Professional Documents
Culture Documents
Susana Mourato, David Pearce, Ece Ozdemiroglu The Price of Virtue The Economic Value of The Charitable Sector
Susana Mourato, David Pearce, Ece Ozdemiroglu 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
Published by
Edward Elgar Publishing Limited
Glensanda House
Montpelier Parade
Cheltenham
Glos GL50 1UA
UK
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
List of figures vi
List of tables viii
A note on contributions x
Acknowledgements xi
Preface xii
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
References 209
Index 219
v
List of figures
vi
List of figures vii
viii
List of tables ix
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
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
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
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
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
6 Measuring the economic value of charities
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 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.
Conceptual foundations 7
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:
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.
Conceptual foundations 11
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.
12 Measuring the economic value of charities
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
Rest of society
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.
14 Measuring the economic value of charities
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.
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.
Conceptual foundations 17
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
CHARITIES
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
20 Measuring the economic value of charities
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.
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.
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)
Table 2.2 Payment ladder used in the general public survey (CV version)
and choices of a hypothetical respondent
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) ✘
Benefits of charities to the public 25
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.
26 Measuring the economic value of charities
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
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
28 Measuring the economic value of 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.
Benefits of charities to the public 29
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.
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
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)
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)
of some importance
important
very important
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
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
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
Benefits of charities to the public 37
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.
38 Measuring the economic value of charities
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
(c) ‘When I give, I don’t really mind what I’m giving to as long as it’s some
kind of good cause’
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
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
42 Measuring the economic value of charities
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.
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
44 Measuring the economic value of charities
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
strongly agree
agree
neutral
disagree
strongly disagree
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
46 Measuring the economic value of charities
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?’
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.
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.
NN NY YN YY
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.
‘Don’t know’ = ‘No’ ‘Don’t know’ = ‘Yes’ ‘Don’t know’ = ‘No’ ‘Don’t know’ = ‘Yes’
48
Tax level –0.9816 –16.203 –1.0687 –17.937 –1.2723 –14.124 –1.4079 –15.649
Notes: Model specification = double-bounded probit; CI = confidence interval calculated using the Krinsky and Robb (1986) procedure.
Benefits of charities to the public 49
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.
50 Measuring the economic value of charities
Table 2.12 Yearly WTP for CVALL and for CVHH from payment ladder
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
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
Benefits of charities to the public 51
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
Benefits of charities to the public 53
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
54 Measuring the economic value of charities
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
Benefits of charities to the public 55
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
Attitudinal variables
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
58 Measuring the economic value of charities
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.
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.
60 Measuring the economic value of charities
CVHH CVALL CR
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.
Housing and
Elicitation method All charities homelessness charities
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
64 Measuring the economic value of charities
Current
Attitudinal CV allocation CR weighting donations
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).
Benefits of charities to the public 65
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.
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.
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)
V = U (P, Y , X, C 0 ) + ε (2.2)
∆U ≥ η (2.6)
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
68 Measuring the economic value of charities
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)
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)
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).
2
exp(bXij )
Pi (Vi1 > Vi 2 > Vi 3 ) = ∏ 3 (2.13)
∑ exp(bXik )
j =1
k = j
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
74 Measuring the economic value of charities
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.
Benefits of charities to users 75
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)?
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.
76 Measuring the economic value of charities
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.
Benefits of charities to users 79
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
Alcohol/drugs/
tobacco
20%
Hostel
30%
Food
21%
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
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-
7–12 mths
13–24 mths
25–36 mths
37–48 mths
49–60 mths
61–120 mths
>120 mths
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.
99%
100 91% 87%
90
Percentage of sample
80
70
60
50
40
30
20
10
0
Washing Laundry Food
100
90
74%
Percentage of sample
80
70
60 51% 51%
50
40
26% 25%
30
17%
20 12%
10
0
Training/finding work
Health problems
Drug/alcohol problems
Total
Finding a home
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
86 Measuring the economic value of charities
Staff attitudes
Opening hours
Hostel rules
Privacy/space
Security/safety
Laundry
Cleanliness
Food
Washing
0 20 40 60 80 100
Percentage of sample
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
Benefits of charities to users 87
Training/finding work
Money/benefits
Finding a home
Drug/alcohol problems
Mental health
0 20 40 60 80 100
Percentage of sample
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.
88 Measuring the economic value of charities
Status quo
Home ownership
Private rental
Institution
Council
B&B
Family
Friends
Squat
Sleeping rough
0 20 40 60 80
Percentage of sample
Long term Medium term Short term
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
Benefits of charities to users 89
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.
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
Don’t mind
Would like it
0 10 20 30 40 50
Percentage of sample
Figure 3.10 Answers to the question ‘Would you like this idea to actually
happen?’
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
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
92 Measuring the economic value of charities
25
20
Percentage of sample
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 £)
Sociological 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 £)
Attitudinal variables
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
Coefficient t-statistic
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
Benefits of charities to users 95
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
96 Measuring the economic value of charities
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%
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
Benefits of charities to users 97
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.
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
98 Measuring the economic value of charities
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
102 Measuring the economic value of charities
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.
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
104 Measuring the economic value of charities
Table 4.1 Summary of Heckman selectivity models used to fit net hourly
wage
Males Females
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
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
Benefits of volunteering 105
Table 4.2 Estimated value of volunteer time using the opportunity cost
approach* (£1993/hour)
Note: * Calculated for that subsample of the IGS for which it was feasible to impute the
opportunity cost.
VC = WG (1 + λ ) − κ
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
108 Measuring the economic value of charities
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)
VB = × PB
O
H
Replacement
cost
Non-wage costs
Volunteer Tax costs
Centre Wage costs
Opportunity
cost
0 2 4 6 8 10
Replacement
cost
Non-wage costs
Volunteer Tax costs
Centre Wage costs
Opportunity
cost
0 2 4 6 8 10
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
112 Measuring the economic value of charities
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.
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.
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
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.
Table 5.1 The net social value of housing and homelessness charities,
1997
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).
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).
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
6.2.1 Giving
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.
128 Capturing the economic value of charities
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.
Providing fiscal incentives for giving 129
6.2.2 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
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.
Providing fiscal incentives for giving 131
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
132 Capturing the economic value of charities
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).
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).
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.
134 Capturing the economic value of charities
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)
50
40
30
20
10
0
0.60 0.75 0.80 1.00
Figure 6.1 Percentage frequency distribution of tax price (£)
Providing fiscal incentives for giving 135
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
136 Capturing the economic value of charities
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.
ϖ
PTi = i (6.3)
υi
Providing fiscal incentives for giving 137
50
40
30
20
10
0
<0.05
0.05–0.10
0.11–0.15
0.16–0.20
>0.20
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.
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 –
Notes: since the time-price variable is observed only for volunteers, the participation rate for
volunteering against the time price cannot be calculated.
Providing fiscal incentives for giving 139
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
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
140 Capturing the economic value of charities
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.
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
Providing fiscal incentives for giving 141
Giving Volunteering
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
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.
144 Capturing the economic value of charities
Table 6.7 Price and income elasticities for different types of philanthropic
activity
Time price
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.
Providing fiscal incentives for giving 145
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.
146 Capturing the economic value of charities
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
148 Capturing the economic value of charities
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 Capturing the economic value of charities
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.
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.
No Yes
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
156 Capturing the economic value of charities
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.
CD CR CD CR
(H02) ∆ C = − =0 (HA2) ∆ C = − >0 (7.2)
ND NR ND NR
45
40
35
30
25
20
15
10
5
0
<£1 £1–2 £2–3 £3–4 £4–5 £5–10 >£10
Direct Remote
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.
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
Choosing fundraising methods 159
Table 7.3 Ordinary least squares regression models for the number of
fundraising approaches
D + 1
Q= (7.3)
R + 1
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.
Table 7.4 Selectivity models for all philanthropic gifts with controlling for
fundraising effort
Participation Contribution
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
Choosing fundraising methods 163
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
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.
Choosing fundraising methods 165
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.
166 Capturing the economic value of charities
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
168 Capturing the economic value of charities
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).
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-
170 Capturing the economic value of charities
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.
Targeting donors 171
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
Targeting donors 173
Global
FoE
GP
WWF
RSPB
Domestic CPRE RA NT
Protection Recreation
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.
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.
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.
176 Capturing the economic value of charities
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
(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
>60
>£25 000
Targeting donors 179
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
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.
Targeting donors 181
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.
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
Targeting donors 183
Recreation- Protection-
oriented oriented
NT RSPB groups groups
P N P N P N P N
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
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
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
Targeting donors 187
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 (%)
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
Targeting donors 189
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
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
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:
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.
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
200 Policy and social capital
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
On social capital 203
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
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.
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.
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
208 The price of virtue
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.
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
210 The price of virtue
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
216 The price of virtue
219
220 The price of virtue