables.
For example, younger generations are more likely to be approached by re-
mote methods, while those in full-time employment are less likely to be
approached by direct methods. Furthermore, there are some significant re-
gional differences. Direct methods are less likely to be used in the southeast
of England, whereas remote methods are less likely to be used in Scotland or
Wales. In both cases, the point of reference is the Midlands region.
A Wald test was conducted to test the null hypothesis that there are no
significant differences between the types of people approached using direct
versus remote fundraising methods (Greene, 1993). This hypothesis was
strongly rejected, indicating that the two different types of methods tend to
reach very distinct constituencies.
A third possible explanation for the results reported in the previous section
is that the use of these two different fundraising methods is not uniform
across different types of charities. For example, if direct fundraising methods
tend to be used disproportionately by more popular charities, this could
provide an alternative explanation of why these techniques tend to meet with
a higher rate of success. The presence of this effect is much harder to
establish, because it is very difficult to observe and measure a subjective
variable such as charity popularity, which may in part be related to the nature
of the good cause and in part to the nature of the specific charitable organiza-
tion (its age, size, reputation and so on).
For all of these reasons, the present study will not attempt to control for the
popularity of the cause in determining the response to different types of
fundraising methods. This position draws some support from a recent study
of fundraising methods (Sargeant and Kaehler, 1998), which finds that the
performance of various different methods does not vary significantly accord-
ing to the nature of the charity.
7.5.2 Controlling for Extraneous Factors
In view of the issues identified above, the subsequent analysis of fundraising

performance will control for differences in the nature of the targeted popula-
tion and differences in the number of fundraising methods. On the basis of
the conceptual framework developed above, this can be done by estimating a
donations equation. The corresponding statistical model is used to isolate the
effect of fundraising method on giving behaviour, while controlling for other
influencing factors, including the number of fundraising approaches and a
range of socioeconomic and demographic factors as well as economic vari-

ables such as price and income. The two price variables used in the giving
equation are described extensively in Chapter 8 below. Briefly, the efficiency
price represents the rate at which charities absorb donations in administrative
expenditures, while the tax price reflects the presence of fiscal incentives for
charitable giving.
The quality of the fundraising methods directed towards each individual
respondent (Q) is measured by an index variable defined in equation (7.3)
below, which gives the ratio of direct (D) versus remote (R) approaches. A
high value of this index (greater than one) indicates that the respondent was
approached predominantly by means of direct fundraising methods, and vice
versa. The average value of the index in the sample population was found to
be 3.39, indicating the preponderant use of direct fundraising methods.
D + 1
Q= (7.3)
 R + 1
Donations are modelled using a Heckman selectivity framework. The prin-

cipal strength of this model is that it takes into account the fact that different
processes may be at work in determining the decision to participate in chari-
table giving and the size of the donation made. This is achieved by estimating
separate equations for the decision to participate in philanthropic giving
versus the decision of how much to contribute (Greene, 1993).
In the context of the Heckman selectivity model, the two hypotheses ad-
vanced above can be restated as hypotheses about the sign of the coefficients
on the quality index (Q). In the case of the first hypothesis, the null (H01)
requires that the coefficient on the quality index in the participation equation
(P) be equal to zero, while the alternative (HA1) entails that it be strictly
positive.
(H01) β QP = 0 (HA1) β QP > 0 (7.4)
In the case of the second hypothesis, the null (H02) requires that the
coefficient on the quality index in the contribution equation (C) be equal to
zero, while the alternative (HA2) entails that it be strictly positive.
(H02) β QC = 0 (HA2) β QC > 0 (7.5)
Table 7.4 reports the results of the corresponding Heckman selectivity

model. The first point to note is that the two variables that capture the number
and quality of fundraising methods are highly significant, both in the partici-
pation equation and in the donations equation. Moreover, in the participation
Table 7.4 Selectivity models for all philanthropic gifts with controlling for
fundraising effort
Participation Contribution
Coeff. t-stat. Coeff. t-stat.
Log (efficiency price) –0.824 –2.077 –0.983 –4.444

Log (tax price) –0.311 –0.569 –0.200 –0.713
Log (household income) 0.143 1.168 0.202 3.234
Years of education 0.006 0.073 0.087 2.372
Sex –0.373 –2.283 0.006 0.074
Age –0.077 –0.946 –0.029 –0.710
Birth year 0.080 0.979 0.032 0.801
Full-time employed 0.041 0.194 0.094 0.908
Part-time employed 0.079 0.289 0.200 1.703
North of England 0.032 0.181 –0.010 –0.106
Scotland or Wales 0.494 2.117 0.002 0.016
Southeast of England 0.309 1.543 –0.010 –0.099
Importance of religion 0.045 0.852
Log (approaches) 1.627 12.407 0.777 11.616
Log (quality index) 0.953 8.422 0.185 2.505
Constant 4.459 0.598 3.060 0.824
Rho 0.27
Sigma 1.08
Log-likelihood –1554.83
Observations 1110
equation, the coefficients for these two variables are substantially larger than
for any of the others.
The positive coefficient on the quality index in the participation equation
indicates a clear rejection of the null hypothesis stated as (H01): controlling
for all other factors, direct fundraising methods do appear to be more suc-
cessful in eliciting charitable gifts. Moreover, the quality index continues to
carry a positive coefficient in the contributions equation, a result that consti-
tutes a clear rejection of (H02). When all other factors are taken into account,
it turns out that the more that people are approached predominantly through
direct fundraising methods, the larger the gifts that they tend to make. This
result is at variance with the conclusion drawn from the analysis of the raw
data. It suggests that the larger average gifts observed from remote fundraising
Table 7.5 Estimated marginal effects of fundraising
Participation (%) Contribution (£) Overall (£)
Aggregate approaches 12.39 0.92 0.81

(0.47) (0.05) (0.05)
Separate approaches
Direct 20.11 1.11 0.93
(0.74) (0.07) (0.05)
Remote 2.38 0.27 0.27
(0.22) (0.03) (0.04)
methods may have been a consequence of the relatively low incidence of

remote fundraising approaches, or of differences in the generosity of the
population that tends to be reached by means of remote methods.
With a view to quantifying the size of these differentials, Table 7.5 reports
the marginal effect for each fundraising approach. In overall average terms,
an additional fundraising approach adds just over 12 per cent to the probabil-
ity of obtaining a charitable gift, and the expected value of that gift is £0.92.
Taking these two effects in combination, the expected return from an addi-
tional fundraising approach is £0.81.
Separating out the marginal effects between direct and remote fundraising
methods reveals that the marginal effect of a direct approach on the probabil-
ity of securing a contribution is of the order of 20 per cent. The corresponding
value for a remote approach is almost an order of magnitude smaller, at
around 2 per cent. The value of the marginal gift is £1.11 for direct methods,
and substantially smaller at £0.27 for remote methods. Thus the overall
expected marginal gain of an additional approach varies between £0.93 for
direct approaches and £0.27 for remote approaches.
These marginal effects, which are reported at the sample averages, conceal
some informative variations across the sample. In particular, from a charity
perspective it is interesting to enquire how rapidly the marginal pay-off
declines as the intensity of fundraising effort is increased. To this end, Figure
7.2 plots the marginal effect on the probability of participation against the
number of approaches of each kind that have already been made to the
corresponding individuals.
A number of interesting conclusions emerge. As might be expected, these
marginal effects decline with additional fundraising approaches. This is no
more than to say that the second derivative of the donation function with
respect to fundraising effort is negative. However, the nature of this decline
differs markedly between direct and remote fundraising methods.
50 Direct
Remote
Marginal effect on probability
40
of giving (%)
30
20
10
0
0 1 2 3 4 5 6 7 8 9 10 >10
No. of approaches
Figure 7.2 Marginal effect of an approach on the probability of a gift

against number of approaches
For direct methods, the probability of obtaining a gift is initially very high.
It is estimated that nearly half of those who have not been approached by a
fundraiser in the preceding month would make a gift if this were elicited
directly. However, this disposition to give tails off dramatically as the number
of approaches increases. By the second request, the probability of success is
little over 10 per cent, while by the fourth request it has tailed off virtually to
zero. The case of remote methods provides a strong contrast. The probability
of success is initially very much lower, at around 10 per cent; however, the
rate of decline is much more gradual, so that even by the tenth approach the
probability of success still falls just short of 5 per cent.
This difference can probably be explained as follows. Owing to the higher
probability of making a gift by direct methods, people are very likely to have
made at least one such gift once they have been approached a couple of
times. They may consequently be satiated with respect to charitable giving
for that time period, and hence would fail to respond to any subsequent
requests. Given the much lower probability of an immediate positive re-
sponse to a remote approach, it is quite possible that even people who have
been approached half a dozen times may not have made any charitable gifts.
Hence the satiation effect is far less prevalent.
7.6 CONCLUSIONS
The current chapter began with the observation that conventional economic
models of philanthropic behaviour have tended to overlook the pervasive
presence of the charity fundraiser. With a view to overcoming this deficiency,
fundraising effort was incorporated as a new component within the conven-
tional analytical framework for charitable giving. The expanded model
postulates that fundraising effort interacts with individual donations in the
production of the warm glow or private benefit associated with philanthropic
behaviour. Two ways in which fundraising effort was thought to enhance the
warm glow were by reducing the transaction costs of giving and increasing
the prestige associated with making a donation.
It was argued that fundraising effort can be measured along two dimen-
sions: the number of times a person is approached by a charity and the
quality of the fundraising method used in that approach. Two types of
method were distinguished: direct methods, which involve face-to-face con-
tact between the fundraiser and the potential donor, and remote methods,
which do not involve face-to-face contact. It was hypothesized that direct
methods had the greater potential for enhancing the warm glow of charita-
ble giving and thus that they would be more successful in eliciting
philanthropic gifts. Access to a unique dataset – containing very detailed
information on exposure to fundraising effort – provided the opportunity to
put this hypothesis to the test.
An analysis of the raw data revealed that direct fundraising approaches
were indeed more likely to be successful, but that the resulting gifts were
on average substantially smaller than those obtained from remote fundraising
drives. However, it was recognized that these results could be biased by a
number of factors. First, there was the fact that direct fundraising methods
are used much more intensively than remote methods. Second, it was found
that the two methods were reaching significantly different populations,
which could be expected to have different underlying dispositions to give.
In order to control for these confounding influences, donations were mod-
elled as a function of both the number and quality of fundraising approaches
as well as a wide range of other socioeconomic and demographic factors. On
the basis of these models, direct methods were still found to be more success-
ful in eliciting contributions, but in this controlled context they were also
found to generate larger contributions. Thus the overall marginal effect of an
additional fundraising approach was found to vary between £0.93 for direct
methods and £0.27 for remote methods. The implication is that the earlier
result derived from a failure to take into account the effect of other influences
on philanthropic behaviour.
These results provide considerable support for the theoretical framework

advanced, suggesting that fundraising expenditures are an essential input in
the creation of warm glows.
8. Targeting donors
8.1 INTRODUCTION
Fundraising expenditures by charities are akin to advertising expenditures by
private firms. Both represent an attempt to persuade individuals to part with
their money, whether in return for consumer goods or more intangible philan-
thropic benefits. Both are typically undertaken in a competitive environment,
where shifting demand from one supplier to another may be as much of an
issue as raising demand overall.
Anecdotal evidence suggests that, like advertising executives, charity man-
agers pay great attention to where they target their persuasive efforts. Some
individuals have a higher predisposition than others towards purchasing cer-
tain types of goods, or contributing to certain types of causes. Where this
predisposition is correlated with observable socioeconomic and demographic
characteristics, an opportunity is created to increase the returns from
fundraising (or advertising activities) by targeting campaigns on particular
segments of the population.
The literature on charitable fundraising to date has focused on the question
of whether charity managers aim to maximize gross or net revenues (Weisbrod
and Dominguez, 1986). However, little attention has been given to the role of
targeting in fundraising activity, and consequently a number of interesting
questions remain unanswered. What factors influence the choice of targeting
strategy adopted by any particular institution? How far is it desirable to take
the targeting process? To what extent does targeting succeed in dissipating
the competition for funds between rival organizations?
This chapter extends the existing literature on fundraising by incorporating
targeting as a basic decision-making variable for the charity manager. Using
a dataset on environmental group membership in the UK, a number of statis-
tical models are estimated which identify how the decision to join an
environmental group correlates with a number of observable characteristics
that could be used as a basis for targeting. The models are then used to
simulate alternative fundraising strategies as well as to examine their poten-
tial implications for competition between charity fundraisers.
Targeting is found to produce a dramatic improvement in the probability
that a given fundraising approach will be successful, but this comes at the
167
expense of reducing the size of the total pool of potential fundraising targets.
The optimum balance between these two factors is shown to depend on the
size of the environmental organization, the objective function of its manag-
ers, and the basic parameters of the cost and revenue functions. Although
there is some evidence of market segmentation among environmental chari-
ties, this does not appear to be sufficient to succeed in softening the competition
for potential members between different types of groups.
The discussion is organized as follows Section 8.2 sets out the conceptual
framework for the analysis. Sections 8.3 and 8.4 provide a descriptive over-
view of UK environmental groups and the data on those groups that are
available from the British Social Attitudes Survey (Brook et al., 1991). Sec-
tion 8.5 presents the statistical models of environmental group membership,
while the simulations are conducted in Section 8.6. Conclusions are provided
in Section 8.7.
A more extensive discussion of the material can be found in Foster (1999a).
8.2 CONCEPTUAL FRAMEWORK

An important area of research in the economics of non-profit institutions has
been to provide models of charitable fundraising activity. One strand of this
literature examines how the objective function of charity managers would
affect their optimal fundraising strategy (Steinberg, 1986; Weisbrod and
Dominguez, 1986). Analogous to advertising expenditures for private firms,
fundraising effort generates costs for non-profit organizations while at the
same time increasing their revenue from donations.
Drawing on the theory of the firm, this literature identifies two polar cases
of managerial objectives. First, there are managers who aim to maximize the
net revenues from fundraising – and thus the overall level of charitable
services provided. Second, there are managers who aim to maximize the total
revenues from fundraising – and thus the overall size of the charitable organi-
zation (Baumol, 1962). Empirical studies of fundraising behaviour have found
evidence of both strategies being adopted by organizations operating in dif-
ferent sectors of charitable activity (Weisbrod and Dominguez, 1986; Posnett
and Sandler, 1989; Khanna et al., 1995).
Indeed, both of these objectives would appear to be plausible depending on
the type of charity. On the one hand, net revenue maximization is particularly
appropriate for charities whose primary aim is to provide direct assistance to
needy groups. These needy groups are effectively analogous to the sharehold-
ers of a private firm in that they have residual claim on the services financed
by the stream of net revenues generated by the charity managers. Thus the
larger the net revenues generated by the charity, the higher the volume of
Targeting donors 169
philanthropic services that can be financed. On the other hand, gross revenue
maximization may make sense for charities which are primarily concerned
with campaigning and whose political clout may be enhanced by the ability
to demonstrate a large following of supporters. Even charities that are pri-
marily service providers may lean towards the maximization of gross revenues
if managerial objectives threaten to eclipse the interests of the needy groups,
or ultimate shareholders.
Clearly, gross revenue maximizers will tend to allocate a higher level of
effort to fundraising than net revenue maximizers. The reason is that gross
revenue maximizers care primarily about raising additional revenues without
much regard to the associated costs, as long as some breakeven constraint is
met. They will consequently continue to pursue fundraising initiatives even
when the additional revenues secured are comparatively low in relation to the
additional costs incurred. By contrast, net revenue maximizers will be more
concerned to minimize the costs of fundraising, which directly reduce net
revenues. Consequently, they will not want to pursue fundraising efforts
beyond the point where the additional costs incurred exactly offset the addi-
tional revenues raised.
Both the costs and benefits of fundraising will depend on the extent to
which resources are well targeted towards those segments of the population
with the highest propensity to make financial contributions. As regards costs,
fundraising expenditures are likely to be proportional to the number of poten-
tial donors targeted. As regards revenues, targeting could be expected to have
two contrary effects on the yield from fundraising expenditures. On the one
hand, the application of more stringent targeting criteria will increase the
probability that each person targeted responds positively with a donation. On
the other hand, as more stringent targeting criteria are applied, the pool of
potential donors will shrink, since fewer and fewer people will satisfy the
corresponding conditions. For example, imagine that the targeting criterion is
gross annual income. The higher the targeting threshold of gross annual
income selected by the charity fundraiser, the higher could be expected to be
the probability of obtaining a positive response, but at the same time the
smaller would be the pool of individuals who met the criterion.
This reality faces fundraisers with a basic trade-off between raising the
probability of successful strike (P) and reducing the number of potential
targets (N). The optimal point of balance between P and N essentially de-
pends on the cost and revenue structure of fundraising. In particular, the
higher the ratio of the marginal cost to the marginal revenue of fundraising,
the more attractive it becomes to target fundraising efforts. The marginal cost
of fundraising is the amount that it costs to approach an additional person (for
example, via a mail shot), while the marginal revenue of fundraising is the
additional amount of money that a new member could be expected to contrib-
ute. When the marginal cost of fundraising is large relative to the marginal
revenue, the cost of failure is high and it is therefore important to focus
efforts on those segments of the population that are likely to be most respon-
sive. However, when the marginal cost of fundraising is small relative to the
subscription fee, the cost of failure is low and so it makes sense to approach
as many potential members as possible.
Furthermore, for any given value of the marginal cost–revenue ratio, it can
be shown that net revenue maximizers will find it in their interest to pursue
targeting to a greater extent than gross revenue maximizers. This is because
maximizing gross revenue is broadly equivalent to maximizing the number of
donors, and hence it is less attractive to apply criteria that reduce the size of
the potential target population. Net revenue maximizers, on the other hand,
are concerned to keep down the costs of fundraising. Targeting helps them to
do this by reducing the number of people that have to be approached to raise
a given amount of revenue.
Finally, the issue of targeted fundraising links with another important
strand in the existing literature. This concerns the problem of excessive
fundraising expenditures when charities must compete for donations from the
same target population (Rose-Ackerman, 1982). The industrial organization
literature suggests that product differentiation can be used to attenuate this
kind of competition (Tirole, 1988). In the traditional models of product
differentiation, consumers have heterogeneous tastes for product characteris-
tics. By locating themselves at a particular point in characteristics space,
firms can act as local monopolists, breaking up the overall market into a
number of distinct and homogeneous segments of consumers (Hotelling,
1929). Successful product differentiation by non-profit organizations would
similarly have the effect of reducing competition for donations by enabling
charities to target their fundraising expenditures on their respective market
segments, thereby raising their returns from fundraising expenditure.
In order to investigate both of these issues, this chapter will focus on one
particular sector of charitable activity, namely environmental groups (Richer,
1995). The probability of joining an environmental group is modelled as a
function of the socioeconomic and demographic characteristics thought to
affect the disposition to give. Using the models, it is possible to conduct
simulations of the effect of alternative targeting criteria on the probability of
obtaining an environmental group member, and on the size of the available
target group. Furthermore, the models shed light on the extent to which
different types of environmental groups may be competing for members from
the same target populations.
8.3 ENVIRONMENTAL GROUPS IN THE UK

The charitable environmental sector in the UK grew rapidly during the 1980s
and is now substantial in size. Recent research on the scale of voluntary
sector activity in the sphere of environmental conservation found that there
were 4000 groups with an aggregate expenditure of £634 million (Fenyo et
al., 1993). To put this figure into context, it is estimated that non-profit
organizations account for between 36 per cent and 69 per cent of national
spending on environmental conservation, depending on which measure of
non-profit expenditure is used (Ecotec, 1993; Fenyo et al., 1993). However,
shares of total expenditure probably understate the true significance of envi-
ronmental groups, since to the extent that the activity of these groups focuses
on campaigning it is designed to bring about an increase in environmental
expenditure by other bodies.
To a greater degree than other voluntary organizations, environmental groups
have tended to organize themselves as clubs, using annual membership fees
as an important fundraising vehicle. Fenyo et al. (1993) estimate that 20 per
cent of the income of environmental organizations derives from this source.
Table 8.1 provides details of membership numbers and subscription income
for some of the largest of the UK environmental charities, and indicates that
they raise 25–50 per cent of their income from subscriptions and assign 10–
25 per cent of their expenditure to fundraising. All of these groups experienced
significant growth in their membership during the 1980s, the most notable
examples being the National Trust (NT), the Ramblers’ Association (RA) and
the Royal Society for the Protection of Birds (RSPB), whose members ap-
proximately doubled over the decade, and the World Wide Fund for Nature
(WWF) whose membership more than tripled over the same period.
Table 8.1 also illustrates how these environmental groups differ with re-
spect to their portfolio of activities and thus provides some indication of the
extent to which they are engaging in product differentiation.
Groups such as the Council for the Protection of Rural England (CPRE),
Friends of the Earth (FoE) and Greenpeace (GP) are largely concerned with
campaigning activities aimed at promoting environmental protection, whether
on a domestic or global level. The WWF also has an important campaigning
function relating to the specific problem of biodiversity, but differs from the
other groups in devoting a considerable portion of its budget to funding conser-
vation projects directly. At the other end of the spectrum, the National Trust
(NT) deliberately eschews campaigning activity with a view to focusing its
efforts on its perceived primary role of direct environmental protection via land
ownership (The National Trust, 1995). By purchasing sites, the NT’s aim is to
preserve them in perpetuity while at the same time providing a recreational
resource to its members. The two remaining groups, the Royal Society for the
Table 8.1 A comparative profile of some of the UK’s major environmental

groups
CPRE FoE GP NT RA RSPB WWF
Members (m) 0.05 0.20 0.40 2.21 0.10 0.89 0.23

Budget (£m) 2.1 3.5 7.2 142.6 1.9 33.9 21.1
Subscriptions (%) 24.8 n.a. n.a. 29 54 33 25
Fundraising (%) 19.5 14.0 18.8 9.5 12.6 18.0 25.0
Campaigns ✔ ✔ ✔ ✔ ✔ ✔
Research ✔ ✔ ✔ ✔
Land management ✔ ✔
Recreational sites ✔ ✔ ✔
Project funding ✔
Notes: data compiled from the 1994/95 Annual Reports for the respective organizations:
Council for the Protection of Rural England (CPRE); Friends of the Earth (FoE); Greenpeace
(GP); National Trust (NT); Ramblers’ Association (RA); Royal Society for the Protection of
Birds (RSPB); World Wide Fund for Nature (WWF). The definitions may not be entirely
consistent across groups due to different accounting conventions. It should be noted that the NT
and the RSPB manage 239 600 and 91 000 hectares of land as well as operating 240 and 130
recreational sites, respectively.
Protection of Birds (RSPB) and the Ramblers’ Association (RA), occupy some-
thing of an intermediate position in that they both supply recreational services
to their members and perform some campaigning activities.
Following Hotelling’s model (1929), the positioning of environmental groups
described above can be understood in terms of the two-dimensional charac-
teristics space illustrated in Figure 8.1. The two axes of the space represent
the extent to which the focus of the group is on the domestic or global
environment, on the one hand, and the extent to which the group is concerned
with environmental recreation or protection, on the other. Those groups lo-
cated towards the top left-hand corner of the figure (global environmental
protection) are supplying what is close to being a pure public good, while
those located towards the bottom right-hand corner of the figure (domestic
environmental recreation) are supplying what is close to being a pure club
good. This suggests that the former may face a more substantial free-rider
problem than the latter. Indeed, the substantive nature of the club facilities
provided by the NT and the RSPB may go some way towards explaining their
relative success in terms of membership numbers and budget sizes.
The positioning of groups in the diagram provides some tentative indica-
tion of the extent to which these organizations may be competing with each
Global
FoE
GP
WWF
RSPB
Domestic CPRE RA NT
Protection Recreation
Figure 8.1 Positioning of some of the UK’s major environmental groups in

characteristics space
other in fundraising. Thus, for example, FoE might be expected to face much
more direct competition for funds with GP and the WWF than with organiza-
tions such as the NT or the CPRE, which have a very different focus of
activity.
8.4 BRITISH SOCIAL ATTITUDES SURVEY

The present analysis is based on data collected as part of the British Social
Attitudes Survey (BSAS), an annual repeated cross-sectional survey covering
a wide range of social issues. In 1985, the BSAS started to run an occasional
section focusing on attitudes towards the countryside, which covered – among
other things – membership of environmental groups. The section was run for
three consecutive years between 1985 and 1987, and was repeated in a more
extensive form in 1990, providing a total of 6142 observations.
The surveys required respondents to indicate whether or not they belonged
to the National Trust (NT), the Royal Society for the Protection of Birds
(RSPB), a recreation-oriented environmental group or a protection-oriented
environmental group. It is important to note that the last two are self-assessed
categories. Therefore, it is not inconceivable – given the composite nature of
environmental group activities as recorded in Table 8.1 above – that two
members of a given organization might assess themselves in one case as
belonging to a recreation-oriented group and in the other case as belonging to

a protection-oriented group. This subjective feature of the data is not neces-
sarily problematic, since what is of primary interest is the preference of
individuals for the recreation or protection activities undertaken by such
organizations.
The patterns of environmental group membership arising in the data are
summarized in Table 8.2. The overall proportion of the sample claiming
environmental group membership is relatively static across the period at
between 15 per cent and 20 per cent – rising slightly during the late 1980s
only to fall back by 1990.
Table 8.2 Environmental group membership patterns from BSAS (%)
1985 1986 1987 1990
NT
BSAS 5.68 7.88 8.53 8.47
Social Trends 2.12 2.49 2.72 3.54
RSPB
BSAS 3.52 5.47 4.39 4.38
Social Trends 0.90 0.90 0.99 1.47
Recreation-oriented group 8.19 6.60 6.55 3.92
Protection-oriented group 4.19 4.52 3.80 6.35
Multiple groups 3.28 3.58 2.91 3.96
All groups 16.81 19.40 19.17 14.78
Notes: all BSAS figures are adjusted by the survey weighting factor; this leads to a slight
reduction in the proportion of members relative to the unadjusted figures. The survey weighting
factor is designed to adjust for the fact that there tend to be some differences between the
number of electors listed on the register and the number of adults actually found at any
particular address.
Source: Brook et al. (1991).
At the beginning of the period, recreation-oriented groups were the most

popular, accounting for just over 8 per cent of the sample. However, this
category exhibited a steep decline in membership numbers throughout the
late 1980s, so that by 1990 it had halved its proportion of members. For the
reasons cited above, this trend need not necessarily be interpreted as a genu-
ine shift in tastes away from recreation-oriented environmental groups. It
could simply reflect a change in individuals’ perceptions of the environmental
organizations they have always belonged to. All the other categories of groups
experience steady growth in membership proportions across the period, par-
ticularly the NT, which tops the membership league in 1990 with well over 8
per cent of the total sample reporting an affiliation.
Finally, about 3 per cent of the sample claim affiliation with multiple
environmental groups. Cross-group membership patterns differ significantly
between groups. The NT and the recreation-oriented environmental groups
tended to have a higher proportion of exclusive membership (in excess of 60
per cent), whereas less than half of the members of the RSPB and the
protection-oriented environmental groups belonged exclusively to those or-
ganizations.
The BSAS is designed with a view to obtaining a representative cross-
section of the UK population, with a sample frame constructed on the basis
of the electoral register (Brook et al., 1991). To verify this representativeness
with respect to environmental group membership, the sample proportions
claiming affiliation with the NT and the RSPB are compared in Table 8.2
with the figures for the population as a whole taken from the government
publication Social Trends (Office of Population Censuses and Surveys, 1992).
The figures reported in Social Trends can be regarded as accurate inasmuch
as they are based on the actual membership rosters of the respective environ-
mental groups. In order to obtain the reported percentages, these membership
statistics are divided by official estimates of the UK adult population.
The figures show that the proportions reported in the BSAS are up to three
times as large as those found in the population for the NT and up to six times
as large as for the RSPB. None the less, it is important to note that the upward
trends over time are broadly consistent, as are the relative membership rates
for the two groups. This indicates that the BSAS data are more reliable when
it comes to relative membership rates than in terms of the absolute level of
membership.
There are a number of possible explanations for this upward bias in the
membership percentages recorded in the BSAS. The first explanation relates
to the low response rates achieved by the survey. For example, in 1990, only
64 per cent of those households originally sampled actually completed the
questionnaire. This overall average masks regional variation in participation
rates of between 52 per cent and 74 per cent. To the extent that participation
in the survey was correlated with environmental group membership, a selec-
tivity bias problem could be expected to arise in the data.
The second explanation relates to ‘yea-saying’ induced by the ordering of
questions within the survey. Since group membership questions were asked
directly after a series of questions on environmental attitudes, it is possible
that respondents who had expressed significant concern about the state of the
environment then felt embarrassed to admit that they did not actually belong
to an environmental group, and were thus led to make false claims about
membership.
Third, respondents may have answered in the affirmative if any member of

their household (apart from themselves) held a subscription to an environ-
mental group, for example, parents whose adolescent children might be more
environmentally conscious than themselves.
Finally, people may have responded in the affirmative if they had at any
time in the past belonged to such an organization, irrespective of whether or
not they were currently paid-up members. This last point highlights a signifi-
cant drawback of the BSAS data, namely the absence of longitudinal
information from which to ascertain a given individual’s membership history.
In addition to providing data on club membership, the BSAS is a rich
source of information on the socioeconomic and demographic characteristics
of respondents. For the purposes of the empirical model, attention will focus
on six respondent characteristics. These are household income, age, news-
paper preference (whether or not the respondent is a broadsheet reader),
region (whether or not the respondent resides in the south of England), years
of full-time education and the presence of children in the household.
These six variables fulfil two important criteria for inclusion in the present
study. First, they are potentially observable (or at least could be inferred) by
charity managers and might therefore plausibly be used as targeting variables
for fundraising campaigns. Second, it is thought that they jointly capture the
main underlying factors determining preferences for environmental preserva-
tion, namely, purchasing power (household income), access to information
about the environment (newspaper preference and education), opportunities
for environmental recreation (age, region and the presence of children) and
tastes (all of the above).
8.5 EMPIRICAL ANALYSIS

Table 8.3 summarizes the coefficient estimates obtained from a series of
binary logit models for membership of the NT, the RSPB, and recreation and
protection-oriented environmental groups respectively (Maddala, 1983). The
purpose of these models is to identify whether there is any statistical associa-
tion between environmental group membership and the six explanatory
variables described above. A time trend is also included in the model in order
to capture any general shift in preferences across the period. The coefficients
indicate the magnitude of the effect that any particular explanatory variable
has on the probability that a person joins an environmental group.
The results show that many of the explanatory variables are statistically
significant in affecting the disposition to become an environmental group
member, as can be seen from the fact that the corresponding t-statistics are
greater than two in absolute value. The chi-squared statistics for the test of
Table 8.3 Summary of coefficient estimates for logit models of group membership
Recreation-oriented Protection-oriented
NT RSPB groups groups
Log (household income) 0.725 6.924 0.642 5.011 0.665 6.090 0.525 4.307
Age 0.035 7.117 0.023 3.849 –0.003 –0.665 0.011 1.910
Broadsheet reader 0.953 6.057 0.466 2.349 0.239 1.352 0.641 3.424
177
South of England 0.467 3.425 0.143 0.877 0.267 1.966 0.454 2.826
Years of education 0.376 7.512 0.209 3.344 0.066 1.202 0.249 4.194
Child in household –0.308 –1.978 0.069 0.380 0.199 1.415 –0.110 –0.641
Year –0.007 –0.210 –0.059 –1.414 –0.180 –4.688 0.059 1.501
Constant –3.722 –0.056 104.084 1.248 346.780 4.555 –130.024 –1.666
Observations 3700 3700 3700 3700

Chi-squared (7df) 381.19 92.76 105.47 141.11
Log-likelihood –831.42 –648.07 –866.41 –668.53
Fitted probability of group membership (%) Fitted probability of group membership
(a)
(b)
178
(%)
0
3
6
9
12
15
0
5
10
15
20
25
Age
<£5000
<30
NT
NT
RSPB
Household income
RSPB
Protection
Recreation
£5001–£10 000
Protection
Recreation
31–45
£10 001–£15 000
46–60 £15 001–£20 000

Capturing the economic value of charities
£20 001–£25 000
>60
>£25 000
30 NT
Fitted probability of group membership (%)
RSPB
Recreation
25 Protection
20
15
10
0
<17
17
18
>18
(c) Age completed full-time education
15 NT
Fitted probability of group membership (%)
RSPB
Recreation
12 Protection
0
1985
1986
1987
1990
(d) Year
Figure 8.2 Fitted probability of group membership against continuous

explanatory variables
overall significance of the regressions are reported at the foot of Table 8.3.
These indicate that while all of the models have significant explanatory
power, the model for NT membership performs significantly better than the
rest, with a test statistic that is more than double the size of any of the others.
The income variable is a highly significant determinant of group member-
ship in all cases. In most cases, membership is also significantly associated
with age, broadsheet readership, residing in the south of England and years of
education. However, the members of recreation-oriented groups stand out as
being substantially different from the rest. In particular, age, broadsheet
readership and residing in the south of England were not found to be signifi-
cantly correlated with belonging to this category of environmental groups.
Furthermore, recreation-oriented groups are the only category to exhibit a
statistically significant time trend, which indicates that the disposition to join
such groups was declining over the latter half of the 1980s. The absence of a
significant time trend in the NT and RSPB membership models stands in
contrast to the growing number of members for these groups reported in
Table 8.2 above. An explanation for this divergence is that the growing
membership may have more to do with a greater prevalence in the population
of the underlying characteristics that predispose towards membership, rather
than to any positive shift in tastes.
A series of Wald tests were conducted in order to test the null hypothesis
that there are no significant differences between the types of people attracted
to joining each of these different categories of environmental groups (Greene,
1993). This hypothesis could not be rejected in the case of the RSPB and the
protection-oriented groups, suggesting that very similar types of people are
attracted to joining both of these organizations. However, in all other cases,
the null hypothesis was strongly rejected, indicating that members of the NT
and the recreation-oriented groups are significantly different in profile from
members of the RSPB and the protection-oriented groups.
The models can be used to estimate the probability that any particular
respondent would join any particular environmental group as a function of
their observable characteristics. In order to illustrate how the probability of
successfully identifying a group member varies with different levels of the
explanatory variables, Figure 8.2 plots the fitted probability of membership
against different levels of the four continuous explanatory variables – house-
hold income, age, years of education and the time trend.
The fitted probability of membership is found to rise substantially with
household income, particularly beyond the £20 000 per year threshold. In-
deed, households with incomes in excess of £25 000 per year are between
five and seven times more likely to join an environmental group than house-
holds with incomes below £5000 per year. The effect of income on the fitted
probability of membership is particularly pronounced in the case of the NT.
The relationship between the age variable and the probability of group
membership is not so pronounced. There is some evidence that this probabil-
ity peaks in late middle age (between 45 and 60), particularly in the case of
the NT. An exception is the recreation-oriented category of environmental
groups, which appear to have a much stronger appeal among younger adults
(under 45).
A positive correlation between educational attainment and the fitted prob-
ability of membership is found for all groups. Specifically, those who remain
in full-time education beyond the age of 18 are three to six times more likely
to belong to environmental groups than those who terminate their full-time
education at the age of 16 or below. This effect is particularly striking for NT
membership.
Finally, the time trend looks somewhat different for each of the four
categories of environmental groups. While the fitted probability of member-
ship for the RSPB is relatively static across the period, it shows some increase
for the NT and the protection-oriented groups. The recreation-oriented groups,
on the other hand, show a clear downward trend in the fitted probability of
membership.
8.6 SIMULATION OF FUNDRAISING STRATEGIES

The binary logit models reported above can be used to address the issues of
prime interest identified in the earlier conceptual discussion. The first of these
was the question of how to design an optimal targeting strategy for environ-
mental group fundraising. The second was the question of whether market
segmentation succeeds in attenuating competition for subscriptions between
different types of groups. However, before turning to these matters, it is
important to take note of two important limitations of the estimated model in
these particular applications.
First, the descriptive discussion of the BSAS dataset suggested that envi-
ronmental group membership in the sample was three to six times larger than
in the general population (see Table 8.2). This indicates that the absolute
levels of the predicted probabilities should be calibrated, by scaling down
with this inflationary factor. However, the descriptive statistics showed a
much closer correspondence in the relative values of membership probabili-
ties both across groups and over time. This suggests that greater confidence
can be attached to the results that relate to the proportionate change in
probabilities of membership under different fundraising strategies.
Second, what is of primary interest to the fundraiser is the conditional
probability of a person taking out a subscription following the receipt of
publicity material from the organization. This is likely to differ, as a function
of individual characteristics, from the unconditional probability of group

membership. Results reported in Chapter 5 indicate that the unconditional
correlations between the probability of group membership and population
characteristics will in all likelihood be confounding two different effects. One
is the disposition of charity fundraisers to approach people with those charac-
teristics. The other is the disposition of people with those characteristics to
respond positively to an approach by a charity fundraiser.
Since the BSAS does not provide information on which of the respondents
had been asked to join an environmental group, it is not possible to separate out
these two effects. Consequently, the subsequent analysis is necessarily based
on the somewhat problematic assumption that the pattern of membership ob-
served in the population is an adequate guide to how individuals with particular
socioeconomic characteristics would respond to a fundraising approach.
These two issues, both of which suggest that the absolute level of predicted
probabilities from the model will overstate the true conditional probabilities
of membership, should be borne in mind when interpreting the results that
follow.
8.6.1 Optimal Fundraising Strategy
The coefficients estimated in the binary logit models identify segments of the
population with a significantly greater predisposition to join particular types
of environmental groups. Charity fundraisers could potentially target their
efforts using any or all of these variables. In order to shed light on the
appropriate choice of targets, Table 8.4 shows the effect which various differ-
ent targeting strategies have both on the probability of a subscription (P) and
on the number of potential fundraising targets (N).
Clearly, there are numerous ways in which the socioeconomic variables
studied in the model could be used to define targeting criteria for fundraising.
The targeting criteria presented here are taken to be illustrative rather than
exhaustive. Part (a) of Table 8.4 examines the effect of targeting on the basis
of each of the variables used in the model individually. For the dummy
variables, the target group is defined as the half of the population exhibiting
the highest propensity to join each corresponding type of environmental
organization. For the continuous variables, there are clearly many ways in
which the target group could be defined. The analysis is based on defining the
target group as the segment of the population exhibiting the highest propen-
sity to join on the basis of the banding used to construct Figure 8.2.
There is no reason why targeting strategies should confine themselves to
the use of a single targeting variable. The information contained in the single-
variable analysis could be combined together so as to develop a multi-variable
targeting approach. In principle, there are any number of ways in which the
Table 8.4 P and N against number of variables used in targeting
Recreation- Protection-
oriented oriented
NT RSPB groups groups
P N P N P N P N
(a) Single-variable targets

No targeting 0.080 3700 0.046 3700 0.067 3700 0.050 3700
Targeting on
Income 0.198 519 0.102 519 0.140 519 0.114 519
Age 0.099 870 0.055 870 0.068 870 0.056 870
Newspaper 0.195 959 0.084 959 0.089 959 0.107 959
Region 0.112 1673 0.056 1673 0.082 1673 0.069 1673
Education 0.251 415 0.109 415 0.115 415 0.136 415
Children 0.092 2348 0.047 2348 0.058 2348 0.052 2348
(b) Multi-variable targets – ‘AND’ basis

No targeting 0.080 3700 0.046 3700 0.067 3700 0.050 3700
Targeting on
best variable 0.251 415 0.109 415 0.139 519 0.136 415
2 best variables 0.311 293 0.154 157 0.162 157 0.163 293
3 best variables 0.404 125 0.169 125 0.166 125 0.209 125
4 best variables 0.441 80 0.175 80 0.179 80 0.235 80
5 best variables 0.535 19 0.203 19 0.166 19 0.255 19
all 6 variables 0.566 15 0.209 15 0.166 15 0.265 15
Note: P: the probability of a subscription. N: the number of potential fundraising targets.
information could be so combined (Foster, 1999a). For the purposes of illus-

tration, attention here will be confined to one particular multi-variable targeting
strategy. This will be referred to as the ‘AND’ strategy, and involves targeting
individuals who simultaneously satisfy two or more of the single variable
criteria. Accordingly, part (b) of Table 8.4 illustrates the effects of using
between two and six targeting criteria simultaneously.
The results for the single-variable approach indicate that substantial in-
creases in the probability of membership (P) can be secured by targeting,
although the magnitude of the benefits varies considerably with the type of
group. Thus the NT can increase its probability of subscription by more than
three times, while the other categories of groups do just better than doubling
their probability of obtaining a subscription. These increases in the probabil-
ity of success come at the expense of substantially reducing the number of
potential fundraising targets. For example, restricting the target population by
income or educational criteria immediately reduces the size of the potential
market to some 10 per cent of the overall sample.
0.6 NT
RSPB
Recreation
0.5
Marginal cost–revenue ratio
Protection
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6
Optimal number of targeting variables
(a) Gross revenue maximizer
0.6 NT
RSPB
Recreation
0.5
Marginal cost–revenue ratio
Protection
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6
Optimal number of targeting variables
(b) Net revenue maximizer
Figure 8.3 Marginal cost–revenue ratio against optimal number of

targeting variables for the ‘AND’ strategy
These trade-offs become even more pronounced when an ‘AND’ strategy is

adopted, as indicated in part (b) of Table 8.4. For example, the NT can raise
its probability of obtaining a subscription from each targeted individual to
over 0.5, but doing so reduces the number of potential targets to less than 1
per cent of the sample. For all groups, the marginal improvements in the
probability of success show a tendency to decline in the number of target
variables applied.
Further to the discussion in Section 8.2 above, it is interesting to enquire
how the optimal degree of targeting might vary according to the marginal
cost–revenue ratio. Figure 8.3 illustrates the case of the ‘AND’ strategy, for
both gross and net revenue maximization. The figures show that targeted
fundraising does not become worthwhile until the marginal cost–revenue
ratio exceeds 5–8 per cent for gross revenue maximizers and 4–6 per cent for
net revenue maximizers. As the marginal cost–revenue ratio increases, the
number of targeting variables needed to maximize the objective function rises
at an increasing rate. For a given value of the marginal cost–revenue ratio, the
NT requires the smallest number of targeting variables and the recreation-
oriented groups the largest number of targeting variables.
Although it is not visually evident from inspecting the graphs, the curves
for net revenue maximizers lie everywhere to the right of the corresponding
curves for the gross revenue maximizers. This is an empirical reflection of the
point described above, that net revenue maximizers will engage in higher
levels of targeting for any particular level of the marginal cost–revenue ratio.
For most of the groups, no amount of targeted fundraising will enable them
to survive financially once the marginal cost–revenue ratios reach the level of
around 20 per cent. The exception is the NT, which can survive even when
marginal fundraising costs absorb more than half of the corresponding mar-
ginal fundraising revenues. This is due to the fact that the relationship between
socioeconomic characteristics and the probability of membership is particu-
larly strong for the NT, making targeting particularly effective.
It is interesting to compare the rates of return (that is the ratio of gross
revenue to fundraising cost) predicted in these simulations with the real rates of
return observed in charity fundraising. A recent survey of charity managers
(Sargeant and Kaehler, 1998) found that fundraising by such methods as mail
shots, telemarketing and press advertisements yielded an average return in the
range 175 per cent to 200 per cent. However, these returns rose to 560 per cent
when fundraising efforts were targeted towards those with a history of giving.
Taking a strategy of single-variable targeting and a plausible marginal cost–
revenue ratio of 0.05, the logit models predict rates of return in the range 217
per cent to 501 per cent depending on the environmental group. Increasing the
marginal cost–revenue ratio to 0.1 lowers the level of returns from this form of
targeted fundraising to the range 109 per cent to 251 per cent.
Clearly strict comparisons between these two sets of values are not possi-
ble given that the Sargeant and Kaehler (1998) study does not report details
of the targeting strategies and marginal cost–revenue ratios faced by the
charity managers. None the less, the fact that the predictions of the binary
logit models are broadly similar in size to those found in the other study
provides some reassurance that the simulations are not seriously overstating
the possible outcomes.
In order to relate the predictions of the binary logit models to real fundraising
situations, it is necessary to have the real-world values for the marginal cost
and marginal revenue of fundraising. Both of these parameters are clearly
choice variables for charity fundraisers in the long run. However, using
information on typical current values for marginal cost and revenue, it be-
comes possible to comment on the degree of targeting that is likely to be
optimal, conditional on this current choice.
A current typical annual subscription fee for an environmental group is
£20. However, it is important to note that once a person subscribes, they are
likely to continue to do so over time and furthermore may make additional
contributions to special appeals. Thus the marginal revenue should ideally
reflect the present value of all these donations, so that the £20 can probably
be regarded as a lower bound. On the other hand, it is also important to recall
the costs that are incurred by environmental groups in order to provide a
range of private benefits that are typically offered to members, including
periodicals and access to special services. The cost of providing these private
benefits would need to be netted out of the subscription fee in order to
provide a clearer reflection of the pure untied revenue generated by an addi-
tional member. In this sense, the figure of £20 is likely to be an overestimate
of marginal revenue.
As regards the appropriate value of the marginal cost parameter, this is
likely to vary substantially according to the chosen fundraising technology.
In the context of a mailshot, which is a fairly traditional method, the marginal
cost could be expected to be particularly low since it would only reflect the
expenditure associated with postage and additional printing. Consequently, in
this case it is hard to imagine that the marginal cost could be any greater than
£1, as an absolute upper bound. These boundary values would give a maxi-
mum ratio for the marginal cost–revenue ratio of 0.05 and thus lie in the
range where targeting is still not an attractive proposition.
This illustrative calculation suggests that targeting may be of limited value
in practice to environmental groups. However, before reaching this conclu-
sion, it is important to note that the analysis presented thus far was premised
on the assumption that the charitable organization would exhaustively ap-
proach everybody in the identified target group, including the entire population
in the case of zero targeting. Whether or not this is a realistic assumption will
depend on the size of the environmental organization relative to the size of

the country in which it is operating. For small groups operating in large
countries it would appear likely that budgetary restrictions could preclude
such an exhaustive approach to fundraising. Indeed, mailshots by the RSPB –
one of the UK’s largest environmental charities – have historically been of
the order of 200 000–400 000, while the WWF’s largest fundraising cam-
paign in recent years encompassed a target population of 3 million. These
figures can be compared with the total number of households in the UK,
which is of the order of 20 million.
It may therefore be more realistic to consider the benefits from targeting in
a context where the charity must maximize the revenues generated from a
fixed fundraising budget. This may, in fact, be true of large groups as well as
small groups in the short run.
The fixed-budget setting carries a number of important implications. First,
the value of population that the charity can afford to reach is essentially fixed
by the size of the budget and is likely to represent only a very small fraction
of the overall population. This means that the loss of market size associated
with targeting may no longer be material, and in the limit could be disre-
garded altogether. Second, since costs are fixed by the size of the budget,
gross and net revenue maximizing behaviour become equivalent and both
entail maximizing the probability that a targeted individual will respond with
a subscription.
The combination of these factors indicates that relatively small charities
operating in relatively large countries should adopt an ‘AND’ strategy, target-
ing to the fullest possible extent on as many variables as possible. The figures
presented in Table 8.4 indicated that, by targeting on six variables, such
charities could raise their gross revenues by between three- and sixfold,
depending on the type of organization. This conclusion is based on the
assumption that the marginal cost of fundraising does not rise with the
number of targeting variables used, as it might, for example, if it became
increasingly difficult to identify members of the target group.
8.6.2 Competition for Funds
In addition to illuminating issues of fundraising strategy, the binary logit

models can be used to identify the extent of competition for funds between
these different types of environmental charities.
The Wald tests performed above provide evidence that environmentalists
loyal to different groups are a segmented population comprising a number of
distinct types of people. However, the coefficients reported in Table 8.3
indicated that the members of different environmental groups also had nu-
merous characteristics in common. Moreover, from examining the benefits of
alternative targeting strategies in Table 8.4, it becomes clear that the charac-
teristics (high income, high educational attainment and broadsheet readership)
which different types of environmental group members have in common have
a much stronger impact on their propensity to subscribe than the characteris-
tics (age, residence in the south of England and the presence of children in
the household) which distinguish them from each other.
The extent to which the different types of groups compete for funds will
therefore depend on the way in which their chosen fundraising strategies link
back to these underlying characteristics of the population. Table 8.5 indicates
the percentage overlap between the target populations for each pair of groups
according to the number of targeting variables used and the chosen strategy.
The higher the percentage overlap, the greater the extent to which groups are
competing for funds from the same segments of the population. The results
indicate that there is only very limited scope for avoiding competition, and
this only when a small number of targeting variables are being used. In the
case of the NT and the protection-oriented groups, targeting does not provide
any means of avoiding competition at all.
Table 8.5 Overlap ratios for target population segments between groups
for the ‘AND’ strategy (%)
NT: NT: NT: RSPB: RSPB: Recreation:

RSPB Recreation Protection Recreation Protection Protection
No targeting 100.0 100.0 100.0 100.0 100.0 100.0

Targeting on
Best variable 100.0 30.3 100.0 30.3 100.0 37.8
2 best variables 79.6 79.6 100.0 100.0 42.7 42.7
3 best variables 100.0 100.0 100.0 100.0 100.0 100.0
4 best variables 100.0 100.0 100.0 100.0 100.0 100.0
5 best variables 100.0 100.0 100.0 100.0 100.0 100.0
All 6 variables 100.0 100.0 100.0 100.0 100.0 100.0
8.7 CONCLUSIONS
This chapter has extended the existing literature on charitable fundraising by
examining to what extent charities can improve their performance by target-
ing effort towards specific segments of the population characterized by
particular socioeconomic and demographic attributes. The conceptual frame-
work suggested that targeting could affect fundraising both positively and
negatively. On the one hand, it could be expected to increase the probability
that a targeted individual would make a donation. On the other hand, it could
be expected to reduce the number of individuals who could potentially be

targeted.
A set of logit models was estimated in order to quantify the impact of a
range of observable characteristics on the probability of subscribing to the
National Trust, the Royal Society for the Protection of Birds, and other
recreation- or protection-oriented environmental groups. Using the models, a
variety of alternative fundraising strategies were simulated. The conclusions
were shown to differ depending on the size of the charitable organization.
For charities that are large relative to the size of the country in which they
operate, the optimal fundraising strategy was found to depend both on the
charity’s objective function (net or gross revenue maximization) and on the
ratio of marginal fundraising costs to the group’s subscription fee. The higher
this ratio, the greater the extent to which the charity should engage in target-
ing. None the less, there is a limit on the extent to which targeting can
overcome the handicap implied by a high marginal cost–revenue ratio. All the
environmental groups considered ceased to be financially viable once this
ratio exceeded a certain threshold.
For plausible values of the marginal cost–revenue ratio, little if any
targeting appeared to be advisable. The intuition behind this result is that
the shrinking size of the market brought about as a result of targeting more
than offsets the greater probability that any targeted individual will sub-
scribe, unless the costs of fundraising are comparatively high. However,
other things being equal, net revenue maximizing charities will tend to
undertake a greater degree of targeting than those whose objective is to
maximize gross revenues.
For charities that are small relative to the size of the country in which they
operate, that is, just about all UK environmental charities, the optimal strat-
egy irrespective of the charity’s objective function and marginal cost–revenue
ratio is to engage in as much targeting as possible. The intuition behind this
result is that small charities are unlikely to be constrained by the shrinking
size of the market, and are thus best advised to maximize the probability that
those whom they approach will choose to subscribe.
For such charities, targeting can increase their returns to fundraising by
between three- and sixfold, although there is substantial variation in the
magnitude of these gains across groups. Groups such as the National Trust,
which succeed in appealing to a sector of the population with very well-
defined observable characteristics, will reap the benefits in terms of the
greater efficacy of their fundraising efforts.
In spite of some demonstrated degree of segmentation in the market for
environmental charities, the model indicated that those factors with the
strongest influence on the propensity to become a group member were in
fact held in common between the different categories of environmental
groups considered. The implication is that targeting does not go very far
towards diluting the intensity of competition for members between environ-
mental groups, as they are all obliged to recruit from a very similar pool of
individuals.
PART III
Policy and Social Capital

9. On social capital
9.1 INTRODUCTION
In this chapter we argue that the phenomenon of giving is part of what has
come to be called ‘social capital’, the ‘glue’ that holds society together and
which plays an as yet generally unquantified role in economic and social
development. Social capital concerns the relationships between individuals,
between institutions (including governmental institutions) and between in-
dividuals and institutions. Social capital includes the norms governing these
relationships. It has been found that different societies can have broadly
equal endowments of other forms of capital, but that some perform better in
terms of economic and social development. The ‘missing link’ is thought to
lie in the fact that the better-performing societies have less conflict between
social groups, more participatory decision-making procedures and greater
trust between economic agents. Thus Putnam (1993) found that one of the
factors explaining northern Italy’s better economic performance compared
to southern Italy was the presence of many more voluntary associations.
Putnam’s focus was on horizontal associations between people and on the
rules of behaviour for those associations, rules designed to ensure that
members secured the full benefit of mutual cooperation. Fukuyama (1996)
identifies successful communities and companies with the presence of reci-
procity and trust. The breakdown of social capital results in more crime,
more violence, family breakdown and more distrust. Others, such as North
(1990), have emphasized institutions generally and political institutions
and the rule of law in particular. Clearly, definitions of social capital vary in
scope.
World Bank (1997) suggests that there are three concepts of social capital.
First, the narrow concept is the one employed by Putnam. It is a set of
horizontal associations between people and the norms governing those asso-
ciations. Associations exist for the mutual benefit of members, and the
associations in turn help sponsor economic efficiency and therefore economic
development.
Second, Coleman’s (1988) concept of social capital expands the narrow
concept above to include vertical associations, that is, linkages between mem-
bers of an association who (or which) are unequal in terms of the power they
193
194 Policy and social capital
possess. Thus hierarchical relationships would be included in this definition

and the agents include firms.
The third definition, building on North (1990), embraces a wider focus
still, and includes government, the rule of law and political structures. This
more comprehensive focus opens up the potential for more empirical indica-
tors, for example of political and civil liberties or corruption as a negative
indicator. To some extent, it would be helpful to define these components as
political capital but, clearly, the distinction between political and social is
very blurred.
In all three cases, it is argued that these forms of social capital are gener-
ally linked in a positive way to economic development, although not necessarily
so (see below).
None of the three definitions stresses the giving relationship, which is our
focus. But trust, cohesiveness, involvement and a wider concern for society
as an entity, and of the disadvantaged in particular, are all features of the acts
of giving and charity. Thus measuring the giving relationship offers one
important insight into the means of quantifying social capital. It cannot be a
comprehensive indicator, of course, but it is a meaningful indicator none the
less. We suggest, then, that the money value of the willingness to give, as
derived in Part I, is at least a candidate indicator of the value of social capital.
In order to set the context we digress for a while on the existing literature on
social capital.
9.2 SOCIAL CAPITAL IN THE THEORY OF

SUSTAINABLE DEVELOPMENT
Social capital is one form of capital asset upon which sustainable develop-
ment depends. In recent years, the capital-based theory of sustainable
development has advanced to a fair degree of sophistication. The briefest
sketch is given here. For a detailed assessment, see Atkinson et al. (1997).
Definitions of sustainable development are not in themselves very interest-
ing, although there is an interesting debate on how ‘development’ might be
measured in terms other than per capita GNP (see, for example, Atkinson et
al., 1997). What matters is what has to be done to secure it. Pearce (1999)
suggests that the conditions for sustainable development are likely to be
invariant with the definition since the conditions will be couched in terms of
capacities and capabilities: that is, sustainable development becomes an en-
abling concept rather than solely a particular path of change.
What determines the ability of a given set of humans to improve their well-
being (utility) is the quantity and quality of capital assets available at the
time. This notion can be traced back to the economic growth theory of the
On social capital 195
1970s. It is elegantly summarized in Solow (1992). The concept of capital

has widened from the classical approach, with its focus on produced goods or
man-made capital (KM in notation), to embrace the skills and knowledge
embodied in humans, or human capital (KH) and natural capital (KN). Natural
capital refers to traditionally defined natural resources, such as oil or gas,
forests and to the stocks of assimilative capacities in the environment. Mod-
ern economic growth theory would add social capital (KS), which is the focus
of concern here.
To complete the discussion of the capital base for sustainable develop-
ment we need to add two further factors. The first is the rate of technological
change, which is a disembodied stock of knowledge and skills. If this rate
of technological change is positive, then the capital stocks listed above can
yield higher and higher flows of services. A declining capital stock may not
therefore be of major concern provided the rate of technological progress
compensates for the loss of that stock. Of course, some technological change
is not ‘good’. Chlorofluorocarbons (CFCs), for example, were thought to be
a major technological advance for the development of aerosols, solvents
and cooling systems. But today we recognize CFCs as the major source of
stratospheric ozone depletion and a risk to human and ecosystem health.
The second additional concept is population change. It is possible that
growth in population can improve human well-being by stimulating techno-
logical change. There is some evidence for this in historical terms, where
changes in invention and innovation appear to be linked to increases in
population, and there is some modern evidence for this relationship too
(Boserup, 1981; English et al., 1994). But population change is more likely
to reduce per capita capital stocks. As population expands, pressure is put
on marginal productive land, for example, and forests are cut down or
burned to make way for agriculture, urban expansion and roads. All this
affects KN especially, but it is also reasonable to argue that rapid population
change reduces KS by increasing the likely conflicts over natural resource
availability, especially land and water.
Absolute population levels and the rationale for large families do, however,
need to be distinguished. There is evidence that large families exist precisely
because of the need to bind the family together. In turn, the focus on the
family as the social unit of concern suggests that the assurances one would
get from community-wide or nation-wide social capital have never existed, or
have broken down. To be more specific, large families provide both labour
(subject to labour laws, if any) and social insurance in old age. Large families
therefore have an economic rationale based on the costs and benefits of
family size to the family itself (Dasgupta, 1995). But the rationale for large
families is greatly reduced if the responsibility for, say, old-age security is
transferred to the state or community through either legislation or tradition.
The negative effects of population change are likely to be more pro-

nounced than any positive effect of population growth in stimulating the
expansion of capital, although this remains a possibility. And if sustainable
development is concerned with per capita utility, then population growth will
mean that more and more output will be needed just to keep pace with
additional numbers of people. Whereas technological change is likely to
make capital more efficient, population is likely to depress some capital
stocks and make sustainable development less feasible.
The intertemporal condition for sustainable development, therefore, amounts
to each generation leaving the next generation a stock of productive capacity,
in the form of capital assets and technology that is capable of producing more
utility or well-being per capita than that enjoyed by the current generation.
Notationally,
dK / dt ≥ 0, where K = KM + KH + KN + KS
This intertemporal requirement is only part of the story: concern with the
poor now is also widely regarded as an important feature of sustainable
development. This provides us with one immediate link to the charitable
sector since so much of this sector is focused on the underprivileged groups
in society.
The analysis above can be applied to this equity issue as well, for the poor
cannot improve their lot without access to productive capacity. If their well-
being is to improve, then they must secure better education as a means of
improving human capital (KH), better technology, more man-made capital
(KM), and more natural capital (KN). Social capital will matter as well in the
sense of the need for more participation in decisions that affect their lives,
more consultation and more concern from others. It is incidentally worth
noting that the questionnaire approaches used in Part I are themselves con-
sultative and participatory since it is the general public and the users who are
involved in expressing their preferences directly rather than through some
political process. Indeed, in the case of the homeless, the political process
may not reflect preferences at all well.
The equation above suggests that the intertemporal part of sustainable
development can be analysed in terms of the conditions necessary for its
achievement, and that those conditions can be interpreted in terms of a
constant capital rule. This rule requires modification in so far as:
● the rate of technological change results in higher capital efficiency (a

higher ratio of generated services to the capital stock). In this case, the
constant capital stock rule can be relaxed in favour of one that ensures
at least constant capital services; and
● the rate of population growth results in less capital per capita.
Of course, population growth may occur together with exogenous techno-

logical change, and this is typically the real-world case. If they work in
opposite directions, as suggested previously, then a rough rule of thumb is
that sustainable development will require that technological change exceed
the rate of population growth. This intuitive result bears a close resemblance
to some of the 1970s theory rules for sustained economic growth, for example
Stiglitz (1979). Note that the links between technological change and social
capital are not clear. Technological change does not obviously augment the
productivity of social capital, although it could be argued that, say, communi-
cation technology vastly improves the chances of being in touch with others
where before a letter or a journey might be involved. In other respects,
technology may actually distance people from each other by substituting for
more traditional communal or even family gatherings.
An implicit assumption in the constant capital rule is that all forms of capital
are substitutable for each other. On this rule, known as the weak sustainability
(WS) rule, any one form of capital can be run down provided proceeds are
reinvested in other forms of capital. Weak sustainability does not imply that
substitution is easy or inexpensive. We may have to surrender a great deal to
obtain one extra unit of some forms of capital, which is a feature of weak
sustainability that tends to be ignored by those who have criticized it. Moreover,
WS requires that the running down of any form of capital is compensated by
investment in some other form of capital. It is not consistent with running down
capital stocks and ‘consuming’ the proceeds. As a ‘weak’ rule, then, WS is not
particularly weak, and empirical tests show that it is quite easy for a country to
fail a weak sustainability test (see Atkinson et al., 1997).
Objections to weak sustainability tend to centre on the assumed substitut-
ability of capital stocks (Victor et al., 1994). Indeed, it can be argued that the
philosophy of sustainable development arose precisely because there were
concerns about the unsustainability of forms of economic development that
sacrificed the environment and social cohesion in the name of economic
growth. The problem can be formalized by saying that the critics regard at
least some forms of capital as having no substitutes. Those who believe in the
non-substitutability of natural capital support strong sustainability (SS). Far
less attention has been paid to strong sustainability based on social capital,
yet opinion polls repeatedly indicate that there are widespread concerns in
society about those aspects of modern life that indicate a breakdown of social
capital: crime, violence, drugs and sheer inhumanity. As noted above, these
are often advanced as negative indicators of social capital.
SS does not imply that WS is irrelevant. What SS requires in addition to
WS is that the stocks of KN or KS or KH (or perhaps all three) should not
decline. ‘In addition to’ is needed because a situation in which any one form
of capital is preserved but other forms of capital are allowed to decline could
hardly be called ‘sustainable development’ (it might be ‘survivable’ but even
that seems very unlikely). SS implies WS, but WS does not imply SS.
9.3 THE NATURE OF SOCIAL CAPITAL

Because the focus on social capital is comparatively new, it is worthwhile
considering in a little more detail what it might mean and whether it neces-
sarily contributes to sustainable development.
Putnam (1993) speaks of social capital as comprising certain features of
social organization – norms of behaviour, networks of interactions between
people and between institutions, and trust between people. Empirical studies
of economic growth have shown that conventional growth accounting models
(stressing labour, capital and technology) explain only a limited amount of
the difference between growth rates in different economies. World Bank
(1997) refers to studies of the ‘Asia Miracle’ economies (as they were before
the Asian economic collapse) which suggest that institutional arrangements
for cooperation and information exchange may be as, if not more, important
than conventional factors. But close interpersonal and interinstitutional ar-
rangements may not always be good for sustainable development. After all,
price-fixing cartels are a form of social arrangement, as is the Mafia. This
suggests that social capital may have positive and negative aspects.
On the positive side it is suggested that social capital contributes to eco-
nomic development because:
● flows of information between economic agents are better and higher if

there are closer social relationships. Such flows may relate to anything
from price information, information on the availability of materials or
labour, through to information on the credit-worthiness of individual
agents;
● trust reduces the need to search out information in order to make a
transaction; that is, transaction costs can be reduced. Trust may also
result in behaviour which avoids the need to make laws and hence to
intervene via government; and
● social links between individuals and organizations and government
also reduce the need for overt public control. Governments may find
it easier and more efficient to operate via established social links than
to legislate. The rise of voluntary agreements as a means to control
environmental problems may be a case in point. Polluters simply
agree to self-regulate and, in turn, self-regulation will be all the more
efficient if the polluters have social arrangements whereby they trust

each other.
Environmental improvement is also regarded as an integral part of modern

sustainable development (Atkinson et al., 1997). Social capital contributes to
environmental improvement by:
● substituting for other forms of capital, especially man-made capital.

Arrangements to share machinery, for example tractors, harvesters,
mean that fewer tractors are needed;
● reducing the high discount rates that often imperil the environment.
This happens because individual insecurity is reduced by ganging to-
gether to fight particular causes and by spreading risks among the
social group;
● reducing external effects, that is the spillover effects of one agent’s
actions on the well-being of another agent. Effectively, such behaviour
is inhibited by the concern for neighbours and third parties arising
from social norms of behaviour;
● resolving the risks arising from common property. Common property
involves a whole community owning and managing a resource, a situa-
tion that has risks of environmental destruction if the resulting communal
management system breaks down. The stronger the social ties, the less
likely the management system is to collapse; and
● inhibiting antisocial behaviour that damages the environment, whether
it is simply the dumping of illicit waste and litter or the perverse
destruction of wildlife.
Social capital could have negative results by keeping contracts with those
within the social circle, when those outside are more efficient. Examples
include price-fixing, closed-contract award systems, and even the require-
ment that small firms institute some social welfare system to look after those
in the social group, imposing costs that impair productivity. One might sum-
marize these problems as the creation of rent by restrictive activity and
through lobbying of government and others. The idea that a capital asset may
have negative impacts on sustainable development perhaps distinguishes so-
cial capital from other forms of capital assets. On the other hand we have
already noted the potential negative effects of some forms of technological
change.
Social capital therefore presents a new and challenging dimension of sus-
tainable development. It may, as some have argued, account for the dynamism
of some economies and even for lower environmental damage and better social
stability than might otherwise occur. Others suggest that it could just as easily
contribute in a negative fashion by restricting economic interests to specific

groups. And, of course, the social group could also be persuaded to destroy the
environment or wider social cohesion in the name of its own group interests.
9.4 INDICATORS OF SOCIAL CAPITAL AND THE VALUE

OF GIVING
The existing literature on indicators of social capital is small and several
commentators have remarked on the absence of a theory to guide model
building (Klitgaard and Fedderke, 1995). As noted above, it is also not very
clear what it is that one is trying to measure. Most of the work on indicators
takes measures of such things as civil and political liberties and then corre-
lates these with some indicator of development, for example GNP per capita.
Efforts are also made to see if there are associations between the different
indicators of social capital, the aim being to see if there are key indicators
that could be used instead of a proliferation of indicators. Indicators also
need to be distinguished from measures: there is no suggestion that indicators
of, say, civil liberties are measures of social capital. Indeed, in the sense
employed in the theory of sustainable development above, the only correct
measure would be a monetary value. To date, this type of measure has been
only indirectly referred to in the literature.
Some writers have focused more on indicators of the absence of social
capital rather than its presence. This is still useful since it reminds us that
what may have been taken for granted in the past is disappearing. Klitgaard
and Fedderke (1995), for example, focus on indicators of social integration
and disintegration, but much of their list of indicators relates to the break-
down of social (and political) cohesion. The positive indicators tend to be
political and civil liberties, thus embracing the much wider definition of
social capital above. The negative indicators include strikes, assassinations,
ethnic separatism, suicides, riots and so on. While the data sources are weak,
they speculate that:
● what they call ‘Factor I’ indicators, relating to the stability of political

institutions, and political and civil rights, are correlated with indicators
such as homicides, corruption and separatist movements. The higher
the instability and the lower the rights, the worse are the other social
indicators:
– ‘Factor II’ indicators, which relate to disorder that does not threaten
wholesale destruction of regimes (riots, assassinations and the like),
are correlated with inequality of land ownership, low voter turnout
and low level of citizen satisfaction;
– Factor I and Factor II indicators are not correlated;

● economic growth is positively correlated with social integration in the
sense that countries with low social scores have low economic growth,
to some extent bearing out the Putnam–Fukuyama hypothesis outlined
earlier. But the authors say that they cannot conclude that social cohe-
siveness causes better economic development;
● those issues that most concern Western countries are not correlated
with either Factor I or Factor II: crime, drugs, rapes, suicide, ethnic
fractionalization and so on; and
● the problems of social integration in the developing world appear to be
very different to those of the developed world.
Cross-sectional analysis of this kind is potentially interesting. Principal

components analysis can be used to reduce the large list of potential indica-
tors to a few key indicators, but, even then, the linkages between indicators
remain complex and there is no suggestion of a general theory linking social
capital to economic development.
World Bank (1997) suggests that data on the types of institutions that exist
can be collected and correlated with the success or otherwise of development
projects. Measuring institutions, rather than relationships, is attractive be-
cause the relationships are, as the Bank argues, already embedded in the
institution. But, even then, institutions will vary substantially in their effec-
tiveness: the existence of an institution does not mean that it is successful.
Data on membership go some way to overcoming this problem since high
membership presumably acts as a surrogate indicator of institutional success.
An alternative procedure, also suggested in World Bank (1997), and which
indirectly aims for measurement in the monetary sense, is to adopt a ‘produc-
tion function’ approach whereby the contributions of man-made and natural
capital to income growth are determined. What is left over (the residual) must
be due to human and social capital. If human capital contributions can be
isolated, then the final residual (the residual of the residual, as it were)
measures social capital. Gaarder and Hamilton (1999) test this approach by
taking various social capital indicators to see if they make any difference to
an explanation of the determinants of human resources per capita. For example,
human resources per capita can be regressed on an educational variable, say
mean years of education per capita, and a social capital variable to see if the
addition of the second variable makes any difference to the explanatory
power of the equation overall. If it does, then social capital is significant.
Gaarder and Hamilton (1999) find that the addition of a corruption index, a
civil/political rights index, and indicators of trust and civic norms (Knack and
Keefer, 1997) makes no significant difference to mean years of education as
the explanation for human resources per capita. What matters, they argue, is
human capital, not social capital. Literacy alone provides the most powerful
explanation of human resource availability.
We suggest that the economic value of giving provides a measure of social
capital, though not a comprehensive one. In other words, the kind of aggre-
gate measure we found in Chapter 5 offers a way of measuring KS. The
attractions of such an approach are (a) that it is a monetary measure rather
than an indicator, and hence is more readily built into the theory of sustain-
able development outlined above; (b) it captures at least part of an absolutely
essential ingredient of social capital (the relationship of care and concern for
others) which is neglected in the existing literature; and (c) potentially, it is
capable of being standardized across countries so as to permit comparisons.
While we are not primarily concerned with the issue of whether large
voluntary sectors are more or less correlated with development, it is easy to
think of various hypotheses linking the two. Moreover, which hypothesis is
correct does affect the extent to which the size of the voluntary sector is a
reasonable measure of what is at least part of social capital.
Thus we might expect a large voluntary sector in economies where state
provision of welfare and other services is low, as in developing economies.
What the state does not, or cannot, provide, has to be provided by voluntary
associations. This would suggest that social capital would be high in low-
income countries. The appropriate indicator would be the money value of
giving (plus consumers’ surplus) relative to gross domestic product (GDP).
Against this, state provision may be minimal and voluntarism may also be
limited if the family unit is strong: social capital is more confined to the
family and a measure of social capital based on the scale of extra-family
giving would miss the critical component of social capital, the intra-family
associations, in such a context.
The more developed an economy, the higher its income per capita, and the
larger the voluntary sector is – when measured in terms of willingness to pay
– simply because willingness to pay is positively linked to ability to pay, that
is, to income. Expressed as a proportion of GDP, however, the money value
of the voluntary sector might be less or more than in the low-income case. We
have no prior expectations about the income elasticity of giving, although
some evidence suggests that it is greater than unity for within-nation analyses
(Ribar and Wilhelm, 1995). If income elasticities are greater than unity, then
one would expect the share of voluntary sector value in GDP to be higher in
rich than in less rich countries. But since our willingness to pay surveys
reported in Part I are the only studies of their kind, we do not have compara-
tive data on which to base any analysis. The best that can be done is to look at
charities’ expenditure as a percentage of GDP. Salamon et al. (1995) report
such a statistic for eight developed and transitional countries. While volun-
tary expenditures do rise as a percentage of GDP as GDP rises, the UK and
the USA appear to behave differently from the other six countries, with very
much higher expenditure shares for income levels that are comparable to
those for France and Germany (the data relate to circa 1990).
The data are clearly still limited but, a priori, we see no particular reason to
suppose that social capital should be higher for a richer nation, even if we
accept the Putnam–Fukuyama hypothesis that social capital figures promi-
nently in the sustainability of the development process. There may, for example,
be an element of take-off, with social capital being essential for the launch
into sustained development rather than being clearly present once develop-
ment has taken place. There is clearly a major research agenda here for the
future. Given the uniqueness of our own study and the limited nature of other
data, we can only speculate here on the possible links between social capital
and development.
10. Conclusions and policy implications
10.1 THE SIZE OF THE CHARITABLE SECTOR

Part I of this volume was dedicated to measuring the size of the voluntary
sector. While most studies have determined size by the measurement of
income and expenditure by designated charities or the amount of employ-
ment associated with them, our own approach departs substantially from
those studies by seeking the willingness to pay (WTP) for charitable services.
The relevant WTP is that of the general public and the users, or beneficiaries,
of the charities. To this end, we conducted extensive questionnaires with the
general public, using stated preference procedures, and with one group of
beneficiaries, the homeless.
Our approach has the following advantages. First, as demonstrated in
Chapter 1, it is rooted in the theory of welfare economics. Second, it enables
us to lay the foundations for an approach to determining the efficiency of
different forms of social provision. Third, it highlights the potential revenues
that, in theory, charities could capture but which currently they do not, that is,
the excess of WTP over actual donations. Fourth, it offers far more scope for
a beneficiary-oriented approach to social provision.
We argue that our approach provides a more accurate measure of the size
of the voluntary sector. Rather than focusing on income and expenditure,
which are, effectively, measures of the cost of the sector, WTP gives some
idea of the benefits of the sector to society at large. Our calculations for the
UK suggest that the social value of charities is some 40 per cent higher than
the costs of providing that output. This result is based on our survey of the
WTP of the general public to ensure that charities continue in existence. This
is not, of course, income to the charities but an upper bound on potential
additional income. The challenge is how to devise strategies for capturing as
much as possible of this excess of true WTP over actual donations.
Because of resource constraints, we were able to test the beneficiary ap-
proach (looking at actual users of charities’ services) only for one charitable
sector, the provision of accommodation and care for the homeless. But we
deliberately chose what we knew would be a difficult sector on the grounds
that, if it works for that sector, it is likely to work for most other sectors.
Questionnaire approaches among the homeless have obvious hazards and
204
Conclusions and policy implications 205
difficulties. But the results were reassuring and they suggest that the benefici-
ary value of homelessness charities is more than 200 per cent; that is, the
social value of homelessness charities is some three times the cost of provid-
ing those services.
Questionnaire approaches have a number of drawbacks, even though the
degree of sophistication now applied to their design and execution is consid-
erable. One obvious drawback is that the answers are hypothetical and we
cannot be sure that people are giving true answers; that is, they might not
actually be willing to pay what they say they are willing to pay. The general
public questionnaire could suffer from this problem but it is notable that we
would expect an upwards bias in WTP over actual donations, and that the
excess WTP is not absurdly high. The homeless survey has one other reassur-
ing factor, namely that the responses (in this case to a WTA question) were
graded according to the quality of the hostel service in the manner we would
expect and that the compensation required was anchored in what the home-
less would actually have had to pay for alternative accommodation. These
considerations give us faith that the answers we secured are reasonable.
10.2 ON THE EFFICIENCY OF CHARITABLE PROVISION

The procedure for estimating the social value of the charitable sector opens
the way for tests of the efficiency of different forms of social provision. For
example, consider the alternative ways in which some form of social care is
provided. It can be provided totally by charities, totally by government (local
or central) or totally by the private sector. Obviously there are also variations
of ownership and management, for example state provision with private
management. From the standpoint of efficient care, which is a central con-
cern of government, what matters is the ratio of output to input, that is the
ratio of social benefit to social cost of provision. Our approach permits these
ratios to be measured. For any form of care, i, output (Qi) can be measured by
the WTP of the beneficiaries for that output (WTPi). WTP captures the per-
ceptions of the beneficiaries, so it allows for the fact that users may prefer
one form of provision over another, even if the essential features of that care
are the same. ‘Quality’ is measured as well as ‘quantity’. The cost of provi-
sion (Ci) can also be measured, so that a cost-efficiency indicator (ei) becomes:
ei = WTPi / Ci
A broader notion of efficiency would include the public’s WTP for particu-
lar forms of provision which, as Part I showed, reflects the public’s option
values and existence values.
206 The price of virtue
While we have not investigated the costs and benefits of different forms of
provision (only the costs and benefits of prevailing charitable institutions),
the potential is clearly there for developing efficiency measures which could
help guide the future provision of the services of the voluntary sector.
10.3 CAPTURING THE SURPLUS

Part I showed that the general public is willing to pay more for the provision
of charitable services than they actually pay. This surplus reflects an un-
tapped source of revenues, although capturing all of the additional 35 per
cent is clearly not feasible. Part II looked at various features of this ‘capture’
problem in more detail. Like most charitable sectors, the UK charitable
sector is financed by volunteer contributions (including volunteer time), sales
of goods and government grants. Including volunteer time in the analysis
shows that voluntary contributions account for about 35 per cent of charitable
sector financing. We have shown that these voluntary contributions could be
higher still if two issues could be tackled: the free-rider problem, whereby
individuals give less than they are willing to because they know others are
giving too, and the transaction costs problem, whereby individuals perceive
that there are costs to the act of giving and that these outweigh the benefits of
giving. Free-riding and transaction costs continue to exist even in the face of
altruistic behaviour by individuals. That is, what look like purely self-inter-
ested features of behaviour, for example reducing contributions because others
are giving, coexist with what may be very altruistic motives for giving.
The existence of free-riding provides one of the traditional arguments for
government financing. Government grants remain a major source of financ-
ing for charities in the UK. However, current government grants to charities
stand at about £35 per person per year, which is well below the additional
£40 to £60 per year that people say they would be willing to pay to support
charities. The existence of the 35 per cent surplus suggests, clearly, that
charities are underfunded, inasmuch as their true social value is 35 per cent
higher than the incomes they receive. The issue is where this extra 35 per cent
should come from. Clearly, one possibility would be to increase government
grants to the charitable sector. However, it is not the only option. Another
possibility would be to sharpen fiscal incentives for giving. Alternatively, the
additional money could be captured by increasing charity fundraising efforts.
It is interesting to enquire, from a social perspective, which of these three
approaches would be the most effective and efficient. Effectiveness can be
gauged in terms of the ability of the method to have a substantial impact on
the flow of resources towards the charitable sector. Efficiency can be meas-
ured in terms of the deadweight loss – or social cost – per pound of additional
Conclusions and policy implications 207
resources yielded to the charitable sector. On the basis of the research pre-
sented in Part II, it is possible to present a tentative ranking of the three
approaches against these two criteria.
In terms of effectiveness, the results reported in Chapters 6–8 suggest that
fundraising initiatives are likely to dominate fiscal incentives as a means of
channelling additional resources towards the charitable sector, since the em-
pirical results indicate that the gains available from improving fundraising
methods are much more substantial than those which could be generated by
sharpening fiscal incentives.
On the one hand, given the limited degree to which giving responds to
changes in fiscal incentives, as shown in Chapter 6, the kinds of changes in
the tax price that might be considered feasible – such as increasing the scope
of tax-efficient giving or altering the tax price in the order of a few pence in
the pound – could not be expected to stimulate tax-efficient giving by much
more than about 10–15 per cent.
On the other hand, the results reported in Chapters 7 and 8 indicate that
shifts between direct and remote fundraising methods or improvements in
targeting techniques can increase the return to fundraising efforts several
times over.
Furthermore, the finding in Chapter 6 that donations were inelastic with
respect to the tax price indicates that a reduction in the tax price of giving
costs more in tax relief than it generates in additional philanthropic contribu-
tions. Thus, by reducing fiscal incentives to give, and spending part of the
additional tax revenue on government grants to the voluntary sector, both
charities and the Exchequer could end up better off than might otherwise
have been the case. Consequently, from the government’s point of view, it is
more efficient to increase direct grants rather than to sharpen fiscal incen-
tives.
The discussion thus far has concluded that both fundraising and govern-
ment grants are likely to dominate improved fiscal incentives as a means of
channelling resources towards voluntary organizations. It remains to compare
the relative merits of fundraising and government grants. In doing so, it is
important to recognize that both methods of financing have their associated
deadweight loss.
In the case of government grants, the raising of tax revenue to finance such
grants imposes a deadweight loss on the economy. The quantification of this
social cost is a controversial issue. However, estimates of the average
deadweight loss per unit of revenue raised for the US economy range from
2.5 per cent (Harberger, 1964) to 30 per cent (Feldstein, 1995), while esti-
mates of the marginal deadweight loss can be as high as 200 per cent
(Feldstein, 1995). In the case of charitable fundraising, the information on
the cost of raising funds is relatively sparse. One recent study suggests an
average cost equal to between 20 per cent and 60 per cent of the revenue
raised (Sargeant and Kaehler, 1998); however, no estimates of the marginal
cost are available. Based on this comparison, a tentative conclusion would be
that the social cost of raising funds is greater for the charity fundraiser than
for the Exchequer.
In summary, the results of Part I suggest that – based purely on free-rider
arguments – there is some case for increasing the flow of resources to the
voluntary sector. The three possible methods identified for achieving this are
sharpening fiscal incentives, increasing government grants and redoubling
fundraising efforts. Based on the research presented in Part II, fiscal incen-
tives would appear to be the least effective of these three options, while
government grants would appear to be the most efficient. Redoubling the
efforts of charity fundraisers seems to be substantially more effective than
sharpening fiscal incentives, but considerably less efficient than increasing
government grants.
10.4 SOCIAL CAPITAL

Finally, in Chapter 9 we speculated that our measure of the economic value
of the charitable sector provided one measure of social capital. Social capital
is widely discussed as one of the essential features of successful economies.
It reflects the degree of trust and stability in personal relationships, and
between people and institutions. While the literature has focused on a number
of indicators, ranging from surveys of the extent to which people trust each
other to measures of civil and political liberties, it has not, to our knowledge,
raised the interesting issue of the measurement of the giving relationship. Yet
our measure of the economic value of the charitable sector is just such a
measure. Because ours is the only study that has taken the full economic
approach, we cannot offer comparative measures of social capital by country
or even over time for one country. But we would like to think we have planted
the idea and that others will take it forward.
References
Abrams, B.A. and M.D. Schmitz (1978), ‘The “crowding-out” effect of gov-
ernmental transfers on private charitable contributions’, Public Choice, 33,
29–39.
Adamowicz, V. (1995), ‘Alternative valuation techniques: A comparison and
movement to a synthesis’, in K.G. Willis and J.T. Corkindale (eds), Environ-
mental Valuation: New perspectives, Wallingford, UK: CAB International.
Adamowicz, W., V. Bhardeaj and B. Macnab (1993), ‘Experiments on the
difference between willingness to pay and willingness to accept’, Land
Economics, 69, 416–27.
Adamowicz, W., J. Louviere and M. Williams (1994), ‘Combining revealed
and stated preference methods for valuing environmental amenities’, Jour-
nal of Environmental Economics and Management, 26(3), 271–92.
Andreoni, J. (1988), ‘Privately provided public goods in a large economy: the
limits of altruism’, Journal of Public Economics, 35, 57–73.
Andreoni, J. (1989), ‘Giving with impure altruism: applications to charity
and Ricardian equivalence’, Journal of Political Economy, 97, 1447–58.
Andreoni, J. (1990), ‘Impure altruism and donations to public goods: a theory
of warm glow giving’, The Economic Journal, 100, 464–77.
Andreoni, J. (1998), ‘Toward a theory of charitable fundraising’, Journal of
Political Economy, 106, 1186–213.
Andreoni, J. and J.K. Scholz (1998), ‘An econometric analysis of charitable
giving with interdependent preferences’, Economic Inquiry, 36, 410–28.
Arellano, M. and C. Meghir (1992), ‘Female labour supply and on-the-job
search: an empirical model estimated using complementary datasets’, Re-
view of Economic Studies, 59, 464–77.
Arrow, K., R. Solow, P. Portney, E. Leamer, R. Radner and H. Schuman
(1993), ‘Report of the National Oceanic and Atmospheric Administration
Panel on contingent valuation’, Federal Register, 58, 4602–14.
Atkinson, G., F. Machado and S. Mourato (2000), ‘Balancing competing
principles of environmental equity’, Environmental and Planning A (in
press).
Atkinson, G., R. Dubourg, K. Hamilton, M. Munasinghe, D.W. Pearce and C.
Young (1997), Measuring Sustainable Development: Macroeconomics and
the Environment, Cheltenham, UK and Brookfield, USA: Edward Elgar.
209
Banks, J. and S. Tanner (1997), The State of Donation: Household Gifts to

Charity, 1974–96, Commentary 62, London, UK: Institute for Fiscal Stud-
ies.
Banks, J. and S. Tanner (1998), Taxing Charitable Giving, Commentary 75,
London, UK: Institute for Fiscal Studies.
Barrett, K. (1991), ‘Panel data estimates of charitable giving: a synthesis of
techniques’, National Tax Journal, 44, 365–81.
Baumol, W. (1962), ‘On the theory of the expansion of the firm’, American
Economic Review, 52, 1078–87.
Becker, G. (1974), ‘A theory of social interactions’, Journal of Political
Economy, 82, 1063–93.
Beggs, S., S. Cardell and J. Hausman (1981), ‘Assessing the potential de-
mand for electric cars’, Journal of Econometrics, 16, 1–19.
Bishop, R.C. and T.A. Heberlein (1979), ‘Measuring values of extra-market
goods: are indirect measures biased?’, American Journal of Agricultural
Economics, 61(5), 926–30.
Boserup, E. (1981), Population and Technological Change: a Study of Long
Term Trends, New York: Aldine Publishing Co.
Boskin, M.J. and M. Feldstein (1977), ‘Effects of the charitable deduction on
contributions by low income and middle income households: evidence
from the National Survey of Philanthropy’, Review of Economics and
Statistics, 59, 351–4.
Brook, L., B. Taylor and G. Prior (1991), British Social Attitudes 1990
Survey: Technical Report, London, UK: Social and Community Planning
Research.
Brown, E. and H. Lankford (1992), ‘Gifts of money and gifts of time: esti-
mating the effects of tax-prices and available time’, Journal of Public
CAF (Charities Aid Foundation) (1999), Dimensions of the Voluntary Sector,
West Malling, Kent: Charities Aid Foundation.
Cameron, T. (1988), ‘A new paradigm for valuing non-market goods using
referendum data: maximum likelihood estimation by censored logistic re-
gression’, Journal of Environmental Economics and Management, 15,
355–79.
Carson, R., T. Groves and M. Machina (1997), ‘Informational and Strategic
Properties of Value Elicitation Methods’, paper presented at the National
Science Foundation Workshop on Valuation and Environmental Policy,
Arlington, VA, USA.
Clotfelter, C.T. (1980), ‘Tax incentives and charitable giving: evidence from
a panel of tax-payers’, Journal of Public Economics, 13, 319–40.
Clotfelter, C.T. (1985), Federal Tax Policy and Charitable Giving, Chicago:
The University of Chicago Press.
References 211
Coleman, J. (1988), ‘Social capital in the creation of human capital’, Ameri-

can Journal of Sociology, 94 (supplement), 95–120.
Coursey, D., J. Hovis and W. Schulze (1987), ‘The disparity between willing-
ness to accept and willingness to pay measures of value’, Quarterly Journal
of Economics, 102, 679–90.
Cummings, R., D. Brookshire and W. Schulze (eds) (1986), Valuing Environ-
mental Goods – A State of the Arts Assessment of the Contingent Valuation
Method, Totowa, NJ: Rowman and Allanheld.
Daniel, W.W. (1990), Applied Non-Parametric Statistics, 2nd edn, Boston,
MA: PWS–Kent Publishing Company.
Dasgupta, P. (1995), ‘The population problem: theory and evidence’, Journal
of Economic Literature, 33, 1875–902.
Day, K.M. and R.A. Devlin (1996), ‘Volunteerism and crowding out: Cana-
dian econometric evidence’, Canadian Journal of Economics, 29, 37–53.
Department of Employment (1991, 1992, 1993), New Earnings Survey, Lon-
don: Her Majesty’s Stationery Office.
Dye, R.F. (1980), ‘Contributions of volunteer time: some evidence on income
tax effects’, National Tax Journal, 33, 89–93.
Ecotec Research and Consulting Ltd (1993), A Review of UK Environmental
Expenditure: A Final Report to the Department of the Environment, De-
partment of the Environment, London.
EFTEC (Economics for the Environment Consultancy Ltd) (1997), The So-
cial Value of the Charitable Sector, report for the Charities Aid Foundation,
London.
EFTEC (Economics for the Environment Consultancy Ltd) (2001), Guidance
on Using Stated Preference Techniques for the Economic Valuation of Non-
Market Effects, report for UK Department of the Environment, Transport
and the Regions, London.
English, J., M. Tiffen and M. Mortimore (1994), Land Resource Management
in Machakos District, Kenya, Environment Department Paper no. 5, Wash-
ington DC: World Bank.
Feldstein, M. (1995), Tax Avoidance and the Deadweight Loss of the Income
Tax, Working Paper no. 5055, Cambridge, MA: National Bureau of Eco-
nomic Research.
Feldstein, M. and C. Clotfelter (1976), ‘Tax incentives and charitable contri-
butions in the United States’, Journal of Public Economics, 5, 1–26.
Feldstein, M. and A. Taylor (1976), ‘The income tax and charitable contribu-
tions’, Econometrica, 44, 1201–22.
Fenyo, A., J. Kendall, M. Knapp, J. Pinner and 6 Perri (1993), The Voluntary
Sector and the Environment: Scope, Contributions, Issues, Research and
Statistics Unit Occasional Paper 5, Tonbridge, Kent: Charities Aid Founda-
tion.
Foster, V. (1997), ‘What value should be placed on volunteering?’, Dimen-

sions of the Voluntary Sector, Tonbridge, Kent: Charities Aid Foundation.
Foster, V. (1999a), ‘How much targeting is optimal? A simulation of fundraising
strategies for environmental charities’, in On the Voluntary Financing of
Public Goods, PhD Thesis, University College London.
Foster, V. (1999b), ‘How are “warm glows” created? The role of fundraising
technology in charitable giving’, in On the Voluntary Financing of Public
Goods, PhD Thesis, University College London.
Foster, V. (1999c), ‘Are British philanthropists price-sensitive? New evidence
from the Individual Giving Survey’, in On the Voluntary Financing of
Public Goods, PhD Thesis, University College London.
Foster, V. and S. Mourato (2000), ‘Measuring the impacts of pesticide use in
the UK: a contingent ranking approach’, Journal of Agricultural Econom-
ics, 51 (1), 1–21.
Foster, V., I. Bateman and D. Harley (1997), ‘Real and hypothetical willing-
ness to pay for environmental preservation: a non-experimental comparison’,
Journal of Agricultural Economics, 48, 123–38.
Freeman, A.M. (1993), The Measurement of Environmental and Resource
Values: Theory and Methods, Washington, DC: Resources for the Future.
Freeman, R.B. (1997), ‘Working for nothing: the supply of volunteer labour’,
Journal of Labour Economics, 15, 140–66.
Fukuyama, F. (1995), Trust: the Social Values and the Creation of Prosperity,
Harmondsworth, UK: Penguin.
Gaarder, M. and K. Hamilton (1999), Estimating the economic returns to
human and social capital: is literacy the whole story?, mimeo, Washing-
ton, DC: World Bank.
Glazer, A. and K.A. Konrad (1996), ‘A signalling explanation for charity’,
American Economic Review, 86, 1019–28.
Greene, W.H. (1993), Econometric Analysis, New York: Macmillan.
Greene, W.H. (1997), Econometric Analysis, 3rd edn, New York: Macmillan.
Halfpenny, P. and C. Pettipher (1992), Individual Giving and Volunteering in
Britain: Who Gives What and Why?, 5th edn, Tonbridge, Kent: Charities
Aid Foundation.
Aid Foundation.
Aid Foundation.
Hanemann, W.M. (1984), ‘Welfare evaluations in contingent valuation ex-
periments with discrete responses’, American Journal of Agricultural
References 213
Hanemann, W.M. (1991), ‘Willingness to pay and willingness to accept: how

much can they differ?’, American Economic Review, 81, 635–47.
Hanemann, W.M. and B. Kanninen (1996), The Statistical Analysis of Dis-
crete-Response CV Data, Working Paper no. 798, Department of Agricultural
and Resource Economics, Berkeley, CA: University of California.
Hanemann, W.M., J. Loomis and B. Kanninen (1991), ‘Estimation efficiency
of double-bounded dichotomous choice contingent valuation’, American
Hanley, N., D. MacMillan, R.E. Wright, C. Bullock, I. Simpson, D. Parsisson
and B. Crabtree (1998), ‘Contingent valuation versus choice experiments:
estimating the benefits of Environmentally Sensitive Areas in Scotland’,
Journal of Agricultural Economics, 49(1), 1–15.
Harbaugh, W.T. (1998), ‘What do donations buy? A model of philanthropy
based on prestige and warm glow’, Journal of Public Economics, 67, 269–
84.
Harberger, A. (1964), ‘Taxation, resource allocation and welfare’, in J. Due
(ed.), The Role of Direct and Indirect Taxes in the Federal Reserve System,
Princeton, NJ: Princeton University Press.
Heckman, J.J. (1979), ‘Sample selection bias as a specification error’,
Econometrica, 47, 153–61.
Hotelling, H. (1929), ‘Stability in competition’, The Economic Journal, 39,
41–57.
Inland Revenue (1998), Inland Revenue Statistics, London: Her Majesty’s
Stationery Office.
Janes, M. and B. Roberts (1990), ‘Labour costs in 1988’, Employment Gazette,
September Issue, Department of Employment, London: Her Majesty’s
Stationery Office.
Jencks, C. (1994), The Homeless, Cambridge, MA: Harvard University Press.
Johnson, F.R. and W.H. Desvousges (1997), ‘Estimating stated preferences
with rated-pair data: environmental, health, and employment effects of
energy programs’, Journal of Environmental Economics and Management,
34, 79–99.
Jones, A. and J. Posnett (1991a), ‘Charitable donations by UK households:
evidence from the Family Expenditure Survey’, Applied Economics, 23,
343–51.
Jones, A. and J. Posnett (1991b), ‘The impact of tax deductibility on chari-
table giving by covenant in the UK’, The Economic Journal, 101, 1117–29.
Jurgen, F. (ed.) (1988), Affordable Housing and the Homeless, New York and
Berlin: de Gruyter.
Kahneman, D. and J. Knetsch (1992), ‘Valuing public goods: the purchase of
moral satisfaction’, Journal of Environmental Economics and Manage-
ment, 22, 57–70.
Kahneman, D. and A. Tversky (1979), ‘Prospect theory: an analysis of deci-

sions under risk’, Econometrica, 47, 263–91.
Kendall, J. (1996), ‘The voluntary sector: international comparisons’, in Chari-
ties Aid Foundation, Dimensions of the Voluntary Sector: How is the
Voluntary Sector Changing, London: Charities Aid Foundation.
Kendall, J. and M. Knapp (1995), Voluntary Means, Social Ends, Canterbury,
Kent: Personal Social Services Research Unit.
Khanna, J., J. Posnett and T. Sandler (1995), ‘Charity donations in the UK:
new evidence based on panel data’, Journal of Public Economics, 56, 257–
72.
Kingma, B. (1989), ‘An accurate measurement of the crowd-out effect, in-
come effect and price effect for charitable contributions’, Journal of Political
Economy, 97, 1197–207.
Klitgaard, R. and J. Fedderke (1995), ‘Social integration and disintegration:
an explanatory analysis of cross-country data’, World Development, 23,
357–69.
Knack, S and P. Keefer (1997), ‘Does social capital have an economic payoff:
a cross-country investigation’, Quarterly Journal of Economics, 112(4),
1251–88.
Knapp, M. (1990), Time is Money: The Costs of Volunteering in Britain
Today, Paper no. 3, Voluntary Action Research, London: Volunteer Centre.
Knapp, M., V. Koutsogeorgopoulou and J. Davis Smith (1995), The Econom-
ics of Volunteering: Examining Participation Patterns and Levels in the
UK, Discussion Paper 1060, Personal Social Services Research Unit, Can-
terbury, Kent: University of Kent.
Krinsky, I. and L.A. Robb (1986), ‘On approximating the statistical proper-
ties of elasticities’, Review of Economics and Statistics, 68, 715–19.
Lareau, T. and D. Rae (1985), ‘Valuing willingness to pay for diesel odour
reduction: an application of the contingent ranking technique’, Southern
Economic Journal, 55(3), 728–42.
Louviere, J. (1988), Analyzing Decision-Making: Metric Conjoint Analysis,
no. 67, Quantitative Applications in the Social Sciences Series, Newbury
Park, CA: Sage University Papers, Sage Publications.
Maddala, G.S. (1983), Limited Dependent and Qualitative Variables in
Econometrics, Econometric Society Monographs no. 3, New York: Cam-
bridge University Press.
McFadden, D. (1994), ‘Contingent valuation and social choice’, American
Menchik, P.L. and B.A. Weisbrod (1987), ‘Volunteer labour supply’, Journal
of Public Economics, 32, 159–83.
Mitchell, R. and R. Carson (1989), Using Surveys to Value Public Goods:
The Contingent Valuation Method, Baltimore, MD: Johns Hopkins Press.
References 215
The National Trust (1995), 1994/95 Annual Report and Accounts, London:
The National Trust for Places of Historic Interest or Natural Beauty.
North, D. (1990), Institutions, Institutional Change and Economic Perform-
ance, Cambridge: Cambridge University Press.
Office of Population Censuses and Surveys (1992), Social Trends, London:
Government Statistical Service, Central Statistical Office.
O’Flaherty, B. (1996), Making Room: The Economics of Homelessness, Cam-
bridge, MA: Harvard University Press.
Olson, M. (1965), The Logic of Collective Action, Boston, MA: Harvard
University Press.
Pearce, D.W. (1999), ‘Sustainable development’, in D.W. Pearce, Economics
and Environment: Economics: Essays in Ecological Economics and Sus-
tainable Development, Cheltenham, UK and Brookfield, USA: Edward
Elgar.
Pharoah, C. (1997), ‘Broad trends: sector size, scope, income and expendi-
ture’, in Charities Aid Foundation (CAF), Dimensions of the Voluntary
Sector, West Malling, Kent: Charities Aid Foundation.
Pharoah, C. and M. Smerdon (eds) (1998), Dimensions of the Voluntary
Sector, West Malling, Kent: Charities Aid Foundation.
Pharaoh, C. and S. Tanner (1997), ‘Trends in charitable giving’, Fiscal Stud-
ies, 18, 427–44.
Posnett, J. and T. Sandler (1989), ‘Demand for charity donations in private
non-profit markets: the case of the UK’, Journal of Public Economics, 40,
187–200.
Putnam, R.D. (1993), Making Democracy Work: Civic Traditions in Modern
Italy, Princeton, NJ: Princeton University Press.
Randall, A. and J. Stoll (1980), ‘Consumer surplus in commodity space’,
American Economic Review, 70, 449–55.
Randall, G. and S. Brown (1996), From Street to Home: An Evaluation of
Phase 2 of the Rough Sleepers Initiative, Department of the Environment,
London: The Stationery Office.
Reece, W.S. (1979), ‘Charitable contributions: new evidence on household
behaviour’, American Economic Review, 69, 142–51.
Reece, W.S. and K.D. Zieschang (1985), ‘Consistent estimation of the impact
of tax deductibility on the level of charitable contributions’, Econometrica,
53, 271–93.
Resource Information Service (1996), London Hostels Directory 1996, Lon-
don: RIS.
Ribar, D. and M. Wilhem (1995), ‘Charitable contributions to international
relief and development’, National Tax Journal, 47(2), 229–44.
Richer, J. (1995), Green Giving: An Analysis of Contributions to Major US
Environmental Groups, Discussion Paper 95–39, Washington, DC: Re-

sources for the Future.
Rose-Ackerman, S. (1982), ‘Charitable giving and “excessive” fundraising’,
Quarterly Journal of Economics, 97, 193–212.
Rose-Ackerman, S. (1986), ‘Charitable giving and excessive fundraising’, in
D. Rose-Ackerman (ed.), The Economics of Non-Profit Institutions, New
York: Oxford University Press.
Salamon, L. (1995), Partners in Public Services: Government – Non-profit
Relations in the Welfare State, London: Johns Hopkins Press Ltd.
Salamon, L., H. Anheier, W. Sokolowski and associates (1995), The Emerg-
ing Sector: a Statistical Supplement, Institute for Policy Studies, Baltimore,
MD: Johns Hopkins University Press.
Samuelson, P. (1954), ‘The pure theory of public expenditure’, Review of
Economics and Statistics, 36, 387–9.
Sargeant, A. and J. Kaehler (1998), ‘Charities and fundraising costs – just
what is the benchmark?’, Dimensions of the Voluntary Sector, West Malling,
Kent: Charities Aid Foundation.
Schiff, J. (1985), ‘Does government spending crowd out charitable contribu-
tions?’, National Tax Journal, 38, 535–46.
Schwartz, R.A. (1970), ‘Personal philanthropic contributions’, Journal of
Political Economy, 78, 1264–91.
Shyamsundar, P. and R. Kramer (1996), ‘Tropical forest protection: an em-
pirical analysis of the costs borne by local people’, Journal of Environmental
Economics and Management, 31, 129–44.
Smith, J., S. Mourato, E. Veneklaas, R. Labarta, K. Reategui and G. Sanchez
(1997), Willingness to Pay for Environmental Services Among Slash-and-
Burn Farmers in the Peruvian Amazon: Implications for Deforestation and
Global Environmental Markets, CSERGE/CIAT/ICRAF Working Paper,
London: University College London.
Smith, V.H., M.R. Kehoe and M.E. Cremer (1995), ‘The private provision of
public goods: altruism and voluntary giving’, Journal of Public Econom-
ics, 58, 107–26.
Solow, R. (1992), An Almost Practical Step Towards Sustainability, Washing-
ton, DC: Resources for the Future.
Steinberg, R. (1986), ‘Should donors care about fundraising?’, in S. Rose-
Ackerman (ed.), The Economics of Non-Profit Institutions, New York:
Oxford University Press.
Steinberg, R. (1987), ‘Voluntary donations and public expenditures in a
federalist system’, American Economic Review, 77, 24–36.
Stiglitz, J. (1979), ‘A neoclassical analysis of the economics of natural re-
sources’, in V.K. Smith (ed.), Scarcity and Growth Reconsidered, Baltimore,
MD: Johns Hopkins University Press, pp. 36–66.
References 217
Sugden, R. (1982), ‘On the economics of philanthropy’, The Economic Jour-

nal, 92, 341–50.
Sugden, R. (1984), ‘Reciprocity: the supply of public goods through volun-
tary contributions’, The Economic Journal, 94, 772–87.
Taussig, M.K. (1967), ‘Economic aspects of the personal income tax treat-
ment of charitable contributions’, National Tax Journal, 20, 1–19.
Tirole, J. (1988), The Theory of Industrial Organisation, Cambridge, MA:
The MIT Press.
Victor, P., H.E. Hanna and A. Kubursi (1994), How strong is weak
sustainability?, paper presented at the International Symposium on Models
of Sustainable Development, Paris, France.
Volunteer Centre UK (1995), ‘The economic value of volunteering’, Volun-
teering Research Bulletin, February, London: Volunteer Centre.
Volunteer Centre UK/Social and Community Planning Research (1991), On
Volunteering: A Qualitative Research Study of Images, Motivations and
Experiences, Berkhamstead, UK: Volunteer Centre UK.
Wang, H. (1997), ‘Treatment of “don’t know” responses in contingent valua-
tion surveys: a random valuation model’, Journal of Environmental
Economics and Management, 32, 219–32.
Weisbrod, B.A. and N.D. Dominguez (1986), ‘Demand for collective goods
in private non-profit markets’, Journal of Public Economics, 30, 83–95.
Willig, R. (1976), ‘Consumer’s surplus without apology’, American Eco-
nomic Review, 66, 589–97.
World Bank (1997), Expanding the Measure of Wealth: Indicators of Envi-
ronmentally Sustainable Development, Washington DC: World Bank.
Index
accommodation, hostels to general public 19–71

expenditure, proposed scenario 97 to homeless 72–100
preferences 87–9 of volunteering 101–13
price of 73 see also aggregate benefits
additional welfare, estimating 13 Benthamite welfare function 10
age bid levels, contingent valuation 23, 47,
environmental group membership 55–8
178, 179, 181 binary logit models, targeting donors
fundraising methods 160 183–6, 187
participation rates of giving 141 British Social Attitudes Survey 173–6
user benefits survey 79–80 broad voluntary sector 5
WTP 55, 56
aggregate benefits, estimating 114–19 campaigning, environmental groups 171
accounting framework 114–16 capital-based theory, sustainable
net social value 116–18 development 194–8
interpretation and policy implications cash donations 30, 32
118–19 charitable sector
alcohol, expenditure, proposed scenario aggregate benefits 114–19
97 indicators of size 5–8
altruistic benefits 11 measurement of output 3–4
donating 36 measuring economic value 9–18
volunteering 43, 44–5 size of 204–5
see also impure altruism social value of 205–6
AND strategy 183–5 charities
Asia Miracle studies 198 benefits to general public survey
associations, social capital 193–4 19–71
attitudes design 20–28
general public survey results 29–61
explanatory variables 54 discussion and conclusions 61–4
payment ladder results 56 statistical appendix 65–71
towards donations 33–42 benefits to homeless survey 72–100
towards volunteering 42–6 design 74–8
WTP 55 results 79–98
towards countryside 173–6 discussion and conclusions 98–100
towards hostels 84–9, 89, 90, 99 net social value 10, 17–18, 102,
116–18
beneficiaries 11 total social cost 16–17, 116
benefits total social value 11–14, 114–15
charity’s outputs 10 Charities Aid Foundation 5, 8, 102
economic theory definition 9 chlorofluorocarbons (CFCs) 195
to beneficiaries 11–14 choice modelling techniques 16
219
club facilities, membership numbers user benefits survey 95–8

172 demographic variables
coefficient estimates BSAS 176
price responsiveness, giving 140–43 user benefits survey 79
targeting donors 176–81 Diana, Princess of Wales, death, WTP
commitment, in giving 39, 40, 41 55
compensation dichotomous choice see double-bounded
hostel closures 89–95, 99 dichotomous choice
WTA measure 13 direct fundraising
competition frequency distribution, donations
environmental groups 172–3 157–8
for funds 187–8 geographic variables 160
product differentiation 170 gift size 165
contingent ranking likelihood of giving 159
general public survey 20 marginal effects 163
design features 25–7 popular charities 160
results 58–9, 62 probability of obtaining a gift 164
statistical and theoretical frame- success rates 158–9
work 69–71 summary statistics 155, 156
contingent valuation 15–16 transaction costs 152
general public survey 20 warm glow benefit 152, 165
questions 22–5 discrete response formats 65–6
results 46–58, 62 donations
statistical and theoretical frame- general public survey
work 65 attitudes towards 33–42
user benefits survey 73 current 29–33
questionnaire 74–8 results 54, 55, 57, 62
results 89–95 in kind 7, 102
contributions equation 142 size of, warm glow 149
cost-effectiveness measures 4 see also giving
costs donations equation 160–62
charitable services 10 donors
fundraising 169–70 opportunity costs 16
see also labour costs; production targeting 167–90
costs; replacement cost British Social Attitudes Survey
approach; total social cost; 173–6
transaction costs conceptual framework 168–70
Council for the Protection of Rural conclusion 188–9
England (CPRE) 171, 173 empirical analysis 176–81
counselling services, use of 84–5, 86 environmental groups 171–3
credit cards 30 simulation, fundraising strategies
cross-membership patterns 175 181–8
cross-tax-price elasticity, volunteering don’t know options, survey questions
129–31 23–4
culture double-bounded dichotomous choice
donations 32, 33 elicitation
WTP 59 questions 22–3
results 46–9, 62
debriefing questions statistical and theoretical framework
general public survey 59–61 65–8
Index 221
drugs, expenditure, proposed scenario expenditure

97 measures of 8
user benefits survey
economic development, social capital breakdown of 80–81
198–9 desired compensation, hostel
economic growth 201 closure 93–5
economic value, charities 9–18 income, proposed scenario 96–7
aggregate benefits 114–19 external benefits 11
general public survey 19–71
homeless survey 72–100 Factor I indicators 200, 201
volunteering 101–13 Factor II indicators 200, 201
economic variables families, large 195
general public survey 53, 54, 56 Family Expenditure Survey (FES) 103
user benefits survey 79, 93 analysis of donations 128
education definition of charitable donations
environmental group membership 132–3
179, 181 favourite charities 42
general public survey, WTP 55, 56 fiscal incentives, for giving 123–46
user benefits survey 79, 80, 95 fixed-budget setting 187
effectiveness, charitable provision 206, food, expenditure, proposed scenario 97
207 foresight, in giving 37, 39, 40, 41
efficiency free-riding
charitable provision 205–6 depression of voluntary contributions
measures of 4 7
efficiency price 135–6 government financing 206
donation equation 161 pure altruism theory 148
elasticities for giving 144 tax exemptions, donations 123
mean monthly donations 139, 140 frequency distributions
participation rates 138, 141 donations, fundraising methods
effort, level of fundraising 151–2 157–8
elasticities, price-responsiveness, giving efficiency price 135, 136
143–5 housing preferences 88
elicitation, giving 149 tax price 134, 135
embarrassment motivation 35, 36, 42 time price 137–8
emergency night shelters 77, 78, 95 Friends of the Earth (FoE) 171, 173
employment fundraising
user benefits survey 79 costs 169–70
volunteering 140 income measures 5
employment related costs, volunteering managerial objectives 168–9
102 methods, valuation 147–66
endogeneity problems 125 motivation 148–50
environment conceptual framework 150–53
donations to 32, 33 hypotheses results 153–8
WTP 59 controlled hypotheses tests 168–4
environmental groups, targeting donors conclusions 165–6
171–3 simulation of strategies 181–8
environmental improvement, social targeting 167–8
capital 199 funds, competition for 187–8
ex-offenders’ projects 78
existence value, donating 36 gender, WTP 55, 56
general public shutdown

benefits survey compensatory amounts 89–95
objectives 19–20 hypothetical scenario 74–5
design 20–28 types 77
results 29–61 use of services 84–9
discussion and conclusions 61–4 hourly value, volunteers’ time 103–6
statistical appendix 65–71 hours, volunteered 30, 32
capturing surplus (WTP) 206–8 house rules, hostels 85–6
incremental WTP 118 housing history, homeless 82–3
total social value, benefits 115 housing and homelessness
giving general public survey
motivation 148–50 contingent ranking results 59
price-responsiveness 123–46, 207 contingent valuation results 47, 48,
private 6–8 49, 51, 55
process of 37–42 donations 32
value of 200–203 importance of 64
see also donations net social value 116, 117
goods user benefits survey 72–100
sale of 6 design 74–8
WTP 14 discussion and conclusions 98–100
government funding 4, 8 results 79–98
government grants housing projects 72
estimation, net hourly wage of housing schemes 77, 78
volunteers 105–6 human capital 201–2
free riding 206 hypotheses see null hypotheses
and general public, incremental WTP
119 importance, of charities 33, 63–4
opportunity costs 17 impure altruism 12, 149, 150–51
Greenpeace (GP) 171, 173 income
gross domestic product (GDP), money environmental group membership
value of giving 202–3 178, 180
gross revenue maximization 169, 184, from government 8
185 general public survey, WTP 53–5, 56
participation rates, giving 141
health and medical research private earned 6
donations 32, 33 success of direct fundraising 158–9
importance of charities 64 user benefits survey
WTP 58, 59 allocation of 81–2
Hekman models expenditure, proposed scenario
net hourly wage, volunteers 103–6 95–6
performance, fundraising methods sources of 80
160–64 income elasticities, for giving 144, 145
price-responsiveness of giving 140–43 income measures 5–6
homelessness incremental WTP
causes 83 all beneficiaries 117
defined 72 general public 118–19
duration of 83–4 indicators
hostels 72 size of charitable sector 5–8
characteristics 78 social capital 1, 200–203
price of accommodation 73 indirect users, benefits to 11–13, 36
Index 223
Individual Giving Survey (IGS) 30, 102 non-tax-efficient donations 31

activity based categories 106–8 null hypotheses, fundraising 152–3
definition of charitable donations 132 controlling extraneous factors 160–64
efficiency price 135 potential biases 158–60
hypothesis testing 153 results 153–8
price and income elasticities 124
tax price 134 open ended questions 24–5
Institute of Charity Fundraising opportunity cost approach 15
Managers 149–50 volunteering benefits survey 102–6,
institutions, data on 201 112
intensity, fundraising 158 optimal fundraising strategy 182–7
option benefits 11
Labour Cost Survey 108 donating 36
labour costs, estimating, volunteers volunteering 43
106–8 output measures, charitable sector 3–4
land ownership 171 output-based approach, volunteering
literature, price-responsiveness of giving benefits survey 109–10, 112
123, 125–32 overseas aid
low-support hostels 77, 78, 92 donations 32, 33
WTP 59
managerial objectives, fundraising
168–9 panel data 125
marginal cost, fundraising 169–70 parametric probability models 67–8
marginal cost-revenue ratio, targeting participation equations
184, 185, 189 net hourly wage estimation 103–4
membership price-responsiveness, giving 143
dues, private income 6 participation rates
environmental groups IGS records 133
fees 171 price variables 138–9
numbers and club facilities 172 socioeconomic variables 141
patterns 174–6 volunteering 140
probability of subscribing 181–2 paternalism 7, 11
moral principles, giving 148 payment ladder elicitation 24, 25
moral satisfaction 12, 149 results 46, 49–53, 56–7, 62
motivations statistical and theoretical framework
for donating 33–7, 148–50 68–9
towards volunteering 42–3 personal enjoyment, volunteering 42
multi-variable targeting strategy 183 policy implications, of benefits 118–19
political capital 194
narrow voluntary sector 5 popular charities 160
National Trust (NT) 171, 173, 175, 185 population change, capital stocks 195,
needy groups, net revenue maximization 196, 197
168–9 preferences
net hourly wage, estimating, volunteers hostels 84–9
103–6, 129 measurement of person’s 9
net revenue maximization 168–9, 184, prestige, giving 149, 151
185 price responsiveness, giving 123–46,
net social value, of charities 10, 17–18, 207
102, 116–18 literature overview 125–32
New Earnings Survey 106, 107 analytical framework 133–40
dataset 132–3 selectivity models 128, 131

empirical results 140–45 selfish motivation
conclusions 145–6 in donating 35–6
principal components analysis 201 volunteering 42–3, 44–5
private benefits 12 semi-supportive projects 77, 78
private earned income 6 services
private giving 6–8 indirect benefits 18
private sector equivalent 14 WTP 14
probability models, parametric 67–8 sex, WTP 56
product differentiation, competition 170 simulation, fundraising strategies 181–8
production costs, output of charities 8 single-variable targeting strategy 183,
production function approach 15, 201 185
protection-oriented groups 174 social capital 208
purchases, donations through 31 concepts 193–4
indicators of 200–203
quality index, fundraising methods measuring 3–4
161–2 nature of 198–200
questionnaires 205 sustainable development 194–8
questions social prestige, giving 149, 151
general public survey social services
contingent ranking method 25–7 donations 32, 33
contingent valuation method 22–5 importance of 64
debriefing 59–61 WTP 58, 59
user benefits survey 74–8 social situations, fundraising 152
debriefing 95–8 Social Trends 175
WTA format 99 social value
estimating 205–6
Ramblers Association (RA) 171, 172 output of charities 6
random utility model approach 66–7, 70 see also net social value; total social
rank-ordered logit model 58–9, 71 value
recreation-oriented groups 174, 180 social welfare functions 9–10
religion, importance in giving 141 society
remote fundraising importance of charities to 33
frequency distribution, donations 157, value of charities to 5
158 well-being of 9
geographic variables 160 socioeconomic variables
likelihood of giving 159 BSAS 176
marginal effects 163 general public survey
probability of obtaining a gift 164 comparison 29
summary statistics 155, 156 description 54
transaction costs 152 valuation functions 56
replacement cost approach 106–9, 112 WTP 55
resources, opportunity costs 18 participation rates of giving 141
Royal Society for the Protection of targeting donors 176–81, 188
Birds (RSPB) 171–2, 175 user benefits survey
compensation, hostel closures 93
sales, income from 6 profile 79–84
satisfaction, with hostels 85, 86 specific needs projects 77, 78
scope, of giving 37, 38, 39–40, 41 spontaneous giving 42
scorn, charitable behaviour 151 stated preference techniques 15, 20
Index 225
statistics, fundraising methods 154–6 volunteering 131

subscription fees, environmental groups United States
186 price responsiveness, giving 123, 125
support services, use of 84–5, 86 tax treatment of donations 127
supportive projects 77, 78, 92 volunteer status 150
surplus, WTP, capturing 206–8 user fees 6
sustainable development, social capital users
194–8 benefits of charities to 72–100
output measures 4
target groups utility, from donations 150–51
benefits 11
opportunity costs 16 valuation
targeting, donors 167–90 general public survey
tax incentives, for giving 123–46 results 46–61
tax price 133–5 of volunteering 43
elasticities for giving 144, 145 returns, fundraising methods 147–66
mean monthly donations 139 user benefits survey, results 89–95
participation rates 138, 139, 141 valuation functions
tax relief, UK 128 general public survey 53, 56–7, 69
tax-efficient donations 31 user benefits survey 93, 94
tax-efficient schemes, awareness of 134 vertical associations, social capital 193–4
technological change, capital stocks Volunteer Centre UK 101
195, 196 volunteering
telescoping effects, IGS 133 attitudes towards 42–6
time benefits survey
hours volunteered 30, 32 reason for valuation 101–2
volunteers’, valuing 7, 101–13 opportunity cost approach 102–6
time price 136–8 replacement cost approach 106–9
elasticities for giving 144 output-based approach 109–10
mean monthly donations 139 comparison of approaches 110–12
participation rates 138, 139 future valuation 112–13
volunteering 140 employment 140
tobacco, expenditure, proposed scenario participation in 3–4
97 price and income elasticities 129–31,
Tobit specification models 128 140
top-down approach, general public volunteers
survey 25 benefits to 12–13
results 51, 62 opportunity costs 17
total social costs status, US 150
of charities 16–17, 116 total social value 115
see also opportunity cost approach valuing time given 7, 101–13
total social value, of charities 11–14,
114–15 wage equation 103, 104–6
traditional hostels 77, 78, 92 Wald tests
transaction costs 7 environmental group membership
fundraising methods 152 180, 187
size of gifts 158 fundraising methods 160
warm glow benefits
United Kingdom, studies fundraising 151, 152, 158–9
price responsiveness 127, 128, 129 impure altruism 12
magnitude of donations 149 general public survey 21–2

selfish motivation 35–6, 42 contingent ranking results 58–9
welfare economics 9–10 contingent valuation results 47–58
well-being debriefing questions 59–61
in economic theory 9 incremental 117, 118–19
impact of charities on 102 World Wide Fund for Nature (WWF)
willingness to accept (WTA) 9 171, 173
in environmental economics 73–4 WTA see willingness to accept
estimating additional welfare 13 WTP see willingness to pay
shutdown of hostels 90–95, 99
user benefits survey 75 younger generations, fundraising
willingness to pay (WTP) 9 methods 160
demand for goods or services 14
estimating additional welfare 13 zero donations, modelling 128

Susana Mourato, David Pearce, Ece Ozdemiroglu The Price of Virtue The Economic Value of The Charitable Sector

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

Susana Mourato, David Pearce, Ece Ozdemiroglu The Price of Virtue The Economic Value of The Charitable Sector

Uploaded by

Copyright:

Available Formats

The Price of Virtue

There is no possibility of making a numerical estimate of the total scale of

All rights reserved. No part of this publication may be reproduced, stored in a

Edward Elgar Publishing, Inc.

A catalogue record for this book

Library of Congress Cataloguing-in-Publication Data

ISBN 1 84064 485 0 (cased)

Typeset by Manton Typesetters, Louth, Lincolnshire, UK.

PART I MEASURING THE ECONOMIC VALUE OF THE

PART II CAPTURING THE ECONOMIC VALUE OF CHARITIES

6 Providing fiscal incentives for giving 123

PART III POLICY AND SOCIAL CAPITAL

9 On social capital 193

10 Conclusions and policy implications 204

1.1 Social benefits from charities 12

6.3 Percentage frequency distribution of time-value price 137

2.1 Bid vectors for CVALL and CVHH 23

3.9 Reasons for not answering the valuation question 97

Our overwhelming debt is to Cathy Pharoah, the Head of Research at the

To our knowledge, the answers to this question, reported in Chapters 2–6

all. We deliberately chose a group of users where we would anticipate serious

amines the relative efficacy of different fundraising methods, in particular

Measuring the Economic Value of the Charitable

1.1 PURPOSE OF THE BOOK

1.2 EXISTING INDICATORS OF THE SIZE OF THE

1.2.1 Income Measures

Private earned income

First, there is a well-known tendency for people to free-ride on the generos-

charities. The problems of free-riding, donations in kind and transaction costs

Income from government

1.2.2 Measures of Expenditure

In addition to income information, the Charities Aid Foundation also pub-

1.3 MEASURING THE ECONOMIC VALUE OF THE

1.3.1 Foundations in Welfare Economics

Welfare economics is concerned with the level of well-being or welfare of

where U is the utility level of individual i from 1 to n and Σ indicates

Change in total social welfare = ∆W = Σ∆Ui

1.3.2 Total Social Value of Charities

Indirect benefits Direct benefits

Warm glow, Volunteer, altruistic, External, altruistic

Donors Volunteers Target groups

Figure 1.1 Social benefits from charities

Figure 1.1 provides a graphical representation of this typology of the total

Private sector equivalent

Opportunity cost approach

Production function approach

Contingent valuation method

straightforward to assess the validity of the estimates produced by CV studies

Choice modelling techniques

1.3.3 Total Social Cost of Charities

Donors Volunteers Government Target groups

Donations Time, skills Grants Fees, price

Figure 1.2 Social costs of charities

1.3.4 Net Social Value of Charities

In conclusion, this introductory chapter has argued that traditional measures

2.2 SURVEY DESIGN

charities, which fund scientific research into presently incurable diseases as

1. Contingent valuation of housing and homelessness charities (CVHH) (sam-

In a preliminary section, common to all versions, respondents were intro-

elicited using increases in taxation as a payment vehicle. In this context, the

Following the description of the hypothetical scenario, part of the sample

2.2.1 Contingent Valuation

The CVALL subsample received a double-bounded dichotomous choice CV