philanthropy in the welfare stated r o o w r o o v 7 introduction: philanthropy in the welfare state...

VU Research Portal

Philanthropy in the welfare state

de Wit, A.

2018

document versionPublisher's PDF, also known as Version of record

document licenseUnspecified

Link to publication in VU Research Portal

citation for published version (APA)de Wit, A. (2018). Philanthropy in the welfare state: Why charitable donations do not simply substitutegovernment support.

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

E-mail address:[email protected]

Download date: 14. May. 2021

https://research.vu.nl/en/publications/8919f034-0ad6-450a-a809-c6532762e5c8

VRIJE UNIVERSITEIT

PHILANTHROPY IN THE WELFARE STATE

Why charitable donations do not simply substitute government support

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad Doctor aande Vrije Universiteit Amsterdam,

op gezag van de rector magnificusprof.dr. V. Subramaniam,

in het openbaar te verdedigenten overstaan van de promotiecommissie

van de Faculteit der Sociale Wetenschappenop dinsdag 30 januari om 15.45 uur

in de aula van de universiteit,De Boelelaan 1105

door

Arjen de Wit

geboren te Beverwijk

promotor:

copromotoren:

prof.dr. R.H.F.P. Bekkers

prof.dr. M.I. Broese van Groenouprof.dr. B.G.M. Völker

Leescommissie : prof.dr. L.C.P.M. MeijsErasmus Universiteit Rotterdam

prof.dr. P.L.H. ScheepersRadboud Universiteit Nijmegen

prof.dr. W.A. TrommelVrije Universiteit Amsterdam

prof.dr. T.W.G. van der MeerUniversiteit van Amsterdam prof. M.O. Wilhelm PhDIndiana University – Purdue University Indianapolis

prof.dr. R.H.F.P. Bekkers

prof.dr. M.I. Broese van Groenouprof.dr. B.G.M. Völker

Voorwoord 7

Introduction: Philanthropy in the welfare state 13

Part I: Crowding-out in contextChapter 1: Exploring crowding-out with cross-country data 35Chapter 2: A meta-analysis of the crowding-out hypothesis 63

Part II: New empirical estimatesChapter 3: Heterogeneity in crowding-out 91Chapter 4: The role of information 113Chapter 5: Look who’s crowding-out! 135

Conclusion: Partners, not substitutes 161

Summary 182Samenvatting 184Acknowledgements 186Funding 188Appendix 189References 195

Contents

7

Voorwoord

Daar stonden we dan, op 25 februari 2015, de avond dat het Maagdenhuis werd bezet. Het protest was gericht op misstanden aan de Universiteit van Amsterdam, maar ik wilde graag laten weten dat het op andere universitei-ten niet veel beter is gesteld.

Aan onze faculteit werd de minor ‘Goede doelen, filantropie en non-pro-fits’ afgeschaft omdat die “niet rendabel” zou zijn. Het is tekenend voor de heersende bestuurderslogica. Aan universiteiten wordt het geld grotendeels verdeeld op basis van de aantallen studiepunten die studenten behalen, waardoor vakken met weinig studenten en/of strenge tentaminering meer kosten dan ze opleveren.

Niet alleen vakken, ook het personeel staat onder zware druk. Medewer-kers met een tijdelijk contract moeten telkens op zoek naar een andere werk-gever. Deels is dit het gevolg van overheidsbeleid dat zijn doel voorbijschiet (de Wet werk en zekerheid), deels van onwil bij universiteitsbesturen om de voordelen van vaste contracten (werknemerstevredenheid, opbouw van expertise, sociale innovatie) te laten opwegen tegen de risico’s (mogelijke begrotingstekorten in het geval van dalende inkomsten).

Ook mij was dit lot beschoren. Omdat ik geen contract had als promoven-dus maar als tijdelijke onderzoeker kon ik, een jaar voor het einde van mijn promotietraject, geen verlenging krijgen.

Na een stuk of honderd e-mails en telefoontjes tussen professoren, finan-cieel directeuren en mijzelf kwam er een oplossing. Ik kon een half jaar in dienst van de UvA treden, waarna er een nieuw tijdelijk contract klaar zou liggen op de VU. Beate Völker werd toegevoegd als copromotor en de pot met promotiegeld zou gedeeld worden tussen de twee universiteiten.

En zo ligt hier toch nog een proefschrift. Het onderwerp van deze dissertatie komt niet uit de lucht vallen met twee

ouders die actief zijn in het welzijnswerk, het speciaal onderwijs, de politiek en het lokale verenigingsleven, en een broer die misschien wel de meest al-truïstische persoon is die ik ken. Mijn interesse voor sociale ongelijkheden werd verder ontwikkeld tijdens mijn studie Politicologie aan de UvA, waar ik leerde over de verschillen tussen verzorgingsstaten. Sociaal-democratische welvaartsstaten zoals de Scandinavische besteden niet alleen veel aan soci-ale voorzieningen, het zijn ook de landen die relatief veel geven aan officiële ontwikkelingshulp. Maar hoe zit het dan met de private geldstromen? Geven

8

inwoners van deze landen dan ook veel aan goede doelen, of juist niet? Via Brian Burgoon (begeleider van mijn bachelorscriptie Kleine overheid, royale burgers: Niet-gouvernementele internationale liefdadigheid in de welvaarts-staat) en Tom van der Meer (begeleider van mijn masterscriptie Mechanis-ms of crowding-out and crowding-in: Voluntary involvement in 30 European countries) kwam ik terecht bij René Bekkers aan de VU, waar een afdeling Filantropische Studies bleek te zijn.

Ik herinner me het kennismakingsgesprek met René toen ik stage kwam lopen, waarbij we meteen verwikkeld raakten in een methodologische dis-cussie. Moet je de clusters van multilevelmodellen zien als waarden op een variabele of als analyse-eenheden? Ik weet niet meer wie er gelijk had, maar het zal René wel weer geweest zijn.

Het karakteriseert de verhouding tussen mijn promotor en mij. René’s visie op goed en verantwoordelijk onderzoek hebben mij sterk gevormd. Een heuse meester-gezelrelatie, maar tegelijkertijd heel gelijkwaardig. Waar buitenlandse promovendi op congressen hun professor met de achternaam aanspreken en soms zelfs alleen de powerpoint van van de presentatie mo-gen doorklikken (dat gebeurt echt!), zitten René en ik vaak als ondeugende schooljongens grapjes te maken.

Toen we een copromotor zochten binnen de faculteit bleek Marjolein Broese van Groenou een leuke en inspirerende collega die zelf uiterst rele-vant onderzoek doet naar verschillende vormen van sociale participatie on-der ouderen, een gebied waar mogelijke verdrijvingseffecten tussen formele zorg, informele hulp en vrijwilligerswerk aan de orde van de dag zijn.

Beate Völker stond meteen klaar om ons te helpen toen we met het aflo-pende contract zaten en heeft me op de UvA met groot enthousiasme bege-leid. Samen met Jochem Miggelbrink en zijn collega’s van het Amsterdams Universiteitsfonds ontwikkelden we een veldexperiment met de jaarlijkse fondsenwervingscampagne onder alumni, dat uiteindelijk te weinig zinnige analyse-eenheden opleverde om op te nemen in dit proefschrift maar waar we veel van geleerd hebben.

Een samenwerking die dit boek wel heeft gehaald is die met Michaela Neumayr, Femida Handy en Pamala Wiepking. Hoofdstuk 1 is het product van een onbaatzuchtige uitwisseling van ideeën en onderzoeksgegevens.

Het moge duidelijk zijn: de academische wereld zit vol met betrokken en gedreven mensen. Dat mijn dissertatie door kon gaan als een samenwerking tussen VU en UvA toont aan dat er, ondanks regels en gewoontes, vaak meer ruimte is dan je denkt om bezig te blijven met datgeen waarvoor we weten-

9

schappers zijn: goed onderzoek en inspirerend onderwijs. Momenteel denk ik samen met mijn lieve vriendin Anika veel na over de

bijdrage die we kunnen leveren aan een mooiere wereld. Langzaam ontstaan de countouren van een postkapitalistische samenleving die niet geleid wordt door werk, prestaties en bezit maar door gelijkheid, duurzame productie en gemeenschappelijk eigendom. Dit uit zich op allerlei gebieden. Zo zien we een discussie over een universeel basisinkomen op macroniveau, een groei aan lokale coöperaties en deelplatformen op mesoniveau en een toenemen-de afkeer van overwerk en prestatiedruk op microniveau.

De invloed van wetenschappers op dit soort maatschappelijke ontwikke-lingen lijkt soms summier maar kan substantieel zijn, zoals Theo Schuyt en andere collega’s van het Centrum voor Filantropische Studies voortdurend laten zien als het gaat om de professionalisering van de filantropiesector in Nederland en Europa.

Laten we niet vervallen in cynisme. Laten we niet in het keurslijf blijven van angstig beleid en overmatige bureaucratie. Laten we beseffen dat de uni-versiteit gemaakt wordt door studenten en docenten, door onderzoekers en ondersteunend personeel, en niet door regels of outputmetingen. Samen-werking, en niet competitie, is de basis voor mooi wetenschappelijk onder-zoek.

11

Philanthropy is marvelous, but it must not cause the philanthropist to overlook the need for working to remove many conditions of economic injustice which make philanthropy necessary

Martin Luther King Jr., On Being A Good Neighbor

“”

Introduction

Philanthropy in the welfare state

14 Introduction Philanthropy in the welfare state

While ageing continues to raise the costs of care and pension provisions, economic volatility and public debts depress government spending. Political choices that are made under these pressures result in smaller budgets for welfare state arrangements, and explicitly stated responsibility for citizens and organizations to participate in social programs like the provision of care and the funding of cultural activities. In 2010, for example, the Dutch govern-ment announced to decrease government funding of nonprofit organizations in the field of arts and culture, combined with a fiscal reform to increase tax deductibility of charitable donations in this area. In hundreds of media articles and public responses from nonprofit organizations, the budget cuts were qualified as “survival of the fittest” (Vos, 2011) and “a terrorist attack on the arts” (theatre director Ivo van Hove in Rijghard, 2010), but also as a call for self-reflection and innovation in the nonprofit sector (Barth, 2011; Pontzen, 2011). The explicit goal of this policy reform was to aim for a small government, with ample room for private generosity and “cultural entrepre-neurship” (Rijksoverheid, 2011). This is a textbook example of policy choices motivated by a supposed zero-sum relationship between government fund-ing and philanthropic activity in the nonprofit sector. But to what degree is philanthropy actually affected by financial government support?

There is a wide array of academic studies dedicated to the idea that in-creasing levels of government support “crowd out” civic engagement, while decreasing government support encourages civic engagement (the crowding-out hypothesis). Conceptually, the idea of crowding-out has its roots in the work of Alexis de Tocqueville (1970[1840]) and Robert Nisbet (1962[1953]), who argued that increasing government power rules out indi-vidual control over one’s own environment as organized in families, church-es and voluntary organizations. Public goods theory predicts that the size and scope of the nonprofit sector is larger when governments fail to meet the needs of heterogeneous societies (Weisbrod, 1977). A more contempo-rary article describes the general claim as: “For every welfare state, if social obligations become increasingly public, then its institutional arrangements crowd out private obligations or make them at least no longer necessary” (Van Oorschot & Arts, 2005: 2). Obligations like providing care or social as-sistance, which often used to be provided through informal networks and charity, are increasingly taken over by insurance schemes or other welfare state arrangements. This argument reflects a perception of society which is consistent with Weber’s (1922[1987]) notion of zweckrationalität, in which social action is driven by its rational, calculated ends. When social needs are

15Philanthropy in the welfare state Introduction

taken over by modern bureaucratic states, private participation fades away.Other scholars have argued for a reverse effect, in which government sup-

port “crowds in” civic engagement. In sociological research, the theoretical foundations for a positive effect of government support come from (neo-)in-stitutionalism. Scholars like Gøsta Esping-Andersen (1990) and Bo Rothstein (1998) argue that the scope and the structure of welfare state institutions determine individual attitudes. Citizens living in extensive welfare states are hypothesized to have a more egalitarian view on social justice (Arts & Gelis-sen, 2001; Svallfors, 1997), more social trust (Uslaner, 2003) and a higher willingness to help others (Kääriäinen & Lehtonen, 2006). This view is con-sistent with Weber’s (1922[1987]) wertrationalität, in which social action is driven by values and beliefs rather than by its calculated consequences.

This thesis investigates the relationship between government support and private charitable donations. Although the current political discourse seems to focus on a more active role of individual participation and informal social support, nonprofit organizations are indispensable actors in welfare state reforms (Schuyt, 2014). Being more or less independent from democratic control and partly funded by philanthropic donations, nonprofits provide a wide range of services. In the Netherlands, Burger, Dekker, Toepler, Anheier, and Salamon (1999) estimate that the Dutch nonprofit sector receives about 60% from public funding and about 3% from private donations. The non-profit sector in the Netherlands, especially in the fields of health, education, social services and international aid, is dominated by public money. Collab-orations between government and nonprofit organizations are prevalent in health, social welfare, international development and many other fields. With the availability of resources being a crucial factor for organizational ef-ficacy and successful collaborations (Ansell & Gash, 2008; Pfeffer & Salancik, 1978), it is important to know how philanthropy is affected by changes in government support.

In contemporary research, crowding-out and crowding-in of charitable donations are investigated in different ways and in different social science disciplines. In behavioral economics, the crowding-out hypothesis has been tested by examining the donation behavior of individuals to collective goods and voluntary organizations in response to tax-funded government contri-butions (e.g. Andreoni, 1993; Eckel, Grossman, & Johnston, 2005; Gronberg, Luccasen, Turocy and Van Huyck, 2012; Steinberg, 1985). Yet another strand of research is dominated by sociology and political science, where differenc-es between countries in levels of civic participation are studied in relation to


different types of welfare state regimes with varying levels of public support for social programs (e.g. Gesthuizen, Van der Meer, & Scheepers, 2008; Pen-nerstorfer & Neumayr, 2017; Van Oorschot & Arts, 2005). The first strand of research, which uses experimental research methods, mainly finds low-er donations in situations of higher government spending, while the second strand, where comparative cross-sectional surveys are more common, typ-ically finds higher donations in situations of higher government spending.

The current thesis aims to bridge the gap between the two strands of re-search. The variety of findings is partly due to the assumptions made in dif-ferent research designs (Payne, 2009; Ribar & Wilhelm, 2002; Tinkelman, 2010). By identifying and examining a number of these assumptions, this dissertation proposes explanations for the inconclusive findings. It does so by adopting a multi-method approach, including both experimental and non-experimental research designs. In the crowding-out literature, it is the first to explore cross-country data on individual amounts donated to non-profit organizations, the first to examine longitudinal survey data, and the first to carry out a content analysis on news media.

The remainder of this Introduction provides (1) the definitions of the main concepts used throughout this dissertation, (2) a brief history of philanthro-py in the Dutch welfare state, (3) the policy context of charitable donations and government support in the Netherlands, (4) a brief discussion of the gaps in the academic literature, (5) the research aims of the dissertation, and (6) the outline of this book.

DEFINITIONS

The original crowding-out hypothesis was formulated as decreasing civic participation in expanding welfare states, but the same line of reasoning predicts increasing donations after decreasing government support. Crowd-ing-out in this thesis refers to any negative association between government support and charitable donations, while crowding-in refers to any positive association.

Charitable donations are defined as voluntary donations by individuals or private households to nonprofit organizations. In the Netherlands, house-holds donate approximately 2.6 billion Euros annually to nonprofit organi-zations, with the largest subsectors in terms of amounts received being re-ligion, international aid and health (Bekkers, Schuyt, & Gouwenberg, 2017).


This thesis examines donations to health, international aid, environment, nature conservation, animal protection, education and research, culture and arts, and public benefits. Religion is not considered because religious con-gregations, at least in the Netherlands, do not receive any direct government support. While philanthropic giving by corporations, foundations and lotter-ies are also important sources of private income for nonprofit organizations, they are excluded here. Informal giving, like financially supporting families or friends, is also not considered.

The main independent variable is financial government support (or: gov-ernment funding), which can take many forms. It includes subsidies to non-profit organizations as well as expenditures that directly target social needs, like unemployment grants. Financial support can be provided by central government or lower levels of government. Government support in this the-sis is restricted to unconditional financial support. This excludes matching schemes that are conditional on other contributions and tax incentives like charitable deductions in the income tax.

PHILANTHROPY IN THE DUTCH WELFARE STATE

Different welfare states have a different historic tradition of philanthropy, which might influence today’s responses of nonprofit organizations and charitable donors to changes in government funding policies. While most research on crowding-out comes from the United States, the empirical work in Chapters 4 and 5 of this thesis use data from the Netherlands, a country with a rich tradition of philanthropic initiatives in modern history.

Rooted in the mixed provision of welfare by churches and local authorities in the Dutch Republic, church boards and town councils were already active in organizing successful charitable collections for poor relief in the 17th and 18th century (Teeuwen, 2014). The ninetieth century was characterized by the local organization of charity through (protestant, catholic or municipal) almshouses, with little influence of central government (Heerma van Voss & Van Leeuwen, 2012; Van Leeuwen, 1999). De Swaan (1988) argues that the development of social programs was a response of wealthy citizens to pro-tect themselves against the undesired consequences of poverty and health problems in industrial cities. The 20th century was characterized by the expansion of the centralized welfare state. Poverty reduction expenditures dramatically rose and the Dutch government introduced different social pro-


grams in the fields of health and social insurance, which made large parts of the work of private charities no longer necessary (Schuyt, 2010; Van Leeu-wen, 1999).

The Central Bureau on Fundraising (CBF) has played an important role in the development of the philanthropic sector from 1925 until now (Wort-mann, 2007). From the late 1980s onwards the philanthropic sector further developed with the formation of umbrella organizations, the introduction of vocational training programs and the start of the Giving in the Nether-lands research project. The professionalization of the philanthropic sector took place in an era of economic prosperity and a political preference for the outsourcing of public services.

Yet, relative to most other Western countries, the government remains very influential in funding public services. Salamon and Anheier (1998) place the Netherlands in the social-democratic model based on the relation-ships between the government and the nonprofit sector. Especially in the fields of social services, health and education, many nonprofit organizations are heavily subsidized with public money (Burger et al., 1999).

CHARITABLE DONATIONS AND GOVERNMENT SUPPORT IN THE NETHERLANDS

To provide a glance of the association between philanthropy and govern-ment spending, Figure 1 shows the development of charitable giving and government support in different subsectors of the Dutch nonprofit sector. Data on household giving are from the Giving in the Netherlands Panel Sur-vey (Bekkers, Boonstoppel, & De Wit, 2016). Government expenditures on health, social protection and international aid (Official Development Assis-tance) are adopted from the OECD, while government expenditures on cul-ture come from Statistics Netherlands. Figure 1 shows the total amounts as a percentage of Gross Domestic Product, with charitable giving multiplied by 10, 100 or 1,000 in order to have both variables on a similar scale.

The health subsector is the part of the nonprofit sector with the largest number of charitable donors, including large fundraising organizations like KWF Kankerbestrijding (Dutch Cancer Society) and Alzheimer Nederland. Still, government expenditures on health care are 100 times higher than the total amount of charitable donations. The Dutch health sector receives a large part of total funding from the government, with hospitals being largely


funded with public money (Burger et al., 1999). Public health expenditures show a steady increase in the period 2000-2010, while charitable giving de-creases in most of the years. The trends are not exactly reversed, however. Government support and charitable giving both increase from 2001 to 2003 and from 2007 to 2009, and both decrease in the years after 2009 until 2015. The rise in public health costs, partly due to ageing, have been a recurrent topic in political debates. In 2006, a new Health Insurance Act came into force, with obligatory basic insurance and a competitive market for health insurance companies, which was partly aimed to reduce total health care spending.

Social expenditures are relatively high and include pensions, disability benefits, unemployment grants and redistributive taxes. These expenditures are 1,000 times higher than the total of charitable donations to social ser-vice organizations. In the years 1997-2003, social expenditures initially de-crease, then stabilize and then increase, while giving follows the opposite pattern. From 2003 till 2008, the trends are similar. After 2008, the trend in giving is opposite to the trend in government expenditures. Recent policy developments include an increased attention to social participation in local communities under the Social Support Act (2007, revised in 2015) and the Participation Act (2015), which could have led to a lower provision of public services and an increased number of clients in nonprofit service organiza-tions like the Salvation Army and De Regenboog.

The Dutch government gives, relative to other countries, a high percentage of Gross National Income to international development. In 2015 the Dutch ranked fifth in Europe with 0.75% of Gross National Income spent on Official Development Assistance. Here, too, a large share of the government spend-ing flows through nonprofit organizations. Changes in giving to international aid organizations seems to precede changes in Official Development Assis-tance (ODA). An increase in giving from 1997 to 1999 was followed by an increase in government support from 1999 to 2000; the decrease from 1999 to 2003 was followed by a decrease in government support from 2000 to 2004. Both philanthropy and government support peaked in 2005, which is largely due to contributions to victims of the December 2004 Tsunami. After 2005, household giving decreased to low numbers, where it stabilized in the period 2005-2015. In 2010, the conservative Rutte administration carried out large cuts in public budgets for organizations like Oxfam Novib, Hivos and ICCO Cooperation. Among international development organizations, lay-offs and hasty organizational reforms were necessary. “We celebrate that we


have to focus more on the private”, an ICCO spokesperson said at the time, “but we hardly get any time for it” (Cats, 2010).

Government expenditures to culture are around 0.6% of GDP, which is 100 times higher than the total amounts donated. Total amounts given to culture organizations are relatively low and volatile. It is not easy to recognize a pat-tern in the relationship between government support and donations. Large budget cuts were announced in 2010, parallel to those on international de-velopment. As stated in the introduction, the government explicitly expected cultural institutions to raise more funds in the private market, supported with an increased tax deductibility of private donations to cultural organi-zations. The lower availability of public funding led to increased fundraising efforts in the sector, although this has not led to a rise in household giving across the board (Franssen & Bekkers, 2016). Museums report decreasing subsidies and, especially among the bigger attractions, higher income from fees and private sources (Museumvereniging, 2015). This is part of a broader trend among social and cultural organizations. Local facilities like libraries, public swimming pools and community centers are increasingly urged to rely on volunteers instead of paid forces, and to find funding sources other than government subsidies.

When looking at the aggregate numbers, the largely decreasing trends in charitable giving do not seem to be strongly affected by changes in govern-ment support. The exact consequences of policy shifts on individual behav-ior are highly uncertain. With budget cuts motivated by a mix of ideological beliefs and budget constraints, it is important to have a better knowledge about the effects of increasing and decreasing government support. When charitable donations are able to (partly) compensate for budget cuts, this would support policy choices with reduced public budgets and a large role for nonprofit organizations.

GAPS IN THE LITERATURE

In the introduction, the contradictive findings in two different strands of re-search were described. The lack of conclusive evidence in previous research is often attributed to methodological choices, which ignores the possibility of substantial differences in the context of different studies that lead to dif-ferent findings (Ribar & Wilhelm, 2002). There is a lack of knowledge on the conditions under which charitable giving is affected by government poli-


Figure 1: Government expenditures and charitable giving to health, social services, international aid and culture, 1997-2015

1997 1998 1999 2000 2001 2002 2003 2004Health expenditures5% 5% 5% 5% 5% 5% 6% 6%Giving (x 100) 5% 6% 7% 6% 5% 6% 6% 5%

0%

2%

4%

6%

8%

10%

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Health expenditures

Giving (x 100)

0%

10%

20%

30%

40%

50%

1997 1999 2001 2003 2005 2007 2009 2011 2013 2015

Social expenditures

Giving (x 1000)

0.0%

0.4%

0.8%

1.2%

Official DevelopmentAssistance

Giving (x 10)

0.0%

0.2%

0.4%

0.6%

0.8%

1997 1999 2001 2003 2005 2007 2009 2011 2013 2015

Culture spending

Giving (x 100)

0.0%

0.2%

0.4%

0.6%

0.8%

1997 1999 2001 2003 2005 2007 2009 2011 2013 2015

Culture spending

Giving (x 100)


cies. Crowding-out is often studied as a constant mechanism across contexts, although it is very likely that the empirical results depend on the assump-tions in the research (Tinkelman, 2010). This dissertation aims to identify the mechanisms and conditions that cause these mixed findings, thereby ad-dressing three issues that are largely overlooked in the literature.

First, mechanisms that explain the relationship between government sup-port and charitable donations (mediating variables) are often not explicitly tested. The main explanation for crowding-out effects is altruism. Research in behavioral economics revealed, under controlled conditions, that donors to some extent give because they care about the recipient. Their contributions can be done mandatory, through tax-funded government support, or volun-tary, through philanthropic donations (Roberts, 1984; Warr, 1982). When altruism is the only motivation, every Euro of government support would crowd out a Euro of donations. This pure altruism model does not hold to the extent that donors derive private benefits from their donation (Andreoni, 1989, 1990). In this line of research, donor motivations are mostly deducted from giving behavior in laboratory experiments, but the explaining process itself is not measured. In sociology and political science, the main argument is that citizens are socialized by the institutions that are surrounding them (Rothstein, 1998; Ingram & Clay, 2000). Cross-country surveys often exam-ine correlations between welfare state characteristics and different types of social and civic participation, without measuring mediating variables. Caus-al mechanisms like the signaling function of government support (Handy, 2000; Heutel, 2014; Schiff, 1990) and organizational behavior (Andreoni & Payne 2003, 2011) are understudied.

Second, there has been little attention to the conditions under which crowding-out does or does not occur (moderating variables). Crowding-out is likely to occur under certain assumptions, including the availability of information (Horne, Johnson, & Van Slyke, 2005), the number of other do-nors (Ribar & Wilhelm, 2002) and the level of government support (Brooks, 2000a). Although different variables have been proposed that might moder-ate the relationship between government support and charitable donations, systematic analyses of the relationship under different conditions is lacking. While the availability of information about government policies seems a pre-requisite for individual donors to be responsive to changes in government support, research in this area often does not take this variable into consid-eration.

Third, there is a lack of attention to individual heterogeneity in responses


to changing government support. If some social groups are crowded out by government support and other groups are crowded in, research on aggregate statistics could find zero correlation on average. There are many ideas about which people donate to charitable causes (Bekkers & Wiepking, 2011a; Wiepking & Bekkers, 2012) and why (Bekkers & Wiepking, 2011b). Hypoth-eses on individual heterogeneity in responses to government support are scarce, however, whereas they would have large implications for fundraisers who may target different segments of the fundraising market. There have been a few attempts to distinguish between income groups (Kingma, 1989) and donor types (Reeson & Tisdell, 2008; Luccasen, 2012), but the findings are inconclusive.

RESEARCH AIMS

In order to contribute to solving these issues in the literature, this research poses five research questions. To examine the main relationship under study, it answers the question to what extent government support affects individu-al charitable donations. In order to test the explanatory power of mediating variables, it examines how government support affects individual charitable donations. Moderating variables are incorporated in the questions where and under which conditions government support affects individual charita-ble donations. Individual heterogeneity is addressed in the question among whom government support affects individual charitable donations.

By answering these five questions, the current thesis adds new evidence with innovative research designs. Every chapter uses fresh data, either self-collected or compiled from different existing data sources. Three studies are conducted in the Netherlands. Empirical crowding-out studies have been non-existent in this country thus far.

RQ1: To what extent does government support affect charitable dona-tions?The main relationship under study is the one between government support and the total provision of private donations. It is important to know the sum of contributions from public and private sources, which determines the quality, nature or even the mere existence of public services. To derive con-clusions about macro phenomena, empirical social science should measure individual behavior (Hedström & Bearman, 2009: 11). Therefore, the main


empirical relationship under study is the one between government support and individual donations.

Many studies in the crowding-out literature use either data from labora-tory experiments, which allow for causal inference but may evoke doubts about external validity. Other studies use data from nonprofit organizations, which measure aggregate private revenues from different sources (including households, corporations and foundations) instead of individual behavior. The empirical studies in this book examine individual donation decisions, either observed or self-reported, in situations outside the research laborato-ries. The dissertation uses a cross-country dataset compiled from different surveys, a sample of crowding-out effect sizes estimates from previous em-pirical studies, a panel dataset that allows for examining changes over time, a scenario module in a larger survey, and a survey experiment that uses actual information to an existing nonprofit organization.

Figure 2: Theoretical model of RQ1


Government support

Private donations

RQ2: How does government support affect charitable donations?The current dissertation examines two possible mediating variables in the relationship between government support and individual donations. First, the fundraising behavior of nonprofit organizations might (partly) explain why donations are affected by government support. On the one hand, it could be that organizations with lower public funding are more urged to invest in campaigns to raise funds in the private market (Andreoni & Payne, 2003, 2011). On the other hand, it could be that organizations use public money to


professionalize their strategies and be more successful in the private market (Bekkers, 2013a).

Second, government policies are a source of information for possible do-nors. In general, public expenditures express political priorities. More spe-cifically, government grants to specific organizations or projects can be a sig-nal that it is trustworthy (Handy, 2000; Heutel, 2014) and serve as a “seal of approval” (Schiff, 1990). This could encourage charitable donations.

Both fundraising and information as mediating variables are tested with organizational-level data. Fundraising expenditures and revenues from gov-ernment subsidies of 19 large nonprofit organization in the Netherlands are adopted from the Central Bureau on Fundraising (CBF). To measure the in-formation variable, the LexisNexis database is used to retrieve newspaper articles about government support to these organizations. These data are matched with micro-level data on donations from the Giving in the Nether-lands Panel Survey from 2002 to 2014.



Government support

Private donations

Fundraising

Information

RQ3: Where does government support affect charitable donations?The effects of government support might depend on the context in which it is provided. First, the relationship might vary across welfare state regimes. While most crowding-out studies come from the United States, it is uncer-tain to what extent the association between government support and chari-table donations is the same in countries with different welfare state arrange-ments. In their classification of nonprofit regime types, which is based on the

26 Introduction Philanthropy in the welfare stateFigure 4: Theoretical model of RQ3

Government support

Private donations

Nonprofit subsector

Nonprofit regime

well-known welfare state regime types as distinguished by Esping-Andersen (1990), Salamon and Anheier (1998) place the United States in the liberal regime. From all nonprofit regime types, it is in the liberal regime that non-profits play the largest role in the provision of public and social services in contrast to the government, resulting in a substituting relationship between government expenditure and philanthropic giving. In the Social-Democratic regime type, to which the Netherlands belongs (cf. Einolf, 2016) and where the government plays a larger role, a complementary relationship is more likely.

Second, the relationship might differ across nonprofit subsectors. There have been multiple studies that tested crowding-out with similar datasets in different subsectors, without conclusive evidence about different effects of government support (e.g. Khanna & Sandler, 2000; Khanna, Posnett, & San-dler, 1995; Okten & Weisbrod, 2000). In a systematic literature review of non-experimental crowding-out findings, Lu (2016) shows that government expenditures and philanthropic donations are generally negatively related in the field of human services, while they are positively related in the fields of health and the arts. This finding is a fruitful basis for theory-building. It could be that crowding-out is strongest in domestic social services. This is supported by cross-national research that finds social protection spending to be negatively related to charitable giving to social services (Pennerstorf-er & Neumayr, 2017; Sokolowski, 2013). The latter two studies propose an additional hypothesis about substitution between subsectors. It could be that domestic social services spending drives donors towards more “expres-sive” subsectors like environment or the arts, an effect that has been labeled “philanthropic flight” (Sokolowski, 2013) or “crosswise crowding-in” (Pen-nerstorfer & Neumayr, 2017).

In this dissertation, differences between nonprofit regimes as well as dif-ferences between nonprofit subsectors are examined using cross-country data from the Individual International Philanthropy Database (IIPD). Using longitudinal data on 19 organizations, adopted from the Central Bureau on Fundraising (CBF) and the Giving in the Netherlands Panel Survey (GINPS), differences between subsectors of the Dutch nonprofit sector are explored. Furthermore, a survey module in which respondents are presented with scenarios about hypothetical budget cuts provides a more systematic test of differences between nonprofit subsectors.

27Philanthropy in the welfare state IntroductionFigure 4: Theoretical model of RQ3

Government support

Private donations

Nonprofit subsector

Nonprofit regime


RQ4: Under which conditions does government support affect charita-ble donations?It is likely that the relationship between government support and charitable donations is not the same under all conditions. Following the well-known Thomas theorem (Thomas & Thomas, 1928), it is important to remember that individual action is driven by how the world is perceived rather than by objective external conditions. As such, the availability of information may determine the existence and magnitude of the association between govern-ment support and private donations. Information may be obtained in one’s social network, via social media, through fundraising materials or through news media (Li & McDougle, 2017; McDougle & Handy, 2014). Information is not only a mediating variable when it follows from government support (see RQ2), it can also be a moderating variable. Although the availability of information seems an obvious prerequisite for donations to be responsive to policy changes, its variability has been largely ignored in the literature thus far. Previous studies showed that many people fail to estimate the correct percentage of public funding to nonprofit organizations (Horne et al., 2005) and often fail to classify firms as public, non-profit or for-profit (Handy et al., 2010). If potential donors are unaware of changes in government support, they will not respond to them.

This dissertation will examine information as a moderating variable in two ways. First, it uses a sample of newspaper articles on nonprofit organi-



Government support

Values Private donations

Resources

zations from 2002 to 2014 to test the availability of information in the Dutch nonprofit sector. Second, it contains a survey experiment in which respon-dents are randomly assigned to a condition of information about actual gov-ernment funding to an existing nonprofit organization.



Government support

Information

Private donations

RQ5: Among whom does government support affect charitable dona-tions?Charitable donations are motivated by a mix of different motivations. Lab-oratory experiments generally find that contributions to public goods are both driven by the consequences for recipients and by their intrinsic value (Andreoni, 1993; Chan, Mestelman, Moir, & Muller, 1996; Eckel et al., 2005; Güth, Sutter, & Verbon, 2006; Isaac & Norton, 2013; Korenok, Millner, & Raz-zolini 2012, 2014). In fMRI scans, mandatory contributions are shown to elicit neural activity in areas linked to reward processing, and this activity further increases when people make transfers voluntarily (Harbaugh, Mayr, & Burghart, 2007). While behavioral experiments often aim to estimate the average response in a sample of individuals, it is likely that different citizens have different responses to government support. In a first step towards the-ory building, existing theories on civic engagement can be used to predict changes in charitable donations as a response to changes in government support.

The Civic Voluntarism Model (Verba, Schlozman, & Brady, 1995) predicts civic voluntarism by individual resources, values and recruitment. While re-


cruitment is covered by the mechanism of fundraising (RQ3), one might ex-pect people with more resources and stronger prosocial values to be more likely to substitute changes in government support. People with more re-sources, who are known to be larger donors (Wiepking & Bekkers, 2012), might be more responsive to government support. They have more financial capacity, and an increase in giving have lower relative costs compared with those with less resources. Also, donors with stronger prosocial values, like empathic concern or the principle of care (Bekkers & Wilhelm, 2016), are more committed to charitable causes. Those with strong values might care more about the consequences of reduced funding of nonprofit organizations, and therefore be more inclined to compensate.

The moderating effects of resources and values are examined with differ-ent studies based on data from the Giving in the Netherlands Panel Survey (GINPS). This survey is representative for the Dutch population and allows for testing responses to changing government support in different social groups.



Government support

Values Private donations

Resources


OUTLINE

The remainder of this dissertation contains five chapters, divided in two parts.

Part I explores the context of the crowding-out hypothesis by describing how government support and charitable donations are associated across countries and across previous academic studies (RQ1).

Chapter 1 compares countries in terms of their levels of charitable giving and government expenditures. It is the first cross-country analysis that uses individual-level data on amounts donated to nonprofit organizations. The study not only examines aggregate levels of public and private contributions but also distinguishes between different nonprofit subsectors. It explores how crowding-out varies between nonprofit regimes and nonprofit subsec-tors (RQ3).

While Chapter 1 provides important insights in the levels of charitable donations in different contexts, it does not tell us much about the causal re-lation between government support and charitable donations. Therefore, Chapter 2 presents a systematic literature review of previous studies on the relationship between government support and charitable donations, aimed to examine how empirical findings are associated with different research de-sign characteristics.

Part II provides original empirical analyses on the association between government support and charitable donations (RQ1), using data from the Netherlands.

Chapter 3 examines private giving and government support to 19 large or-ganizations in the Netherlands across nonprofit subsectors (RQ3). By adding organizational data and a content analysis on newspaper reports about these organization, fundraising and information are examined as possible mediat-ing variables (RQ2). Using survey data on individual background character-istics, the moderating effects of resources and values are examined (RQ5).

The role of information is further investigated in Chapter 4. In an experi-mental design, participants are presented with information about an actual change in government support to a large organization in the field of health care. By examining levels of donations as well as the perceived change in government support, this enables a test of the availability of information as a condition under which charitable donations are responsive to government policies (RQ4). Once again, resources and values are included as moderating variables (RQ5).


Chapter 5 examines a survey module in which respondents are asked whether they would increase their giving in a scenario of hypothetical bud-get cuts. Also, the actual change in giving is examined two years later. By comparing the responses in different scenarios, this study provides a test of crowding-out in different nonprofit subsectors (RQ3). The correlates of back-ground characteristics allow for testing the moderating effects of resources and values (RQ5).

With these two parts, this thesis both explores the current state of the crowding-out hypothesis and examines new empirical evidence. After Part II, a Conclusion provides a summary of the findings and their implications for theory, research, the nonprofit sector, and policy.

Part One

CROWDING-OUT IN CONTEXT

This chapter makes a start with exploring the association between gov-ernment support and charitable donations across countries. It is the first cross-country study to correlate government expenditures with the level of individual private donations to different fields of social welfare. Using the new Individual International Philanthropy Database (IIPD), this chapter explores the association between government expenditures and philan-thropic donations to different social welfare sectors across 19 countries. The results of the descriptive and multilevel analyses support the hypoth-esis that in countries where government expenditures in health and social protection are higher, there are more donors in “expressive” sectors like environment, international aid and the arts. People in generous welfare states are more likely to donate, but they donate amounts that are similar to those made by donors in less generous welfare states. The results thus reject the crowding-out hypothesis and give a nuanced picture of the rela-tionship between public funding and philanthropic giving across different fields of social welfare.

Exploring crowding-out with cross-country data

Chapter 1

Arjen de Wit, Michaela Neumayr, Pamala Wiepking and Femida Handy

MN, AdW and FH designed the study; PW and FH collected the micro-level data; AdW collected the macro-level data; AdW and MN carried out data analysis; AdW, MN, FH and PW contributed to writing the article.

Accepted for publication as: De Wit, A., Neumayr, M., Wiepking, P., & Handy, F. (forth-coming). Do Government Expenditures Shift Private Philanthropic Donations to Par-ticular Fields of Welfare? Evidence from Cross-Country Data. European Sociological Review.

Supplementary materials are available through the Open Science Framework at http://osf.io/6jwh2/.

36 Chapter 1 Exploring crowding-out with cross-country data

INTRODUCTION

The relationship between the welfare state and civic engagement is a top-ic of recurrent discussions (Andreoni & Payne, 2011; Van Oorschot & Arts, 2005). Tracing back to the theoretical foundations of Alexis de Tocqueville (1970[1840]), it is hypothesized that the development of the modern wel-fare state “crowds out” citizens’ own, private, initiatives such as informal caring relations and self-help (e.g. Künemund & Rein, 1999; Suanet, Broese Van Groenou, & Van Tilburg, 2012). Additionally, critics argue that more gen-erous government expenditures creating public goods and services discour-age citizen’s involvement in the creation of these public goods and services. For example, when the local municipality provides shelter for the homeless, there is less need for citizens to contribute to nonprofit organizations that target homelessness. To what extent the increase of higher government so-cial expenditures is associated with lower private contributions to the public good is widely studied in the economic literature as the crowding-out hy-pothesis (Abrams & Schmitz, 1978; Andreoni, 1993; Brooks, 2004; Kingma, 1989; Roberts, 1984; Warr, 1982).

However, researchers from a range of disciplines have rejected the crowd-ing-out hypothesis by arguing that “a well-developed welfare state creates the structural and cultural conditions for a thriving and pluralist civil so-ciety” (Van Oorschot & Arts, 2005: 6). They posit that more generous gov-ernment expenditures promote civil society organizations and encourage private engagement in the form of philanthropic contributions of money and time (Anheier & Toepler, 1999; Khanna & Sandler, 2000: 1544; Rose-Acker-man, 1981). Another explanation arguing in favor of this “crowding-in”-hy-pothesis is that the support of nonprofit organizations by governments acts as a signal of the organizations’ quality and thus crowds-in private funding (Handy, 2000; Heutel, 2014).

Despite a large number of empirical studies, the debate is unsettled. A meta-analysis of empirical studies of the relationship between government expenditures and private contributions shows that two thirds of all findings reveal a negative association, supporting the crowding-out hypothesis, while one third finds a slightly positive association, in support of the crowding-in hypothesis (De Wit & Bekkers, 2017).

Challenged by these diverse findings, we cannot state unequivocally that government social welfare expenditures crowd out private contributions. Systematic literature reviews show that estimates of the effects of govern-

37Exploring crowding-out with cross-country data Chapter 1

ment expenditures on philanthropic giving are widely disparate and argue to establish a “contextual dependence” to validate the legitimacy of the crowd-ing-out hypothesis. In other words, the findings depend on the measures used, the nature of the government expenditures, the nonprofit subsectors involved, or other moderating factors (De Wit & Bekkers, 2017; Lu, 2016). The large majority of studies refer to data from the U.S. and less is known about the relationship in countries with different traditions of the welfare state (De Wit, 2016; Bekkers, 2016).

There are few studies that investigated the crowding-out effect across countries, with two drawbacks. First, due to the lack of data availability, these studies could only investigate measures on the decision to give or not (Bredt-man, 2016; De Wit, 2016; Gesthuizen et al., 2008; Pennerstorfer & Neumayr, 2017), or an aggregate measure of private nonprofit revenues rather than individual donations (Sokolowski, 2013). In order to study how individu-al philanthropic donations respond to government programs, information about the level of these individual donations is required. Second, many em-pirical studies examine aggregate measures of total government funding and philanthropic donations, and it is unlikely that a relationship is unidirection-al across all fields to which governments make contributions (Brooks, 2004: 173; Khanna et al., 1995; Khanna & Sandler, 2000; Lu, 2016). For example, it is likely that correlations are positive in one subsector and negative in the other, which might be the reason for inconclusive findings in prior studies.

Against this background, our research question states: To what extent are government social expenditures associated with philanthropic giving in dif-ferent nonprofit subsectors? By using a new cross-country database we are able to examine for the first time and in different subsectors how govern-ment expenditures are associated with the incidence and level of individual philanthropic donations, across a range of 19 countries with a large diversity in welfare state traditions. Thus, to answer our research question, we first examine the relationship between government expenditures and individual philanthropic donations across countries. Second, we test the crowding-out hypothesis across different nonprofit subsectors to understand how govern-ment support of one subsector may result in either crowding-out in some subsectors or crowding-in in other subsectors. We also examine “crosswise crowding-in”: whether an increase in government expenditure in one sub-sector may lead to an increase of individual philanthropic donations in other subsectors.

Our paper contributes to the important debate about the role of philan-


thropic donations in the light of changing government support, and whether private philanthropic donations can be seen as supplementary or comple-mentary to government support for different public goods and services (Lecy & Van Slyke, 2013; Salamon, Sokolowski, & Anheier, 2000; Young, 2000). In an era of changing relations between public and private actors, and of con-tinued pressure on governments to decrease public social expenditures, the results of this study may provide important insights on the capability and willingness of citizens in different countries to engage in the voluntary pri-vate funding of different types of public goods. The findings of this study thus have significant consequences for public policy depending on if, and to what extent, philanthropic giving is displaced by government expenditures in var-ious nonprofit subsectors (Andreoni & Payne, 2011; Bekkers, 2016; Bonoli, George, & Taylor-Gooby, 2000; European Commission, 2013).

THEORY

Crowding-outIndividual contributions to the public good can be made either mandato-ry, through government taxes, or voluntarily, by philanthropic donations to nonprofit organizations providing that public good. If and how government expenditures attenuates philanthropic giving (or not) is debatable based on the evidence in the literature to date, and remains one of the most exten-sively discussed questions in public economics (Andreoni & Payne, 2011: 334). Early economic studies of the voluntary private provision of public goods argue that “preferences are assumed to be purely altruistic” (Andre-oni, 1988: 57) that is, individuals making philanthropic donations receive no utility from the very act of giving the gift, as utility is related only to the consumption of private goods and the total supply of the public good. If peo-ple are purely altruistic, an increase in tax-financed government spending leads to a concomitant reduction of private donations, thereby keeping the total individual contribution (voluntary and involuntary) at the same level (Roberts, 1984; Warr, 1982). After all, pure altruists do not care whether the public good is realized through voluntary or involuntary contributions; they just care about realizing the public good. The prevailing assumption thus suggests a full crowding-out: that an increase in public expenditures by for instance one dollar persuades purely altruistic donors to decrease their own philanthropic contributions by one dollar – and vice versa (Brooks, 2004:


168; Payne, 1998: 324). In addition to this crowding-out explanation, there are other reasons to

expect crowding-out. Donors might hesitate to make donations to organi-zations receiving government subsidies, in those contexts wherein such or-ganizations are seen either as not viable, or as the long arm of the govern-ment (Brooks, 2004: 172). Other scholars suggest that when organizations receive government subsidies, they decrease their fundraising efforts among the public, resulting in decreased individual donations (Andreoni & Payne, 2011; Khanna & Sandler, 2000: 1545).

However, there are a couple of reasons to expect partial crowding out but not a full crowding out. If individuals are incentivized to give because of oth-er motivations than altruism, such as to feel good about oneself, to enhance one’s reputation or to conform with social norms or social pressure (Bek-kers & Wiepking, 2011b), they will give regardless of who else contributes or does not contribute to the public good (Andreoni, 1989, 1990). To the extent that donors derive private benefits from the act of donating, like the warm glow and reputation gain, their donations would not be responsive to chang-es in contributions from a third party like the government (Payne, 2009).

Crowding-inBesides arguments for crowding-out, there are reasons to expect that gov-ernment expenditures and philanthropic donations are positively associated. The findings of crowding–in of private donations on the heels of increased public support rely on the signaling value of government expenditure. Phil-anthropic donors generally prefer to give to organizations that are well-es-tablished, which they perceive as being trustworthy and under information uncertainty; government subsidies, in some contexts, is seen as a “seal of approval” of the nonprofit organization (Handy, 2000; Schiff, 1990). In addi-tion, non-profits may gain significant scaling advantages in their operations due to government support, which might increase their scope and motivate donors who care about impact (Anheier & Toepler, 1999; Khanna & Sandler, 2000: 1544; Rose-Ackerman, 1981).

Another argument for a positive correlation between government expen-ditures and philanthropic giving is provided by (neo)institutionalist theo-ries, which posit that people adopt values and norms from the institutions surrounding them (Rothstein, 1998; Ingram & Clay, 2000). In this line of lit-erature, attitudes towards social policies are shaped by the way a welfare state is structured (Arts & Gelissen, 2001, Jæger, 2006). Countries with a


higher productivity growth are able to spend more on health care, education and other social issues (Baumol, 1996), and it could be that generous and universal welfare states “socialize” people to be more benevolent. Hence, people in generous welfare states would develop stronger pro-social values that encourage philanthropy.

The causal relationships between social values, welfare state generosity and philanthropy are hard to disentangle. If the median voter theorem holds, political outcomes are in line with preferences at the center of the elector-ate’s political spectrum. As such, choices in welfare state spending express social values. When philanthropic giving is partly an expression of the same values, government expenditures and philanthropic giving will be positively correlated.

Empirical evidenceThe majority of prior studies find that there is some form of partial crowd-ing-out, meaning that a dollar of public grants crowds-out donations by less than a dollar (Brooks, 2004: 173). Some studies find no significant relation-ship between government expenditures and private giving (Brooks, 1999), and other studies find a crowding-in effect, i.e. that the level of government grants is positively correlated with private donations (Andreoni & Payne, 2011; Hughes & Luksetich, 1999; Payne, 1998). A recent meta-analysis, that systematically reviews previous studies on crowding-out, shows that the results are strongly shaped by methods used; for example, in experimental studies a one dollar increase in government expenditures is associated with an average decrease of about 0.64 dollars, while non-experimental data anal-yses find a crowding-in effect of about 0.06 dollars on average (De Wit & Bekkers, 2017).

The vast part of the empirical literature is based on within-country vari-ance in government spending. It is the question whether these findings tell us something about differences between countries. Most previous cross-coun-try studies find either positive correlations or no statistically significant re-lationship between measures of government expenditures and philanthrop-ic donations (De Wit, 2016; Einolf, 2016; Gesthuizen et al., 2008; Nguyen, 2015; Pennerstorfer & Neumayr, 2017; Sokolowski, 2013).

Some cross-country studies examine only Western countries (De Wit, 2016; Gesthuizen et al., 2008; Pennerstorfer & Neumayr, 2017). However, it could be that effects found in Western countries do not apply to other welfare state contexts. Using broader samples of developed and developing


countries, both Sokolowski (2013) and Einolf (2016) find positive correla-tions across the board. The latter two studies show correlations based on aggregate country-level statistics of individual giving behavior, which makes them vulnerable to the ecological fallacy (Piantadosi, Byar, & Green, 1988).

Based on the theoretical reasoning and empirical evidence related to both crowding-out and crowding-in, we formulate two rival hypotheses:

Crowding-out hypothesis: Higher levels of total government expenditures are associated with lower levels of total private donations across nations.Crowding-in hypothesis: Higher levels of total government expenditures are associated with higher levels of total private donations across nations

Nonprofit regime typesEmpirical evidence gives reason to assume that the relationship between government expenditure and private donations is much more complex, de-pending not only on the motivations of the donors but also on institutional settings (Sokolowski, 2013: 359). Referring to social origins theory, Salamon and Anheier (1998) point out that the relationships between government (social) expenditures, the size of the nonprofit sector and the role of philan-thropy within a country are not related in a linear way, but that those rela-tionships differ depending on the nonprofit regime of a country. Based on the classifications used by Esping-Anderson (1990), Salamon and Anheier (1998) identified a liberal, a social-democratic and a corporatist nonprofit regime and add the so called statist regime.

From all nonprofit regime types, it is in the liberal regime that nonprofits play the largest role in the provision of public and social services in contrast to the government, resulting in a substituting relationship between govern-ment expenditure and philanthropic giving. In addition, in the liberal regime philanthropic income is arguably the largest source of funding for the non-profit sector, next to government subsidies and fees for services (Salamon & Anheier, 1998: 243). In contrast, in the corporatist regime, government and nonprofit sector expenditure and philanthropic giving are much more com-plementary. In this type of nonprofit regime, both are responsible for creat-ing public goods and services. In the social-democratic regime, government provides the majority of public goods and services, and the nonprofit sector derives the largest part of income from public expenditure. Statist regimes are characterized by low social public expenditure and a small nonprofit sec-tor.


Based on this reasoning, we suggest stronger crowding-out effects in countries of the liberal nonprofit regime compared to the other regimes:

Liberal regime hypothesis: Crowding-out effects of government expendi-ture and aggregate levels of philanthropic giving are higher in countries belonging to the liberal nonprofit regime compared to all other regimes.

Different subsectorsUse of highly aggregative data may conceal substantively different crowd-ing-out effects for different sectors. For example, an aggregate finding of significant crowding-out does not preclude the possibilities that in one subsector donations have been completely crowded-out while in the other subsector there is partial crowding-out and in a third subsector there is no impact of increases in government expenditures.

It can be hypothesized that crowding-out is more likely in the area of social welfare. In a study on volunteering, Stadelmann-Steffen (2011) ar-gues that crowding-out is most likely in sectors where public and private contributions are in direct competition, like health care and social protec-tion, where nonprofits and governments often provide similar public goods. Young (2000: 155) argues that governments and nonprofits are most likely to be substitutes in the area of social services, where public service delivery is often complex and target groups are heterogeneous, making it more likely that governments will leave service provision to nonprofit organizations. In “expressive” areas (Salamon et al., 2000), on the other hand, like environ-ment, the arts or international aid, philanthropic donors are less likely to be discouraged by government programs. In these sectors, the goods that are produced are different. Klamer (2004) argues that arts is not a public good but a common good, to which value is added by enjoying it, and to which the free rider problem does not apply. For environment and international aid, it holds that the public goods provided (e.g. a clean environment, less world poverty) can only indirectly be enjoyed. Donating to these sectors is there-fore an expression of one’s values rather than a contribution to a public good in the standard economic meaning.

There is some empirical evidence that the relationship between govern-ment expenditures and philanthropic donations varies across subsectors. Indeed, in a systematic literature review of non-experimental crowding-out findings, Lu (2016) shows that government expenditures and philanthrop-ic donations are generally negatively related in the field of human services,


while they are positively related in the fields of health and the arts. In his cross-national study, Sokolowski (2013: 375) found crowding-in for social services, health and education, but no effect in other fields. Empirical analy-ses on volunteering show that government expenditures discourage volun-tary participation in social services and education, while it stimulates partic-ipation in recreation and culture (Day & Devlin, 1996; Stadelmann-Steffen, 2011).

Regarding differences between subsectors, we formulate the following hy-pothesis:

Social welfare crowding-out hypothesis: The association between govern-ment expenditures and private donations is more strongly negative in the subfields of health and social services than in other subsectors.

Cross-wise crowding-in Based on the empirical evidence showing that changes in government ex-penditures affect private donations to different types of non-profit subsec-tors differently (Brooks, 2004: 173; De Wit & Bekkers, 2017; Lu, 2016), we argue that expenditures in one subsector may be associated with increases in philanthropic giving to other subsectors, with the aggregate level of giving remaining constant. This effect has been labelled “philanthropic displace-ment” (Sokolowski, 2013) or “cross-wise crowding-in” (Pennerstorfer & Neumayr, 2017).

Underlying this assumption is the argument that people are impure altru-ists, who are motivated for personal reasons as well as altruistic reasons and who have preferences for public good provision in more than one subsector. If multiple public goods have value in the eyes of donors, higher government support to one subsector could lead donors to decrease donations to this subsector, but increase donations to other subsectors. This is also a reason-able assumption if we believe that individuals have a philanthropic budget, or a mental account for philanthropic giving (Thaler, 1999). Nevertheless, it is possible that purely altruistic donors exist who only care about one type of public good such as social welfare services but not about another type of public good (arts, environment, education etc.). In this case we will not see cross-wise crowding-in. Donors would simply reduce their total donations in response to increased government expenditures to one subsector.

Supporting the notion of philanthropic displacement, Sokolowski (2013: 369) notes that high levels of government expenditures in the “service”-sub-


sectors of education, health or social assistance lead to higher private dona-tions in the “expressive”-subsectors such as arts and entertainment, human rights, environmental issues, and religion. Based on similar grounds, Pen-nerstorfer and Neumayr (2017) argue that people, when public funding cov-ers core-welfare fields, may not necessarily reduce total giving, but instead donate to other, non-core welfare issues, such as international aid. Results of a historical analysis on private donations in Sweden concur with these find-ings, concluding that increases in welfare state expenditure do not damp-en private initiatives per se but rather displace civic engagement, resulting in higher levels of private giving in other subfields (Vamstad & Von Essen, 2013).

We thus hypothesize:

Crosswise crowding-in hypothesis: Higher levels of government expendi-tures to the subfields of social services and health are associated with higher levels of private donations to the subfields of environment, inter-national aid and arts and culture.

RESEARCH DESIGN

Data and measuresThe Individual International Philanthropy Database (IIPD) is a novel dataset, composed of synchronized and merged micro-level datasets from multiple countries. We use data on 126,923 respondents from 19 countries to esti-mate the correlation between government expenditures and philanthropic giving: Australia, France, UK, the Netherlands, US, Canada, Norway, Finland, Mexico, South Korea, Japan, Austria, Indonesia, Taiwan, Ireland, Israel, Rus-sia, Germany and Switzerland. The datasets of different countries were col-lected between 2005 and 2011. This is a wide range, in which there were large economic and political change and thus, differences between countries might be the result of variation over time instead of between-country varia-tion. Furthermore, since the data were collected using different designs, dif-ferences between countries should be interpreted with caution1. People with an altruistic orientation, who are more likely to donate, are also more likely to take (voluntary) surveys (Abraham, Helms, & Presser, 2009). This might imply that a higher non-response leads to lower estimates of donations. In

1 More information on the IIPD can be found in Wiepking and Handy (2016) and IIPD (2016).


questionnaires, it has been shown that survey prompts helps respondents to recall their donations, which leads to higher estimates (Bekkers & Wiepking, 2006; Rooney, Steinberg, & Schervish, 2001, 2004). Because different sam-pling methods and questionnaires are used in different countries, this might explain a part of the variance between countries.

The level of individual philanthropic donations, the amounts donated, are calculated in 2012 US dollars. Donations are strongly skewed, so large dona-tions would have a disproportionate influence on the regression results. It is unlikely that government expenditures have a similar linear effect on do-nations at the very top of the distribution than they have on the bottom and the middle of the distribution. Therefore we take the natural logarithm of the amounts as dependent variables in the regression models. Total amounts donated to philanthropic organizations are available for all countries. For a smaller number of countries we were able to distinguish the amounts do-nated in the sectors (1) environment and animals, (2) arts and culture, (3) education and research, (4) international (relief), (5) social services/welfare and (6) health.

Data on public funding are adopted from the IMF’s Government Finance Statistics. The numbers for Korea do not appear in the IMF data and are ad-opted from the OECD, which uses the same operationalization. We use expen-ditures in the year 2003 in order to have the independent variables precede the outcome variables. Expenditures in the local currency are calculated in US Dollars using the exchange rates as of January 1, 2003 and are divided by the population in order to have the expenditures per capita. Besides total government expenditures, we use expenditures on (a) environment protec-tion, (b) education, (c) social protection and (d) health, which we match with giving in sectors 1, 3, 5 and 6, respectively. In the analyses on the likelihood of donating, government expenditures are divided by 1,000 in order to let the range of the different variables not be different from each other. In the analyses on the influence of the nonprofit regime, we assign Australia, Cana-da, UK and US to the liberal regime, Germany, Austria, France, Ireland, Israel and South Korea to the corporatist regime, Norway, the Netherlands, Finland and Switzerland to the social-democratic regime and Russia, Indonesia, Tai-wan, Mexico and Japan to the statist regime (see Einolf, 2016: 514).

Both philanthropy and government efforts might be driven by a country’s economy. Therefore we take GDP in US Dollars per capita as a control vari-able, also adopted from the IMF Government Finance Statistics. Control vari-ables at the individual level include age, education, gender, marital status


and the natural logarithm of income in US Dollars.

Analytical strategyWe explore the theoretical ideas as lined out in the previous section in two ways. First, we graphically explore our data, examining the correlation be-tween government expenditures and aggregated, average philanthropic donations. The average philanthropic donation per country is calculated based on both donors and non-donors, whose donation value is 0. Second, we run multilevel regression analyses to examine contextual effects while controlling for individual characteristics and allowing slopes to vary across countries.

The decision to give or not may differ from the decision how much to give. For example, financial considerations are likely to be more decisive for amounts donated than for the decision to make a donation (Petrovski, 2017). Therefore we deploy separate Probit regression models on the probability to donate and linear regression models on the amount donated, conditional on donating.

In the analyses of total giving and total government expenditures, we take the sum of donations to different sectors for each respondent. Respondents are clustered in countries, so random intercepts are added when estimating the association between government expenditures and philanthropic dona-tions. For the probability to donate and the amount donated, respectively, the following mixed effects regression models are deployed:

P(Yij) = β0 + u0j + β1Gj + β2Cj + β3Ii + εij

and

ln(Yij)= = β0 + u0j + β1Gj + β2Cj + β3Ii + εij

in which Y is the amount donated by respondent i in country j, u0 is the coun-try-specific intercept, G is government expenditures in US Dollars per cap-ita divided by 1,000, C is the control variable on the country level, GDP per capita divided by 1,000, and I refers to the individual control variables age, education, gender, marital status and income. The natural logarithm of the amounts donated are used.

For the analyses on giving in subsectors, a dataset is constructed in which the units of analysis are combinations of respondents and sectors. A respon-


dent can donate to multiple sectors and therefore appear in the data more than once. Random intercepts are added for each country-sector combina-tion:

P(Yijs) = β0 + u0js + β1Gjs + β2Cj + β3Ii + εijs

and

ln(Yijs) = β0 + u0js + β1Gjs + β2Cj + β3Ii + εijs

in which Y is the amount donated by respondent i to sector s in country j, and u0 is the country-sector specific intercept.

There is an ongoing debate about the problems associated with multilevel models in comparative research (Bryan & Jenkins, 2016). With a number of countries below 20, we should be cautious with strong conclusions that hold for the total population of countries. The results can be taken as a first at-tempt to explore cross-country differences in the relationship between gov-ernment expenditures and philanthropic giving.

RESULTS

Aggregate givingFigure 1 plots the average amount donated per country with total govern-ment expenditures as US Dollars per capita (upper panel) and as percentage of GDP (lower panel). In Indonesia, Russia, Mexico, Taiwan and Korea, coun-tries with relatively low government spending per capita, donations are low too. The United States and the United Kingdom have a moderate government spending and relatively high donations. The average amount donated in the US and the UK is higher than in countries with high government spending per capita, like Switzerland and Norway.

Models 1 to 3 in Table 1 provide a statistical test of the relationship be-tween government expenditures and philanthropic donations. Because re-spondents are nested in countries, we run regression models with random intercepts for countries. Intra-class correlations (Rho) from empty models (not shown) indicate that about 8% of the variance in the likelihood to do-nate and 41% of the variance in the amounts donated can be explained by country level characteristics. A Rho of 8% for the likelihood to donate is low


Figure 1: Average philanthropic donations and government expenditures (Sources: IIPD, IMF)


Tabl

e 1:

Pro

bit a

nd L

inea

r mul

tilev

el re

gres

sion

mod

els

on to

tal g

ivin

g (S

ourc

es: I

IPD

, IM

F, O

ECD

)

Prob

abili

tyAm

ount

(ln)

12

34

12

34

Govt

exp

endi

ture

s / 1

,000

0.00

90.

019

0.04

0-0

.139

0.06

7**

-0.0

93-0

.053

-0.3

17(0

.015

)(0

.039

)(0

.041

)(0

.117

)(0

.032

)(0

.076

)(0

.089

)(0

.297

)Re

gim

e: L

iber

al

Ref

ref.

Regi

me:

Soc

ial-D

emoc

ratic

-1.9

73-6

.277

(2.2

89)

(5.8

18)

Regi

me:

Cor

pora

tist

-3.3

46*

-4.3

32(2

.002

)(5

.089

)Re

gim

e: S

tatis

t-3

.369

*-5

.018

(2.0

09)

(5.1

06)

Soc-

Dem

* Go

vt e

xpen

ditu

res /

1,0

000.

145

0.32

4(0

.139

)(0

.354

)Co

rpor

atis

t * G

ovt e

xpen

ditu

res

/1,0

000.

180

0.27

8(0

.131

)(0

.334

)St

atis

t * G

ovt e

xpen

ditu

res /

1,0

000.

127

0.13

3(0

.139

)(0

.353

)GD

P /

1,00

0-0

.007

-0.0

36-0

.047

**0.

110*

*0.

057

0.04

5(0

.025

)(0

.026

)(0

.023

)(0

.048

)(0

.057

)(0

.058

)Co

nsta

nt0.

457*

0.49

0*-0

.186

3.25

03.

601*

**3.

082*

**1.

911*

**6.

832

(0.2

41)

(0.2

68)

(0.2

80)

(2.0

29)

(0.5

23)

(0.5

21)

(0.6

13)

(5.1

59)

Indi

vidu

al-le

vel c

ontr

ols

Yes

Yes

Yes

Yes

Obse

rvat

ions

126,

923

126,

923

126,

923

126,

923

72,0

7672

,076

72,0

7672

,076

Num

ber o

f cou

ntry

1919

1919

1919

1919

Rho

0.08

20.

082

0.08

80.

043

0.37

30.

323

0.42

10.

316

Indi

vidu

al-le

vel c

ontr

ols:

Age

, Sec

onda

ry e

duca

tion,

Ter

tiary

edu

catio

n, M

ale,

Mar

ried

, Inc

ome

(Ln)

St

anda

rd e

rror

s in

pare

nthe

ses

*** p

<0.0

1, **

p<0

.05,

* p<

0.1


compared with similar studies (Gesthuizen et al., 2008; Pennerstorfer & Neumayr, 2017). The 41% Rho for amounts is much higher, although there are no similar multilevel studies on amounts donated to compare this result with.

The left panel displays results of Probit models on the likelihood to be a donor. There is no significant association between government expenditures and the likelihood to donate, with the coefficient being ß=.04 in the model with full individual-level controls.

The right panel displays the coefficients from linear models on the amount donated. Model 1 shows a positive correlation between government expen-ditures and donations. When controlled for GDP, which is positively cor-related with both variables of interest, the association becomes negative and non-significant (Model 2). Adding individual-level controls makes the main effect less strongly negative (Model 3). The coefficient is ß=-.05, which means that a USD 1,000 increase in government expenditures is associated with a USD 1 decrease in donations, albeit non-significant.

Nonprofit regime typesThe role of nonprofit organizations varies between countries, and we hy-pothesized that a negative relationship between government expenditures and donations is stronger in countries with a liberal nonprofit regime type. Figure 2 shows scatter plots in which each regime type is distinguished with a different marker symbol, both for government expenditures in US Dollars per capita (upper panel) and as a percentage of GDP (lower panel). Although the number of countries per regime type is small, the picture provides a first attempt to explore their heterogeneity. There is a negative correlation among liberal countries. In the bottom panel of Figure 2, which takes into account the size of a country’s economy by looking at government expenditures as a percentage of GDP, the overall correlation is weakly positive (r=0.04, p=0.00) but the picture is different when we examine the associations within each of the regime types. Correlations are negative among countries with a lib-eral nonprofit regime (n=36,103, r=-0.19, p=0.00), a social-democratic re-gime (n=11,346, r=-0.17, p=0.00), a corporatist regime (n=27,756, r=-0.24, p=0.00) and a statist regime (n=53,300, r=-0.13, p=0.00). The correlation is most strongly negative for corporatist countries, which is contrary to what we would expect from theory.

Model 4 in Table 1 explores this argument with interaction terms between regime type and government expenditures in multivariate models. We take


Figure 2: Average philanthropic donations and government expenditures per nonprofit regime type


the liberal regime type as the reference category, since we expect crowd-ing-out in this type to be stronger than in all other regime types. Neither for the likelihood to donate, nor for the amount donated we find statistically significant interactions. Although there is large country heterogeneity, there is no robust evidence for crowding-out to be stronger in liberal nonprofit regimes.

Nonprofit subsectorsHow is government spending in a certain sector related to philanthropic giv-ing in the same sector? Figure 3 shows a scatter plot in which each point is a country-sector combination, with the average amount donated in this sector on the y-axis and the government spending in the same sector on the x-axis. Both government spending and philanthropic donations are relatively low in the environment sector. In some sectors there is high government spending and low donations, like in the social sectors in the Netherlands, France and Norway. In other sectors, low government spending is related to high dona-tions, like the health sector in Canada, the educational sector in Australia and the social sector in the US.

Table 2 provides a more systematic test of the association. Across all sec-tors, government expenditures are positively associated with the likelihood of donating, which is statistically significant (ß=.13 in a model with full in-dividual-level controls). Model 4 adds interactions with sectors. Compared with government expenditures on environment, expenditures on education, health and social services are more strongly negative correlated with the probability of giving. The interaction terms of health and social services with government expenditures are most strongly negative, which is in line with the social welfare crowding-out hypothesis.

The right panel of Table 2 shows coefficients on the amount donated. Gov-ernment expenditures and donations are negatively associated, but this is not statistically significant. The coefficient is ß=-.07 in the model with full controls, which is equivalent to a decrease of USD 1 with every increase in 1,000 USD government expenditures. The relationship is less strongly nega-tive in the fields of social services and health, which is opposite to the expec-tation in the social welfare crowding-out hypothesis. None of the interaction terms are statistically significant.


Figure 3: Average philanthropic donations and government expenditures per sector


Tabl

e 2:

Pro

bit a

nd L

inea

r reg

ress

ion

mod

els

on g

ivin

g to

non

profi

t sub

sect

ors

(Sou

rces

: IIP

D, I

MF,

OEC

D)

Prob

abili

tyAm

ount

(ln)

12

34

12

34

Govt

exp

endi

ture

s per

sect

or /

1,0

000.

127

***

0.12

0 **

0.12

9 **

2.70

0 **

*-0

.022

-0.0

87-0

.068

-1.4

53(0

.043

)(0

.056

)(0

.059

)(0

.535

)(0

.055

)(0

.063

)(0

.071

)(1

.741

)Se

ctor

: Env

iron

men

tre

fRe

fSe

ctor

: Edu

catio

n-1

.050

1.36

7(1

.283

)(1

.634

)Se

ctor

: Hea

lth0.

461

-0.3

63(0

.500

)(0

.878

)Se

ctor

: Soc

ial s

ervi

ces

1.85

2 **

*-0

.357

(0.5

66)

(1.1

07)

Educ

atio

n * G

ovt e

xpen

ditu

res /

1,0

00-1

.913

**0.

594

(0.8

15)

(1.8

71)

Hea

lth *

Govt

exp

endi

ture

s / 1

,000

-2.4

35 **

*1.

382

(0.5

41)

(1.7

41)

Soci

al *

Govt

exp

endi

ture

s / 1

,000

-2.7

41 **

*1.

409

(0.5

36)

(1.7

63)

GDP

/ 1,

000

0.00

40.

129

-0.0

030.

034*

0.01

70.

023

(0.0

10)

(0.0

59)

(0.0

16)

(0.0

18)

(0.0

20)

(0.0

29)

Cons

tant

- 0.9

05**

*- 1

.005

***

- 1.6

01 **

*-2

.234

***

3.87

8***

3.08

2***

1.91

9***

2.05

2**

(0.1

14)

(0.1

87)

(0.4

46)

(0.4

37)

(0.2

10)

(0.4

55)

(0.5

05)

(0.9

26)

Indi

vidu

al-le

vel c

ontr

ols

Yes

Yes

Yes

Yes

Obse

rvat

ions

157,

392

157,

392

157,

392

157,

392

49,7

2549

,725

49,7

2549

,725

Num

ber o

f cou

ntry

-sec

tor

3939

3939

2626

2626

Num

ber o

f res

pond

ents

40,8

9940

,899

40,8

9940

,899

27,4

5327

,453

27,4

5327

,453

Rho

0.17

70.

177

0.17

70.

132

0.22

50.

196

0.20

80.

242

Indi

vidu

al-le

vel c

ontr

ols:

Age

, Sec

onda

ry e

duca

tion,

Ter

tiary

edu

catio

n, M

ale,

Mar

ried

, Inc

ome

(Ln)

St

anda

rd e

rror

s in

pare

nthe

ses

*** p

<0.0

1, **

p<0

.05,

* p<

0.1


Crosswise crowding-inNext, we look at the argument of crosswise crowding-in, which states that public funding of domestic welfare state issues drives donations towards other sectors. Figure 4 plots social protection and health expenditures with philanthropic giving to organizations in the fields of social services, health, environment, international relief or arts and culture. Again, plots are dis-played both for government expenditures in US Dollars per capita (upper panel) and as a percentage of GDP (lower panel). Markers with the symbol + represent donations in the field of social services and health, the dots repre-sent donations in the three other sectors, environment, arts and culture and international aid. We would expect that government expenditures for social protection and health are negatively related to donations in the field of social services and health but positively related to donations in the other fields. There seems to be some empirical support for this argument. Countries with high domestic social welfare expenditures tend to have lower donations to social services and health but higher donations to sectors like international aid and environment.

Table 3 provides a statistical test of crosswise crowding-in. We expect health and social protection expenditures to be associated with donations in “expressive” subsectors. Here, we take donations to environment, arts and culture, and international aid as dependent variable.

Health and social protection expenditures are positively associated with the likelihood to donate to environment, arts or international aid (ß=.15 in a model with full controls), which is in line with the crosswise crowding-in hy-pothesis. The amount donated to these sectors, however, is not significantly affected (ß=-.02 in the full model). This suggests that stronger social welfare programs may drive donors towards these other sectors, but do not lead to higher amounts donated by these donors to those sectors.


Figure 4: Average philanthropic donations per sector and government expenditures to social protection and health


Tabl

e 3:

Pro

bit a

nd L

inea

r reg

ress

ion

mod

els

on g

ivin

g to

env

ironm

ent,

arts

and

cul

ture

, and

inte

rnat

iona

l aid

(S

ourc

es: I

IPD

, IM

F, O

ECD

)

Prob

abili

tyAm

ount

(ln)

12

31

23

Soci

al p

rote

ctio

n an

d he

alth

exp

endi

ture

s / 1

,000

0.15

4 **

*0.

108

*0.

146

***

-0.0

32-0

.077

-0.0

16(0

.030

)(0

.057

)(0

.056

)(0

.055

)(0

.067

)(0

.046

)GD

P /

1,00

00.

026

0.01

10.

031

0.01

4(0

.020

)(0

.023

)(0

.026

)(0

.018

)Co

nsta

nt-2

.342

***

-2.6

95 **

*-3

.193

***

4.32

6***

3.77

8***

2.49

7***

(0.2

39)

(0.3

02)

(0.4

34)

(0.4

69)

(0.6

64)

(0.4

77)

Indi

vidu

al-le

vel c

ontr

ols

Yes

Yes

Obse

rvat

ions

115,

825

115,

825

115,

825

11,2

4511

,245

11,2

45N

umbe

r of C

ount

ry-s

ecto

r com

bina

tions

2828

2817

1717

Num

ber o

f res

pond

ents

40,8

9940

,899

40,8

999,

180

9,18

09,

180

Rho

0.12

30.

119

0.11

50.

175

0.16

90.

181

Indi

vidu

al-le

vel c

ontr

ols:

Age

, Sec

onda

ry e

duca

tion,

Ter

tiary

edu

catio

n, M

ale,

Mar

ried

, Inc

ome

(Ln)

St

anda

rd e

rror

s in

pare

nthe

ses

*** p

<0.0

1, **

p<0

.05,

* p<

0.1


Robustness analysesIn multilevel models, one influential cluster can drive the results in a certain direction. As a robustness check, we re-run each model excluding one coun-try, or a cluster of countries (Van der Meer, Te Grotenhuis, & Pelzer, 2010) at the time.

In our data, the UK and especially the USA seem to be influential cases in the Probit models on the likelihood to donate. Among countries other than the USA and the UK there is a positive correlation between government ex-penditures and philanthropic giving at the country level (β = 0.087, p<.05), only a weak correlation on the country-sector level (β = 0.107, p=ns), and no evidence of crosswise crowding-in (β = 0.009, p=ns).

DISCUSSION AND CONCLUSION

This paper contributes to the broad literature on different aspects of civic participation in the welfare state. Given the large differences between coun-tries in rates of donors and volunteers (Bekkers, 2016; Salamon & Sokolows-ki, 2001), one of the challenges for the literature is to examine contextual explanations (Wiepking & Handy, 2015). This study explores government spending as correlate of philanthropic giving. On the one hand, government expenditures might be expected to displace philanthropic giving, e.g. be-cause altruistically motivated donors reduce voluntary giving when the gov-ernment is already providing public goods (Andreoni, 1993; Roberts, 1984; Warr, 1982). On the other hand, government expenditures might be expect-ed to encourage giving, e.g. because it sends positive signals about (the goals of) nonprofit organizations (Handy, 2000; Schiff, 1990).

The results of this study shows that government spending and philan-thropic giving is most likely to go hand in hand. In countries with high gov-ernment expenditures, there is likely to be a large proportion of philanthrop-ic donors. This confirms earlier findings with cross-national datasets on the likelihood to donate (De Wit, 2016; Pennerstorfer & Neumayr, 2017). Our analysis goes a step further, though, by examining government spending and giving in specific nonprofit subsectors. There is stronger crowding-in in the field of education and research, and, most strongly, environment. In the sub-sectors social services and health, on the other hand, government spending does not strongly affect the number of donors. Government expenditures in these areas are associated with a higher number of donors in other fields,


like environment, arts and culture, and international aid, suggesting that high levels of social welfare spending in “service” subsectors drives donors towards “expressive” areas. In previous research, this has been labeled “phil-anthropic displacement” (Sokolowski, 2013) or “crosswise crowding-in” (Pennerstorfer & Neumayr, 2017).

Among donors, the amounts donated to philanthropic causes are not as-sociated with government spending. There is no significant relationship when looking at aggregate giving, nor is there any evidence for correlations in specific sub-sectors. This has never been studied before and can be con-sidered a very important null-finding. If there would have been evidence for levels of philanthropic giving to be crowded out by welfare state efforts, this would have supported arguments for the nonprofit sector as substitute to the government. With the current results, there is no reason to believe that governments and philanthropic donations to nonprofit organizations are competitive.

The use of cross-sectional comparative data with less than 20 countries is contested. First, the results are hardly generalizable to a larger population of countries. One of a few exceptional countries can drive the results in a cer-tain direction. The robustness checks showed that the United States and the United Kingdom are influential countries in our sample. The database that is used for this study poses further problems because it is compiled from dif-ferent national surveys. Different sampling methods (Abraham et al., 2009) and questionnaires (Bekkers & Wiepking, 2006; Rooney et al., 2001, 2004) may lead to differences in estimated donations. Second, it is difficult to de-duct conclusions about the direction of causality. Both government support and philanthropic donations might be driven by the same underlying vari-ables, which produces upwardly biased estimates. Previous studies dedicat-ed a lot of effort to reduce this bias (Payne, 2009), although a meta-analysis did not find systemically lower estimates with techniques that account for endogeneity and omitted variable bias (De Wit & Bekkers, 2017).

We are very well aware of the problems associated with cross-sectional research and multilevel analyses with a low number of clusters. However, the topic of philanthropy in different welfare states is too important to ne-glect. This study is the first comparative analysis that (1) relates individual amounts donated to government spending and (2) is able to examine differ-ent correlations in a number of specific nonprofit subsectors, where differ-ent effects may exist.

In doing so, it rejects the hypothesis of governments and philanthropic


donations as substitutes. Amounts donated to philanthropic causes in dif-ferent sectors are not crowded out by government spending, and the asso-ciation between government expenditures and the percentage of donors is robustly positive. In the light of the mixed evidence on welfare state effects on different forms of civic participation (Bredtman, 2016; De Wit, 2016; Gesthuizen et al., 2008; Pennerstorfer & Neumayr, 2017; Sokolowski, 2013; Stadelmann-Steffen, 2011; Van Ingen & Van der Meer, 2011; Van Oorschot & Arts, 2005), this study delivers important insights by exploring philanthrop-ic giving in different subsectors of social welfare. Although the evidence is still not conclusive with a sample of 19 countries, there is reason to be opti-mistic about productive government-nonprofit collaborations.

While Chapter 1 gave important insights in the variation of government support and charitable donations across countries, its cross-sectional de-sign did not provide much evidence about the causal relationship. Previ-ous studies aimed to address causality in different ways. Therefore, the current chapter systematically reviews previous studies on the crowd-ing-out hypothesis. It finds that about two-thirds of previous estimates find a negative correlation (crowding-out), while one third of the esti-mates find a positive correlation (crowding-in). The results are strong-ly shaped by the research methods that are used. In experiments, a $1 increase in government support is associated with an average $0.64 de-crease in private donations, while non-experimental data analyses find an average increase of $0.06. Random-effects regression models show that, contrary to arguments that are prevalent in the literature, studies that take subsidies to organizations as a measure of government support are less likely to estimate crowding-out than studies that use a measure of di-rect government expenditures. Central government support is associated with higher charitable donations, while measures that include multiple levels of government tend to find negative correlations. The results chal-lenge the consistency of prior research findings and demonstrate the con-textual dependence of the validity of the crowding-out hypothesis.

A meta-analysis of the crowding-out hypothesis

Chapter 2

Arjen de Wit and René Bekkers

RB designed the study; AdW collected the data and carried out data analysis; AdW and RB contributed to writing the article.

This chapter is published as: De Wit, A., & Bekkers, R. (2017). Government Support and Charitable Donations: A Meta-Analysis of the Crowding-out Hypothesis. Journal of Public Administration Research and Theory, 27(2), 301-319.

Data, syntax and supplementary materials are available through the Open Science Framework at http://osf.io/6jwh2/.

64 Chapter 2 A meta-analysis of the crowding-out hypothesis

INTRODUCTION

How does the level of fundraising income of nonprofit organizations respond to changes in government funding? Over the last few years, nonprofit rev-enues in Western democracies have been pressured due to the economic downturn and unreliable government funding. At the same time, govern-ment policies both in the US and abroad seek to increase the role of profit and nonprofit actors in the private sector. Forms of governance that received a lot of attention include the outsourcing of public services through con-tracting (Smith & Lipsky, 1993), the involvement of non-state actors in con-sensus-based decision making (Ansell & Gash, 2008) and the emergence of interorganizational networks to deliver public services (Milward & Provan, 2003). There has been much debate about the effectiveness of different gov-ernment-nonprofit collaborations. Besides internal characteristics like the institutional structure and management styles, an important condition for effective collaborations is the availability of resources in the organizational context (Ansell & Gash, 2008; Milward & Provan, 2003; Pfeffer & Salancik, 1978). Public goals can be funded through government support in the form of expenditures, subsidies, contracts or tax incentives, but also through non-profit fundraising income. Despite the large body of governance literature, it is still unsure how different funding streams interact.

There is a wide array of studies dedicated to the crowding-out hypothesis, which claims that increasing government contributions, financed through taxes, are associated with reducing charitable donations from private do-nors. In earlier literature reviews Steinberg (1985, 1997) concludes that there is evidence for partial crowding-out. Payne (2009) discusses how dif-ferent studies find different results and concludes that “crowdout exists—at least sometimes” (Payne 2009: 181). From a sample of 46 published and un-published non-experimental studies, Tinkelman (2010: 24) concludes that “the results vary tremendously” and that the effect of government support depends on a number of assumptions, like full information and the costs of providing public goods. The variety of findings raises the question which conditions influence the estimated relationship between government fund-ing and private contributions.

The current meta-analysis examines estimations of crowding-out as well as methodological and contextual characteristics in previous empirical arti-cles. This contributes to the crowding-out literature in two ways. First, map-ping methodological differences is extremely useful for further research in

65A meta-analysis of the crowding-out hypothesis Chapter 2

this area. A better understanding of the consequences of different method-ologies allows for a sensible comparison between previous results and more careful future research design choices. Second, mapping contextual differ-ences yields theoretically useful insights on the conditions under which high government support is associated with lower charitable donations. Our meta-analysis builds upon earlier literature reviews (Steinberg, 1985, 1997; Payne, 2009; Tinkelman, 2010) by mapping differences between empirical findings in a more systematic way, providing robust evidence on contextual characteristics that are often hypothesized to be moderating variables but never tested as such. A meta-analysis is suitable for testing the conditions under which a relationship occurs. However, although we test a variety of possible moderators, there are many other theoretically relevant conditions that we are not able to test here.

Both public and nonprofit managers benefit from robust information about the effects of different types of government funding. Policy makers need to know how policy programs can be funded through effective public-private networks. Evidence that high levels of public funding are detrimental for charitable giving would support ideas about government programs with small roles for public actors and large roles for nonprofit organizations that are dependent on private funding. From the side of nonprofits, it is import-ant to know how revenue streams interact. Organizations that heavily rely on government subsidies are likely to have a lower organizational autonomy (Froelich, 1999; O’Regan & Oster, 2002; Verschuere & De Corte, 2014). More refined knowledge about the effects of public funding on fundraising income would enable nonprofit managers to better position their organizations be-tween government, local communities and other private actors.

The outline of this paper is as follows. In the next section we present hy-potheses on the correlates of crowding-out estimates in previous research. In the Data and Methods section we present the methodology of the me-ta-analysis, while the Results section contains Analyses of Variance (ANOVA) and multivariate regression models to show how different study characteris-tics are correlated with the direction and magnitude of crowding-out that is estimated. The article closes with a discussion and conclusion.


THEORY AND HYPOTHESES

In this section we formulate hypotheses on the correlates between charac-teristics in research design and the crowding-out estimate. We distinguish between hypotheses on data source, sample country, regression model and specification, and operationalization of the independent variable.

Data sourceFour types of data are used to test the relation between government support and charitable donations: laboratory experiments, survey experiments, ar-chival (financial information) data and micro-level survey data. Lab exper-iments differ from real-world settings in “the nature and extent of scruti-ny, the emphasis on the process by which decisions are made, the artificial limits placed on the action space, the imposition of task, the selection rules into the environments, and the stakes typically at risk” (Levitt & List, 2007: 168). However, the defining characteristics of laboratory experiments do not necessarily bias their outcomes in a systematically positive or negative di-rection. Camerer (2015) argues that laboratory and field experiments often find the same results and that the problems with generalizability of lab ex-periments are exaggerated.

In the case of donors’ reactions on government support, we hypothesize that laboratory experiments create a controlled environment with settings that make it more likely for crowding-out to occur. First, participants typical-ly receive full information on the behavior of the “government” as simulated by the researchers. Most of the crowding-out experiments have a repeat-ed-measure design in which participants not only are aware of the level of government support but also of changes therein, making it more likely that they change their giving behavior in different treatments. Horne et al. (2005) show that in reality many donors do not know how much public subsidies organizations receive. Second, participants are more sensitive to social cues because they know that they take part in a study. If people see changes in government support they suspect that this is supposed to affect their giving and, aware of being watched, they will change their donations. Especially people whose preferences are supportive of private donations as a substi-tution for public expenditures may be sensitive for such information. Third, participants in crowding-out experiments are almost always undergraduate students, arguably non-representative samples scoring lower on different measures of prosocial behavior and being more responsive to experimental


manipulations (Henrich, Heine, & Norenzayan, 2010). Carpenter, Connolly, and Myers (2008) show that students, especially males, give considerably lower amounts in experiments than a sample drawn from a broader popula-tion, but other studies showed that students and non-students do not differ in their level of giving in a dictator game with charities as recipients (Bek-kers, 2007) and in their change in giving as a reaction to changes in other participants’ donations in trust games (Falk, Meier, & Zehnder, 2013). Fourth, participants in experiments receive an endowment from the researchers, making it easier to change levels of giving than in situations where they de-cide on their own expenditures. Although the relative financial impacts in ex-perimental conditions might be large (e.g. a 25% tax on a $20 endowment), it is easier to spend money that you have not yet earned.

At least the first two characteristics of lab experiments also hold for sur-vey experiments. A survey experiment is a randomized control trial that is part of a questionnaire, in which respondents receive different questions or pieces of information. In contrast to lab experiments, survey experiments are often carried out among a sample that is representative of the popula-tion. The only published survey experiment on crowding-out that we know of is a vignette experiment without any earnings for the participants (Kim & Van Ryzin, 2014).

Crowding-out can also be tested with archival data, for example when adopted from the U.S. Internal Revenue Service 990 financial informa-tion forms. Despite serious doubts about the accuracy of reported informa-tion on 990 forms, data that organizations report in these forms are highly correlated with those in audited financial statements (Froelich, Knoepfle, & Pollak, 2000). Organizations’ income from private donors is a relatively valid measure of aggregate real-world charitable donations.

As a final data source, crowding-out studies can use survey data on indi-vidual donations that are paired with financial data on government support from other sources. Although survey research has its own issues like sample selectivity and social desirability, self-reported micro data approximates do-nations that are made in absence of the conditions in experimental designs.

A major concern for empirical crowding-out research is endogeneity. As described by Payne (2009), government support and private donations may be jointly determined by unobserved variables. Voter preferences for public goods might drive donations to these goods as well as government funding through the political process, generating an upward bias in the association between government support. Also, places where the need is more urgent


(e.g. high poverty) are likely to receive both high levels of public and pri-vate funding. Finally, a predictor of nonprofit revenues might be previous government grants as organizations have growing success over time (Fos-ter & Fine, 2007). Omitted variables such as voter preferences, (changes in) the need for public goods and previous government funding upwardly bias the relation between government support and charitable donations, which would result in estimations of (in the case of crowding-out) a less strongly negative or (in the case of crowding-in) a more strongly positive association between government support and charitable donations. These concerns ap-ply to studies using archival or survey data. Well-designed experiments are not affected because the treatment (i.e., the level of government support) is randomly assigned and participants generally cannot affect levels of govern-ment support (exceptions are experiments that allow voting, such as Blanco, Lopez, & Coleman, 2012; Isaac & Norton, 2013; Sutter & Weck-Hannemann, 2004).

In both archival and survey data, donors do not necessarily receive infor-mation on the actions of the government, the researcher demand effects are absent or weaker, the samples are generally less selective and participants report on decisions about their own (rather than the experimenters’) mon-ey. Experiments are able to measure the relation between two variables in a controlled environment, while studies using financial data from surveys or archives have to deal with other factors that interfere.

H1: Studies using experimental data are more likely to find crowding-out and find stronger crowding-out than studies using non-experimental data.

Sample countryAlmost all published crowding-out studies come from Western countries, which is an important caveat of the literature since the effect of government policies might be different in developing countries. But even among West-ern countries people and organizations may react in systematically differ-ent ways to changing policies. People from different countries differ in their stance towards social problems as requiring action from private citizens and charitable organizations or government intervention. Citizens in different countries show systematically different levels of support for extensive pro-vision of public services by the government (Andress & Heien, 2001; Svall-fors, 1997). People who are used to extensive welfare state arrangements expect the government to take care of public services and might be reluctant


to compensate for changing levels of government provision of public goods. In countries where public services are considered a shared responsibility for public and private actors, on the other hand, donors might be more willing to raise the level of donations to nonprofit organizations in order to reach the desired goals.

A possible explanation for country differences is that the marginal utility of donations decreases with the extensiveness of welfare state programs. It has been argued that the marginal increase in well-being derived from in-come is high for poor countries but diminishes with economic prosperity (Inglehart, 2000). The need for public or private provision of public services is more urgent in countries with more severe social problems. Welfare states differ in size and inclusiveness, and thus in their efficacy when aiming to alleviate problems like poverty, hunger and homelessness. Given that social needs are higher in countries with smaller welfare states, an additional dol-lar of contributions to alleviate those needs has a higher value for recipients compared to countries with extensive welfare states and less urgent social needs. It is likely that donors are more inclined to compensate for changing government support when the stakes are higher.

Also, the nature of collaborations between governmental and nonprofit actors is different in different countries. Discussing the development of “gov-ernance regimes” in Western Europe, Bode (2006: 355) perceives “a growing distance between voluntary agencies and both the welfare state and civil so-ciety; with more volatile public–private partnerships; and with a dispersed involvement of volunteers and donors.” Smaller government involvement may cause more volatile nonprofit management with a stronger focus on fundraising (Froelich, 1999; O’Regan & Oster, 2002), so organizations should be better able to respond to changing government policies.

In sum, people in countries with smaller welfare states, where the needs are more urgent, public goods are less strongly perceived as government responsibility and nonprofit management is more volatile, should be more likely to compensate government support than countries with extensive gov-ernment arrangements.

H2: Studies are more likely to find crowding-out and find stronger crowd-ing-out in less generous welfare states than in more generous welfare states.


Regression model and specificationAs explained above, studies using non-experimental data are most likely to suffer from endogeneity bias. A first issue is omitted variable bias. Causal claims that are inferred from regression analyses rely on what Angrist and Pischke (2009) call the conditional independence assumption, also known as selection on observables, meaning that only observed variables account for the correlation between the independent variable and the error term. When regressing levels of charitable donations on levels of government support, this assumption is unlikely to be met, since omitted variables such as the need for public goods might upwardly bias the estimated relationship be-tween government support and charitable donations. A related issue is the endogeneity that occurs when both the independent and dependent variable are jointly determined. If government policies reflect the same political pref-erences that underlie charitable donations, it is problematic to treat govern-ment support as an exogenous variable (Payne, 2009). We expect regression models and specifications that deal with omitted variable bias and endoge-neity to estimate more and stronger crowding-out effects.

We test two hypotheses on regression models and model specification. First, we expect that crowding-out estimates are stronger in empirical spec-ifications that account for time-invariant omitted variables. A simple OLS regression estimates the relation between both the level and the change in government support and private donations. Fixed-effects specifications in-clude dummies for the units of analysis, holding all time-invariant factors constant. Most of these specifications include fixed effects for organizations (reducing bias caused by organizational size, mission, etc.), but studies with other units of analysis can include fixed effects for states or districts (reduc-ing bias caused by population characteristics, geographical features, etc.). A first-difference estimation, regressing the changes in donations on the changes in government support, is a similar way to deal with endogeneity. Note that fixed-effects and first-difference specifications do not account for omitted variables that change over time. This can be solved by including a lagged dependent variable as a predictor in the model, but estimating both fixed effects and lagged dependents in one model comes with new (and prob-lematic) assumptions (Angrist & Pischke, 2009: 245). Simply using a lagged government support measure as independent variable might mitigate, but not solve the bias caused by time-variant omitted variables.

Second, Payne (2009) argues that empirical specifications measuring only the exogenous part of government support, including two-staged least


squares regression (2SLS), lead to less biased estimates. Instrumental vari-able regression is a way to deal with the endogeneity problem, using predic-tor variables that correlate with the independent but not with the dependent variable or its error term (Morgan & Winship, 2007). In these models, gov-ernment support is regressed on one or more instrumental variables (like region characteristics, organizational characteristics or measures of political power) to model the part of government support that is exogenous. In the second stage of the regression, private donations are regressed on the ex-ogenous part of government support, hereby reducing the upward bias that is due to organizations receiving both high government support and high private donations.

H3: Studies using fixed-effects models and first-difference specifications are more likely to find crowding-out and find stronger crowding-out than studies using other model specifications.H4: Studies using instrumental variable regression models are more likely to find crowding-out and find stronger crowding-out than studies using other regression models.

Government supportOur final set of hypotheses concerns the operationalization of the indepen-dent variable in primary studies. In experimental designs, researchers most-ly simulate a government tax by imposing an involuntary contribution from participants. In non-experimental designs, we distinguish two dimensions that can raise differences.

First, measures of government support are either expenditures directly targeted at the need in society or subsidies to nonprofit organizations. Gov-ernment support may have a direct effect on individual donations because people derive utility from the total amount that they contribute to the public good, either through taxes or through their own voluntary donations. Howev-er, it is unlikely that people change their behavior when they are not aware of (changes in) government support (Horne et al., 2005). Government support may also have an indirect effect on donations through the behavior of orga-nizations, who play a crucial role because they collect donations and may increase their fundraising efforts when government support is lowered or vice versa. The latter effect has been labeled “fundraising crowd-out” and is a plausible explanation of the negative relation between government support and private donations (Andreoni & Payne, 2003, 2011; Hughes, Luksetich,


& Rooney, 2014). If organizational behavior explains changes in donations, studies that take subsidies to organizations as an independent variable pro-vide more precise estimates that capture this effect and are more likely to yield crowding-out. Direct government expenditures include a wide range of government programs that benefit public goals, either through direct spending or through mediating organizations. Subsidies to nonprofit orga-nizations are a more precise measure of public funding through nonprofit organizations, including contracts, the purchase of services, matching grants and unconditional subsidies. While some studies estimate crowding-out with aggregated measures of public and private funding in districts or sec-tors, studies that use organizational-level data are expected to find more and stronger crowding-out effects.

Second, both the central government and lower levels of government can provide support for nonprofit organizations. In the case of the US, federal grants are likely to not only have an effect on individual private donations but also on spending of lower levels of government, and both private do-nors and lower governments are responsive to one another. The term “joint crowd-out” refers to the collective effect of federal grants on both private and lower government support, while the direct effect of federal support on pri-vate donations is referred to as “simple crowd-out” (Steinberg, 1989, 1991; Lindsey & Steinberg, 1990). State and local governments tend to match fed-eral grants, especially when those are targeted at specific needs and thus, private donors would not only substitute a decreasing federal government grant but also the decreasing local government support that sticks to federal money. Studies that only use a measure of central government spending or only a measure of spending at lower levels could overestimate the effect of government support because a part of the change in private donations is due to the change in spending by other levels of government. Studies that include a measure of total government support, or use a model that controls for oth-er levels of government, are expected to provide weaker crowding-out esti-mates. Disentangling the effects of different levels of government is import-ant because many governance networks are found on local levels and the provision of public services is increasingly decentralized (Klijn, 2008).

H5: Studies that measure subsidies to organizations are more likely to find crowding-out and find stronger crowding-out than studies that measure government expenditures.H6: Studies that measure only central or only lower levels of government


support are more likely to find crowding-out and find stronger crowd-ing-out than studies that measure all levels of government support.

Other moderatorsIn addition to the characteristics of studies that we have discussed thus far, there are many other characteristics that could affect estimates of the crowd-ing-out effect. First, different types of private giving can be distinguished. Empirical studies that use financial archival data often use aggregated mea-sures of private nonprofit revenue, which include donations from individ-uals, companies, foundations and other organizations. Much giving is reli-giously orientated, which is often not directly substitutable for government provision. Second, government support can be further specified. There is a variety of grants, purchases, subsidies and vouchers that may have different effects, and while the implicit assumption in many studies is that govern-ment funding is unconditional, aggregate measures of government support often include matching grants. Government grants may have differential ef-fects when they are publicly announced or when they are part of a larger policy shift. Third, there might be differences across organizations. Due to the small number of studies and estimates per field we cannot distinguish between different parts of the nonprofit sector, nor can we align organiza-tions on the extent to which they are subsidy-dependent.

Furthermore, it has been argued that crowding-out effects vary with the level of government support (Borgonovi, 2006; Brooks, 2000a, 2003a), the salience of the tax (Eckel et al., 2005), the number of other donors (Ribar & Wilhelm, 2002), the difference between public goods that are generally provided by public funding and public goods that are generally provided by private funding (Tinkelman, 2010), the linearity of the cost function of public good production (Tinkelman, 2010), the number of people that initially do not contribute to a public good (Chan, Godbyb, Mestelman, & Muller, 2002; Tinkelman, 2010) and substitution between nonprofit organizations or be-tween sectors (Sokolowski, 2013; Tinkelman, 2010).

Due to data limitations or research design choices, not all of these con-ditions have been systematically tested. Table A in the Appendix shows the moderators that are often, sometimes of not often distinguished in previ-ous empirical work, and whether or not these moderators are tested in this meta-analysis. The large number of possible moderators in the right-bottom cell shows that the crowding-out literature still has a long way to go after this meta-analysis.


DATA AND METHODS

The meta-analysis we present relies on a sample of previous studies on the crowding-out effect. To ensure comparability we limit our review to studies with the amount of donations of money as the dependent variable, either self-reported in surveys or observed in experiments or in archival (financial information) data, and the amount of government support as independent variable. Governments can enhance donations by matches or rebates (Eckel & Grossman, 2003; Peloza & Steel, 2005), but our analysis is restricted to unconditional government grants.

A meta-analysis is a good way to examine differences between studies on the crowding-out effect. The term “meta-analysis” has been proposed by Glass (1976: 3) as referring to “the statistical analysis of a large collection of analysis results from individual studies for the purpose of integrating the findings.” Such analyses have become quite common in educational research, psychological research and especially in medical research, and are increas-ingly used in several other social science areas. Meta-analyses are useful in calculating an average of effect sizes that are found in a number of studies, and in examining differences among studies by running meta-regressions of study characteristics on the effect size.

Data were collected in two stages. In the first stage we used EndNote X7 to retrieve studies in the Web of Science database. We search for studies (1) with the term “crowding-out” in the title, keywords or abstract, or (2) that use a pair of possible formulations of the dependent and independent vari-able in title, keywords or abstract.1 In the second stage we browsed the refer-ence lists of the studies in the sample that we obtained from Web of Science to look for additional peer-reviewed journal articles that suited our criteria.

The sample contains studies with quantitative empirical research on the relation between government support and private charitable donations. We include studies with charitable donations by individuals or households, ei-ther observed or self-reported, as dependent variable. Donations should be charitable in the sense that the donors do not have a personal relation with the recipients, so studies on private transfers between households are not included in this meta-analysis. Studies measuring the incidence of donating are excluded, as we are interested in the amounts donated. The independent

1 The search command used is: “(crowding-out OR crowding out OR crowd-out OR crowdout OR crowd out) OR ((donations OR giving) AND (government OR subsidies OR tax OR taxing OR taxes OR matching OR rebate OR rebates OR altruism)).” This command yielded 4,930 records on February 26, 2015.


variable of interest is the amount of government support to a goal or organi-zation, either real or simulated. Government support should be uncondition-al, so matches, rebates and tax and price elasticities of private donations are excluded. Many studies use an aggregated measure of government support, which is one of the major weaknesses of a part of the crowding-out litera-ture. Studies that take an aggregate measure of government expenditures are in our sample, even though these expenditures often include matching grants.2

Most studies contain multiple estimates of the relation between govern-ment support and private donations, using different regression models or specifications, treatment groups, or sub-samples. We code every estimate separately, so we obtain a sample of estimates that are clustered within studies. Besides a dichotomous variable on finding a negative or a positive correlation, we calculate a standardized crowding-out estimate: what is the change of private donations in the case of a $1 increase in government con-tributions?3

Our search resulted in a set of 70 studies that matched the criteria, of which the main study characteristics and findings are displayed in the Ap-pendix. Because most studies report different estimates of the association between government contributions and private donations we extracted a total of 422 findings of crowding-out or crowding-in. It is not possible to calculate a standardized effect size estimate for every finding, so the sample of standardized crowding-out effects includes 325 results from 54 studies that estimate the effect on private donations of a $1 increase in government contributions.4

The sample of effect sizes contains a number of extreme values. To prevent 2 We excluded 18 estimates from 6 studies that use only subsidies from the American National Endow-ment for the Arts (NEA) as independent variable, because those are matching grants by nature. Studies that use contributions from other private donors as independent variables, like large gifts from famous lead donors intended to increase fundraising success, are also excluded because we are theoretically interested in the effect of government policies.3 In the case of an unstandardized regression or correlation coefficient of 0.5 and independent and dependent variables measured in absolute values, the estimate equals 0.5. When a treatment group donated $20 on average while the government contribution was $25, and the control group donated $10 by a government contribution of $5, the estimate equals (20-10)/(25-5)=0.5. We do not compute an es-timate in the case of transformed variables like logarithmic variables (58 estimates from 11 studies) or relative measures (6 estimates from 2 studies), neither do we include an estimate if the model includes a quadratic term of government support (13 estimates from 2 studies).4 Missing values on the standardized estimate are not randomly distributed in the sample. Independent samples t-tests show that studies with non-experimental data, studies from Europe, other specifications than fixed-effect or first-differences, regression models without instrumental variables and studies that use only one level of government as independent variable are less likely to report a standardized crowding-out effect size estimate.


these outliers from having a disproportionally large influence on the results the one percent lowest values are given the value of the first percentile while the one percent highest values are set on the value of the ninety-ninth per-centile. This procedure is known as “Winsorizing.” As opposed to trimming, where the lowest and highest values are deleted, this method treats the data for outliers while leaving all relevant data points in the sample, making the descriptive and regression results more robust (Tukey, 1962: 17-19).

The sample includes 262 findings of a negative correlation between gov-ernment support and charitable donations, and 160 findings of a positive correlation. Figure 1 graphically displays all standardized crowding-out es-timates after treating the data for outliers, each horizontal line representing one study. Table 1 contains descriptive statistics for findings of crowding-out or crowding-in, the crowding-out effect estimate and the study character-istics that are used in the analyses. The median is -0.18 and the robust un-weighted mean is -0.17, with a 95 percent confidence interval between -0.25 and -0.09, indicating that a $1 increase in government support is associated with a $0.17 decrease in private charitable donations across all studies.

We test our hypotheses in two stages. First, we examine H1 in a comparison of mean findings in experimental and non-experimental studies. An Analysis of Variance (ANOVA) is used to test whether differences between the groups are statistically significant.

Experiments differ from other studies in many ways, so in the second stage we test the remaining hypotheses for experimental and non-experi-mental research designs separately. We run logistic regression analyses on the binary variable of crowding-out (value 0) vs. crowding-in (1) as well as linear regression analyses on the smaller sample of standardized effect size estimates. H3 to H6 are only tested with non-experimental studies because experimental designs do not vary on these dimensions.

The probability of finding a positive association between government sup-port and charitable giving is estimated with a logit model. Because estimates are clustered within studies we allow intercepts to vary across studies, ex-amining the model

P(crowding-in)ij / (1 – P(crowding-in)ij) = β0 + β1X1ij + β2X2j + … + βkXkij + uj + eij

where P(crowding-in)ij is the probability of finding a positive correlation of


Figure 1: Dot graph of crowding-out effect estimates per study

Table 1: Descriptive statistics

N Mean SD Min MaxFull sampleCrowding-out (0) vs. crowding-in (1) 422 0.379 0.486 0 1Crowding-out effect estimate 325 -0.170 0.707 -1.580 2.707Experimental study (no/yes) 422 0.268 0.443 0 1Experimental studiesLess generous welfare state (no/yes) 113 0.575 0.497 0 1Year of publication 113 2003.425 7.250 1993 2014Sample size (ln) 113 4.748 0.696 3.611 6.908Non-experimental studiesLess generous welfare state (no/yes) 306 0.941 0.236 0 1Fixed-effects or first-difference (no/yes) 306 0.412 0.493 0 1Instrumental variable (no/yes) 306 0.265 0.442 0 1Subsidies to organizations (no/yes) 306 0.814 0.390 0 1Only central government (no/yes) 306 0.101 0.302 0 1Only lower government (no/yes) 306 0.052 0.223 0 1Year of publication 306 2001.324 8.591 1978 2014Sample size (ln) 306 5.540 2.191 0.693 14.864


the ith estimate in the jth study, β0 the baseline intercept, βk the regression coefficient of the kth independent variable, uj the study-specific intercept, and eij the error term for each estimate. We report odds ratios, to be inter-preted as the ratio between the odds of finding crowding-in vs. the odds of finding crowding-out. An odds ratio of 1 means that the probabilities are equal, an odds ratio below 1 means a higher probability of finding crowd-ing-out, an odds ratio higher than 1 means that the probability of finding crowding-in is higher.

Correlates of standardized crowding-out effect estimates are estimated by linear Generalized Least Squares (GLS) regression models with the crowd-ing-out estimate as the dependent variable and different study characteris-tics as the independent variables,

Yij = β0 + β1X1ij + β2X2j + … + βkXkij + uj + eij

where Y is the effect of a $1 increase in government support on the amount donated.

Note that some Xs only vary across studies (e.g. welfare state type) and some vary both across and within studies (e.g. the use of fixed-effects re-gression). Hausman tests are not statistically significant, suggesting that a random-effects specification is appropriate here.

The sample includes estimates in different parts of the voluntary sector. The sample includes 18 studies that estimate effect sizes in the field of arts and culture, 10 in the field of education, 1 study in the field of environment and animals, 7 in the health sector, 8 on international aid, 12 studies that have estimates on social services, 3 on religion, 21 studies that estimate ef-fect sizes on an aggregated measure of giving in different sectors, and 15 studies where the receiving sector is undefined. Comparing the differences between those fields would increase our understanding of varying effects of government efforts, but the numbers of studies and estimates in each field are too small to make reliable claims.

In order to test H2 on differences between welfare state regime types we classify the United States, Canada, the United Kingdom and Australia as less generous welfare states. The only cross-country study in the sample (Sokolowski, 2013) is excluded from the regression analyses.

We include two control variables. The first control is the year of publi-cation because the correlations of our variables of interest could be due to period effects. The second control variable is the sample size, which is often


used in meta-analyses as an indicator of the statistical power of the estimate (Borenstein, Hedges, Higgins, & Rothstein, 2009).5 We take the natural loga-rithm because the distribution of sample sizes is highly skewed.

RESULTS

Data sourceTable 2 displays the means of our dependent variables for experimental and non-experimental studies. In line with H1, estimates from experiments show more and stronger crowding-out estimates. There are only 5 crowding-in estimates with experimental data in the sample, all from different studies, representing 4 percent of all experimental estimates. In non-experimental studies, there are as many crowding-out estimates as crowding-in estimates. In experiments a $1 increase in government support is associated with a $0.64 decrease in private donations on average (which is significantly dif-ferent from zero with a 95 percent confidence interval between -0.70 and -0.58), while archival or survey data analyses find a mean increase of $0.06 (not significantly different from zero with a 95 percent confidence interval between -0.04 and 0.15). The differences between experiments and non-ex-periments are statistically significant.

5 Some studies do not report sample sizes for each estimate because it uses sub-samples for different es-timates. In those cases we calculated an approximate sample size based on the size of the whole sample.

Table 2: Mean findings for experimental and non-experimental research designs

Experimental studies

Non-experimental studies Sig.

Crowding-out (0) vs. crowding-in (1) 0.044 0.502 ***(0.019) (0.028)

Crowding-out effect estimate -0.643 0.056 ***(0.031) (0.049)

Differences between groups are tested with one-way ANOVA. SD values are given in parentheses.***p < .01.


Sample countryTable 3 reports the odds ratios of the logistic regression models, and Table 4 displays the regression coefficients from the GLS models. Non-experimental research designs in less generous welfare states are less likely to find posi-tive associations between government support and private donations. The odds ratio becomes 0.60 and 0.46 when controlling for regression model and specification (Table 3, Model VI) and type of government support (Table 3, Model V), indicating that studies from less generous welfare states have a predicted probability of 19 percent to estimate crowding-out. The differenc-es are not statistically significant due to the large standard errors. Regarding the effect size (Table 4), experimental estimates from less generous welfare states are 0.16 more strongly positive than estimates from more generous welfare states (Table 4, Model III), and non-experimental estimates from less generous welfare states are 0.19 higher (Table 4, Model IX). The GLS coeffi-cients are in the opposite direction of what we expected.

Regression model and specificationOur hypotheses predict that crowding-out is stronger in models and specifi-cations that account for endogeneity. Fixed-effects or first-difference specifi-cations are between 3.5 and 4 times more likely to find a positive association between government support and charitable donations (Table 3), which is contrary to the expectation. In line with our hypothesis, instrumental vari-able models more often find crowding-out. In the full model (Model VIII) the odds ratio is 0.46, indicating that instrumental variable analyses have a predicted probability of 19 percent to find crowding-out, which is not statis-tically significant. In the linear regression (Table 4), the differences between models and specifications are small and not statistically significant. There is a large variance in instrumental variable regression estimates: the standard deviation of crowding-out effect size estimates is 0.91 for these models. It is likely that crowding-out findings strongly depend on the instruments that are used.

The unexplained between-study variance ρ does not substantially de-crease in models including variables on regression model and specification. The use of fixed-effects, first-difference and instrumental variables varies both between and within studies and does not explain much of the heteroge-neity in crowding-out estimates across studies.


Government supportThe expectation in H5 that government subsidies to organizations have a stronger negative effect than direct expenditures must be rejected with our data. Estimates obtained for levels of subsidies as independent variables are 7.9 times more likely to find crowding-in than estimates that use direct gov-ernment expenditures (Table 3, Model VII), which is contrary to the hypoth-esis. In the linear regression model the coefficient is positive too, although it is not significantly different from zero.

Contrary to the expectation, crowding-out estimates are not stronger when different levels of government spending are measured. Instead, esti-mates with a measure of only central government spending find more (Table 3, Model VIII) and stronger (Table 4, Model VIII) positive effects. H6 is re-jected.

The intraclass correlation ρ does not substantially decrease when differ-ences between measures of government support are included in the model.

Robustness checkAs a robustness check we reran our analyses several times, each time exclud-ing one study. The data are already treated for outliers (see under Data and Methods), but studies with extreme values can still have a disproportionally large influence on the results.

The mean effect size estimate of x̅=-0.17 has a 95 percent confidence in-terval from -0.25 to -0.09, and excluding influential studies does not result in a mean outside this range.

The differences between the means of experimental and non-experimen-tal designs are large and robust. Most results from the random-effects mod-els are robust against excluding one of the studies in the sample too, with two exceptions. First, when excluding a study by Hughes et al. (2014) GLS regression coefficient of fixed-effects and first-difference models becomes more strongly negative (β=-0.20 in the full model, p=0.09). Hughes and col-leagues find strong positive coefficients in their fixed-effects models with archival data on symphony orchestras. Second, excluding one of the studies by Brooks (2000b) makes the GLS regression coefficient of subsidies to or-ganizations moderately negative (β=-0.19, p=0.60). Using longitudinal data, Brooks estimates coefficients close to zero but also one positive coefficient of 0.73 among arts and cultural organizations.

In sum, there is robust evidence that experimental designs find more and

82 Chapter 2 A meta-analysis of the crowding-out hypothesisTa

ble

3: L

ogis

tic re

gres

sion

resu

lts o

n cr

owdi

ng-o

ut (0

) vs.

cro

wdi

ng-in

(1),

rand

om e

ffect

s

Expe

rim

enta

l stu

dies

Non

-exp

erim

enta

l stu

dies

III

IIIIV

VVI

VII

VIII

IX

Gene

rous

wel

fare

stat

ere

f.re

f.re

f.re

f.re

f.re

f.re

f.Le

ss g

ener

ous w

elfa

re st

ate

1.22

90.

988

0.93

50.

604

0.46

40.

436

0.44

4(1

.405

)(1

.182

)(1

.024

)(0

.664

)(0

.480

)(0

.464

)(0

.477

)Ot

her s

peci

ficat

ions

ref.

ref.

ref.

ref.

Fixe

d-ef

fect

s or f

irst

-diff

eren

ce4.

012*

*3.

653*

*3.

547*

*3.

460*

(2.4

45)

(2.1

12)

(2.0

91)

(2.1

97)

No

inst

rum

enta

l var

iabl

esre

f.re

f.re

f.re

f.In

stru

men

tal v

aria

bles

0.71

40.

530

0.46

00.

460

(0.3

58)

(0.2

63)

(0.2

34)

(0.2

36)

Dire

ct g

over

nmen

t exp

endi

ture

s re

f.re

f.re

f.Su

bsid

ies t

o or

gani

zatio

ns7.

940*

**9.

778*

**9.

388*

*(5

.890

)(8

.405

)(8

.399

)Bo

th le

vels

of g

over

nmen

tre

f.re

f.On

ly ce

ntra

l gov

ernm

ent

3.54

0*3.

555*

(2.5

18)

(2.5

43)

Only

low

er g

over

nmen

t0.

953

0.94

7(1

.284

)(1

.300

)Ye

ar o

f pub

licat

ion

1.29

3*1.

008

(0.1

81)

(0.0

38)

Sam

ple

size

(ln)

1.45

51.

029

(0.8

29)

(0.1

21)

(Con

stan

t)0.

038*

**0.

034*

**0.

000*

0.79

00.

840

0.84

80.

252

0.22

00.

000

(0.0

27)

(0.0

34)

(0.0

00)

(0.2

53)

(0.8

79)

(0.8

70)

(0.2

70)

(0.2

46)

(0.0

00)

Betw

een-

stud

y SD

1.02

61.

026

0.00

11.

786

1.78

51.

738

1.53

41.

559

1.57

1Rh

o0.

242

0.24

20.

000

0.49

20.

492

0.47

90.

417

0.42

50.

429

No.

of s

tudi

es20

2020

4949

4949

4949

Obse

rvat

ions

113

113

113

306

306

306

306

306

306

Odd

s rat

ios a

re re

port

ed. S

E va

lues

are

giv

en in

par

enth

eses

; * p

< .1

0; **

p <

.05;

*** p

< .0

1


Tabl

e 4:

GLS

regr

essi

on re

sults

on

crow

ding

-out

effe

ct s

ize

estim

ate,

rand

om-e

ffect

s

Expe

rim

enta

l stu

dies

Non

-exp

erim

enta

l stu

dies

III

IIIIV

VVI

VII

VIII

IX

Gene

rous

wel

fare

stat

ere

f.re

f.re

f.re

f.re

f.re

f.re

f.Le

ss g

ener

ous w

elfa

re st

ate

0.05

30.

155

0.17

10.

140

0.17

00.

168

0.19

3(0

.122

)(0

.120

)(0

.424

)(0

.429

)(0

.431

)(0

.435

)(0

.433

)Ot

her s

peci

ficat

ions

ref.

ref.

ref.

ref.

Fixe

d-ef

fect

s or f

irst

-diff

eren

ce0.

083

0.07

50.

049

-0.0

69(0

.129

)(0

.128

)(0

.134

)(0

.151

)N

o in

stru

men

tal v

aria

bles

ref.

ref.

ref.

ref.

Inst

rum

enta

l var

iabl

es-0

.019

-0.0

42-0

.041

-0.0

05(0

.116

)(0

.119

)(0

.121

)(0

.122

)Di

rect

gov

ernm

ent e

xpen

ditu

res

ref.

ref.

ref.

Subs

idie

s to

orga

niza

tions

0.11

60.

116

0.04

7(0

.183

)(0

.235

)(0

.280

)Bo

th le

vels

of g

over

nmen

tre

f.re

f.On

ly ce

ntra

l gov

ernm

ent

0.35

2*0.

359*

(0.2

09)

(0.2

08)

Only

low

er g

over

nmen

t-0

.052

0.07

9(0

.343

)(0

.355

)Ye

ar o

f pub

licat

ion

0.01

40.

011

(0.0

09)

(0.0

10)

Sam

ple

size

(ln)

-0.1

75**

-0.0

60*

(0.0

81)

(0.0

35)

(Con

stan

t)-0

.612

***

-0.6

43**

*-2

7.49

70.

048

-0.1

19-0

.118

-0.2

32-0

.235

-21.

081

(0.0

58)

(0.0

94)

(17.

635)

(0.0

66)

(0.4

19)

(0.4

19)

(0.4

57)

(0.4

83)

(20.

483)

Betw

een-

stud

y SD

0.18

30.

194

0.17

50.

228

0.23

50.

224

0.21

40.

244

0.22

9Rh

o0.

285

0.30

90.

264

0.11

50.

122

0.11

20.

103

0.13

10.

116

No.

of s

tudi

es18

1818

3636

3636

3636

Obse

rvat

ions

105

105

105

220

220

220

220

220

220

SE v

alue

s are

giv

en in

par

enth

eses

; * p

< .1

0; **

p <

.05;

*** p

< .0

1


stronger crowding-out, that fixed-effects of first-difference models less of-ten estimate crowding-out, that studies using subsidies to organizations as measure of government support are more likely to find crowding-out, and that studies using a measure of central government support find more and stronger crowding-out estimates.

DISCUSSION AND CONCLUSIONS

In previous research, questions have been posed about the effectiveness of new forms of governance with larger roles for nonprofit organizations in the creation and implementation of public services (Ansell & Gash, 2008; Mil-ward & Provan, 2003; Smith & Lipsky, 1993). In order to understand the con-textual dynamics of effective governance, there is a need for robust evidence on the effects of changing government spending on fundraising income. De-spite a large number of empirical studies there is no decisive evidence for government support to crowd out private charitable contributions. About two-thirds of the findings in our meta-analysis show a negative correlation between government support and charitable donations, while one third finds a positive correlation.

Payne (2009) argues that research on the relation between government support and charitable donations suffers from endogeneity. One way to es-tablish causality is through experimental research designs, and our analy-sis shows that these designs find more and stronger crowding-out effects than studies using archival or survey data. While experiments show that each dollar of government support crowds out $0.64 of private donations, a dollar increase in government support in non-experimental data from sur-veys, financial information forms or other archival data is associated with a slight increase in voluntary contributions on average. Our analysis shows that there is incomplete crowding-out and that the pure altruism model, in which each dollar of mandatory contributions leads to a dollar reduction in voluntary contributions, should be reconsidered. The pure altruism model makes a number of assumptions about the situation in which the govern-ment and private donors contribute to a public good, and crowding-out find-ings depend on the extent to which empirical studies relax these assump-tions (Tinkelman, 2010). In experiments people have full information on the level of government contributions, decide on money that is not their own, are sensitive to social cues because they are aware of taking part in a study,


and are often undergraduate students that differ in their prosocial behavior and reactions to experimental manipulation (Henrich et al., 2010). There is an ongoing debate about the extent to which findings from laboratory ex-periments can be generalized to natural settings (Camerer, 2015; Galizzi & Navarro-Martinez, 2015; Levitt & List, 2007) and the large difference in our meta-analysis sample between the estimates obtained in experiments and other types of data emphasizes the importance of this debate.

One could argue that experimental designs provide cleaner estimates of the causal relation because they rule out the interference of other variables. If endogeneity explains why experimental findings differ from other find-ings, we would observe that regression models and specifications that ef-fectively deal with this issue produce stronger crowding-out estimates than other regression models. Our results do not confirm this line of reasoning. Neither fixed-effects or first-difference specifications nor the use of instru-mental variables are robustly linked with stronger crowding-out. It is likely that findings in instrumental variable models are highly dependent on the measures that are used as instruments. Similar measures of organizational output and region characteristics are used by some studies as instruments for government support (Brooks, 1999; Khanna & Sandler, 2000; Payne, 2001) and by another study as instruments for private giving (Becker & Lindsay, 1994). Hughes and Luksetich (1999) use the same set of variables as instruments for both public and private funding sources in different 2SLS regression models. If a prerequisite for a valid instrumental variable is that it is correlated with X but not with Y or its error term (Morgan & Winship, 2007), it is striking that the same kind of variables are used for both govern-ment support and charitable giving. Researchers should be very careful in applying these techniques, and preferably use a range of different models, specifications and instrumental variables to estimate the effect of govern-ment support in a certain dataset.

Our results also challenge the argument of indirect crowding-out, which means that the fundraising behavior of organizations partly explain why people change their donations after government investments or budget cuts (Andreoni & Payne, 2003, 2011; Hughes et al., 2014). Subsidies to organiza-tions are much more likely to crowd in donations than direct government ex-penditures, but they do not lead to stronger crowding-in effects on average. A possible explanation for this result is that the effect is non-linear, with small-er subsidies enhancing donations and larger subsidies discouraging them (Borgonovi, 2006; Brooks, 2000a, 2003a). This also means that subsidizing


does not make organizations dependent on public funding, but rather seems to encourage revenue diversification at the organizational level. Considering previous arguments that governance networks benefit from resource-rich environments (Milward & Provan, 2000) and that organizations with a di-versified revenue mix tend to be financially stable (Carroll & Stater, 2009), subsidizing the nonprofit sector could strongly improve the chances of fruit-ful public-private partnerships.

Our analyses show that measures of central government support are pos-itively related to charitable donations, while measures of multiple govern-mental levels are negatively related to giving. This is contrary to what we expect from models developed by Steinberg (1989, 1991) and Lindsey and Steinberg (1990). If there would be a “flypaper effect,” meaning that federal funding induces support from lower levels of government to the same public good, studies of central government support would underestimate the total crowding-out effect. The results from this meta-analysis contradict this ar-gument. Federal policy programs turn out to be effective in stimulating pri-vate giving, while policy programs on local levels, which often involve non-profit actors, are difficult to fund through a mix of public and private funding. This raises the question how effective local policy makers are in establishing fruitful collaborations with nonprofit actors, which is an emerging topic in an era in which public services are increasingly decentralized (Klijn, 2008).

Our analysis suffers from a few limitations. First, there are more differ-ences between research designs than we accounted for in this paper. A com-mon critique on meta-analyses is that they compare apples and oranges by including findings that diverge in many more ways than can be tested for (Borenstein et al., 2009: 379-380; Petticrew & Roberts, 2006: 203-204; Wolf, 1986: 14-15). Without doubt, different measures of the dependent and in-dependent variable lead to different findings. In future research, comparing the effects of different types of government support and on different types of organizations would add much to our understanding about nonprofit fi-nancing across society. The most important difference we found is the one between experimental and non-experimental studies, and there are numer-ous differences between these two approaches that cannot all be examined by meta-analytical techniques. In the current analysis we cannot establish with certainty to what extent the stronger crowding-out results in experi-ments are due to the information that is provided, the endowment partici-pants receive, demand effects, subject pool composition effects or reduced endogeneity. Systematic comparisons between data that vary on dimensions


that are not often tested in previous empirical research (see the Appendix, Table A) could provide more insight. Furthermore, the current analysis con-cerns the relationship between funding sources without paying attention to the actors involved or the governance processes behind it. Although the en-vironmental dynamics are important for organizations and the analysis is valuable as such, reactions on funding streams may depend on many other factors like institutional characteristics, management styles and relations between the actors involved (Ansell & Gash, 2008; Milward & Provan, 2003). More research would be necessary to shed light on other factors that moder-ate crowding-out effects.

A second important limitation is that the estimates in our meta-analysis are not necessarily a random sample. Weak or non-significant results are generally more likely to remain unpublished (Borenstein et al., 2009: 277-292; Francis, 2012; Petticrew & Roberts, 2006: 230-235; Rosenthal, 1984: 125; Stanley, 2005) and our search technique excludes findings from books and “grey literature.” Although the findings presented here are robust, our analyses concern a possibly biased sample of all crowding-out estimates that empirical research is able to measure. Being a generally recognized prob-lem of scientific publishing, publication bias is less likely to be a problem in crowding-out research because null findings in this area have important policy implications. An analysis of unpublished studies could be added in order to examine this bias, which is beyond the scope of the current article.

Despite these limitations, this paper makes an important contribution to the literature on the interaction between organizations and their environ-ment. In field research situations, where different environmental process-es are at play, individual giving is generally not strongly affected by varying levels of government support. However, private donors are responsive to changing government support under certain circumstances. When people are aware of government budgets they might change their donations, so the effects of public policy largely depend on information flows. In general we advise policy makers to carefully consider the societal context before decid-ing to reduce public spending, since budget cuts mostly decrease total fund-ing for public goods. This has important consequences for governance styles in which the government collaborates with nonprofit actors, like nonprofit contracting (Smith & Lipsky, 1993), “collaborative governance” (Ansell & Gash, 2008) and interorganizational networks (Milward & Provan, 2003).

There is a widespread belief among politicians and intellectuals that gov-ernment expenditures suppress private participation, an assumption that


lies behind policy decisions in which the government cuts its spending and aims to shift public services towards nonprofit organizations that are largely dependent on private funding. The current meta-analysis shows that in most situations, private charitable donations are not likely to be crowded out by government support and that each dollar of extra public funding increases total contributions to the public good. When governments are able to main-tain high levels of public funding, they may continue to seek collaborations with nonprofit actors as complementary in the funding and implementation of public services. Instead of substituting each other, there is ample opportu-nity for government and nonprofits to jointly enhance the scope and quality of public services in different organizational arrangements.

Part Two

NEW EMPIRICAL ESTIMATES

Part II provides new empirical estimates of the association between government support and charitable donations with data from the Netherlands. The current chapter examines mediating and moderating factors in the relation between government support and charitable do-nations through an innovative mixed-methods design. A unique dataset is obtained, matching individual-level survey data from the Giving in the Netherlands Panel Survey (GINPS) with media coverage of government support from LexisNexis and organizational-level information from the Dutch Central Bureau on Fundraising (CBF) from 2002 to 2014. An in-terpretative analysis shows the ways in which people are informed about changes in public funding, which is assumed to be a prerequisite for do-nors to change their donations. Media coverage often does not reflect actual changes in government support. Additionally, regression analyses are deployed to examine how changes in government support and media reports are associated with changes in donations. The results show that responses to public funding are dependent on the nonprofit context. Do-nations in the fields of social services, health and nature are displaced by government support, while crowding-out does not occur in the field of in-ternational development. Even in fields where crowding-out is more likely to occur, the increase in donations does not offset the decrease in public support. The conclusions nuance popular beliefs about the direct conse-quences that policy changes have for public awareness and participation.

Heterogeneity in crowding-out

Chapter 3

Arjen de Wit, René Bekkers and Marjolein Broese van Groenou

AdW, RB and MBvG designed the study; AdW collected the data and carried out data analysis; AdW, RB and MBvG contributed to writing the article.

This chapter is published as: De Wit, A., Bekkers, R., & Broese Van Groenou, M. (2017). Heterogeneity in Crowding-out: When Are Charitable Donations Responsive To Government Support? European Sociological Review, 33(1), 59-71.

Data, syntax and supplementary materials are available through the Open Science Framework at http://osf.io/yu735/.

92 Chapter 3 Heterogeneity in crowding-out

INTRODUCTION

How do government efforts change the landscape of the voluntary sector? Previous studies have examined the effects of government support on the financial and managerial practice of nonprofit organizations (Andreoni & Payne, 2003; Froelich, 1999; O’Regan & Oster, 2002; Verschuere & De Corte, 2014) as well as on individual participation, networks and social trust (Sta-delmann-Steffen, 2011; Van Oorschot & Arts, 2005). The current paper fo-cuses on private charitable giving. Charitable donors are indispensable for many organizations across the nonprofit sector and it is important to know how they react on contextual changes. An often formulated expectation is that donations are “crowded out” by increasing levels of government sup-port to public goals. The empirical foundations of the crowding-out hypoth-esis are ambivalent, however, as a recent meta-analysis shows that previ-ous findings on the relationship between government financial support and private donations have not been conclusive and depend strongly on the re-search design (De Wit & Bekkers, 2017).

Given the large number of theoretical and empirical publications on the public good crowding-out hypothesis, it is striking that three factors have been understudied in this literature. First, there has been little attention to the information that charitable donors receive about government funding. While Horne et al. (2005) show that most donors do not know how much money governments grant to organizations, the assumption in many studies is that people have perfect information and that they base their decisions on this information. Second, it is likely that there is a wide variety in people’s reactions to varying levels of government funding, but only a small number of empirical studies examined individual heterogeneity. Third, only a few studies examined the role of nonprofit organizations as intermediary actors whose behavior might explain the relationship between government sup-port and private donations.

The question that this paper seeks to answer is how government support and private charitable donations are related, and to what extent this relation can be explained by individual reactions of donors, organizational strategies and media coverage of government policies. Using a unique and innovative research design, the paper formulates and explores relevant mediating and moderating effects that spring from behavioral and institutionalist theories, thereby enhancing our understanding of the ways in which the government, private donors, nonprofit organizations and the media affect each other.

93Heterogeneity in crowding-out Chapter 3

LITERATURE REVIEW

Government support and private donationsThe central argument in the crowding-out debate is that a large government is detrimental for civic life. This claim can be traced back to Alexis de Toc-queville (1970[1840]), who argued that democratic government diminish-es rather than oppresses social action, ruling out private control over the small things in life. In contemporary research the crowding-out hypothesis is investigated in two strands of research. The first line of research takes a rather sociological approach. Incorporating welfare state regime theo-ries and analyzing survey data, studies in this area investigate the effect of cross-national characteristics on different forms of individual participation like volunteering or organizational memberships (Gesthuizen et al., 2008; Kaariainen & Lehtonen, 2006; Koster, 2007; Scheepers & Grotenhuis, 2005; Stadelmann-Steffen, 2011; Van Oorschot & Arts, 2005). The second strand of research consists largely of work of economists and concerns private char-itable giving. Here, crowding-out is mostly translated as individuals com-pensating with donations what the government does not provide (Andreoni, 1993; Andreoni & Payne, 2003, 2011; Okten & Weisbrod, 2000; Payne, 1998; Ribar & Wilhelm, 2002).

The findings in this literature are mixed. While some studies find posi-tive relations between government funding and private donations (Brooks, 2003b; Khanna & Sandler, 2000; Okten & Weisbrod, 2000; Sokolowski, 2013), most studies find a negative correlation (Andreoni & Payne, 2003, 2011; Dokko, 2009; Isaac & Norton, 2013). In a cross-national analysis with Eurobarometer data, Scheepers and Grotenhuis (2005) find that in liberal welfare states more people give to alleviate poverty than in other welfare state regimes.

Individual behaviorHow do charitable donors react on changes in government funding? The main hypothesis here is that government financial support displaces individ-ual donations. Economic crowding-out theories (Roberts, 1984; Warr, 1982) follow a rational choice perspective on social behavior, assuming that a do-nor’s utility function includes a certain contribution to the public good. This individual contribution can be provided either mandatory, through govern-ment expenditures that are financed by taxes, or voluntarily, in the form of donations to nonprofit organizations. When the government funds the pre-


ferred public good with tax money, an increase of government contributions would allow charitable donors to reduce donations without consequences for the nonprofit output.

However, one could argue that public funding has a positive impact on in-dividual donations. Government funding might serve as a “seal of approval” indicating the quality and efficacy of nonprofit output (Schiff, 1990), which would lead donors to increasingly contribute to organizations that are suc-cessful in attracting public funding.

It could also be that there is no causal relationship between government support and donor behavior. In Max Weber’s notion of substantive rational-ity, the ends of social action rather than its goals are leading in driving in-dividual behavior (Weber 1922[1987]: 85-86). Following this perspective, charitable donors are mainly driven by their (political or ethical) values and not by the ultimate economic consequences of their decisions. Donations would then be driven by the content of a nonprofit’s portfolio rather than by its financial revenues.

A recent meta-analysis shows that laboratory experiments generally find partial crowding-out, while studies with organizational or survey data find an average correlation close to zero (De Wit & Bekkers, 2017). This suggests that the rational choice theory holds under controlled circumstances in the lab, while other mechanisms suppress a negative correlation in the field.

Organizational behaviorAnother explanation of a negative relation between government support and private donations is the behavior of voluntary organizations. Sources of nonprofit revenues may affect financial volatility, the extent to which organi-zations change the goals they target, the extent to which organizational pro-cesses and procedures are formalized and professionalized, and the autono-my of nonprofit boards (Froelich, 1999; O’Regan & Oster, 2002; Verschuere & De Corte, 2014).

It is yet unsure how organizations with different levels of dependence from government support differ in their fundraising efforts. On the one hand, organizations could be inclined to invest in fundraising as a compensation strategy when they receive lower government funding (Andreoni & Payne, 2003, 2011). It is especially likely that organizations change their strategies after radical decreases in government funding, as Randall and Wilson (1989) show for the budget cuts of the Reagan administration.

On the other hand, it could be that organizations use different ways of


funding to further increase and diversify their revenues. Extra government support could be used to develop better and more professionalized fundrais-ing techniques, so that higher government support increases private income, too.

While some studies show that fundraising expenditures can be an import-ant explanation of the negative association between government support and private donations (Andreoni & Payne, 2003, 2011; Hughes et al., 2014), a meta-analysis shows that subsidies to organizations are unlikely to displace charitable donations (De Wit & Bekkers, 2017).

WHERE CROWDING-OUT SHOULD OCCUR

Informed donorsA prerequisite for giving as a reaction on changes in government support is availability of information on government actions, because people will not change their donations when they are not aware of any changes in govern-ment support. In an experimental design, Horne et al. (2005) show that most donors do not know how much government support charitable organizations receive, and that estimates of levels of public funding are highly inadequate. Even if people are not aware how much income organizations receive from the government, they could still be informed on policy changes. News media will report budget cuts because they have important consequences for an organization and its goals, as they will report it when an organization gets a large grant for a certain project. People get most of their information on government policies from news media, and government grants are likely to have an effect on individual decisions because they are covered in the media. News items might especially affect donor behavior when they discuss prob-lems within an organization that may need additional funding, like financial concerns or issues regarding personnel, and when they describe (the work and output of) nonprofit organizations on a generally positive tone.

To date, only a handful of studies have empirically examined the effect of media coverage on charitable giving. Both after the 2004 Tsunami and the 2010 Haiti earthquake, more extended coverage on T.V. and in the newspa-pers was associated with higher private giving (Brown & Minty, 2008; Lobb, Mock, & Hutchinson, 2012).


Individual heterogeneityPrevious crowding-out studies have given little attention to individual het-erogeneity in reactions to government policies. Some studies have looked at different income groups (Chan et al., 1996; Güth et al., 2006; Kingma, 1989) or different donor groups (Reeson & Tisdell, 2008), with no conclusive find-ings. In a public good experiment, Luccasen (2012) find complete crowd-ing-out among different player types, genders and social classes. How people react on government policies and information about these policies as depict-ed in the media might depend on their ability to donate and their prosocial values.

First, people who are able to donate might also be better able to change their donations. It is known that people with a paid job and more wealth donate higher amounts than people who are not in paid labor or with low-er wealth, and the higher educated donate more than the lower educated (Bekkers & Wiepking, 2011a; Wiepking & Bekkers, 2012). More financial resources also enable people to change their donations more easily because they decrease the marginal value of a dollar that can be spent on a public goal. Not only is a donor’s spending budget higher with more financial re-sources, the price of giving is also lower in a progressive income tax system including a charitable deduction.

Second, people with stronger prosocial values are more likely to change their giving. People who find it important to help others, who are more em-pathic and who have more confidence in voluntary organizations are gener-ally larger donors (Bekkers & Wiepking, 2011b; Wilhelm & Bekkers, 2010), are expected to be more committed to the output of nonprofit organizations, and may change their donations after changes in government policies.

Organizational heterogeneityThe voluntary sector is unique because of its large diversity. Do changes in government support have the same effect on nonprofit organizations across society? Previous studies showed that the magnitude and direction of crowd-ing-out estimates differ strongly between subsectors of the voluntary sector (Khanna & Sandler, 2000; Yetman & Yetman, 2003) or even between organi-zations within subsectors (Payne, 2001). Two dimensions of organizational heterogeneity are discussed here.

First, there might be stronger crowding-out effects for organizations that receive relatively large amounts of public funding. Multiple studies found an inverted U-shaped relationship between government support and pri-


vate donations (Borgonovi, 2006; Brooks, 2000a, 2003a; Nikolova, 2015). This could be due to donor perceptions, as Borgonovi (2006) suggests that low levels of government support serves as a signal of efficacy while donors start to perceive public funding as undesirable government control above a certain threshold. It could also be that subsidy-dependent organizations are more financially stable and less strongly focused on fundraising activities (Froelich, 1999; O’Regan & Oster, 2002).

Second, there might be differences between subsectors because of the na-ture of the public good that is provided. While not all charitable giving is directly substitutable for government funding, crowding-out is most likely to occur in areas where they are in direct competition (Stadelmann-Steffen, 2011). While a shelter for homeless people is a tangible service where in-vestments have immediate consequences for nonprofit output, international development aid is a goal where the need is practically infinite. Both donors and professionals in nonprofit organizations might be more responsive to government support if the public good can be equally provided by public or private funding and if the consequences of a change in total public good provision are more visible.

DATA AND STRATEGY

To examine the responsiveness of donors to changes in government support, a dataset has been created matching individual donor behavior to specific organizations with organizational-level data from annual reports and media archives. The units of analysis in this study are dyads of individuals and or-ganizations. Individual-level data were used from 6 waves (2002-2014) of the Giving in the Netherlands Panel Survey (Bekkers et al., 2016), a biennial survey which is nationally representative.

In the survey respondents were asked whether or not their household do-nated in the previous calendar year to a list of the largest charitable organi-zations in the Netherlands, and if yes, what amount. In 2006 four health care organizations (Alzheimer, Longfonds, Diabetes Fonds and Nierstichting) were not in the list of organizations, so they were attributed missing values for these years. The phenomenon under study is the change in donations compared to the previous wave.

To measure media coverage of government support to organizations the LexisNexis database was searched for articles in seven national subscribed


newspapers in the Netherlands, collecting articles published within two years, the year in which donations are measured and the preceding year, that include both the name of the organization and the Dutch word “subsidie” or “overheidssubsidie” in the title and/or text. Only articles on government sup-port to an organization were included, so articles were omitted when they concern grants that are given by an organization and when the government support is actually unrelated to the organization. Each article was coded on (1) whether it mentioned increasing government funding, budget cuts or no change in government funding, (2) whether it mentioned internal problems within the organization (e.g. issues regarding finances or personnel) or not, and (3) whether it described the (work of) the nonprofit organization as generally positive, negative or neutral. By reading and coding the content of all articles the media analysis contains both an interpretative analysis and a quantitative measure that was used in the regression analyses.

As measures of resources, three dummy variables indicate whether a re-spondent achieved higher (tertiary) education, has a paid job (either part-time or full-time), and owns a home.

Values were measured by scales of altruistic values and empathic concern, as well as a single-item measure of trust in charitable organizations in the Netherlands (Bekkers et al., 2016). All answers were recoded from Likert scales to dichotomous variables where 1 means a high score.

Information on government funding of the organizations under study was adopted from the Central Bureau on Fundraising (CBF), a nongovernmental accreditation organization that monitors income and expenditures of Dutch charities (Bekkers, 2003). The amounts for each year were divided by the number of households in that year as indicated by Statistics Netherlands (CBS) in order to have all variables on the level of the household.

Large changes in donations, government support and media coverage have a disproportionally large influence on the results from the analysis. To mitigate the effect of extreme values the change variables were treated for outliers by setting the 5% most negative values on the border of the 5th per-centile and the 5% most positive values on the border of the 95th percen-tile. This procedure has been labeled “Winsorizing”, after Charles P. Winsor (Tukey, 1962: 17-19).

In the pooled dataset (23,094 observations among 2,175 respondents) ev-ery unique combination of a respondent i and an organization j represents a dyad with various observations at different years t. Table 1 displays descrip-tive statistics. Note that respondents who did not donate to an organization


were excluded. To explore the validity of different arguments in the crowd-ing-out debate two methods were used that complement and strengthen each other. First, the time trends of donations, government financial support and fundraising expenditures were examined for all organizations and each organization separately, and the content of media coverage was analyzed for three organizations with different trends in government support. Second, re-gression models were deployed to explore the main relation between dona-tions and government support as well as possible mediating and moderating effects.

The following mixed-effects regression model is deployed:

ΔYijt = β0 + u0j + vi + β1ΔGjt-1 + u1j ΔGjt-1 + β2ΔEt-1 + β3ΔEt + β4ΔOjt-1 + β5ΔPt-1 + β6ΔTt-1 + εijt

in which ΔY is the change in charitable donations by donor i to organi-zation j from year t-2 to year t, u0 is the organization-specific intercept, v is the individual-specific intercept, ΔG is the change in government support to the organization from year t-3 to year t-1 and u1j is the organization-specific random slope. Control variables include the change in GDP per capita ΔE, the change in the organization’s total expenditures on its mission ΔO, the change in the presence of the Labor Party (PvdA) in the national government coali-tion ΔP and the change in total government social transfers ΔT.

The data are cross-nested on three levels, and random intercepts were added for respondents and organizations to account for this structure. Fur-thermore, the model allows slopes to vary between organizations, because government support might have different effects across organizations giv-en the large variety in missions, management structures and donor bases. The first difference regression provides estimates of changes in time, ruling out the between-individuals and between-organization effects. To estimate a lagged effect, changes in government support were measured one year pre-ceding the year of donating. However, there may still be confounding factors that influence the coefficient of government support, and four control vari-ables are included in order to reduce omitted variable bias due to the overall economic cycle, the growth of an organization’s budget, a government that is more supportive of social programs and the overall change in government spending.


Tabl

e 1:

Des

crip

tive

stat

istic

s

Vari

able

Mea

suri

ngM

ean

Std

dev

Min

Max

Dona

tions

Amou

nt d

onat

ed (€

)12

.718

36.3

080

1696

.353

Δ Do

natio

nsCh

ange

in a

mou

nt d

onat

ed (€

)-0

.888

40.8

63-2

424.

242

1128

.589

Δ Do

natio

nsCh

ange

in a

mou

nt d

onat

ed (€

), tr

eate

d fo

r out

liers

-0.4

2510

.694

-24.

917

24.2

95Go

vt su

ppor

tGo

vern

men

t sup

port

per

hou

seho

ld (€

)2.

280

6.31

60

34.3

83Δ

Govt

supp

ort

Chan

ge in

gov

ernm

ent s

uppo

rt p

er h

ouse

hold

(€)

0.12

71.

009

- 5.1

684.

363

Δ Go

vt su

ppor

tCh

ange

in g

over

nmen

t sup

port

per

hou

seho

ld (€

), tr

eate

d fo

r out

liers

0.10

00.

683

-1.4

332.

592

Δ N

ews i

tem

s on

govt

supp

ort

Num

ber o

f new

spap

er a

rtic

les o

n go

vern

men

t sup

port

0.20

45.

885

-31

32

Δ N

ews i

tem

s on

govt

supp

ort

Num

ber o

f new

spap

er a

rtic

les o

n go

vern

men

t sup

port

, tre

ated

for

outli

ers

0.27

34.

217

-10

13

Δ N

ews i

tem

s on

budg

et cu

tsN

umbe

r of n

ewsp

aper

art

icle

s on

decr

easi

ng g

over

nmen

t sup

port

0.20

33.

072

-24

26

Δ N

ews i

tem

s on

budg

et cu

tsN

umbe

r of n

ewsp

aper

art

icle

s on

decr

easi

ng g

over

nmen

t sup

port

, tr

eate

d fo

r out

liers

0.14

52.

050

-66

Δ N

ews i

tem

s on

prob

lem

sN

umbe

r of n

ewsp

aper

art

icle

s on

orga

niza

tiona

l pro

blem

s0.

246

4.45

0-2

218

Δ N

ews i

tem

s on

prob

lem

sN

umbe

r of n

ewsp

aper

art

icle

s on

orga

niza

tiona

l pro

blem

s, tr

eate

d fo

r ou

tlier

s0.

365

3.75

2-1

014

Δ Po

sitiv

e ne

ws i

tem

sN

umbe

r of n

ewsp

aper

art

icle

s tha

t are

pos

itive

ly fr

amed

0.11

61.

086

-46

Δ Po

sitiv

e ne

ws i

tem

sN

umbe

r of n

ewsp

aper

art

icle

s tha

t are

pos

itive

ly fr

amed

, tre

ated

for

outli

ers

0.12

10.

825

-13

Hig

her e

duca

ted

Achi

eved

tert

iary

edu

catio

n (n

o/ye

s)0.

274

0.44

60

1Pa

id jo

bH

avin

g a

paid

job

(no/

yes)

0.55

40.

497

01

Own

hom

eOw

ning

a h

ome

(no/

yes)

0.63

10.

483

01

Altr

uist

ic v

alue

sSc

ore

on a

ltrui

stic

val

ues s

cale

0.25

80.

438

01

Empa

thic

conc

ern

Scor

e on

em

path

ic co

ncer

n sc

ale

0.52

50.

499

01

Trus

tTr

ust i

n Du

tch

char

ities

0.41

00.

492

01


Tabl

e 1

(con

tinue

d): D

escr

pitiv

e st

atis

tics

Vari

able

Mea

suri

ngM

ean

Std

dev

Min

Max

Fund

rais

ing

expe

nditu

res

Fund

rais

ing

expe

nditu

res p

er h

ouse

hold

(€)

0.78

30.

489

0.10

71.

960

Δ Fu

ndra

isin

g ex

pend

iture

sCh

ange

in fu

ndra

isin

g ex

pend

iture

s per

hou

seho

ld (€

)0.

021

0.13

9-0

.601

0.41

9

Δ Fu

ndra

isin

g ex

pend

iture

sCh

ange

in fu

ndra

isin

g ex

pend

iture

s per

hou

seho

ld (€

), tr

eate

d fo

r ou

tlier

s0.

013

0.09

8-0

.181

0.18

6

Subs

idy-

depe

nden

cyGo

vern

men

t sup

port

/ T

otal

inco

me

0.11

70.

212

00.

922

Soci

al/h

ealth

Orga

niza

tion

in th

e fie

ld o

f soc

ial s

ervi

ces o

r hea

lth (n

o/ye

s)0.

565

0.50

00

1N

atur

eOr

gani

zatio

n in

the

field

of n

atur

e co

nser

vatio

n (n

o/ye

s)0.

180

0.38

40

1In

tern

atio

nal

Orga

niza

tion

in th

e fie

ld o

f int

erna

tiona

l dev

elop

men

t (no

/yes

)0.

255

0.43

60

1

Tabl

e 1:

Des

crip

tive

stat

istic

s

Vari

able

Mea

suri

ngM

ean

Std

dev

Min

Max

Dona

tions

Amou

nt d

onat

ed (€

)12

.718

36.3

080

1696

.353

Δ Do

natio

nsCh

ange

in a

mou

nt d

onat

ed (€

)-0

.888

40.8

63-2

424.

242

1128

.589

Δ Do

natio

nsCh

ange

in a

mou

nt d

onat

ed (€

), tr

eate

d fo

r out

liers

-0.4

2510

.694

-24.

917

24.2

95Go

vt su

ppor

tGo

vern

men

t sup

port

per

hou

seho

ld (€

)2.

280

6.31

60

34.3

83Δ

Govt

supp

ort

Chan

ge in

gov

ernm

ent s

uppo

rt p

er h

ouse

hold

(€)

0.12

71.

009

- 5.1

684.

363

Δ Go

vt su

ppor

tCh

ange

in g

over

nmen

t sup

port

per

hou

seho

ld (€

), tr

eate

d fo

r out

liers

0.10

00.

683

-1.4

332.

592

Δ N

ews i

tem

s on

govt

supp

ort

Num

ber o

f new

spap

er a

rtic

les o

n go

vern

men

t sup

port

0.20

45.

885

-31

32

Δ N

ews i

tem

s on

govt

supp

ort

Num

ber o

f new

spap

er a

rtic

les o

n go

vern

men

t sup

port

, tre

ated

for

outli

ers

0.27

34.

217

-10

13

Δ N

ews i

tem

s on

budg

et cu

tsN

umbe

r of n

ewsp

aper

art

icle

s on

decr

easi

ng g

over

nmen

t sup

port

0.20

33.

072

-24

26

Δ N

ews i

tem

s on

budg

et cu

tsN

umbe

r of n

ewsp

aper

art

icle

s on

decr

easi

ng g

over

nmen

t sup

port

, tr

eate

d fo

r out

liers

0.14

52.

050

-66

Δ N

ews i

tem

s on

prob

lem

sN

umbe

r of n

ewsp

aper

art

icle

s on

orga

niza

tiona

l pro

blem

s0.

246

4.45

0-2

218

Δ N

ews i

tem

s on

prob

lem

sN

umbe

r of n

ewsp

aper

art

icle

s on

orga

niza

tiona

l pro

blem

s, tr

eate

d fo

r ou

tlier

s0.

365

3.75

2-1

014

Δ Po

sitiv

e ne

ws i

tem

sN

umbe

r of n

ewsp

aper

art

icle

s tha

t are

pos

itive

ly fr

amed

0.11

61.

086

-46

Δ Po

sitiv

e ne

ws i

tem

sN

umbe

r of n

ewsp

aper

art

icle

s tha

t are

pos

itive

ly fr

amed

, tre

ated

for

outli

ers

0.12

10.

825

-13

Hig

her e

duca

ted

Achi

eved

tert

iary

edu

catio

n (n

o/ye

s)0.

274

0.44

60

1Pa

id jo

bH

avin

g a

paid

job

(no/

yes)

0.55

40.

497

01

Own

hom

eOw

ning

a h

ome

(no/

yes)

0.63

10.

483

01

Altr

uist

ic v

alue

sSc

ore

on a

ltrui

stic

val

ues s

cale

0.25

80.

438

01

Empa

thic

conc

ern

Scor

e on

em

path

ic co

ncer

n sc

ale

0.52

50.

499

01

Trus

tTr

ust i

n Du

tch

char

ities

0.41

00.

492

01


DESCRIPTIVE ANALYSIS

Figure A1 in the Appendix shows how donations, government support and organizational fundraising expenditures developed over the years. Dona-tions, not treated for outliers, peaked in 2005 at 4.4 Euro per household per year and then declined to 3.4 in 2009, 3.7 in 2011 and 3.4 in 2013. Gov-ernment support generally increased between 2002 and 2012, peaking at 3.5 Euro per household in 2008 and then slightly declining in the years that the economic recession hit the Netherlands and a right-wing administration came in charge. To the extent that fundraising expenditures changed they follow a similar curve as government support, with a peak in 2009. Across all organizations there is not much change on average in donations, govern-ment support and fundraising expenditures, but more pronounced patterns are visible when organizations are examined separately.

Several organizations (Dierenbescherming, Longfonds, Nierstichting, Plan Nederland, Red Cross) had to cope with decreasing levels of donations whereas others (Amnesty International, KWF Kankerbestrijding) seemed successful in attracting more private donations over time. Government sup-port substantially decreased for international aid organizations Doctors without borders, Oxfam Novib and Plan Nederland, but the Salvation Army received more and more government grants over the years. Two organiza-tions, Greenpeace and health care association Hartstichting, did not receive any government funding at all but still experienced volatile fundraising rev-enues. A clear picture of crowding-out or crowding-in does not emerge from the graphs.

The description below shows how media coverage developed for three organizations with varying revenue patterns: one with no clear trend in gov-ernment support, one that gained increasing public funding over the years and one that faced heavy budget cuts.

No clear trend: the Red CrossBeing one of the most well-known international nonprofit organizations, the Red Cross provides aid both in the Netherlands and abroad. Government support increased from 2002 to 2008, after which it fell down until 2010 and then was raised again. Donations followed a declining trend from 2003 to 2013. The drop in government support from 2008 to 2010 was followed by a slight increase in donations. Note that the fundraising expenditures, which are generally very stable over time, decreased in these two years, in contrast


with the idea of “fundraising crowd-out” (Andreoni & Payne, 2003, 2011).How did the Red Cross appear in the media? The graphs on the left hand

of Figure 1 show the total number of articles in seven newspapers and the number of articles on decreasing and increasing government support. The right-hand graphs display the number of articles discussing problems in the organization, like financial problems, as well as the number of news items with a generally negative or positive tone towards the (work of) the organi-zation. There are two clear peaks. The organization was often named when the Minister of Health, Well-Being and Sports (VWS) announced a number of budget cuts in 2003, where the Red Cross ultimately escaped the cuts after the plans were discussed in parliament. Around 2010 the organization was named in a series of critical articles on top salaries of board members, which is visible in the peak in negatively framed items.

The actual increases and decreases in government support are not men-tioned in the newspapers, making it less likely that they had an effect on individual donor behavior.

Increasing government support: the Salvation ArmyThe Salvation Army is a large service provision organization, based on a Chris-tian identity, and, at least in the Netherlands, heavily subsidy-dependent (in 2012, public funding accounted for 90% of Salvation Army’s total revenues). The Dutch government provides grants for each client that is helped by orga-nizations like the Salvation Army, so the amount of public funding increases with the number of people that are served. The steady increase in govern-ment support from 2002 to 2012 went together with decreasing donations on average from 2003 to 2013.

The number of newspaper articles on the Salvation Army is shown in Fig-ure 1. The Salvation Army appears in the media quite often. There are news items about public funding and fundraising in general and about specific projects that received government grants across the years. A number of crit-ical articles in 2001 discussed the organization’s definition of “homeless”, which was said to include as many people as possible in order to claim more public money. Problems for the organization appeared in the news in 2003 when the national government announced to cut budgets on a number of nonprofit organizations. Also in 2003, an Amsterdam-based project lost its local government funding. In 2005, the Minister of Social Affairs announced to withdraw a 200,000 Euros grant because the organization refused to hire two Muslim women (the Salvation Army aims to hire Christians only). In a


Figu

re 1

: Rev

enue

s fro

m a

nd n

ewsp

aper

arti

cles

on

the

Red

Cro

ss, t

he S

alva

tion

Arm

y an

d O

xfam

Nov

ib (2

002-

2013

)

012345

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Red

Cros

s (in

tern

atio

nal)

Dona

tions

Gove

rnm

ent s

uppo

rt

Fund

raisi

ng e

xpen

ditu

res

051015202530

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Red

Cros

s (in

tern

atio

nal)

New

s ite

ms o

n go

vern

men

t sup

port

New

s ite

ms o

n cu

ts

New

s ite

ms o

n in

crea

sing

fund

ing

024681012141618

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Red

Cros

s (in

tern

atio

nal)

New

s ite

ms o

n pr

oble

ms

Posit

ive

new

s ite

ms

Nega

tive

new

s ite

ms

010203040

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Salv

atio

n Ar

my (

socia

l)

Dona

tions

Gove

rnm

ent s

uppo

rt

Fund

raisi

ng e

xpen

ditu

res

01020

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Salv

atio

n Ar

my (

socia

l)

New

s ite

ms o

n go

vern

men

t sup

port

New

s ite

ms o

n cu

ts

New

s ite

ms o

n in

crea

sing

fund

ing

0246810

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Salv

atio

n Ar

my (

socia

l)

New

s ite

ms o

n pr

oble

ms

Posit

ive

new

s ite

ms

Nega

tive

new

s ite

ms

05101520

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Oxfa

m N

ovib

(int

erna

tiona

l)

Dona

tions

Gove

rnm

ent s

uppo

rt

Fund

raisi

ng e

xpen

ditu

res

0102030405060

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Oxfa

m N

ovib

(int

erna

tiona

l)

New

s ite

ms o

n go

vern

men

t sup

port

New

s ite

ms o

n cu

ts

New

s ite

ms o

n in

crea

sing

fund

ing

0510152025303540

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Oxfa

m N

ovib

(int

erna

tiona

l)

New

s ite

ms o

n pr

oble

ms

Posit

ive

new

s ite

ms

Nega

tive

new

s ite

ms


similar debate, in 2009, a number of articles discuss a proposal by the city council of Amsterdam to stop subsidizing organizations that discriminate in their employee policy. Some media gave voice to arguments in favor of continuing public funding, which is represented by the spike in positively framed items.

In sum, the media coverage does not reflect the general trend in govern-ment support. Although there has been some reporting on the reasons for the Salvation Army to acquire government support, a general increase in public funding does not withhold newspapers from writing about the gov-ernment cutting specific grants.

Budget cuts: Oxfam NovibOxfam Novib receives a large share of its funding from governments, al-though not as much as the Salvation Army (public funding accounted for 52% of Oxfam’s total revenues in 2012). Government support and donations follow a similar trend in time. There is a clear drop in government support after 2008. Donations increased between 2005 and 2009, after which they decreased. Here, private donations seem to follow government support.

This might be due to media coverage on changing government policies. Figure 1 shows the number of newspaper articles on Oxfam Novib. A first peak in the years 2003-2004 reflects a discussion about the government setting new rules before international aid organizations could receive pub-lic funding, resulting in some news items with a rather negative tone. An even higher peak is shown after a right-wing administration came in charge. Dramatic budget cuts on international aid organizations were announced in 2010 and resulted in a lot of media attention for the organization’s problems. Oxfam anticipated on decreasing government funding by firing employees and abandoning all of its projects in Latin-America, resulting in even less government funding.

Private donations decreased in the years after the budget cuts, suggesting that donors follow the government in its policy choices, which they were likely to know about since they have been reported across all newspapers in the sample.

A more systematic analysis of the suggested mechanisms is provided in the regression analyses below.


REGRESSION ANALYSES

Government support and private donationsA formal test of the relation between government support and private do-nations is presented in the regression models on the change in donations in Table 2. The coefficient indicates that each Euro extra government support is generally associated with a 0.09 Euro decrease in donations, which is not statistically significant (Model I).

Media coverageThere is no clear relationship between the number of newspaper articles that are published on an organization and the amount that donors give to this organization (Model II). To examine the effects of media coverage with a different content, Model III shows the coefficients of changes in news items about budget cuts, news items about organizational problems and positively framed news items. The coefficients are positive but none of them is statisti-cally significant. Interestingly, the coefficient of a change in total news items is negative and significant in this model, indicating that there is content in the news other than budget cuts, organizational problems and positive fram-ing that discourage donors.

Individual heterogeneityDo reactions to changes in government support depend on individual char-acteristics like financial resources, educational level or one’s values? We in-cluded interaction effects with six individual characteristics, and graphically show the interactions that are statistically significant in Figure 2.

People with stronger altruistic values are more inclined to follow govern-ment support with their donations (see the top panel of Figure 2). This is contrary to our reasoning that people with stronger prosocial values would substitute government support. It rather suggests that those are the people that perceive changes in government support as a signal of nonprofit quali-ty. However, although the slopes are significantly different between groups, the separate coefficients for government support in each group are not sta-tistically significant. This means that there is no significant crowding-in or crowding-out among people with lower or higher altruistic values.

The association between the number of newspaper articles and charitable donations is significantly negative among the higher educated (the marginal effect among the higher educated is β=-0.069 with p=0.024, see the center


Table 2: Maximum Likelihood estimation on Δ Donations

I II III IVΔ Govt support -0.089 -0.113 -0.157 -0.120

(0.195) (0.184) (0.178) (0.192)Media coverageΔ News items on govt support -0.024 -0.074**

(0.019) (0.034)Δ News items on budget cuts 0.047

(0.051)Δ News items on problems 0.027

(0.030)Δ Positive news items 0.123

(0.110)FundraisingΔ Fundraising expenditures 2.378***

(0.772)(Constant) 0.057 0.061 0.090 0.007

(0.153) (0.153) (0.154) (0.152)AIC 206,662 206,663 206,666 206,655BIC 206,725 206,731 206,751 206,723Observations 27,284 27,284 27,284 27,284Organizations 19 19 19 19Respondents 2,201 2,201 2,201 2,201

Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1; Controlled for changes in GDP per capita, whether the Labor Party is in the government coalition, total social transfers from government and total expenditures on the organization’s mission.

panel of Figure 2). The higher educated are generally larger donors and are more likely to read the newspapers, so this finding suggests that informed donors are more responsive to changes in government support.

FundraisingFundraising efforts might explain a part of the relationship between govern-ment support and charitable donations. As one might expect, fundraising expenditures are positively related to the amount people donate to an orga-nization (Model IV). Compared to Model I, the coefficient of government sup-port is more strongly negative in this model, indicating that fundraising is positively correlated with both government support and donations. Rather


Figure 2: Linear prediction for Δ Govt support and Δ News items on govt support

-1.2

-0.8

-0.4

0

0.4

0.8

-1.3 -1 0 1 2 2.5Change in government support

Altruistic valuesLow High

-1.6

-1.2

-0.8

-0.4

0

0.4

-9 -5 0 5 10 12Change in news items

EducationLow High

-3.5-3

-2.5-2

-1.5-1

-0.50

0.51

-1.3 -1 0 1 2 2.5Change in government support

Nonprofit sectorInternational Social Nature


than to support the idea of “fundraising crowding-out”, the Dutch data sug-gest that organizations use government funding to increase their fundraising success in the private market.

Organizational heterogeneityNext, we examine the extent to which the impact of changing government support systematically differs across organizations.

We included an interaction between government support and the degree to which organizations are dependent of public funding in the year under study, which is positive and not statistically significant (not shown).

The bottom panel of Figure 2 shows interaction effects between nonprofit sectors and changing government support. Among organizations in the field of health and social services (β=-0.349, p=0.010) as well as in the field of na-ture (β=-0.991, p=0.006), government support is negatively associated with charitable donations. In the field of international development the associa-tion is positive and not significant (β=0.271, p=0.174). These results are in line with the expectation that crowding-out is more likely in sectors where both public and private money fund similar public goods.

RobustnessBecause the results in the regression analyses can be mainly driven by one exceptional organization, all models have been re-estimated excluding one organization each time and excluding the two organizations that did not receive any government funding. Not surprisingly, the sample of 19 orga-nizations is not large enough to draw robust conclusions about systematic effects across the nonprofit sector. Full results of the robustness checks are available at https://osf.io/yu735/.


There is much uncertainty about the effects of government efforts on the fundraising income of nonprofit organizations. Despite the large body of lit-erature on crowding-out there is no conclusive evidence, and the availability of information, individual donor characteristics and organizational charac-teristics are understudied. The current study offers a mixed-method design in which longitudinal micro-level data is matched with data on media cov-erage and financial information from annual reports of voluntary organiza-


tions. Although previous studies combined survey data with data on firms at one point in time (Kingma, 1989; Manzoor & Straub, 2005), we are not aware of any previous study that uses longitudinal micro-level data to test for crowding-out effects.

We found no significant crowding-in or crowding-out among any of the social groups. This can be interpreted as evidence for behavioral models based on substantive rationality (Weber, 1922[1987]), in which social action is mainly driven by values and donors are not responsive to changes in orga-nizational finances. However, it could also be that some donors are willing to substitute government support while others perceive it as a signal of organi-zational quality, and that both effects rule each other out. Further research could examine individual heterogeneity in crowding-out effects across more dimensions than we tested for here.

The validity of the crowding-out hypothesis is largely dependent on the organizational context. In the field of social services and health as well as in the field of nature, donations substitute government support, suggesting that crowding-out is most likely to occur in sectors that are close to the individ-ual donor and where public and private revenues are in direct competition (Stadelmann-Steffen, 2011). In the field of international development, on the contrary, crowding-out is not likely to occur. This is in line with previous crowding-in findings in international development (Herzer & Nunnenkamp, 2013; Nunnenkamp & Öhler, 2012). A striking example is Oxfam Novib. After the central government announced large budget cuts on several internation-al aid organizations, which were widely reported in newspapers, donations to Oxfam decreased.

Across all organizations, donors are not responsive to media coverage of policy changes. This confirms previous findings among charitable donors who are informed about a national fundraising campaign (Yörük, 2012) and public funding to nonprofit organizations (Horne et al., 2005). However, a multivariate analysis controlling for media content suggests that an increase in neutral information about nonprofit funding is associated with declining levels of giving to those organizations. Furthermore, there are some excep-tional social groups that might be more responsive to information about pol-icy changes. The higher educated, who are larger consumers of newspapers, are more likely to reduce donations when more articles are published about public funding. Information that is channeled through news media only af-fect a small group of interested donors, which calls for more research on how media coverage of nonprofit organizations affects different segments of


charitable donors. It would be highly interesting to see whether these findings can be rep-

licated with similar research designs in other contexts. Most crowding-out research comes from the United States. Although it is likely that differences in legislation and culture account for different findings between countries, it might very well be that replications of the current study in other countries confirm the heterogeneity in crowding-out effects.

Although the data has considerable quality, the sample suffers from a few limitations. The analysis only concerns people who participate in at least one wave of the study and people who donated at least once to an organization, so the sample under study consists of people who are willing to participate in surveys and to donate to charitable organizations. The analysis only esti-mates changes in amounts donated and does not allow for conclusions about people who start and stop donating, neither does it include organizations that did not receive any government funding over the years. With 19 organi-zations in the regression sample it is hard to make strong claims about the entire population of charitable organizations in the Netherlands. Also, there are aspects of media coverage that are associated with donations other than those in our analyses, and future research on media coverage and charitable giving should be more fine-grained.

Despite these limitations, the findings offer valuable conclusions for man-agers in the nonprofit and public sector. To the extent that policy changes have direct consequences for public awareness and participation, their ef-fects are highly dependent on the organizational context. In the fields of na-ture, health and social services, there is partial crowding-out. This means that decreasing government spending leads to decreasing total contribu-tions to nonprofit output, because the overall increase in donations do not offset the overall decrease in public support. In the field of international de-velopment, donations are not likely to substitute government support at all. Governments should be careful with large budget cuts like the one on de-velopment aid in the Netherlands, which was widely covered in news media and followed by decreasing donations to development aid organizations.

Chapter 3 showed that news coverage in the media often does not reflect actual changes in government support. Yet, the availability of informa-tion is a prerequisite for citizens to change their donations as a response to government support. The current chapter examines the moderating effect of information. Are private donors willing to replace cuts in gov-ernment funding if they are provided with relevant information? A survey experiment was conducted (n=2,458) to examine how information about government funding affect decisions to donate money to a large charita-ble organization in the Netherlands. Providing information about actual budget cuts increases the number of donors, attracting donors from oth-er organizations but also some who otherwise would not have donated. Exploratory analyses reveal that the magnitude of the effect is stronger for citizens with lower levels of empathic concern. The conclusions of this paper show that policy information not only shapes attitudes towards government, but also civic engagement.

The role of information

Chapter 4

Arjen de Wit and René Bekkers

AdW and RB designed the study; AdW carried out data analysis; AdW and RB con-tributed to writing the article.

Submitted for publication as: De Wit, A., & Bekkers, R. Can Charitable Donations Compensate for a Reduction in Government Funding? The Role of Information. A previous version of this chapter was presented at the 44th Annual Meeting of the As-sociation for Research on Nonprofit Organizations and Voluntary Action in Chicago, IL (USA), November 2015.

Data, syntax and supplementary materials are available through the Open Science Framework at http://osf.io/qf2py/.

114 Chapter 4 The role of information

INTRODUCTION

Information availability has become an important topic in the debate on cit-izen participation. Recent studies have examined the effects of exposure to information on trust in government (Grimmelikhuijsen & Klijn, 2015; Grim-melikhuijsen & Meijer, 2014; Kim & Lee, 2012; Tolbert & Mossberger, 2006), political participation (Gerber, Karlan & Bergan, 2009; Lassen, 2005; Worthy, 2010, 2015) and support for welfare state programs (Lergetporer, Schwerdt, Werner, & Woessmann, 2016, Slothuus, 2007). A recent review of the liter-ature shows that the effects of government transparency on legitimacy, cit-izen participation, trust in government and satisfaction tend to be positive, but that findings are inconclusive (Cucciniello, Porumbescu & Grimmelikhu-ijsen, 2017).

This paper examines the effects of policy information on charitable dona-tions. Because governments and nonprofit organizations often work in the same fields, it is likely that charitable donations are affected by information about policy content. Do citizens who are aware of decreasing government spending to a large health organization increase their giving to this organiza-tion, because their donations can compensate for the budget cuts?

There are many empirical studies dedicated to the relationship between government spending and individual private donations (see for a review De Wit & Bekkers, 2017). Laboratory experiments have provided support for the hypothesis that taxes and voluntary donations are partial substitutes: participants demonstrate a tendency to give more to a nonprofit organiza-tion when they know that the organization receives less funding from sub-sidies financed by taxes. The assumption in virtually all these studies is that people have perfect information about government policies. Thus far, only a handful of studies examined how charitable giving and volunteering are af-fected by different levels of exposure to knowledge about public policies (De Wit, Bekkers & Broese van Groenou, 2017, Horne et al., 2005, Jones, 2015, Yörük, 2012). Given that behavioral responses to government policies are partly dependent on the available knowledge, we should shift our attention to the information on which social preferences are based.

Our research aim is threefold. First, we examine how information about government support affects indidivual charitable giving. We provide a ran-dom selection of participants in a large representative panel survey with information about a reduction in government funding to a well-known non-profit organization in the Netherlands and observe the change in their be-

115The role of information Chapter 4

havior by comparison to a control group in which we provide no information. Second, we examine to what extent such information has the ability to

draw non-donors into donating. It might be that information about a specif-ic nonprofit organization attracts existing donors from other organizations rather than non-donors, which would not increase the total size of the fund-raising market (Ek, 2017; Reinstein, 2006, 2007).

Third, we examine how the effect of providing information differs across social groups. Effects of government policies on charitable donations are likely to be heterogeneous across individuals (De Wit et al., 2017). Because the experiment is part of a larger panel survey, this research design provides us with the possibility to analyze how the effects of information varies across groups of citizens with different characteristics.

The results of this study show how information about public policy shapes civic engagement. Evidence that public budget cuts can lead to a larger sum of charitable income would have important consequences for public policy. It would show that information about government policy can not only change attitudes, but also actual citizen participation outside the political sphere. Nonprofit organizations have increasingly important roles in the implemen-tation of public policy through contracting, collaborations and partnerships (Ansell & Gash, 2008; Milward & Provan, 2003; Smith & Lipsky, 1993). Espe-cially in times of fiscal stress, governments seek to outsource services (Geys & Sørensen, 2016). It is therefore important to know how information about public policies affects engagement in the nonprofit sector.

THEORY

Information on government fundingIn collaborative and networked governance, nonprofit organizations are im-portant actors (Ansell & Gash, 2008; Milward & Provan, 2003). A distinct fea-ture of nonprofits is that many of them are partly or even fully dependent on income from charitable donors. They use income from government support, charitable donations and other sources to produce the desired outcomes. However, these revenue streams are not independent from each other.

The argument that extensive government programs “crowd out” chari-table donations is formulated and tested in different disciplines across the social sciences (for a review, see De Wit & Bekkers, 2017). In the current academic debate, the crowding-out hypothesis is mostly based on econom-


ic theories that argue that, at least to some extent, people donate to public goods because they care about the welfare of the recipients (Andreoni 1989, 1990; Roberts, 1984; Warr, 1982). This is what behavioral economists call altruism, a motivation for charitable giving that is based on the need of the recipient.

Most experimental research suggests that private donors are sometimes willing to replace cuts in government financial support, as people give higher voluntary donations to a public good when their mandatory contribution is lower (Andreoni, 1993; Chan et al., 2002; Eckel et al., 2005; Hsu, 2008; Kore-nok et al., 2014; Reeson & Tisdell, 2008) and when the beneficiary organiza-tion is not funded by public money (Kim & Van Ryzin, 2014).

Experiments find much stronger crowding-out effects than studies using survey data or archival data, which is likely due to the assumptions that are made in different research designs (De Wit & Bekkers, 2017; Tinkelman, 2010). In laboratory experiments, participants are presented with the choice to give money (partly) away or not. In different experimental conditions, a part of their endowment is transferred to the recipient organization before they make a choice, which simulates a government tax. This procedure makes it clear to participants that a third party is also funding the organization, en-suring that the assumption of full information about levels of government taxation is satisfied.

It is an empirical question to what extent this assumption is true in re-al-world donation decisions. The empirics are not in favor of the assump-tion of full information. In Germany, only 2.7% estimated total government spending on education within 10% of the actual value (Lergetporer et al., 2016). In the United Kingdom citizens strongly underestimate the amount of public funding that goes to medical research (Shah, Sussex, & Hernandez-Vil-lafuerte, 2015). A survey in the United States in which people were asked to estimate the percentage of public funding to nonprofit organizations showed that 45% answered “don’t know”, whereas 28% guessed within 10 percent-age points of the correct percentage (Horne et al., 2005). The majority of Canadians who are asked to classify firms as public, non-profit or for-prof-it fail to do so correctly for most organizations (Handy et al., 2010). Given that popular knowledge about public funding is often not accurate, it is the question to what extent decisions in charitable giving are based on actual government spending.

Up to now a couple of studies examined how donations are affected by different levels of exposure to information about government support. How-


ever, findings are inconclusive. In their follow-up analyses on people who estimated the levels of public funding correct and people who did not, Horne et al. (2005) find no differences in reported charitable giving. Yörük (2012) shows that people who are informed about a nation-wide fundraising cam-paign in the US tend to do more voluntary work, but do not increase their charitable giving. In the Netherlands, there is no overall association between newspaper coverage on government funding and charitable donations (De Wit et al., 2017). Hence, the effect of policy information on charitable dona-tions is not certain.

In an experimental design that most closely resembles the one as present-ed here, Shah et al. (2015) provide respondents with scenarios about levels of government spending on medical research and about hypothetical chang-es in spending. They find that citizens are more likely to increase private giving when they are informed about actual government spending, and that the willingness to donate further increases in the case of hypothetical budget cuts.

In sum, exposure to information about decreasing levels of government funding equals a condition of full information that is present in many labora-tory experiments. When citizens are provided with information about gov-ernment budget cuts, they are expected to give more than when they do not have such information.

Substitution between organizationsNonprofit organizations do not operate isolated from each other. The strong competition in the fundraising market might raise the expectation that do-nations to one organization go to the expense of the other. This is likely espe-cially if citizens have a mental account for charitable giving, which separates the decision to give from other financial decisions (Thaler, 1999).

There is considerable empirical support for substitution between char-itable organizations both with longitudinal survey data (Reinstein, 2006) and in laboratory experiments (Ek, 2017; Reinstein, 2007). When the price of giving to one organization decreases, for example through a matching scheme, the increase in donations goes to the expense of giving to other or-ganizations (Reinstein, 2007). Especially when charitable organizations are similar to each other and/or serve the same purpose, donations to these or-ganizations are likely to be substitutes (Ek, 2017; Reinstein, 2007).

New information or shocks in social needs may draw new donors into giving. The campaign after the 2004 Tsunami, for example, attracted many


donors who previously did not donate to international aid (Bekkers et al., 2017).

In the current research design, respondents have the possibility to give away a reward to four existing nonprofit organizations. They receive infor-mation about only one of these organizations, which is expected to increase giving. This closely resembles experimental designs that find substitution between organizations in the U.S. (Reinstein, 2007) and Sweden (Ek, 2017). Thus, the expectation here is that increasing donations to one organization goes to the expense of donations to the other organizations.

Moderating variablesCitizens differ in their demands for different types of government informa-tion (Piotrowski & Van Ryzin, 2007), which makes it likely that behavioral responsiveness to public information varies between social groups, too. Al-though previous studies have examined how crowding-out effects differ by income, gender, social class and prosocial values (Chan et al., 1996; De Wit et al., 2017; Güth et al., 2006; Kingma, 1989; Luccasen, 2012), there are no strong theoretical grounds from which we would expect different groups of citizens to react different to changing levels of public funding, so we examine individual heterogeneity in an explorative way. We examine four moderating factors.

First, we examine the information effect among a group of individuals from relatively wealthy households. Because of the sizeable value of their donations, High Net Worth (HNW) donors have received increasing atten-tion (e.g. Bekkers, 2013b; Center on Philanthropy, 2011; Rooney et al., 2014; Schervish & Havens, 1995) and it is interesting to see whether reactions to changing government funding are different for this socio-economic group. The marginal value of a Euro is smaller for citizens with larger wealth, so it is less costly for them to give money away. Because a change in donations by a relatively small group of wealthy donors can have an important influence on total amounts donated, it is important to study how wealthy donors re-spond to information about government funding to nonprofit organizations. The design of our survey poses a rare opportunity to compare responses in a group of relatively wealthy respondents with responses in a representative sample.

Secondly, we examine the difference between citizens who previously do-nated to a nonprofit organization and citizens who did not. Previous donors are more committed to the goals of the organization and value the need ad-


dressed by the organization as more important than non-donors. The crowd-ing-out hypothesis would predict that donors who care more strongly about the needs addressed by a nonprofit organization are more responsive to changes in funding by third parties, including government. Thus, providing information about government funding could have a stronger effect on pre-vious donors.

A third possible moderator is empathic concern, the “other-oriented emo-tion elicited by and congruent with the perceived welfare of someone in need” (Batson, 2010). Citizens who are more empathic are more touched by what recipients go through. Bekkers (2008) shows that giving to the Dutch Heart Association is higher among people who know someone with a cardio-vascular disease, who are the people that are more exposed to the needs of possible recipients. He shows that the association between knowing a sick person and charitable giving is stronger when having a high empathic con-cern, which suggests that empathic citizens change their preferences more strongly when exposed to a need.

The fourth and final possible moderator is a moral principle to care about others. While empathic concern is a psychological reaction to others in need, the principle of care refers to the moral standard that helping is the right thing to do. Bekkers and Wilhelm (2016) and Wilhelm and Bekkers (2010) show that the principle of care is a strong predictor of different helping be-haviors and that the principle of care mediates the relationship between em-pathic concern and helping. If the principle of care is the motivation to give, it is likely that information about decreasing public funding has a strong im-pact. Budget cuts will affect all recipients, irrespective of their relationship with possible donors, increasing a general awareness of need.

RESEARCH DESIGN

ContextThe nonprofit organization under study is the KWF Kankerbestrijding (the Dutch Cancer Society). KWF Kankerbestrijding funds medical research relat-ed to cancer, patient care and prevention programs and is the largest fund-raising organizations in the Netherlands with a private income of 137 mil-lion Euros in the year 2012 (Central Bureau on Fundraising [CBF], 2014). The organization receives no structural government funding. Rather than deliberate policy shifts, funding changes are the result of incidental project


subsidies. Incidental government funding forms a very small share of KWF’s total revenues and although the financial information is publicly available on the internet, it is likely that they are unknown to the larger audience.

The main channel through which citizens could have heard about govern-ment subsidies to KWF is news media. To examine the available information, we carried out an analysis on seven large national subscribed newspapers in the 2012-2014 period1. This analysis shows no mention of actual govern-ment funding to the organization, which confirms our assumption that citi-zens are not likely to know about the existence of these government subsi-dies.

In 2013, the organization appeared in the news with a large controversy about the invoices of the founder of Alpe d’HuZes, a popular sponsor ride to collect money for cancer research. The organization operated independent-ly but was largely funded by KWF Kankerbestrijding to make the sponsor ride happen. Articles about the controversy started to appear from the sum-mer of 2013, a year before respondents took our survey experiment, until December 2013. This might have affected respondent’s perception of the or-ganization, although KWF’s overall fundraising income has not structurally suffered from the incident (CBF, 2014).

DataIn an experimental design we examine the effect of providing information about actual government subsidies to KWF Kankerbestrijding.

Data were collected in May and June 2014 as a part of the Giving in the Netherlands Panel Survey (GINPS), a nationally representative survey on giving and volunteering among Dutch households (Bekkers et al., 2016). The 2014 wave of the GINPS included an experiment that allows for examining the effect of information about government funding on donations. The sam-ple consists of a group (n=1,271) that is representative for the population in terms of socio-demographic characteristics, and a group (n=1,187) of High Net Worth (HNW) individuals who are disproportionally wealthy2. Because a

1 We conducted a search query in the LexisNexis database on seven large national subscribed news-papers from 2012 till 2014, collecting articles with both the name of the organization (KWF Kanker-bestrijding) and the Dutch words “subsidie” or “overheidssubsidie” in the title and/or text. This query resulted in a total of 31 newspaper articles in 2012 (5), 2013 (21) and 2014 (5). A few articles consider (the development of) different income sources of charitable organizations in general, mentioning KWF Kankerbestrijding as an example of a large fundraising organization, but there is no specific information about actual government funding to KWF, nor about changes in such funding.2 Average household wealth is 271,693 Euros in the HNW sample versus 72,273 Euros in the represen-tative sample, excluding the value of one’s primary residence and Winsorized at 99%.


part of the HNW sample filled out a shorter version of the questionnaire, not all moderating variables are measured in the full sample.

Participants in the survey received a reward that, depending of the num-ber of questions they had to fill in, had a value of about 3.5 Euros. The reward came in points which respondents could divide between five vouchers for personal use, Air Miles and donations to four major charitable organizations. A similar version of this dictator game is described in Bekkers (2007). The most popular charitable organization is KWF Kankerbestrijding. The oth-er organizations were Aidsfonds (HIV/AIDS Foundation), Rode Kruis (Red Cross) and Nederlandse Hartstichting (Dutch Heart Foundation), which we grouped in the analyses as “other organizations”. The different possibilities to keep the reward as a voucher or Air Miles were grouped as “kept reward”.

First experimental treatment: Real decisionWhen respondents arrived at the end of the survey, they could see how many points they had earned with filling out the questionnaire. They were offered the possibility to divide the reward between vouchers, Air Miles and chari-ties. The awareness of need among all respondents was evoked by the sen-tence “The charities could use your support”. While the control group made their decision right after that sentence, the treatment group additionally re-ceived information on the amount of government funding that KWF lost. The complete text they were shown was: “The charities could use your support. KWF Kankerbestrijding, for example, received € 361,000 government sub-sidies in 2011, but received no subsidies at all from Dutch government in 2012.” These are actual numbers, adopted from annual reports as collected and published on the website of the Dutch Central Bureau on Fundraising (CBF, 2014). The control group received no information about government funding.

Manipulation check: Perceived change in fundingAfter the donation decision, the perceived change in funding was measured with the question “What do you think, did KWF Kankerbestrijding receive more, an equal amount of, or less government subsidies in 2012 compared with 2011?”. This question was the same for all respondents. For respon-dents in the treatment group, who received information, this serves as a ma-nipulation check.


Second experimental treatment: Scenario decisionAfter the real donation decision and the knowledge question, respondents were exposed to an extra scenario experiment. Respondents were given a second, hypothetical choice what they would have done with their reward if they would have had different information. Respondents in the treatment group either received a scenario in which funding increased or a scenario in which funding did not change. The control group received one of three sce-narios in which funding either decreased, increased or did not change.

The question was: “Imagine you would have heard that government subsi-dies to KWF Kankerbestrijding [decreased/did not change/increased], what would you have done with your reward?”. Respondents could divide their reward in exactly the same way as the actual donation decision, this time without consequences.

We should be careful with generalizing these results to real-life situations, since previous research has shown that people are more generous with hy-pothetical than with real money (Bekkers, 2005a).

Other moderating variablesThe other moderating variables were adopted from questions asked earlier in the questionnaire.

To measure whether respondents are previous donors, they were asked “To which of the following charitable causes did your household donate in 2013?”, followed by a list of 30 large fundraising organizations in alphabeti-cal order, one of which is KWF Kankerbestrijding. A part of the HNW sample filled in a shorter questionnaire in which this question was not asked, so those respondents were excluded from the analysis on this moderating ef-fect.

As in previous research (Bekkers & Wilhelm, 2016), empathic concern is measured by four statements, with answer categories on a 5 points Likert scale ranging from “Totally disagree” to “Totally agree”:

- “I often feel concern for people who are less fortunate materially than me”- “Other people’s problems do not usually bother me”- “Other people’s misfortune does not usually bother me”- “I am often touched by what other people go through”The four items have a high Cronbach’s Alpha (0.791) and are recoded in

a 1 to 5 scale. A dummy variable is created for people scoring higher than 3, indicating a high empathic concern.


Also following Bekkers and Wilhelm (2016), the principle of care is mea-sured by four statements, with answer categories again on a 1 to 5 scale:

- “People should be prepared to help others who are less fortunate mate-rially than themselves”- “Everyone has the responsibility to help others when they need it”- “It is important to help people who are less off, also when they are very different from us”- “Personally helping people who have problems is very important for me”Again, a scale is formed with an high internal reliability (Alpha = 0.866)

and a dummy measures whether respondents score higher than 3 to mea-sure a high principle of care.

Table 1 displays descriptive statistics.

Table 1: Descriptive statistics

N Min Max Mean SDDonated to KWF (real decision) 2,458 0 1 0.101 0.301Amount donated to KWF (real decision) 248 0.15 4.65 2.254 1.107Donated to KWF (scenario decision) 2,458 0 1 0.107 0.310Amount donated to KWF (scenario decision) 264 0.15 4.65 2.205 1.083Thinks funding has increased 2,458 0 1 0.040 0.197Thinks funding has remained the same 2,458 0 1 0.384 0.486Thinks funding has decreased 2,458 0 1 0.576 0.494Reward 2,458 0.60 5.10 3.228 0.629Wealthy individual 2,458 0 1 0.483 0.500Donating in 2013 1,753 0 1 0.67 0.471Amount donated in 2013 754 0 2000 22.53 85.356Empathic concern scale 2,458 1 5 3.624 0.719High empathic concern 2,458 0 1 0.754 0.431Principle of care scale 2,458 1 5 3.529 0.706High principle of care 2,458 0 1 0.715 0.452


RESULTS

Manipulation checkIn our first experimental treatment, respondents receive information about an actual decrease in government funding. Table 2 shows the perceived change in government funding in the control and treatment group. Among those who received no information, a majority (51.8%) believes that KWF has lost funding, while only 5.1% thinks that there has been an increase.

Providing information about the actual change should increase the per-centage who give the right answer. The second row of Table 2 shows that it does. In the information group, 63.7% says that KWF lost funding, while 2.9% says that funding increased. Providing information seems to work, al-though there are still 7 out of 20 respondents who give the wrong answer. An analysis on background characteristics (not shown) reveals that respon-dents who are younger and higher educated are more likely to give the right answer after being exposed to information.

Table 2: Percentage of respondents who think that funding increased, did not change or decreased

Thinks funding increased

Thinks funding did not change

Thinks funding decreased

No information 5.08 43.13 51.79Information 2.92 33.36 63.72

First experimental treatmentThe theoretical expectation is that providing information increases the num-ber of donors as well as the mean amount given. Table 3 shows the percent-age of people who donated (a part of) their reward to KWF, donated the full reward to KWF, donated the full reward to other organizations, kept the full reward, split the reward between KWF and another organization, split the reward between KWF and oneself, and split between KWF, another organi-zation and oneself. The last three columns display the conditional average amounts that were donated to KWF, donated to other organizations, or kept as vouchers or Air Miles.


Tabl

e 3:

Per

cent

age

resp

onde

nts

that

don

ated

or k

ept t

he re

war

d, a

nd c

ondi

tiona

l mea

n am

ount

don

ated

(bet

wee

n-su

bjec

ts)

%€

(con

ditio

nal)

Thin

ks th

at

fund

ing…

Dona

ted

to

KWF

Dona

ted

100%

to

KWF

Dona

ted

100%

to

othe

r org

Kept

10

0%

Split

be

twee

n KW

F an

d ot

her o

rg

Split

be

twee

n KW

F an

d se

lf

Split

be-

twee

n KW

F, ot

her o

rg

and

self

Amou

nt

dona

ted

to

KWF

Amou

nt

dona

ted

to

othe

r org

Amou

nt

kept

No

info

rmat

ion

...inc

reas

ed/

did

not c

hang

e7.

584.

123.

7987

.97

1.15

1.81

0.49

2.19

2.44

3.21

...dec

reas

ed10

.28

6.29

3.68

84.9

70.

922.

450.

612.

442.

643.

16To

tal

8.98

5.24

3.73

86.4

21.

032.

140.

562.

342.

553.

18

Info

rmat

ion

...inc

reas

ed/

did

not c

hang

e5.

984.

372.

0791

.72

0.00

1.15

0.46

2.50

2.64

3.16

...dec

reas

ed14

.27

7.07

2.62

82.2

01.

573.

272.

362.

112.

143.

15To

tal

11.2

66.

092.

4285

.65

1.00

2.50

1.67

2.19

2.23

3.15

Tabl

e 4:

Per

cent

age

resp

onde

nts

that

don

ated

or k

ept t

he re

war

d, a

nd c

ondi

tiona

l mea

n am

ount

don

ated

(with

in-s

ubje

cts)

%€

(con

ditio

nal)

Dona

ted

to

KWF

Dona

ted

100%

to

KWF

Dona

ted

100%

to

othe

r org

Kept

10

0%

Split

be

twee

n KW

F an

d ot

her o

rg

Split

be

twee

n KW

F an

d se

lf

Split

be-

twee

n KW

F, ot

her o

rg

and

self

Amou

nt

dona

ted

to

KWF

Amou

nt

dona

ted

to

othe

r org

Amou

nt

kept

Real

No

info

rmat

ion

8.43

5.10

3.33

87.5

81.

111.

550.

672.

482.

493.

18Sc

enar

ioFu

ndin

g de

crea

sed

11.5

36.

872.

6685

.14

1.77

2.44

0.44

2.52

2.23

3.12

Real

Info

rmat

ion

11.2

66.

092.

4285

.65

1.00

2.50

1.67

2.10

2.23

3.15

Scen

ario

Fund

ing

incr

ease

d/di

d no

t cha

nge

10.4

35.

422.

9286

.16

1.25

2.25

1.50

2.05

1.98

3.14


In line with the crowding-out hypothesis, providing information increas-es the total share of respondents who donated the reward fully or partly to KWF from 9% to 11.3% (X2(1, 2,458) = 3.53, p=0.06). The amount donated to KWF among donors is not significantly different in the treatment group (2.2 Euros) compared with the control group (2.3 Euros). The net effect of providing information is an increase of 17% in the total amount donated.

The percentage of respondents who donated the full reward to KWF in-creases from 5.2% to 6.1% when providing information (X2(1, 2,458) = 0.82, p=0.36) and the share of respondents who keep the full reward decreases from 86.4% to 85.7% (X2(1, 2,458) = 0.30, p=0.56), but those differences are not statistically significant. The largest differences occur in the donations to other organizations. In the information condition, less respondents (2.4%) donate the full reward to one of the three other nonprofit organizations com-pared with the no information condition (3.7%). The difference is significant at 10% (X2(1, 2,458) = 3.54, p=0.06). There are significantly more respon-dents (1.7% versus 0.6%) who divide their reward between KWF, another organization and themselves (X2(1, 2,458) = 6.99, p=0.01). Also, the average amount donated to other organizations is somewhat lower here (F(1, 144) = 3.24, p=0.07). This suggests substitution between organizations, with re-spondents giving a part of their reward to KWF when they are provided with information instead of donating the full reward to other organizations.

Next, we take the manipulation check into account. We compare the control group with those in the treatment group who think that funding decreased, which is the correct answer that respondents could have known after having read the information. Among these respondents, the share of donors to KWF is even higher (14.3%), providing stronger support for the crowding-out hy-pothesis under the condition of full information. Furthermore, the share of respondents who keep the full reward is substantially lower here (82.2%). This changes the picture of substitution. Rather than substitution between organizations, it suggests that the largest part of the increase in KWF do-nors can be attributed to the decrease in respondents who would not have donated without the information. The difference between the control group and this part of the treatment group is statistically significant for donating at least something to KWF (X2(1, 2,023) = 13.63, p=0.00), donating 100% to KWF (X2(1, 2,023) = 2.84, p=0.09), keeping 100% of the reward (X2(1, 2,023) = 6.57, p=0.01) and dividing the reward between KWF, another organiza-tion and oneself (X2(1, 2,023) = 12.62, p=0.00). These results suggest that information about government funding, when it is effectively communicated,


might draw non-donors into donating.3

Second experimental treatmentAfter the actual donation decision and the knowledge question, respondents were asked what they would have done with their reward in case they had heard about decreasing, increasing or equal funding. Comparing decisions in the actual experiment with those in the scenario experiment allows us to do an extra within-subjects test of the effect of information about government funding.

Table 4 shows the differences between the real and the hypothetical deci-sion for different experimental groups.

First, we examine respondents in the control group who did not receive information and then were asked what they had decided when they would have heard that government funding to KWF Kankerbestrijding decreased. The results largely confirm the between-subjects analyses. When informed about budget cuts, some respondents would decide to start donating a part of their reward to KWF, which is significant in a paired samples t-test (t(450) = 3.80, p=0.00). 3.1% indicate they would start donating if they would have heard that subsidies to this organization decreased. This is slightly more than the 2 percentage points difference between the treatment group and control group in the between-subjects design. Furthermore, some respon-dents change their decision to donating the full reward to KWF (t(450) = 2.85, p=0.01), no longer donating the full reward to another organization (t(450) = -1.74, p=0.08), no longer keeping everything for themselves (t(450) = -3.35, p=0.00), starting to split their reward between KWF and another organization (t(450) = 1.74, p=0.08) or starting to split between KWF and themselves (t(450) = 2.01, p=0.05). Again this points to both substitution between organizations and attracting new donors.

Second, we examine the crowding-out effect the other way around. What if respondents who know that government funding decreased would have heard that it increased or did not change? The percentage of KWF donors goes from 11.3% in the information treatment to 10.4% in the hypotheti-cal decision (t(450) = -1.72, p=0.09). The percentage of respondents who change their minds toward splitting the reward between KWF and another organization increases from 1% to 1.25% (t(450) = 1.73, p=0.08). The other

3 The numbers in the row of respondents in the treatment group who think that subsidies increased or did not change are very interesting. In this group, more people kept their reward for themselves and less people donated their reward fully or partly to KWF compared with the control group. It is possible that those respondents read, but misinterpreted the information.


differences are in the expected direction, but not statistically significant. The evidence for crowding-out is weaker here.

In both groups, respondents who change their decision are a minority and over 95% would not be affected by information about government funding.

Moderating variablesTo explore individual heterogeneity, Figure 2 shows the percentages of re-spondents who (partly) donated their reward among different groups in the first experimental treatment.

Panel 1 of Figure 2 shows the information effect in the regular sample and in the sample among wealthy individuals. The information effect is stronger and statistically significant among wealthy individuals (X2(1, 1,187) = 3.67, p=0.06) but is not statistically significant in the regular sample.

Panel 2 shows the information effect among respondents who donated to KWF in the year preceding the survey and those who did not. Providing information about government funding has a stronger effect among non-do-nors. In this group, the information effect is statistically significant (X2(1, 581) = 4.66, p=0.03).

People with low empathic concern are more sensitive to information. Pan-el 3 shows that the information effect is stronger among respondents with low empathic concern, which is contrary to the expectation. Among those who score relatively low on empathic concern, the effect of information is statistically significant (X2(1, 606) = 9.62, p=0.00).

The fourth panel in Figure 2 shows the interaction between information and the principle of care. The information effect is somewhat stronger for re-spondents with a higher moral principle to care about others. In this group, the effect is significant at the 10% level (X2(1, 1,757) = 3.09, p=0.09).

Regression analysisIn order to check results together in a regression model, we obtained a data-set in which the real and scenario decisions are pooled together so that every respondent appears twice in the data. Table 5 shows the odds ratios from a logistic regression model on the probability to donate, indicating the ratio between the odds to donate and the odds to donate nothing. A coefficient higher than 1 means that the variable is associated with a higher probability to be a donor, while an odds ratio lower than 1 indicates a lower probability.

Providing information is significantly associated with a higher likelihood to donate (Model I). Controlled for the level of the reward and being in a


0%

5%

10%

15%

No information Information

Regular Wealthy

0%

5%

10%

15%


Non-donors Donors

0%

5%

10%

15%


Low EC High EC

0%

5%

10%

15%


Low POC High POC

Figure 1: Percentage respondents that donated to KWF Kankerbestrijding among peo-ple who received information and people who did not, interacted with (1) being from the sample of disproportionally wealthy households, (2) being a regular donor, (3) empathic concern and (4) the principle of care.

real or a scenario experiment, providing information increases the odds to donate with 20.9%, which confirms the increase from 9% to 11.3% donors in Table 3 and is in line with the crowding-out hypothesis.

The information effect is stronger for respondents who think that funding decreased (manipulation check), which is not statistically significant (Model II). The information effect is not different for real and hypothetical decisions (Model III).

The regression analysis confirms that respondents with lower empathic concern are more sensitive to information than those with higher empathic concern (Model IV), but the interaction term is not significant when all in-teractions are included (Model VI). Note that the sample size is smaller here because the variable for being a donor in 2013 is not measured in the full sample. The other interaction terms are not statistically significant.

The odds ratio of the size of the reward that respondents received, which


Table 5: Odds ratios of the probability to donate

I II III IV V VIInformation: Decrease 1.209* 0.977 1.229 1.605* 1.570* 2.108*

(0.125) (0.172) (0.166) (0.439) (0.367) (0.891)Thinks funding decreased 1.435*** 1.579*** 1.573*** 1.602*** 1.631***

(0.175) (0.158) (0.157) (0.199) (0.205)Information * Thinks funding decreased 1.364

(0.289)Information * Scenario 0.942 1.294

(0.202) (0.337)Wealthy household 1.365** 1.074

(0.174) (0.178)High empathic concern 2.112*** 3.322***

(0.369) (0.924)High principle of care 0.925 0.831

(0.137) (0.158)Donating in 2013 1.515** 1.407*

(0.267) (0.249)Wealthy * Information 0.926 1.074

(0.182) (0.285)EC * Information 0.607* 0.535

(0.167) (0.222)POC * Information 1.229 1.127

(0.302) (0.346)Donating * Information 0.751 0.781

(0.203) (0.214)Reward 0.872* 0.862** 0.956 0.893 1.202 1.086

(0.064) (0.064) (0.077) (0.073) (0.185) (0.171)Scenario decision 1.138 1.156 1.161 1.136 1.183 1.076

(0.113) (0.116) (0.143) (0.114) (0.148) (0.166)(Constant) 0.171*** 0.131*** 0.076*** 0.055*** 0.024*** 0.016***

(0.041) (0.034) (0.023) (0.018) (0.013) (0.009)N 4,916 4,916 4,916 4,916 3,506 3,506

Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1


is dependent on the time it took them to fill out the survey, is below 1. This suggests that respondents who decide over a larger endowment are less generous.


Charitable giving can fundamentally change the output of public policy. Giv-en that the availability of resources is an important condition for successful governance collaborations between public, nonprofit and for-profit organi-zations (Ansell & Gash, 2008; Milward & Provan, 2003; Pfeffer & Salancik, 1978), it is important to know how charitable donations, as a source of in-come for many nonprofit organizations, are shaped by public policies.

Many papers in the literature on the relationship between government support and charitable giving ignore the fact that information is often not available. This paper took information uncertainty as a starting point, giv-en the lack of accuracy of popular knowledge about public policy towards the nonprofit sector (Horne et al., 2005; Handy et al., 2010; Lergetporer et al., 2016; Shah et al., 2015). Charitable giving was examined in a context in which citizens were unlikely to know the actual change in government fund-ing. This provides a good opportunity to test the effect of changing knowl-edge.

Providing information about decreasing government funding increases the number of donors. This confirms the findings from a similar scenario experiment in the United Kingdom (Shah et al., 2015). In our study, this in-formation increased the share of donors with about 20%, which can make a considerable difference for nonprofit organizations who are dependent on fundraising income. This effect is statistically significant among citizens who are very wealthy, who do not regularly donate to the organization under study, who have a relatively low level of empathic concern and who have a relatively high principle of care.

An important question for the nonprofit sector is whether incentives leads donors from one organization to the other, or that they increase the total size of the fundraising market (Ek 2017; Reinstein 2006, 2007). Our results show support for both effects. There is substitution between organizations but in-formation about government funding also has the ability to draw non-do-nors into donating. This is a very important finding and shows the potential of information. Attracting new donors, even when they give low amounts ini-


tially, may contribute to building a sustainable donor base in the long term.Although the analyses showed significant effects on the percentages of cit-

izens who donate, it should be noted that the vast majority does not change its giving behavior in response to information. This is in line with surveys in which more than 7 out of 10 people say they would not change their do-nations as a reaction on changing government funding (Bekkers & De Wit, 2014; Horne et al., 2005; Horne, Van Slyke, & Johnson, 2006; Shah et al., 2015).

Interestingly, information about government funding affects the number of donors, but does not substantially change the amount donors give. This re-sult goes against experimental findings in the crowding-out literature, which on average shows that a $1 increase in government support is associated with a $0.64 decrease in charitable giving (De Wit & Bekkers, 2017). How-ever, there are three reasons to be cautious in drawing strong conclusions from the current findings. First, the rewards that respondents could donate in this experiment were low, so there was not much room to increase or de-crease the amounts of giving. Secondly, respondents were not made aware of the fact that government support is funded by their own tax money, which is common in crowding-out experiments and is known to have a strong ef-fect on charitable giving (Eckel et al., 2005). Thirdly, the experimental design only enabled a test of the effect of a fixed amount of government funding. It could be that information on higher amounts would in fact lead citizens to change the amounts they give.

This paper showed that information about government policies can have consequences for non-political civic engagement. Three lines of future re-search are promising in this area. First, it is interesting to not only look at exposure to information, but also investigate the content of this information. Different types of information, for example on decision-making processes and policy effects instead of simple policy content, are expected to have different effects on citizen attitudes and behavior (Heald, 2006). Previous studies have touched upon effects of framing (Eckel et al., 2005) and news content (De Wit et al., 2017) on donations, but more research will give bet-ter insights in the effects of different messages. The information provided to respondents in the current research design was very specific, and more research is needed to test information effects in different contexts. Second-ly, future studies should investigate effects of information about changes in funding versus information about levels of funding. Whereas the current study examined effects of information about policy changes, many other


(experimental) studies test effects of levels of government funding. Thirdly, more research is needed on individual heterogeneity in responses to gov-ernment policies. In our analyses, we found a significant interaction effect for empathic concern. Low empathic citizens, who are less likely to donate to begin with, are more responsive to new information. Those are the ones who do not have a strong intrinsic motivation. Apparently, external incentives can encourage them to increase their giving towards levels comparable to those with high empathic concern. No strong theoretical expectations existed here, and this finding might contribute to further theory building. It is possible, for example, that empathy is related to the “warm glow” of giving, which makes citizens less responsive to changes in government contributions to the pub-lic good (Andreoni, 1989, 1990).

The benefits of government transparency on citizen participation are widely studied yet still contested. This paper contributes to the literature on government information by hypothesizing and testing effects on non-po-litical participation. Given the important roles of nonprofit organizations in governance processes, changes in civic engagement can have large conse-quences for public policy. Previous research has shown that the right fram-ing can increase popular support for retrenchments (Elmelund-Præstekær & Emmenegger, 2013; Rodriguez, Laugesen, & Watts, 2010). Our study goes a step further by examining whether information about budget cuts can lead to different civic behavior. If citizens are aware of budget cuts, they may com-pensate for them with their charitable donations. Professionals in the non-profit sector might use information about public funding as a tool not only to raise support, but also to raise money.

Chapters 3 and 4 explored resources and values as possible moderators in the relationship between government support and charitable giving, sug-gesting that responses to changes in government support vary across so-cial groups. The current chapter examines an even wider range of possible correlates of the willingness to compensate for government budget cuts. A large sample of the Dutch population is asked how they would respond to a reduction of government funding to organizations in different parts of the nonprofit sector, and their actual change in donations is examined two years later. Citizens who have a higher education, who have stronger prosocial values, who have more confidence in charitable organizations, who have lower trust in government and who receive more donation so-licitations are more likely to increase donations after government budget cuts. The findings provide useful tools for fundraising in the context of decreasing revenues.

Look who’s crowding-out!

Chapter 5

René Bekkers and Arjen de Wit

RB designed the study; RB and AdW carried out data analysis; RB and AdW contrib-uted to writing the article.

A previous version of this chapter was presented at the 42nd Annual Meeting of the Association for Research on Nonprofit Organizations and Voluntary Action in Hart-ford, CT (USA), November 2013.

Data, syntax and supplementary materials are available through the Open Science Framework at http://osf.io/zqay5/.

136 Chapter 5 Look who’s crowding-out!

INTRODUCTION

Who is willing to donate more when the government is lowering financial support? There is a wide array of studies dedicated to the crowding-out hy-pothesis, stating that higher (lower) government contributions lead to a de-crease (increase) in donations from private donors. In the academic debate the relationship between government contributions and private donations is studied from different angles and with different research designs, not yield-ing conclusive results (reviews by De Wit & Bekkers, 2017; Lu, 2016). The inconclusive findings ask for more empirical research on factors that might have a moderating effect on the relationship between government support and charitable donations. Here, we focus on the question among whom crowding-out occurs.

Previous studies often do not examine individual heterogeneity in re-sponses to varying levels of government funding. Only a handful of studies explored individual differences in psychological traits, socioeconomic status and ethnicity (Luccasen, 2012), resources and prosocial values (De Wit et al., 2017), income groups (Kingma, 1989) or different donor groups (Reeson & Tisdell, 2008), yielding no conclusive findings. The lack of tests of individu-al heterogeneity with large study samples is an important lacuna in crowd-ing-out research, while it has important consequences for fundraising and nonprofit research.

This study is the first to offer a theoretical framework that might explain individual heterogeneity in responses to changing government funding. Based on the Civic Voluntarism Model (Verba et al., 2005), we propose that citizens who have more resources, who are more engaged and who are more sensitive for recruitment are most likely to substitute government budget cuts. With a large sample of survey data from the Dutch population we are able to examine individual heterogeneity in willingness to increase dona-tions in the case of government budget cuts in different subsectors. First, we present respondents with scenario questions about hypothetical budget cuts in different nonprofit subsectors, and examine correlates of the responses to those scenarios. Second, we examine the correlates of actual changes in do-nations over the course of two years.

The conclusions of this study have important implications for profession-als in the nonprofit sector. It is uncertain to what extent nonprofit organi-zations succeed in receiving more revenues from private donations when income from government support decrease. With more knowledge about

137Look who’s crowding-out! Chapter 5

the profiles of donors who are willing to adapt their giving as a response to changing government support in different parts of the nonprofit sector, fundraising might be better targeted at specific social groups.

THEORY AND HYPOTHESES

Civic voluntarismWe base our answer to the question which characteristics of donors deter-mine their responses to changes in government support for nonprofit or-ganizations on the Civic Voluntarism Model (Verba et al., 1995). The Civic Voluntarism Model explains civic voluntarism – citizens’ voluntary contribu-tions of money and time to public and club goods – with three requirements: resources, engagement and recruitment. Here we use the model to explain changes in levels of donations to nonprofit organizations in the Netherlands facing cuts in government support. Further support for our hypotheses comes from a review of the literature on philanthropy, which identified eight mechanisms as the major drivers of philanthropy (Bekkers & Wiepking, 2011b). The civic voluntarism model includes several of these mechanisms.

Mechanisms in resources, engagement, and recruitmentResources influence giving through the mechanism of costs. Wealth and in-come are well-known correlates of charitable giving, and they might also in-crease the capability to change giving. For citizens with a higher financial capacity, an increase in their donations of a certain amount represents a smaller proportion of their expenditure budget. Also a higher financial ca-pacity lowers the costs of donating for citizens in progressive income tax systems that allow deduction of charitable donations. There has not been many crowding-out studies that examined resources as a moderating factor in the relationship between government support and charitable donations. Kingma (1989) shows partial crowding-out in all income groups, with the strongest association among lower middle incomes. De Wit et al. (2017) do not find significantly different correlates among citizens with a higher edu-cation, a paid job or an own home.

Resources in the Civic Voluntarism Model also include the availability of time and civic skills. While the available time is less likely to explain finan-cial donations, civic skills might be important in explaining changes in do-nations. Civic skills are developed through education, which is argued to be


related to charitable giving partly because it expands one’s information set and because it has a socializing effect, which make children develop a more prosocial world view (Brown, 2005; Wiepking & Maas, 2009).

Engagement with the cause incorporates the mechanisms altruism, awareness of need, psychological benefits, values and efficacy. In the clas-sical crowding-out literature, the mechanism of altruism is the foundation for the hypothesis that citizens will crowd out government support for non-profit organizations. It is assumed that citizens will reduce their donations when they learn that others are contributing to a cause because the contri-butions of others are already enabling the organization to reach their objec-tives (Roberts, 1984; Warr, 1982). A reduction of government support for a specific organization will lead donors motivated by altruism to increase their donation levels. In this context, the mechanism of altruism is closely related to the awareness of need. If social needs increase because budget cuts reduce total contributions to the public good, citizens are more likely to increase their donations.

In economic theory, citizens will not easily change their giving behavior in response to changes in government support to the extent that they derive private benefits from their donation (Andreoni, 1989, 1990). The assump-tion in this prediction is that changes in government support do not affect the psychological or social rewards obtained by giving. However, non-altru-istic motives such as a concern for one’s self-image could also lead citizens to increase their donation levels when they learn about a reduction in gov-ernment support for a cause they support. Citizens who raise their giving as a response to budget cuts derive increasing psychological benefits from their donation, because giving is more desirable in this situation. Psychological benefits thus motivate increased donations: the joy of giving that donors ex-perienced upon an initial donation will increase with further contributions.

In a similar way, citizens who give as an expression of their values might increase donations after government budget cuts. Examples are altruistic values (Bekkers & Schuyt, 2008), social trust (Brown & Ferris, 2007) and the moral principle of care (Bekkers & Wilhelm, 2016). Citizens who are endorse such values are more strongly engaged with nonprofit organizations, and are likely to increase their giving as a response to budget cuts.

The perceived efficacy of organizations is another mechanism that can be shared under the heading of engagement. Political efficacy is one of the indicators that Verba et al. (1995) use to measure political engagement. In the context of charitable giving, “[e]fficacy refers to the perception of donors


that their contribution makes a difference to the cause they are supporting” (Bekkers & Wiepking, 2011b: 942). The more citizens believe their donation can make a difference, the more effect a change in government funding to the same cause will have. In this context, confidence in the government might also enhance perceived efficacy. If citizens trust their government, they are likely to belief that changing government policies will have consequences for the provision of public goods, and more likely to adapt their giving behavior accordingly.

Recruitment of citizens refers directly to the mechanisms of solicitation and reputation. Citizens who receive more solicitations inviting contribu-tions to an organization tend to increase their donation levels. We also in-clude the reputation mechanism here, referring to the approval of giving in the social environment. The same reputational concerns that motivated giv-ing to a certain cause can also motivate increasing the level of donations to that cause after a reduction in government support, if giving to that cause is still socially rewarded.

Different mechanisms in different nonprofit subsectorsA well-known distinction is the one between instrumental and expressive nonprofit organizations (Gordon & Babchuk, 1959). Referring to the expres-sive dimension of the nonprofit sector, Frumkin (2002, p. 24) states: “For do-nors, volunteers, and particularly staff, the very act of attempting to address a need or fight for a cause can be a satisfying end in itself, regardless of the ultimate outcome”. Government expenditures can affect social needs, but are less likely to affect the engagement that donors express. Thus, government support less likely to displace charitable giving in parts of the nonprofit sec-tor where the expressive dimension prevails than it is in parts of the non-profit sector where the instrumental dimension is more important.

This is consistent with the argument that crowding-out is more likely in service subsectors like social assistance or health, where organizations are mostly instrumental, than it is in “expressive” subsectors like culture and in-ternational aid (Pennerstorfer & Neumayr, 2017; Sokolowski, 2013). This is only partly supported by a systematic literature review of non-experimental crowding-out findings, however, which shows that government expenditures and philanthropic donations are generally negatively related in the field of human services, while they are positively related in the fields of health and the arts (Lu, 2016).

The current research design provides an excellent opportunity to com-


pare (correlates of) responses to government support in service subsectors and expressive subsectors of the nonprofit sector.

DATA AND METHODS

We test our expectations using data from the 2012 wave (n=2,459) of the Giving in the Netherlands Panel Survey (Bekkers et al., 2016).

Scenario questionsThe 2012 wave of the Giving in the Netherlands Panel Survey (GINPS) in-cluded scenario questions designed to capture the responsiveness of citizens to changes in government support for nonprofit organizations. Respondents first read a general introduction, after which the online survey software se-lected a subsector in which nonprofit organizations in the Netherlands are actively raising funds: religion, health, international aid, environment, na-ture conservation, animal welfare, education and research, culture and arts, sports and recreation, and public benefits. Respondents were shown their response to a previous question about the amount their household donat-ed to nonprofit organizations in that particular subsector and were shown a potential cut in government funding of organizations in this sector, with randomized sizes: 5%, 10%, 20% or 33%. An example is: “With your house-hold you donated €100 to health in the past year. If the government cuts 5% in this area, how would you react?” Participants could respond “I will give the same as last year”, “I am willing to give more” or “I will also give less”. In the latter two cases, the participants answered the follow-up question “What will be the new amount?” Participants that did not report donations by their household in a particular sector received the question “With your household you did not donate to [sector] this year. If the government cuts 5% in this area, how would you react?” Participants could respond “I will give the same as last year” or “I am willing to give more” – giving less than nothing is im-possible.

Because religious organizations do not receive government funding in the Netherlands, the answers on the religion subsectors are excluded from the analyses. The total number of observations is 6,580 scenarios. It is important to note that the scenarios are not randomly allocated over respondents. Par-ticipants who had donated to nonprofit organizations in three or more non-profit subsectors (56% of all GINPS12 respondents) responded to scenarios


about three randomly selected sectors in which they had made donations. The remaining participants, who had donated to organizations in fewer than three subsectors, responded to scenarios about randomly selected sectors. This non-random distribution of subsectors over respondents complicates inferences for the total population in the Netherlands. In the way we treat the answers to the scenario questions in the current study, they are a regular survey module rather than an experiment.

Actual change in donationsBesides the willingness to increase donations after cutbacks as measured in the scenario questions, we also examine the actual change in donations by using the panel nature of the dataset. Respondents who participated both in the 2012 wave and the 2014 wave of the GINPS reported their household do-nations for 2011 and 2013 respectively. Although issues like social desirabil-ity and recall bias might play a role here, self-reported donations tend to be pretty accurate (Bekkers & Wiepking, 2011c). We adopt a dichotomous vari-able measuring whether household donations increased or not in service subsectors (health and public benefits) and expressive subsectors (interna-tional aid, environment, nature conservation, animal welfare and culture).

Other measuresFrom the responses to the question on the highest level of education achieved we created three dichotomous variables: primary education, secondary ed-ucation and tertiary education. Household income was measured with a se-ries of questions on the respondents’ income and the spouses’ income (if present) from different sources. Income quintiles were calculated. Income from wealth is included as a dummy variable, measuring whether the re-spondent’s household obtains income from equity, real estate, shares etc. A dummy variable measures whether the respondent is a home owner, which is another indicator of wealth. Current financial security is measured with the question “How financially secure do you feel on a scale from 1 (finan-cially insecure) to 10 (financially secure)?” Expected financial situation is the response the question: “What do you expect the financial situation of your household to be in the coming 12 months? (1) Will be much better; (2) Will be better; (3) About the same; (4) A bit worse; (5) Much worse”.

Perception of need and knowledge about need are two variables created from responses to four questions testing the participants’ knowledge about societal needs: (1) What proportion of the Dutch population lives in pover-


ty?; (2) What proportion of the world population lives in poverty?; (3) What proportion of the Dutch population suffered from depressive symptoms in the past year?; (4) Which disease is the most common cause of death among the Dutch? For each of the first three questions we recoded the proportions mentioned by the participants into three categories: an underestimation (-1), a correct answer (0) and an overestimation (1). The sum of these values is the perception of need variable, ranging from -3 to 3. Knowledge about need is the number of correct answers in response to all four questions.

The principle of care is the average score of responses to four statements expressing the moral principle that one should help people in need. The al-pha coefficient for the reliability of the four items is .881. Altruistic values were measured using statements like: “I prefer to work for my own welfare rather than for that of others” and “I strive to work for the welfare of society” (six items, Cronbach’s α = .727). Empathic concern is the average score of re-sponses to four statements expressing compassionate reactions in response to the needs of others (Cronbach’s α = .792). The joy of giving is the average score of responses to three statements expressing the intrinsic reward that people experience by giving (three items, Cronbach’s α = .738).

High charitable confidence and high confidence in government are dichot-omous variables obtained by recoding “quite a lot” (4) and “very much” (5) to the questions “How much confidence do you have in charities in the Neth-erlands?” and “How much confidence do you have in the government in the Netherlands?” respectively.

The number of areas in which donations were made is a variable counting the number of areas in which donations were made in the past calendar year (2011). Responses ranged from 0 (no donations made) to 10 (donations in all areas mentioned in the questionnaire). On average, the respondents re-ported donations in 3 different areas. Total amounts donated were obtained by summing the amounts donated in the past calendar year as reported af-ter the questions on donations in different areas. We created two separate variables distinguishing the amount donated to health and public benefits (service giving) and the amount donated to international aid, environment, nature conservation, animal welfare, and culture and arts (expressive giving). The amounts were log-transformed after adding 1 to avoid taking the log of zero.

Social pressure is the average response to four statements about the so-cial norm on philanthropic behavior in the respondent’s social network. The alpha coefficient for the reliability of the five items is .778. Solicitations are


measured by asking respondents whether they were asked for a charitable donation in the two weeks prior to taking the survey. Respondents were pre-sented a list of 15 possible solicitation methods, like “door-to-door collec-tion”, “direct mail” or “phone call”. A dummy variable is obtained indicating whether respondents received more than three solicitations.

Control variables include gender, age (measured with dummy variables for being born before World War II and being born after 1980), religious denom-ination (measured with dummy variables for Roman-Catholic, Protestant or another religious denomination) and whether or not respondents belonged to the selective subsample from the OCW study (Bekkers et al., 2016).

RESULTS

Vast majority is not responsive to changes in government supportA vast majority of GINPS respondents state that they would not change giv-ing after government cutbacks. Participants responded “I will give the same as last year” in 86% of the scenarios. In 8.6% of the scenarios participants responded “I will also give less”, and in 5.7% of the scenarios participants responded “I am willing to give more”.

Crowding-in is more likely than crowding-out in most subsectorsIn service subsectors (health and public benefits), 12% of the respondents say they will give less and 5% respond that they will give more. In expres-sive subsectors (international aid, environment, nature conservation, animal welfare, and arts and culture), crowding-out seems a bit more likely. In these sectors, 7% might give less while 6% is willing to give more after budget cuts.

Overall, the responses in the scenarios show that in most subsectors the participants responding they are willing to give more are outnumbered by participants responding they will give less in response to government cuts (see Figure 1). Only for animal welfare and arts and culture the proportion of citizens willing to give more is higher than the proportion saying they will donate less.

Interestingly, we find the largest proportions of respondents who say they will donate less after budget cuts in health and public benefits. These are the service subsectors in which we would have expected crowding-out to be more likely.

The net effect of cuts, however, depends not only on the prevalence of


0 2 4 6 8 10 12 14

Sports and recreationEducation and research

EnvironmentInternational aidArts and culture

HealthAverage

Public benefitsNature protection

Animal welfare

Willing to donate more Will donate less

Figure 1: Incidence of responses to government cutbacks in the Netherlands by sector (in percentages)

-3 -2 -1 0 1 2 3

Public benefits

International aid

Education and research

Environment

Sports and recreation

Average

Health

Animal welfare

Nature protection

Arts and culture

Figure 2: Net effects of government cutbacks on amounts donated in the Netherlands relative to previous donation levels by sector (in percentages)

willingness of citizens to give more and the tendency to reduce donations, but also on the changes in the amounts that respondents will give. Figure 2 shows that the net effect is positive for nonprofit organizations in the fields of arts and culture, nature conservation and animal welfare. Amounts donat-


ed to organizations in these sectors would increase by 2.2%, 2.1%, and 1.4%, respectively. For health and environmental organizations the effects are near zero. The other nonprofit organizations should expect losses in donations as a result of declining contributions by citizens.

Amounts donated to organizations in service subsectors would decrease by 0.8%, while donations to organizations in expressive sectors would in-crease by 0.3%. This is in line with the argument that crowding-out is more likely among service organizations, although both average effects are very small.

Donations to animal welfare organizations increase most stronglyWe test the difference between nonprofit subsectors in a more systematic way in a regression analysis (not shown). Here, the data of the different pos-sible scenarios on different nonprofit subsectors are stacked, and fixed ef-fects on individuals are included. The fixed effects model specification rules out potential effects of participant characteristics as a result of non-random allocation of scenarios to participants.

Respondents are more likely to say that they will give less (crowding-in) when evaluating scenarios about international aid than when evaluating sce-narios about other sectors. Formal tests are significant at 5% for contrasts with animal welfare, education and research, and sports and recreation. Par-ticipants were most likely to say they would give more (crowding-out) to ani-mal welfare, and least likely to say they would give more to the environment. Donations increase most strongly in animal welfare (+4.6%) and in culture and arts (+3%).

The willingness to increase donations is larger when budget cuts are larg-er, although the differences are not statistically significant. Furthermore, re-spondents were more likely to say that they were willing to increase dona-tions and less likely to say they were going to reduce donations in the second and especially the third scenario rather than the first.

The higher educated are willing to donate moreTables 1 and 2 display the results of multivariate regression analyses on the likelihood that respondents are willing to increase donations after budget cuts in service (health and public benefits) and expressive (international aid, environment, nature conservation, animal welfare and culture) subsectors, respectively. Because scenarios are nested within respondents, we add ran-dom intercepts for respondents.


The willingness to increase contributions after government cutbacks is much higher among those with higher education (college/university) com-pared with those with lower levels of education, which is more strongly so in expressive subsectors (Model 1).

The willingness to contribute more also increases with household income. Among households in the top quintile of the income distribution, the will-ingness to increase contributions after government cutbacks in at least one of the scenarios is 135% (Table 1) and 22% (Table 2) higher than in the lowest quintile. The differences are not statistically significant. In expressive subsectors, the middle incomes are less likely to increase donations than the lowest incomes. There seems to be a U-curve in the willingness to compen-sate for budget cuts, with the lowest and highest income quintiles being the most likely to increase donations.

In expressive subsectors, financial security is significantly associated with the willingness to increase donations. Contrary to the expectation, home ownership is negatively related to the willingness to increase donations.

Committed donors are more willing to increase donationsThe results also provides evidence supporting the hypothesis that engage-ment is positively correlated with the willingness to increase donations af-ter government cutbacks. Model 2 in Tables 1 and 2 stepwise add different measures of engagement, showing that there are different correlates for the willingness to increase donations in service and expressive subsectors, re-spectively.

In service sectors, the willingness to increase donations is weakly and not significantly associated with how severe people perceive problems like pov-erty and depression to be, and how accurate their knowledge about these problems is. The joy of giving – a trait measure of the “warm glow” people feel by giving to charity – is associated with responses in the scenarios on service subsectors. Respondents who experience more joy of giving are about 78% more likely to increase donations. Respondents with a high charitable confi-dence are more willing to increase donations, which is a large and significant difference (Table 1, Model 2c). Interestingly, trust in government correlates negatively. Service donations in the previous year are not associated with the willingness to compensate budget cuts.

In expressive subsectors, the awareness of need also correlates with the willingness to increase donations (Table 2, Model 2a), but not statistically significant. The principle of care, altruistic values and the joy of giving are


all positively associated with the willingness to contribute more, albeit not significant either (Model 2b). Those who have much confidence in charities are about three times more likely to compensate for budget cuts, which is significant (Model 2c). This suggests that the perceived efficacy of nonprofit organizations might help to raise funds after budget cuts. The willingness to increase donations is significantly associated with the amount donated to expressive organizations in the past year (Model 2d).

Donors targeted already are willing to donate more, especially in ser-vice subsectorsThe coefficients in Model 3 from Tables 1 and 2 show that the recruitment mechanism helps to predict the willingness to increase donations. In service subsectors, experiencing strong social pressure is associated with a 37% higher likelihood to compensate budget cuts, which is not statistically sig-nificant (Table 1, Model 3a). Citizens who are currently targeted by nonprof-it organizations are more likely to increase donations if asked. Participants who received more than three solicitations in the past two weeks are 175% more likely to compensate than those who received less than three solicita-tions, which is strongly significant (Table 1, Model 3b). The odds ratios for expressive subsectors are in the same direction, but weaker and not statisti-cally significant (Table 2, Model 3).

In expressive subsectors, resources influence willingness to contribute through engagementWhat are the interactive effects between the mechanisms of resources, en-gagement and recruitment? In expressive subsectors, resources influence the willingness to contribute more in response to government cutbacks mainly through engagement, and not so much through recruitment. The relation-ships between resources and the willingness to contribute more are reduced when the level of engagement with nonprofit organizations in the past year are included as controls (Table 2, Models 2b, 2c and 2d). Controlling for the number of solicitations received, however, does not diminish the association between resources and the willingness to contribute more (Table 2, Model 3b).

The mediating effect of engagement only holds for expressive subsectors. In service subsectors (Table 1), adding measures of engagement and recruit-ment to the model does not substantially change the coefficients of educa-tion and income.


Tabl

e 1:

Mul

tilev

el lo

gist

ic re

gres

sion

of t

he w

illing

ness

to c

ontri

bute

mor

e af

ter g

over

nmen

t cut

back

s in

se

rvic

e su

bsec

tors

Reso

urce

sEn

gage

men

tRe

crui

tmen

t1

2a2b

2c2d

3a3b

Seco

ndar

y ed

ucat

ion

1.54

71.

580

1.50

11.

529

1.55

61.

546

1.61

1(0

.563

)(0

.574

)(0

.533

)(0

.547

)(0

.563

)(0

.561

)(0

.592

)Te

rtia

ry e

duca

tion

5.05

7***

5.26

1***

4.74

0***

4.74

2***

4.80

8***

4.75

5***

4.84

0***

(2.2

02)

(2.2

97)

(1.9

79)

(2.0

07)

(2.0

90)

(2.0

71)

(2.1

29)

Seco

nd in

com

e qu

intil

e0.

717

0.72

50.

727

0.74

80.

740

0.75

90.

822

(0.3

82)

(0.3

85)

(0.3

79)

(0.3

94)

(0.3

93)

(0.4

04)

(0.4

40)

Thir

d in

com

e qu

intil

e1.

101

1.11

21.

089

1.15

31.

143

1.14

31.

181

(0.5

64)

(0.5

68)

(0.5

50)

(0.5

87)

(0.5

86)

(0.5

88)

(0.6

11)

Four

th in

com

e qu

intil

e1.

437

1.47

41.

487

1.44

21.

415

1.42

31.

547

(0.7

44)

(0.7

63)

(0.7

56)

(0.7

38)

(0.7

32)

(0.7

38)

(0.8

09)

Fifth

inco

me

quin

tile

2.34

62.

398

2.49

4*2.

402

2.40

12.

446

2.55

6*(1

.286

)(1

.312

)(1

.339

)(1

.298

)(1

.312

)(1

.341

)(1

.411

)In

com

e fr

om w

ealth

1.04

61.

051

1.15

71.

273

1.24

61.

246

1.15

3(0

.492

)(0

.495

)(0

.527

)(0

.580

)(0

.572

)(0

.574

)(0

.537

)H

ome

owne

r0.

542*

0.57

5*0.

597

0.64

80.

651

0.64

10.

605

(0.1

75)

(0.1

85)

(0.1

88)

(0.2

04)

(0.2

07)

(0.2

04)

(0.1

95)

Fina

ncia

l sec

urity

1.15

51.

172*

1.14

81.

145

1.14

61.

138

1.13

0(0

.105

)(0

.107

)(0

.102

)(0

.102

)(0

.103

)(0

.103

)(0

.103

)Fu

ture

fina

ncia

l situ

atio

n0.

842

0.84

80.

852

0.83

10.

827

0.82

70.

839

(0.1

40)

(0.1

40)

(0.1

38)

(0.1

36)

(0.1

37)

(0.1

37)

(0.1

40)

Perc

eptio

n of

nee

d1.

183*

1.14

71.

149

1.15

91.

156

1.15

0(0

.113

)(0

.106

)(0

.107

)(0

.109

)(0

.109

)(0

.109

)Kn

owle

dge

abou

t nee

d1.

081

1.08

31.

095

1.10

91.

109

1.11

8(0

.185

)(0

.181

)(0

.184

)(0

.188

)(0

.188

)(0

.191

)Pr

inci

ple

of ca

re1.

435

1.35

11.

357

1.23

71.

200

(0.4

01)

(0.3

81)

(0.3

88)

(0.3

64)

(0.3

54)


Tabl

e 1

(con

tinue

d)

Altr

uist

ic v

alue

s1.

473

1.36

41.

389

1.34

71.

367

(0.5

77)

(0.5

35)

(0.5

49)

(0.5

34)

(0.5

45)

Empa

thic

conc

ern

0.95

80.

928

0.93

50.

938

0.90

4(0

.266

)(0

.260

)(0

.265

)(0

.267

)(0

.259

)Jo

y of

giv

ing

1.78

1**

1.62

8*1.

649*

*1.

558*

1.55

8*(0

.447

)(0

.407

)(0

.419

)(0

.401

)(0

.402

)Ch

arita

ble

conf

iden

ce2.

945*

**2.

979*

**2.

859*

**2.

951*

**(0

.916

)(0

.955

)(0

.917

)(0

.958

)Tr

ust i

n go

vern

men

t0.

528*

0.52

5*0.

514*

0.51

5*(0

.181

)(0

.181

)(0

.179

)(0

.180

)N

umbe

r of d

onat

ion

area

s0.

912

0.90

40.

881

(0.0

90)

(0.0

91)

(0.0

90)

Serv

ice

dona

tions

(ln)

1.06

61.

065

1.06

0(0

.095

)(0

.095

)(0

.095

)Ex

pres

sive

don

atio

ns (l

n)1.

024

1.02

11.

037

(0.0

85)

(0.0

85)

(0.0

87)

Soci

al p

ress

ure

1.37

81.

327

(0.3

47)

(0.3

36)

Mor

e th

an th

ree

solic

itatio

ns2.

743*

*(1

.129

)(C

onst

ant)

0.00

2***

0.00

1***

0.00

0***

0.00

0***

0.00

0***

0.00

0***

0.00

0***

(0.0

02)

(0.0

01)

(0.0

00)

(0.0

00)

(0.0

00)

(0.0

00)

(0.0

00)

No.

of s

cena

rios

2,56

42,

564

2,56

42,

564

2,56

42,

564

2,56

4N

o. o

f res

pond

ents

1,98

71,

987

1,98

71,

987

1,98

71,

987

1,98

7O

dds R

atio

s are

repo

rted

; * p

< .1

; ** p

< .0

5; **

* p <

.01;

Con

trol

led

for g

ende

r, ag

e, re

ligio

us d

enom

inat

ion,

scen

ario

ord

er, s

ize

of th

e bu

dget

cut

, and

OC

W su

bsam

ple


Tabl

e 2:

Mul

tilev

el lo

gist

ic re

gres

sion

of t

he w

illing

ness

to c

ontri

bute

mor

e af

ter g

over

nmen

t cut

back

s in

ex

pres

sive

sub

sect

ors

Reso

urce

sEn

gage

men

tRe

crui

tmen

t1

2a2b

2c2d

3a3b

Seco

ndar

y ed

ucat

ion

1.68

31.

732

1.66

01.

607

1.55

81.

554

1.58

7(0

.570

)(0

.588

)(0

.563

)(0

.548

)(0

.537

)(0

.536

)(0

.549

)Te

rtia

ry e

duca

tion

2.90

5***

2.98

3***

2.73

9***

2.59

5**

2.19

6**

2.18

0**

2.16

5**

(1.1

07)

(1.1

39)

(1.0

46)

(1.0

03)

(0.8

58)

(0.8

52)

(0.8

47)

Seco

nd in

com

e qu

intil

e0.

428*

0.42

9*0.

439*

0.44

8*0.

436*

0.43

8*0.

440*

(0.2

08)

(0.2

09)

(0.2

14)

(0.2

19)

(0.2

15)

(0.2

16)

(0.2

17)

Thir

d in

com

e qu

intil

e0.

902

0.89

70.

888

0.90

50.

895

0.89

30.

882

(0.3

99)

(0.3

98)

(0.3

97)

(0.4

06)

(0.4

07)

(0.4

06)

(0.4

01)

Four

th in

com

e qu

intil

e0.

875

0.87

90.

896

0.86

20.

821

0.82

30.

828

(0.4

01)

(0.4

03)

(0.4

12)

(0.3

97)

(0.3

86)

(0.3

87)

(0.3

89)

Fifth

inco

me

quin

tile

1.22

01.

242

1.27

21.

183

1.07

11.

073

1.07

0(0

.570

)(0

.581

)(0

.596

)(0

.558

)(0

.515

)(0

.516

)(0

.514

)In

com

e fr

om w

ealth

1.42

81.

427

1.40

01.

557

1.43

41.

440

1.41

7(0

.630

)(0

.634

)(0

.620

)(0

.695

)(0

.647

)(0

.650

)(0

.640

)H

ome

owne

r0.

620

0.64

90.

668

0.73

90.

762

0.75

70.

741

(0.1

88)

(0.1

97)

(0.2

03)

(0.2

26)

(0.2

37)

(0.2

36)

(0.2

31)

Fina

ncia

l sec

urity

1.19

4**

1.20

7**

1.19

5**

1.17

6*1.

153*

1.15

0*1.

152*

(0.0

98)

(0.1

00)

(0.0

99)

(0.0

98)

(0.0

97)

(0.0

97)

(0.0

97)

Futu

re fi

nanc

ial s

ituat

ion

1.22

11.

228

1.20

61.

170

1.20

41.

205

1.21

0(0

.183

)(0

.184

)(0

.181

)(0

.177

)(0

.185

)(0

.185

)(0

.186

)Pe

rcep

tion

of n

eed

1.14

91.

123

1.13

31.

144

1.14

31.

146

(0.1

04)

(0.1

02)

(0.1

04)

(0.1

07)

(0.1

07)

(0.1

07)

Know

ledg

e ab

out n

eed

1.11

51.

117

1.14

81.

141

1.13

91.

149

(0.1

86)

(0.1

86)

(0.1

93)

(0.1

94)

(0.1

94)

(0.1

96)

Prin

cipl

e of

care

1.36

81.

283

1.24

31.

215

1.19

2(0

.375

)(0

.355

)(0

.347

)(0

.348

)(0

.341

)


Tabl

e 2

(con

tinue

d)

Altr

uist

ic v

alue

s1.

489

1.40

01.

326

1.31

51.

335

(0.5

48)

(0.5

17)

(0.4

97)

(0.4

93)

(0.5

02)

Empa

thic

conc

ern

0.91

60.

874

0.84

60.

847

0.85

2(0

.254

)(0

.244

)(0

.239

)(0

.240

)(0

.241

)Jo

y of

giv

ing

1.30

91.

152

1.11

81.

100

1.08

2(0

.309

)(0

.274

)(0

.271

)(0

.271

)(0

.267

)Ch

arita

ble

conf

iden

ce3.

124*

**2.

618*

**2.

590*

**2.

634*

**(0

.968

)(0

.825

)(0

.818

)(0

.835

)Tr

ust i

n go

vern

men

t0.

572*

0.58

80.

587

0.59

5(0

.187

)(0

.194

)(0

.193

)(0

.196

)N

umbe

r of d

onat

ion

area

s0.

912

0.91

00.

894

(0.0

79)

(0.0

79)

(0.0

79)

Serv

ice

dona

tions

(ln)

0.94

10.

942

0.94

3(0

.079

)(0

.079

)(0

.080

)Ex

pres

sive

don

atio

ns (l

n)1.

333*

**1.

331*

**1.

340*

**(0

.118

)(0

.118

)(0

.119

)So

cial

pre

ssur

e1.

087

1.06

4(0

.256

)(0

.251

)M

ore

than

thre

e so

licita

tions

2.01

9(0

.888

)(C

onst

ant)

0.00

1***

0.00

1***

0.00

0***

0.00

0***

0.00

0***

0.00

0***

0.00

0***

(0.0

01)

(0.0

01)

(0.0

00)

(0.0

00)

(0.0

00)

(0.0

00)

(0.0

00)

No.

of s

cena

rios

3,23

93,

239

3,23

93,

239

3,23

93,

239

3,23

9N

o. o

f res

pond

ents

2,11

22,

112

2,11

22,

112

2,11

22,

112

2,11

2O

dds R

atio

s are

repo

rted

; * p

< .1

; ** p

< .0

5; **

* p <

.01;

Con

trol

led

for g

ende

r, ag

e, re

ligio

us d

enom

inat

ion,

scen

ario

ord

er, s

ize

of th

e bu

dget

cut

, and

OC

W su

bsam

ple


Willingness predicts actual increase in expressive donationsRespondents might respond that they are willing to increase donations after budget cuts, but will they actually do so? To explore this question, we make use of the panel nature of the data set to examine the actual change in do-nations from 2011 to 2013. This was a period with budget constraints for the Dutch government and decreasing government expenditures in different areas, including large and controversial budget cuts in the subsectors of in-ternational development and culture.

Figure 3 shows the percentage of respondents that increased, decreased or did not change their actual donations from 2011 to 2013. The percent-age of respondents that actually increased donations is 46% for those who said to be willing to increase donations in the scenario questions, and 40% for those who were not. The difference only occurs in expressive subsectors (42% vs. 30%) and not in service subsectors (37% vs. 37%). Surprisingly, the percentage of respondents who decreased donations is also larger among those who said to be willing to compensate budget cuts in the scenarios. This holds for both donations to service organizations and expressive organiza-tions.

Resources and engagement explain actual increases in givingTables 3 and 4 show regression models on the likelihood of increasing actu-al donations to organizations in service and expressive subsectors, respec-

0 10 20 30 40 50 60

Not willing to increase donations

Willing to increase donations in at least one scenario





Expr

essiv

eSe

rvice

Tota

l

Increased donations 2011-2013 Decreased donations 2011-2013

Figure 3: Proportion of participants that increased or decreased actual donations 2011-2013 by response in scenario experiment (in percentages)


tively. Here, respondents that participated in two waves of the panel survey (n=916) are the units of analysis.

In service subsectors, the willingness to increase donations as indicated in the scenarios is negatively related to the likelihood to increase actual dona-tions, which is not statistically significant (Table 3). Similar to the responses in the scenario, a higher education and a higher income are positively asso-ciated with increasing donations. However, respondents who perceive their future financial situation as more bright are less likely to increase their do-nations to service organizations. Stronger altruistic values and a higher char-itable confidence are positively related to increasing donations. The amount donated in 2011 is negatively related to the likelihood of increasing dona-tions, which might be the result of a ceiling effect. Contrary to the results of the scenario responses, the joy of giving and the number of solicitations are not strongly associated with an actual increase in donations. Resources and engagement seems the strongest mechanism to explain actual increasing do-nations to service organizations.

In expressive subsectors, the willingness to increase donations in the sce-narios is positively related to actual increases (Table 4). Respondents who indicated they would increase donations after budget cuts are about 50% more likely to actually increase donations. The higher educated and home owners are more likely to increase donations. Similar to the correlates of the scenario responses, the lowest and highest income quintiles are the most likely to increase donations. From the engagement variables (Model 2), char-itable confidence correlates strongly and significantly with the likelihood to increase actual donations. The previous level of donations to expressive or-ganizations decreases the likelihood to increase giving, which can be a ceil-ing effect again, but the amount donated to service organizations increases this chance. Apparently, those who initially donate high amounts to service organizations are more willing to compensate for budget cuts in expressive subsectors. There is no evidence for mediating effects of engagement in these analyses.

It is striking that the mechanism of recruitment seems a stronger expla-nation for hypothetical increases than for actual increases. If organizations would become more active in the fundraising market after government bud-get cuts (Andreoni & Payne, 2003, 2011), we would expect that citizens who are targeted more by fundraisers are more likely to increase donations from 2011 to 2013.


Tabl

e 3:

Log

istic

regr

essi

on o

f inc

reas

ing

dona

tions

to s

ervi

ce s

ubse

ctor

s 20

11-2

013

Reso

urce

sEn

gage

men

tRe

crui

tmen

t1

2a2b

2c2d

3a3b

Will

ingn

ess t

o in

crea

se in

scen

ario

s0.

938

0.90

70.

872

0.83

90.

852

0.85

60.

841

(0.2

92)

(0.2

84)

(0.2

75)

(0.2

65)

(0.2

77)

(0.2

78)

(0.2

76)

Seco

ndar

y ed

ucat

ion

1.62

1**

1.63

8**

1.55

5**

1.53

2**

1.59

0**

1.59

3**

1.60

6**

(0.3

16)

(0.3

20)

(0.3

08)

(0.3

05)

(0.3

26)

(0.3

27)

(0.3

31)

Tert

iary

edu

catio

n1.

456*

1.48

9*1.

433

1.38

61.

427

1.44

51.

449

(0.3

32)

(0.3

41)

(0.3

35)

(0.3

31)

(0.3

57)

(0.3

64)

(0.3

65)

Seco

nd in

com

e qu

intil

e1.

620*

1.64

7*1.

812*

*1.

820*

*1.

909*

*1.

896*

*1.

928*

*(0

.467

)(0

.477

)(0

.530

)(0

.534

)(0

.576

)(0

.573

)(0

.588

)Th

ird

inco

me

quin

tile

1.37

41.

411

1.50

21.

476

1.54

11.

533

1.55

0(0

.408

)(0

.421

)(0

.451

)(0

.444

)(0

.481

)(0

.479

)(0

.486

)Fo

urth

inco

me

quin

tile

1.38

41.

413

1.49

51.

463

1.63

21.

622

1.64

4(0

.414

)(0

.425

)(0

.452

)(0

.444

)(0

.511

)(0

.508

)(0

.518

)Fi

fth in

com

e qu

intil

e1.

244

1.24

81.

387

1.35

51.

443

1.43

31.

446

(0.3

98)

(0.4

01)

(0.4

51)

(0.4

42)

(0.4

85)

(0.4

82)

(0.4

88)

Inco

me

from

wea

lth1.

020

1.01

11.

017

1.04

11.

067

1.06

61.

056

(0.2

87)

(0.2

86)

(0.2

90)

(0.2

98)

(0.3

16)

(0.3

16)

(0.3

14)

Hom

e ow

ner

1.03

61.

067

1.03

11.

069

1.15

41.

155

1.15

1(0

.182

)(0

.190

)(0

.185

)(0

.194

)(0

.217

)(0

.217

)(0

.217

)Fi

nanc

ial s

ecur

ity0.

995

0.99

81.

005

1.00

10.

998

1.00

00.

999

(0.0

48)

(0.0

49)

(0.0

50)

(0.0

50)

(0.0

51)

(0.0

52)

(0.0

52)

Futu

re fi

nanc

ial s

ituat

ion

0.80

2**

0.80

8**

0.80

3**

0.79

6**

0.76

7***

0.76

6***

0.76

8**

(0.0

78)

(0.0

79)

(0.0

79)

(0.0

79)

(0.0

79)

(0.0

79)

(0.0

79)

Perc

eptio

n of

nee

d1.

076

1.06

31.

065

1.05

31.

053

1.05

3(0

.058

)(0

.058

)(0

.058

)(0

.059

)(0

.059

)(0

.059

)Kn

owle

dge

abou

t nee

d1.

052

1.03

01.

043

1.03

61.

039

1.04

0(0

.106

)(0

.105

)(0

.107

)(0

.109

)(0

.110

)(0

.110

)Pr

inci

ple

of ca

re0.

933

0.92

20.

939

0.95

40.

947

(0.1

50)

(0.1

49)

(0.1

57)

(0.1

62)

(0.1

62)


Tabl

e 3

(con

tinue

d)

Altr

uist

ic v

alue

s1.

915*

**1.

925*

**1.

865*

**1.

876*

**1.

882*

**(0

.431

)(0

.435

)(0

.436

)(0

.439

)(0

.441

)Em

path

ic co

ncer

n1.

031

1.01

61.

038

1.03

71.

035

(0.1

77)

(0.1

75)

(0.1

84)

(0.1

83)

(0.1

83)

Joy

of g

ivin

g1.

009

0.97

20.

982

1.00

01.

003

(0.1

48)

(0.1

44)

(0.1

51)

(0.1

58)

(0.1

58)

Char

itabl

e co

nfid

ence

1.40

4*1.

569*

*1.

586*

*1.

594*

*(0

.261

)(0

.310

)(0

.316

)(0

.318

)Tr

ust i

n go

vern

men

t0.

883

0.90

10.

907

0.90

7(0

.184

)(0

.194

)(0

.196

)(0

.196

)N

umbe

r of d

onat

ion

area

s1.

083

1.08

51.

083

(0.0

65)

(0.0

65)

(0.0

65)

Serv

ice

dona

tions

(ln)

0.70

2***

0.70

2***

0.70

1***

(0.0

42)

(0.0

42)

(0.0

42)

Expr

essi

ve d

onat

ions

(ln)

1.08

61.

087

1.08

9(0

.059

)(0

.060

)(0

.060

)So

cial

pre

ssur

e0.

924

0.92

3(0

.142

)(0

.142

)M

ore

than

thre

e so

licita

tions

1.13

1(0

.339

)(C

onst

ant)

0.47

9*0.

396*

*0.

292*

0.34

50.

462

0.50

50.

510

(0.1

85)

(0.1

63)

(0.2

03)

(0.2

44)

(0.3

40)

(0.3

82)

(0.3

86)

No.

of r

espo

nden

ts74

274

274

274

274

274

274

2O

dds R

atio

s are

repo

rted

; * p

< .1

; ** p

< .0

5; **

* p <

.01;

Con

trol

led

for g

ende

r, ag

e, re

ligio

us d

enom

inat

ion,

scen

ario

ord

er, s

ize

of th

e bu

dget

cut

, and

OC

W su

bsam

ple


Tabl

e 4:

Log

istic

regr

essi

on o

f inc

reas

ing

dona

tions

to e

xpre

ssiv

e su

bsec

tors

201

1-20

13

Reso

urce

sEn

gage

men

tRe

crui

tmen

t1

2a2b

2c2d

3a3b

Will

ingn

ess t

o in

crea

se in

scen

ario

s1.

496

1.51

2*1.

491

1.43

21.

567*

1.57

3*1.

584*

(0.3

68)

(0.3

73)

(0.3

70)

(0.3

58)

(0.4

01)

(0.4

02)

(0.4

06)

Seco

ndar

y ed

ucat

ion

2.01

4***

2.01

0***

1.97

7***

1.97

4***

1.97

6***

1.98

0***

1.96

3***

(0.4

23)

(0.4

24)

(0.4

19)

(0.4

20)

(0.4

26)

(0.4

28)

(0.4

25)

Tert

iary

edu

catio

n1.

495*

1.48

21.

428

1.40

21.

419

1.39

51.

398

(0.3

60)

(0.3

58)

(0.3

51)

(0.3

51)

(0.3

65)

(0.3

61)

(0.3

61)

Seco

nd in

com

e qu

intil

e0.

801

0.78

20.

794

0.79

90.

741

0.74

80.

736

(0.2

21)

(0.2

18)

(0.2

22)

(0.2

24)

(0.2

11)

(0.2

14)

(0.2

12)

Thir

d in

com

e qu

intil

e0.

699

0.67

80.

676

0.66

80.

626

0.63

00.

626

(0.1

98)

(0.1

93)

(0.1

94)

(0.1

92)

(0.1

83)

(0.1

85)

(0.1

84)

Four

th in

com

e qu

intil

e0.

464*

*0.

458*

**0.

458*

**0.

453*

**0.

415*

**0.

420*

**0.

417*

**(0

.139

)(0

.138

)(0

.138

)(0

.137

)(0

.128

)(0

.130

)(0

.129

)Fi

fth in

com

e qu

intil

e1.

112

1.10

91.

142

1.12

71.

077

1.08

91.

082

(0.3

30)

(0.3

30)

(0.3

41)

(0.3

38)

(0.3

29)

(0.3

33)

(0.3

32)

Inco

me

from

wea

lth0.

950

0.94

40.

934

0.99

21.

033

1.04

01.

043

(0.2

63)

(0.2

63)

(0.2

61)

(0.2

79)

(0.2

94)

(0.2

96)

(0.2

97)

Hom

e ow

ner

1.45

4**

1.39

4*1.

404*

1.44

8*1.

349

1.34

71.

351

(0.2

69)

(0.2

61)

(0.2

65)

(0.2

75)

(0.2

59)

(0.2

59)

(0.2

60)

Fina

ncia

l sec

urity

1.02

71.

024

1.02

81.

024

1.03

01.

024

1.02

5(0

.051

)(0

.050

)(0

.051

)(0

.051

)(0

.053

)(0

.053

)(0

.053

)Fu

ture

fina

ncia

l situ

atio

n0.

871

0.86

30.

856

0.85

20.

853

0.85

50.

851

(0.0

85)

(0.0

85)

(0.0

85)

(0.0

85)

(0.0

87)

(0.0

87)

(0.0

87)

Perc

eptio

n of

nee

d0.

925

0.91

90.

919

0.91

30.

913

0.91

2(0

.050

)(0

.050

)(0

.051

)(0

.051

)(0

.051

)(0

.051

)Kn

owle

dge

abou

t nee

d1.

017

1.02

51.

040

1.04

01.

035

1.03

3(0

.105

)(0

.107

)(0

.109

)(0

.110

)(0

.109

)(0

.109

)Pr

inci

ple

of ca

re1.

228

1.21

01.

190

1.16

11.

172

(0.2

05)

(0.2

03)

(0.2

02)

(0.2

01)

(0.2

04)


Tabl

e 4

(con

tinue

d)

Altr

uist

ic v

alue

s0.

999

1.01

20.

985

0.97

50.

975

(0.2

23)

(0.2

27)

(0.2

24)

(0.2

22)

(0.2

22)

Empa

thic

conc

ern

0.93

40.

908

0.88

60.

885

0.88

5(0

.167

)(0

.163

)(0

.162

)(0

.161

)(0

.161

)Jo

y of

giv

ing

1.15

91.

118

1.07

71.

055

1.05

4(0

.173

)(0

.169

)(0

.165

)(0

.165

)(0

.165

)Ch

arita

ble

conf

iden

ce1.

464*

*1.

434*

1.41

2*1.

404*

(0.2

78)

(0.2

82)

(0.2

79)

(0.2

78)

Trus

t in

gove

rnm

ent

0.76

80.

719

0.71

50.

713

(0.1

64)

(0.1

56)

(0.1

56)

(0.1

55)

Num

ber o

f don

atio

n ar

eas

1.10

4*1.

102*

1.10

5*(0

.060

)(0

.060

)(0

.061

)Se

rvic

e do

natio

ns (l

n)1.

166*

**1.

165*

**1.

166*

**(0

.064

)(0

.064

)(0

.064

)Ex

pres

sive

don

atio

ns (l

n)0.

887*

*0.

885*

*0.

884*

*(0

.048

)(0

.048

)(0

.048

)So

cial

pre

ssur

e1.

117

1.11

8(0

.172

)(0

.172

)M

ore

than

thre

e so

licita

tions

0.85

8(0

.273

)(C

onst

ant)

0.32

7***

0.37

5**

0.15

4***

0.18

6**

0.18

7**

0.16

7**

0.16

6**

(0.1

27)

(0.1

54)

(0.1

06)

(0.1

32)

(0.1

35)

(0.1

24)

(0.1

23)

No.

of r

espo

nden

ts78

778

778

778

778

778

778

7O

dds R

atio

s are

repo

rted

; * p

< .1

; ** p

< .0

5; **

* p <

.01;

Con

trol

led

for g

ende

r, ag

e, re

ligio

us d

enom

inat

ion,

scen

ario

ord

er, s

ize

of th

e bu

dget

cut

, and

OC

W su

bsam

ple



The vast majority of citizens in the Netherlands is not willing to change do-nations after changes in government support, which is similar to survey re-sults from the U.S. (Horne et al., 2005) and the U.K. (Shah et al., 2015). Both among service and expressive organizations, decreasing subsidies are not likely to be made up with fundraising income. Yet, there is a small group of citizens who are willing to compensate for government budget cuts with their charitable donations. In the present paper we sought to explore a pro-file of this group.

Until now, only a few studies examined individual heterogeneity in re-sponses to changes in government support, without a strong theoretical ba-sis for the expected results (De Wit et al., 2017; Kingma, 1989; Luccasen, 2012; Reeson & Tisdell, 2008). This paper offers a first attempt to provide a theoretical framework for the question which groups of citizens are more willing to compensate for public funding than others. By introducing and exploring the Civic Voluntarism Model (Verba et al., 1995) in this context, we tested whether predictors of civic voluntarism are also associated with changes in charitable donations upon reductions of government support for nonprofit organizations.

The results provide considerable support for the predictions from the Civic Voluntarism Model on resources, recruitment, and engagement. Citi-zens with more resources like education are more likely to compensate for budget cuts. Engagement is generally correlated with a higher willingness to compensate for decreasing government funding. Citizens with a high level of general confidence in charitable organizations are more willing to com-pensate for budget cuts, which suggests that efforts to increase confidence in the nonprofit sector can also make citizens more willing to increase their contributions. In service subsectors, those with stronger altruistic values are also more likely to increase giving. The joy of giving – a trait measure of the “warm glow” people feel by giving to charity – is positively related to the willingness to increase charitable giving after government budget cuts in service subsectors, which is contrary to expectations from the impure al-truism model (Andreoni, 1989, 1990). Regarding recruitment, citizens who are more frequently solicited for charitable contributions are more likely to say they are willing to donate more when government funding to nonprofit organizations is reduced.

In expressive subsectors, resources influence the willingness to compen-


sate partly through the mechanism of engagement, and not so much through recruitment. This finding is consistent with prior research on solicitation practices by nonprofit organizations in the Netherlands (Bekkers, 2005b), which showed that the association between resources and donations cannot be explained by solicitations.

Some findings do not confirm expectations from the Civic Voluntarism Model. In expressive subsectors, the correlation between household income and the willingness to contribute more after budget cuts takes a U-shape, with the bottom and the top of the distribution being more likely to increase donations. Expressive organizations tend to have lower number of donors in general, especially among lower income groups. It could be that the lower in-comes are more sensitive to the awareness of need. This would explain why they generally tend to favor organizations that focus on social needs in their own environment, and why they are more sensitive to external signals about increasing need caused by governmental budget cuts in expressive areas. Second, citizens with high trust in government are found to be less respon-sive to changes in government support. This is a highly interesting finding. A possible explanation is that those citizens attach a different value to public goods as provided by the government than to public goods as provided by nonprofit organizations (Tinkelman, 2010). Third, we find empathic concern to be not strongly related to increases in private giving when other prosocial values are included in the model, which is consistent with findings on chari-table giving by Bekkers and Wilhelm (2016). This shows that our measures of engagement do not perfectly predict changes in donations after decreas-ing government support.

For nonprofit professionals facing unstable revenues from public funding, the findings provide useful tools. In order to increase fundraising income, organizations should target specific social groups that are more likely to change their giving behavior. Citizens who are higher educated, who have stronger prosocial values, who have more confidence in charitable organiza-tions and who are more frequently solicited for charitable contributions are the most interesting groups to target with fundraising appeals after govern-ment budget cuts. Service provision organizations should consider enhanc-ing the “warm glow” of giving in order to reach out to possible donors who feel good by doing good. Organizations in “expressive” areas could approach those with more resources by appealing to their previous engagement with the cause. Building personal, long-term relationships seems especially fruit-ful in this area. In this way, organizations can use donor profiles in the spe-


cific context of changing government funding in order to be better capable of compensating financial uncertainties with increases in fundraising income.

Conclusion

Partners, not substitutes

162 Conclusion Partners, not substitutes

Ever since Alexis de Tocqueville published is seminal De la démocratie en Amérique (1970 [1840]), conservative thinkers have proposed that exten-sive governments crowd out civic engagement. The idea is appealing and still resonates in political debates on the consequences of policy interventions. In academic research, too, the crowding-out hypothesis has been studied extensively in economics (Andreoni, 1993; Andreoni & Payne, 2003, 2011; Korenok et al., 2014), sociology (Gesthuizen et al., 2008; Kääriäinen & Leh-tonen, 2006; Khanna & Sandler, 2000; Van Oorschot & Arts, 2005), public ad-ministration (Brooks, 1999, 2000a, 2000b, 2003a, 2003b, 2004; Horne et al., 2006), fiscal studies (Heutel, 2014; Hsu, 2008; Payne, 2001; Sutter & Weck-Hannemann, 2004) and philanthropic studies (Hughes et al., 2014; Kim & Van Ryzin, 2014). This thesis examined the relationship between financial government support and charitable giving to nonprofit organizations, while exploring mediating and moderating factors that may explain its contextual dependence.

SUMMARY OF FINDINGS

The empirical work in this dissertation provides elaborate answers to the questions to what extent, how, where, under which conditions and among whom government support affects charitable donations.

RQ1: To what extent does government support affect charitable dona-tions?Across countries (Chapter 1), this thesis finds zero correlation between gov-ernment support and amounts donated, and a positive correlation between government support and the number of donors. However, the cross-section-al nature of this study does not tell us much about the causal relationship. The meta-analysis in Chapter 2 shows that the empirical evidence for the crowding-out hypothesis in previous studies is weak. Much of the evidence comes from laboratory experiments, while findings outside the research laboratory tend to find zero correlation or positive correlations. This causes doubts about the validity of the crowding-out hypothesis in real-life situa-tions. This is confirmed by the results on data from the Netherlands in Part II of this dissertation. Chapter 3 shows that citizens typically are not provided with information about actual changes in government support. Looking at donations to specific organizations over time, there is no correlation across

163Partners, not substitutes Conclusion

all organizations in the sample and large differences between organizations. The studies in Chapters 4 and 5 provide respondents with information about (potential) budget cuts, which resembles the assumption of full information that is made in most experimental studies (see RQ4 below). When asked about their change in charitable giving in response to budget cuts, only 8% of citizens indicate that they would increase donations (Chapter 5). The stron-gest support for the crowding-out hypothesis is found in the survey exper-iment in Chapter 4. Providing citizens with information about budget cuts might increase an organization’s donor base with over 20%.

In theory, the null findings in non-experimental research can also be due to a non-linear relation between government support and private donations. When organizations receive a small part of their funding from the govern-ment, it might be perceived as a signal of quality which encourages dona-tions; when they are largely subsidy-dependent, this might be perceived as undesirable, which drives out donations (Brooks, 2000a, 2003a; Borgonovi, 2006; Nikolova, 2015). In Chapter 3 of this dissertation, an interaction effect was tested between government support and the degree to which organiza-tions are subsidy-dependent, which was not statistically significant. Chapter 5 shows that when respondents are randomly presented with 5%, 10%, 20% or 33% budget cuts, the number of donors increases with the size of the cut. Thus, the data show no indication of non-linearity.

Another question is whether increases in government support have the same effect as decreases. Originally, the crowding-out hypothesis is formulat-ed in an era in which welfare states were developing. Most studies formulate crowding-out as a simple negative relationship between government sup-port and charitable donations, with increases (decreases) in the first associ-ated with decreases (increases) in the latter. Yet, it could be that government expansion depresses civic engagement, while budget cuts do not lead to in-creases in private giving. This is not supported by the data in this thesis. On the contrary, the survey experiment in Chapter 4 shows that crowding-out is slightly stronger after decreasing government support (leading to more donations) than it is after increasing government support (leading to less donations).

In sum, the results show that the validity of the crowding-out hypothesis is highly context-dependent. The vast majority does not change its giving be-havior in response to changing government funding. This is in line with sur-veys in which more than 7 out of 10 people say they would not change their donations as a reaction on changing government funding (Horne et al., 2005;


Shah et al., 2015). Across the board there is no clear direct association be-tween government support and charitable donations, but there is evidence that specific groups of people, under specific circumstances, are responsive to government policies.

RQ2: How does government support affect charitable donations?Regarding mediating processes, this thesis does not find support for the idea of “fundraising crowding-out” which has been suggested with organizational data from the US (Andreoni & Payne, 2003, 2011). Fundraising efforts lead to higher private income, but there is no negative relationship between govern-ment funding and fundraising expenditures in the Netherlands (Chapter 3).

Information can explain associations between government support and charitable giving (Chapter 3). In some cases, changes in government policies are reported in news media, which leads to a larger availability of informa-tion about nonprofit organizations and their goals. How this process works depends on the context, however. Often, media coverage is rather unrelated to changes in government support. Media only report on the public funding of nonprofit organizations when government policies suddenly take a differ-ent direction or when there is another newsworthy feature. An example are newspaper articles about the Salvation Army. This organization appeared in the media in relation to government subsidies mainly when there were de-bates about whether an organization that possibly discriminates in its em-ployee policy (the Salvation Army aims to hire Christians only) should be funded with public money. This example shows that media coverage follows incidental issues rather than long-term trends. Hence, the mediating effect of information highly depends on the content. In the case of Oxfam Novib, a large budget cut on international development grants was widely reported in the news, followed by a decrease in charitable donations. This can be tak-en as explorative evidence for the argument that government funding serves as a sign of quality – and thus, that a decrease in government support signals a loss of quality (Handy, 2000; Heutel, 2014; Schiff, 1990).

RQ3: Where does government support affect charitable donations?Across different nonprofit regime types (Salamon & Anheier, 1998), this the-sis finds relatively high proportions of charitable donors in countries with high government spending. Within countries, however, government support does not have similar effects in different nonprofit subsectors. Overall, char-itable giving is most likely to substitute government support in the field of


nature conservation. The strongest evidence of crowding-in, on the other hand, is found in the fields of environment, education and research, and in-ternational aid. Findings on health and social services are ambivalent. These are complex subsectors with a wide variety of organizations which are fund-ed by different types of government grants, contracts and project subsidies. Although the overall relationship between government support and charita-ble giving is unclear in these sectors, Chapter 4 showed that it is possible to attract donors with information about specific budget cuts.

Furthermore, the results of this thesis provide support for the argument that there is substitution between different parts of the nonprofit sector. Budget cuts may draw donors from other organizations or subsectors. On the macro level (Chapter 1), government spending on health and social protec-tion seem to drive donors towards “expressive” subsectors like culture, en-vironment and international aid. This effect has been labeled “philanthropic displacement” (Sokolowski, 2013) or “crosswise crowding-in” (Pennerstor-fer & Neumayr, 2017). In more precise analyses, there is evidence that de-creasing support to one organization partly attracts donors who otherwise would have donated to other organizations (Chapter 4), and that donors who previously donated to service organizations are more likely to increase giv-ing after budget cuts in expressive subsectors (Chapter 5).

RQ4: Under which conditions does government support affect charita-ble donations?

Information is not only a possible mediating variable (RQ2), it can also be a moderating variable. The availability of information is a prerequisite for people to change their giving behavior after changes in government policies. Although this might seem like stating the obvious, the role of information is largely overlooked in the debate.

Previous findings largely depend on the assumptions that are made in the research design (Payne, 2009; Ribar & Wilhelm, 2002; Tinkelman, 2010), which is confirmed by the meta-analysis in Chapter 2. In experimental re-search designs, a 1 Dollar increase in government support is associated with a 0.64 decrease in charitable donations, while non-experimental studies that use archival or survey data find zero change in donations on average. Al-though this difference may be due to the number of other donors in different research designs (Ribar & Wilhelm, 2002) or the endogenous nature of gov-ernment support (Payne, 2009), it is likely that a large part of the difference can be attributed to the assumption of full information which is present in


the laboratory experiments, but not in most field research. Information about government funding can be used to increase donations.

Chapter 4 showed that providing citizens with information about budget cuts might attract donors, both those who otherwise would donate to other organizations and those who otherwise would not donate. However, in many circumstances citizens rely on other sources than fundraising materials in making their donation decision (Li & McDougle, 2017; McDougle & Handy, 2014), which makes it harder to diffuse information. News media often do not reflect actual changes in government support, which makes it unlikely that donations change in most daily-life situations (Chapter 3).

RQ5: Among whom does government support affect charitable dona-tions?Responses to changes in government support vary across social groups. In this dissertation, there is some evidence that the higher educated and the wealthy are more responsive to changes in government support. With re-sources being a well-known correlate of civic engagement in the Civic Volun-tarism Model (Verba et al., 1995) and in donor profiles (Wiepking & Bekkers, 2012), they are also an important factor in explaining which social groups are likely to response to changing government support (Chapter 5). It could be that the costs of changing donations are lower for people with more re-sources, so that they are better able to change their giving behavior after changes in external circumstances. It could also be that the better-off are the ones who feel most responsible for providing public services in situations of little government presence.

The role of prosocial values is not straightforward. An extensive analysis on background characteristics based on the Civic Voluntarism Model (Chap-ter 5) shows that citizens with a higher generalized confidence in charitable organizations are more likely to substitute change in government funding with their donations. In nonprofit subsectors where organizations mainly provide services, those with stronger altruistic values are more likely to in-crease giving. The joy of giving is positively related to the willingness to in-crease charitable giving after government budget cuts in service subsectors.

Values might also suppress crowding-out effects. Citizens with a high trust in the government are substantially less responsive to changes in govern-ment support (Chapter 5). In different areas of the nonprofit sector, empath-ic concern is found to be not or even negatively related to the willingness to increase donations after budget cuts (Chapters 4 and 5). This suggests that


low-empathic citizens, who are less likely to donate, can be drawn into do-nating through extrinsic incentives.

Some findings in this dissertation seem to contradict each other. While those with high altruistic values say they would be more likely to increase donations after budget cuts (Chapter 5), their donations to specific organiza-tions are positively related with government funding (Chapter 3). A possible explanation is the availability of information. When not informed about the changes in government subsidies to specific nonprofit organizations, those with high altruistic values tend to follow the government. But when they are informed about the actual consequences of government funding for nonprof-it organizations, which evokes the awareness of need, they are inclined to compensate for budget cuts. More research is needed to test such possible interactions between information and values.

LIMITATIONS

This thesis suffers from a few important limitations. First, the assumption that government support is an exogenous variable is problematic in most empirical analyses. As stated by previous evaluations of evidence on the crowding-out hypothesis (Payne, 2009; Steinberg, 1985, 1997; Tinkelman 2010), endogeneity is a major concern for empirical crowding-out research. Government support and private donations may be jointly determined by unobserved variables like voter preferences and (changes in) the need for public goods, which would upwardly bias the relationship between the two. Also, there might be reversed causality when policy makers respond to lev-els of philanthropy in society (Sav, 2012). Especially the cross-country study in Chapter 1 does not allow for strong statements about the causal effects of government spending. It could very well be that social preferences for re-distributive justice in a country have led, through the political system and the creation of institutions over time, to expansive welfare states. The same preferences for justice are associated with prosocial behavior outside the political sphere, because people donate to organizations that work on these issues. If this holds, the size of the government and philanthropic donations are both expressions of the same values.

The comparative analysis should be taken as a descriptive analysis. Yet, no matter how the causal relation is, the positive correlation between social welfare expenditures and donations to environment, arts and international


aid indicate that countries with a strong domestic welfare state generally experience high donations to “expressive” subsectors, as opposed to expec-tations as formulated in the traditional crowding-out hypothesis.

The longitudinal analysis in Chapter 3 and the survey experiment in Chap-ter 4 are better suited to address the issue of causality. Chapter 3 looks at changes over time and takes a lagged measure of government support, with findings being highly dependent of the nonprofit sector and individual do-nor characteristics. The experimental design in Chapter 4 shows that, when providing a random group of people with the right information, budget cuts can attract donors.

A second limitation is that the causal model does not provide a com-prehensive picture of all possible moderating and mediating mechanisms. Regarding moderating variables, there are many conditions under which charitable donations might be affected by government support. To improve further research in this field, this dissertation offers a list of possible moder-ators in the Appendix.

Regarding mediating variables, important avenues for further research are the possible effects of government support on values and resources. (Neo-)institutionalist theories predicting more prosocial attitudes in exten-sive welfare states (Kääriäinen & Lehtonen, 2006; Rothstein, 1998; Svallfors, 1997) might offer an explanation for positive correlations between govern-ment support and charitable donations, especially between countries. The research designs in this thesis did not allow for testing the effect of welfare state efforts on values and resources, but future research could investigate such mechanisms. Furthermore, it would be interesting to further examine the mediating role of information. This dissertation measured information in newspapers, but government support can also signal information through social media, campaign materials and other channels.

A third issue is the limited generalizability of the results across countries and organizations. This thesis contributes to the literature by adding evi-dence on the relationship between government support and charitable do-nations in the Netherlands, but it is not sure to what extent these findings are applicable to other contexts. Different forms of civic engagement has different meanings across countries. Gesthuizen, Scheepers, Van der Veld, and Völker (2013) showed that informal and formal social capital are com-plementary forms of engagement in Western Europe while they are com-partmentalized in the largest part of Eastern Europe, calling “to re-address explanatory questions relating macro-characteristics of national institutions


[…] to individual level pro-social behavior” (Gesthuizen et al., 2013, p. 920). Although the comparative study in Chapter 1 included some non-Western countries, this thesis paid no attention to the developing world, in which the proposed mechanisms might play out differently. Chapters 1 and 2 examined whether crowding-out effects are systematically different in different coun-try contexts, but the samples of countries are not large enough to provide conclusive evidence that context does not matter.

Since the effects of government support varies between organizational contexts, the findings are not easily generalizable to all nonprofit organiza-tions. Analyses in Chapters 1, 3 and 5 show that there are strong differences between nonprofit subsectors, which is supported by a systematic review of findings from organizational-level studies (Lu, 2016). This questions the generalizability of the findings in Chapter 3, which is restricted to a sample of 19 organizations, and Chapter 4, which examines only one organization. The scenario questions in Chapter 5 provide more valid estimates of differ-ences between subsectors of the nonprofit sector.

A fourth limitation is the limited operationalization of welfare state indi-cators and civic engagement. This thesis only examined individual private donations. To asses a complete picture of the consequences of changes in public funding for nonprofit organizations, it would be helpful to incorporate the changes in income from foundations, corporations, fees and commercial activities. Besides fundraising as possible mediator in Chapter 3, it would be interesting to look at the effects of public funding for organizations in terms of financial stability, mission drift, employee policy and other governance strategies (Froelich, 1999; O’Regan & Oster, 2002; Verschuere & De Corte, 2014). Also, further research should examine voluntary contributions of time (volunteering), which can be another substitute for government fund-ing (Day & Devlin, 1996; Simmons & Emanuele, 2004; Stadelmann-Steffen, 2011).

On the independent variable side, unconditional financial government support was the only welfare state indicator in this thesis. Tax incentives, which might be perceived as a form of conditional government support, have large consequences on private giving, but are not considered here. Also, oth-er aspects of the welfare state, like de-commodification and institutional his-tory, can have different effects on civic engagement (Ferragina, 2017).

A fifth limitation is that the numbers of donors and the amounts donated to specific organizations or sectors under study were low. The low baselines make it less likely to observe substantial changes and to obtain statistical


significance for the results in the survey studies. Furthermore, citizens might behave differently in situations where the stakes are higher. Decisions with an earned reward of a few Euros (Chapter 4) and a hypothetical change in donations (Chapter 5) are not necessarily generalizable to real and larger giving decisions with actual consequences for one’s own budget. More re-search should test the external validity of the results from these studies.

IMPLICATIONS FOR THEORY

Sokolowski (2013) refers to Max Weber’s theory of social action to explain the relationship between government support and charitable donations. Weber (1922[1987]) argued that formal rationality takes over large parts of modern bureaucratic societies, driving out traditional, value-driven behav-ior. Philanthropy is an area where both rationalities still exist. Charitable do-nations are often an expression of values, which is consistent with Weber’s notion of Wertrationalität. Under specific circumstances, however, Zweck-rationalität can become more dominant, which makes donors consider the consequences of their contributions.

The literature on crowding-out is dominated by the individual-level ex-planations of altruism and warm-glow in economics, and the macro-level explanations of neo-institutionalism and welfare state theory in sociology. This dissertation goes beyond established explanations and proposes dif-ferent individual and contextual factors that drive the relationship between government support and charitable giving. Individuals, organizations and countries are heterogeneous, which makes it desirable or even necessary to formulate arguments for different effects in different contexts. Based on the insights in this dissertation, I provide four propositions for further theory building.

Proposition 1: Information is a prerequisite for government support to af-fect charitable donations. While behavioral experiments often aim to make predictions about macro effects, they generally fail to take the availability of information into account. Citizens adapt their giving behavior only when they are aware of external changes like changing government support.

Theories on civic engagement in relation to government activities can learn from public administration research, where an increasing number of studies pay attention to the effects of government transparency on trust, le-gitimacy and participation (Cucciniello et al., 2017). It has been proposed,


for example, that government transparency has stronger effects on political trust among citizens with lower levels of prior knowledge (Grimmelikhui-jsen & Meijer, 2014; Grimmelikhuijsen & Klijn, 2015; Tummers, Weske, Bou-wman, & Grimmelikhuijsen, 2015) and that citizens who initially underesti-mate levels of government spending change their policy preferences more strongly when exposed to actual information (Lergetporer et al., 2016).

Potential donors are always solicited with a different framing, leading to different decisions. The wording in the survey experiment in Chapter 4 in-cluded the phrase “The charities could use your support”, which might have primed an encouraging effect of the financial information that followed. Oth-er ways of framing might lead to different behavioral responses. It could be, for example, that information that shows (potential) losses following budget cuts are most likely to increase giving (Lee, Fraser, & Fillis, 2017).

Proposition 2: It is not the joy of giving, but specific intrinsic values that make donors less responsive to government support. This thesis finds the joy of giving to be positively related to crowding-out, which questions “warm glow” (Andreoni, 1990) as the most appropriate term when explaining the part of donations that is not responsive to government support. Theories on charitable giving should go beyond the concept of warm glow and distin-guish types of donors based on the values that drive them to give. This would provide a sociological basis for explaining why social groups are responsive to government policies and other groups are not, besides the economic mod-els that dominate the academic literature.

Some first ideas in this direction are provided by the findings in this the-sis, which suggest that some moral values are associated with crowding-out, but that citizens with high trust in government and high empathic concern are less sensitive to changes in government support. Responses to changing government policies might depend on how citizens perceive government re-sponsibilities. Who thinks that the government should provide services in a certain area is not likely to increase charitable giving to organizations in this field. This could explain why crowding-out is unlikely in the education subsector, which is, at least in the Netherlands, widely regarded as an area for which the government should take responsibility.

Proposition 3: There is a hierarchy of charitable causes. Philanthropy re-search traditionally appraises the diversity of the philanthropic sector. Ev-ery donor has his or her own preferred charity and every charity should be equally valued. However, some causes are more popular than others. Body & Breeze (2016) made an important contribution by exploring the concept


of “unpopular causes”, drawing attention to the question “why some causes appear to more easily attract widespread support whilst others struggle to raise any significant donated income” (Body & Breeze, 2016, p. 58).

In the Netherlands, the causes that are named as most important for soci-ety to some extent resemble the top causes to which people donate (Bekkers et al., 2017). The cause that is most often named as important – health – is also the area to which most people donate. Health donors often know some-one who suffered from a disease (Bekkers, 2008). The two causes that are least mentioned as being important for society – culture and sports – are also the subsectors with the lowest percentages of donors. This almost per-fectly correlates with the areas in which citizens desire government influ-ence. Education is an exception. 56% of the Dutch thinks that education and research are important for society, but only 12% donates to organizations in this area. In the Netherlands, where education is largely state-funded, it is likely that citizens think this is government responsibility. Furthermore, it could be that educational institutes have no need for fundraising because their level of services is on a sufficient level with the current (public) funding (Body & Breeze, 2016).

We might speak of a hierarchy in charitable causes. Of course, different people have different preferred causes (Bennett, 2003; Wiepking, 2010), but in general, areas of domestic service provision, like health, education and so-cial services, are currently perceived as the most important areas for society. These areas are characterized by an instrumental rationale, with a focus on the ultimate output of nonprofit organizations (Frumkin, 2002). Economic development and more extensive welfare state arrangements in social ser-vice areas could drive donor priorities towards other causes. When service provision is sufficiently funded, citizens may shift their attention to more “ex-pressive” fields like arts and culture. This is consistent with Maslow’s (1954) hierarchy of needs and Inglehart’s (1997) theory of the increasing social and political importance of postmaterialistic values in Western societies.

An additional argument is that different types of goods are provided in different parts of the nonprofit sector. Klamer (2004) argues that the arts is a common good rather than a public good, and that donations are an expres-sion of values rather than “giving”. For environment and international aid, it holds that the public goods provided (e.g. a clean environment, less world poverty) can only indirectly be enjoyed. Budget cuts have salient conse-quences for public service provision in one subsector (e.g. social protection, where the recipients can be your neighbors), whereas they remain largely


unnoticed in the other (e.g. international aid, where the recipients live in countries far away). Evans, Evans & Mayo (2017) argue that international aid and culture are luxury goods, with the total amount of donations increasing faster than a society’s income, while health and public benefits are inferior goods, with donations decreasing with increasing income. When a country gets richer, there is less demand for social protection because the bottom of society is better off while at the same time, the supply for “luxury goods” increase.

The causal relations between economic development, welfare state ar-rangements, individual values and charitable giving are complex and often reciprocal. Yet, large cultural changes have been detected from traditional societal structures to a modern rationality-bureaucratic society (Weber, 1922[1987]) and from modernization to postmodernization (Inglehart, 1997). If it is true that “expressive” causes become increasingly important, this would have large consequences for the future of the nonprofit sector.

There is surprisingly little attention to explanations of differences be-tween charitable causes in the philanthropy literature, and this dissertation provides important suggestions for theory building in this regard. The hi-erarchy of causes could explain why “charity begins at home”, meaning that most citizens have a preference for local rather than international charities (Knowles & Sullivan, 2017). It can also explain why countries with high ser-vice expenditures tend to have more donors in subsectors like culture and international aid (Chapter 1). For the future, this argument predicts that higher government expenditures in health, social services, education, inter-national aid, environment, nature conservation and animal welfare would in the end lead to increased donations to the most “unpopular” and “unimport-ant” causes, like culture and sports.

Proposition 4: Crowding-out is most likely to occur in the liberal welfare state regime type. Looking at the scatter plots in Chapter 1, it seems that the United States and the United Kingdom are outliers in the sense that they have moderate government spending and relatively high levels of philanthropic donations. In countries other than the US and the UK, a positive correlation is found between government support and charitable donations. This leads to the prediction that crowding-out depends on the welfare state context. No systematic differences are detected between welfare state regime types, which can be due to the small sample of countries. Still, the signs are in the expected direction, with crowding-out being most likely in the liberal wel-fare state regime type which is characterized by relatively low government


influence. Especially the US is a country with a culture of giving that is very different from European countries. Three lines of reasoning can explain why crowding-out is more likely in the US and other liberal welfare states than in the Netherlands.

First, the US has a highly professionalized fundraising regime (Wiepking et al., 2016). The hypothesis of “fundraising crowd-out”, which states that or-ganizations are less inclined to invest in fundraising when they obtain high-er revenues from governments, received considerable support in samples of American nonprofit organizations (Andreoni & Payne, 2003, 2011; Hughes et al., 2014) but none in the Dutch context (Chapter 3). It could be that or-ganizations in less developed fundraising regimes have lower capabilities to change their fundraising efforts as a response to changing government support. This could also partly explain differences between organizations within countries. Large organizations with professional fundraisers have more fundraising capacity than small, volunteer-based organizations. A re-cent evaluation of policy shifts in the Dutch cultural sector shows that larger organizations are better able to increase private income after budget cuts (Franssen & Bekkers, 2016).

More developed fundraising regimes are also characterized by more availability of information about nonprofit organizations. In the US, there are specialized media on the nonprofit sector like Nonprofit Quarterly and The Chronicle of Philanthropy, there are watchdogs like Charity Navigator, and there are many academic and non-academic research institutes with a unique focus on the voluntary sector, including the Urban Institute’s Nation-al Center for Charitable Statistics. Although similar initiatives exist in coun-tries like the Netherlands, they are not as large and professionalized as their American counterparts. Media, watchdogs and research centers may publish publicly available information about nonprofit organizations which is, as ar-gued in this thesis, a prerequisite for citizens to adapt their donations as a response to changing government support.

Second, citizens in liberal welfare state regimes have a more critical atti-tude towards government interventions (Andress & Heien, 2001; Svallfors, 1997). Citizens who are convinced that the government is responsible for reducing income differences and providing shelter for the homeless could be less inclined to donate to social service organizations, even if public funding is reduced. Citizens who think this is a shared responsibility between gov-ernments and nonprofits, on the other hand, are more likely to adapt their donations depending on the level of government intervention. If such atti-


tudes are partly shaped by the welfare state regime (Rothstein, 1998), this could explain why crowding-out differs between countries.

A third explanation could be that the marginal utility of donations is small-er in liberal welfare state regimes. It has been argued that the marginal in-crease in well-being derived from income is high for poor countries but di-minishes with economic prosperity (Inglehart, 2000). Welfare states differ in size and inclusiveness, and thus in their efficacy when aiming to alleviate problems like poverty, hunger and homelessness. If social needs are higher in more restrictive welfare states, an additional dollar of contributions to alleviate those needs has a higher value for recipients compared to countries with extensive welfare states and less urgent social needs. It is likely that donors are more inclined to compensate for changing government support when the stakes are higher.

An important question here is the extent to which differences between countries are due to the composition of their population or due to macro characteristics. Crowding-out effects vary across individuals within coun-tries, but also between countries. Differences between different types of wel-fare states have large historic continuity. To some extent, however, they may change over time, when a series of policy reforms adapts the structure of the welfare state. A structural reduction in the scope of the welfare state could lead to a different association between government support and charitable giving. The empirical estimates from the Netherlands in Chapters 3 to 5 are close to 0 across the board, which may become more strongly negative when policy choices brings the country closer to the liberal regime type.

IMPLICATIONS FOR RESEARCH

The academic literature on crowding-out is divided by methodological pref-erences. Researchers from behavioral economics carry out laboratory exper-iments in which participants are randomly assigned to different conditions of tax-funded government support. Given the problematic assumption of government support as an exogenous variable in survey and field research, such experimental designs are arguably the best way available to estimate causal effects.

If endogeneity explains why experimental findings differ from other find-ings, we would observe that regression models and specifications that ef-fectively deal with this issue produce stronger crowding-out estimates than


other regression models. The results of the meta-analysis in Chapter 2 do not confirm this line of reasoning. Neither fixed-effects or first-difference specifications nor the use of instrumental variables are robustly linked with stronger crowding-out. It is striking that similar variables are used as instru-ments for government support or private giving in different studies, which violates the assumptions of valid instrumental variables (Morgan & Winship, 2007). Researchers should be very careful in applying these techniques, and preferably use a range of different models and specifications to estimate the effect of government support in a certain dataset.

It is more likely that the assumption of full information, which is almost al-ways made in experimental research, explains the large difference between experimental and non-experimental findings. The research in this thesis shows that this assumption is not realistic in daily life. News media often do not cover actual changes in government support, which makes it unlikely that those changes lead to changing donor behavior.

Empirical crowding-out estimates from the Netherlands in Part II of this dissertation are close to 0. The results from the survey data confirm the aver-age null finding in non-experimental research as found in the meta-analysis (Chapter 2). The strongest coefficient is found in the information experiment in Chapter 4, with information about budget cuts leading to a 17% increase in the total amount donated. This is still far from the average experimen-tal finding of -.64 in previous studies (Chapter 2). An explanation could be that respondents in the survey experiment, in contrast to most laboratory experiments, are not aware of being part of an experiment. Also, the gov-ernment contribution is presented as an exogenous factor, while most lab experiments present it as a mandatory contribution from the participant’s endowment. Making respondents aware of a government tax is likely to en-courage crowding-out (Eckel et al., 2005).

The systematic analysis of findings and assumptions in different research contexts in this dissertation adds to the ongoing debate on the validity of findings from laboratory vs. field data (Camerer, 2015; Henrich et al., 2010; Levitt & List, 2007). Analyses on organizational data or survey data generally fail to find crowding-out not only because of causality issues (Payne, 2009), but also because they look at people who are not aware of every change in government policies. Media coverage follows incidental issues rather than long-term trends. This thesis strongly encourages future experimental and non-experimental research to take the role of information into consider-ation. One avenue of future research here is the framing of such information,


because different ways of phrasing can have large consequences for behav-ioral decisions (Meier, 2006). If research has the aim to make valid state-ments about social processes in daily life, the availability and framing of in-formation should be included in empirical research. Information as provided by mass media, social media, face to face contacts and fundraising materials can all be relevant in influencing giving decisions.

Besides information, crowding-out effects depend on the organizational context and individual characteristics. Future research should examine dif-ferences between nonprofit subsector as well as possible substitution be-tween subsectors (Sokolowski, 2013; Pennerstorfer & Neumayr, 2017) and between organizations (Ek, 2017; Reinstein, 2006, 2007). Also, it should examine individual heterogeneity in responsiveness to changes in govern-ment policies. All possible moderating variables of the relationship between government support and charitable donations, as provided in the Appendix, should be systematically tested to examine their relative importance.

New research designs could overcome the methodological divide. While laboratory experiments might provide more valid estimates of causality, their external validity is low. Studies that use organizational revenue data, on the other hand, have their own problems. They often use an aggregate measure of income from private sources, making it impossible to make strong statements about the behavior of individual donors. This thesis offers important innovations in terms of research design. In the field of empirical crowding-out research, it is the first to carry out a cross-country analysis using individual-level data on amounts donated to nonprofits organizations (Chapter 1), the first to use longitudinal survey data (Chapter 3) and one of the few that explicitly examine the role of information (Chapters 3 and 4).

IMPLICATIONS FOR THE NONPROFIT SECTOR

The main conclusion that government support will generally not crowd out private donations is a positive message for nonprofit organizations that are partly funded with public money, which is the case for many organizations in the Dutch nonprofit sector (Burger et al., 1999). The share of total reve-nues that comes from fundraising income is relatively small, and changes in charitable giving are not likely to compensate for reductions in government support. Other organizations, like the large health foundations that fund research on specific diseases, are almost exclusively dependent on private


income. Those organizations have to mobilize donors anyway to maintain their financial position, regardless of possible government subsidies. Dif-ferent revenue streams often exist relatively separate from each other and substitution effects are only likely to occur when public services are con-cerned that are not perceived as sole government responsibility, nor as sole nonprofit responsibility. There is only a small share of total nonprofit reve-nues which is vulnerable to substitution between government subsidies and private donations. In such areas, nonprofit organizations can employ differ-ent strategies to be prepared for governmental budget cuts. One way is to diversify the revenue mix, which make organizations more financially stable (Froelich, 1999).

The results of this thesis show two additional ways in which nonprofits can cope with budget cuts. First, reductions in government funding can be used in fundraising materials to show possible donors the urge of donating. This dissertation showed how an appeal that included information on a bud-get cut on a large health organization increased the proportion of donors by 22%. This may attract those who otherwise do not donate, for example be-cause they are relatively less empathic. An increase in the number of donors can be very fruitful in the long term, since donors who initially start with a small donation might develop into more generous donors (Sargeant & Lee, 2004). One fundraising strategy might be to show the consequences of bud-get cuts, because donations tend to increase when the (possible) losses after funding cuts are shown (Lee et al., 2017).

Second, fundraisers may target specific segments of possible donors who are more responsive to changes in government support. Those with a higher income, a higher education and more confidence in the charitable sector are more likely to increase donations after budget cuts. Those with high trust in the government, on the other hand, are less likely to compensate. Low-em-pathic citizens can be drawn into donating by informing them about decreas-ing subsidies. In the context of budget cuts, service provision organizations might reach out to possible donors who feel good by doing good, while orga-nizations from expressive areas might approach those with more resources by appealing to their previous engagement with the cause. By applying such donor profiles in fundraising, organizations can be more effective in using in-formation about budget cuts that appeal to certain social groups. While such strategies may attract new donors, they will not immediately lead to high-er donations. Establishing stable donor relationships, for example through “Friends of” organizations where members receive material and non-mate-


rial benefits, is necessary for increasing fundraising income in the long term.

IMPLICATIONS FOR POLICY

Two areas in which the Dutch government reduced spending during the last years are the arts and international development. In the arts sector, dona-tions decreased despite an increased tax deductibility of donations to cul-tural organizations (Franssen & Bekkers, 2016). In international aid, budget cuts were followed by decreasing donations (Chapter 3). Evaluations of the Big Society policy in the UK noted that the combined efforts of governments and the voluntary sector did not reach the people who needed it most (Civil Exchange, 2015). This does not make a positive picture for policy choices in which the state reduces budgets while aiming to give a larger role to non-profits in providing public services.

The results of this thesis show that overall, charitable donations will gen-erally not substitute public funding. Even if charitable giving increases, it can never make up for reductions in government support. The general advice for policy makers is clear: be careful with budget cuts if you aim to encour-age private funding of a flourishing nonprofit sector. Yet, in specific circum-stances, budget cuts can be used to draw citizens into donating. Consulting nonprofit organizations, examining the policy context and taking notice of the available research can shed light on the possible consequences of budget choices.

A first remark here is on equal collaborations between governments and private actors. Too often, nonprofit organizations are overlooked by central and local governments when revising policies (Schuyt, 2014), while there is often ample opportunity for collaborations with public, for-profit and nonprofit organizations to provide services (Ansell & Gash, 2008; Milward & Provan, 2003). In the city of Amsterdam, for example, an increasing num-ber of volunteer-based organizations like De Regenboog, UVV and Buren-netwerk employ buddy projects to help deprived people. With yearly gov-ernment funding for each organization separately, an increasing number of organizations working in the same area poses the risk of lower government funding per organization, which they are not likely to make up with private donations. Here, collaborations between local government bodies and dif-ferent nonprofit organizations can make a huge difference in the reduction of poverty and loneliness while using the available public money in a more


efficient way.A second remark is on the substitution between different policy fields.

Policy choices not only affect the area that they target, they also affect other parts of the nonprofit sector. Budget cuts may attract private donors who otherwise would have donated to other nonprofit organizations, which leads to undesired side-effects in seemingly unrelated areas. Interventions should not only target one area of society, but should be designed as inclusive pol-icies that consider all possible economic, social and cultural consequences.

Third, policy makers could be more creative in finding ways to encourage donations. In March 2017, the Dutch Minister for Foreign Trade and Devel-opment contributed 2 million Euros to the national campaign targeting fam-ine in Africa. Most likely, gestures like this help to signal the importance of a project, especially in an area like international aid. Government support may be used as “seed money” that encourages citizens to contribute to certain projects or organizations. An even more fruitful way to encourage private giving is through matching schemes, which have not been examined in this thesis but which are proven to be successful in increasing donations (Bek-kers, 2015; Eckel & Grossman, 2003, 2008).

CONCLUSION

Theo Schuyt (2010) refers to the core principles of the French revolution to characterize the role of philanthropy in a society. While the market and the state are driven by liberté and égalité respectively, the nonprofit sector is characterized by fraternité. While each of these three sectors have their own and unique logic and merit, they are strongly interwoven. Morally, the ques-tion which responsibilities can legitimately be left to philanthropic organiza-tions is a recurrent one. In the important bundle Philanthropy in Democratic Societies, Beerbohm (2016) warns for the lack of solidarity and democratic control when providing public responsibilities through philanthropic orga-nizations (the “free provider problem”). This makes a case for the argument that public good provision through philanthropy and through democratic governments both have a unique intrinsic value.

Welfare state expenditures are much higher than total amounts donated to nonprofit organizations. In the Netherlands and many other European countries, the provision of basic social services, like unemployment and ba-sic health insurances, are in the hands of the government, which is largely


undisputed among citizens and politicians. Moreover, there is large historic continuity in institutional arrangements

and practices. In the 17th and 18th century Dutch Republic, fundraising was deployed through door-to-door collections, recommendations by celebrities, social information about other people’s donations, incidental campaigns in the case of a natural disaster elsewhere, and charitable bequests in testa-ments (Teeuwen, 2014). It is surprising to notice that all these aspects are still relevant in today’s philanthropic sector.

Tocqueville (1970[1840]) observed fundamental differences between modern democracies in terms of the individual freedom to control one’s own environment. Small adaptations in welfare state arrangements in the course of a few years, however, are not likely to change the institutional practices or social preferences on which society is built. Across the board, charitable do-nations and government support are not substitutes and should not be treat-ed like that by policy makers and nonprofit professionals. There is ample room for governments to invest in public services while collaborating with nonprofit organizations. In most contexts, governments and their citizens are partners in giving rather than substitutes.

182

Summary

There is a wide array of studies dedicated to the idea that increasing levels of financial government support “crowd out” charitable giving, and that de-creasing government support leads to more giving. However, the validity of the crowding-out hypothesis is yet unsure. Much of the evidence comes from laboratory experiments in behavioral economics, while sociological studies tend to find zero correlation on average. This dissertation aims to bridge the gap between behavioral economics and sociology by identifying and exam-ining mediating and moderating factors that may explain the diversity of findings in the literature. It does so by adopting a multi-method approach, including both experimental and non-experimental research designs. In the crowding-out literature, it is the first to explore cross-country data on in-dividual amounts donated to nonprofit organizations, the first to examine longitudinal survey data, and the first to carry out a content analysis on news media.

Generally speaking, the empirical evidence for the crowding-out hypoth-esis is weak. Analyses on cross-country data, a panel survey and scenario questions find either no correlation or positive correlations between gov-ernment support and charitable donations across the board.

Data from the Netherlands do not show support for the argument that the fundraising behavior of organizations partly explains the association between government support and charitable donations. The analyses do suggest that in some contexts, government support serves as a signal of the quality of a charitable cause.

Charitable giving is most likely to substitute government support in the field of nature conservation. The strongest evidence for a positive associa-tion between government support and charitable donations, on the other hand, is found in the fields of environment, education and research, and in-ternational aid. Findings on health and social services are ambivalent. The results show substitution between subsectors, supporting the argument that government support in health and social services drive donors to “expres-sive” subsectors like international aid and the arts.

The availability of information is a prerequisite for people to change their giving behavior after changes in government funding. Although this might seem like stating the obvious, the role of information is largely overlooked in the academic debate. Changes in government support are often not covered

183

in news media, making it unlikely that they will affect donations. Providing citizens with information about actual budget cuts might increase an organi-zation’s donor base with over twenty percent.

Furthermore, not all social groups are equally responsive to changes in government support. Citizens who are higher educated, who have stronger prosocial values, who have more confidence in charitable organizations and who are more frequently solicited for charitable contributions are more like-ly to compensate for reductions in government funding. Those with a high empathic concern and high trust in the government are less likely to increase donations after government budget cuts. When faced with decreasing gov-ernment subsidies, fundraisers might use this information to target specific social groups in order to increase fundraising income.

For policy makers, the take-away message is clear: be careful with budget cuts if you aim to encourage private funding of a flourishing nonprofit sector. Even if charitable giving increases, it can never make up for reductions in government support. Yet, in specific circumstances, budget cuts can be used to draw citizens into donating. Consulting nonprofit organizations, examin-ing the policy context and taking notice of the available research can shed light on the possible consequences of budget choices.

In sum, empirical evidence for the crowding-out hypothesis is not con-vincing and it is only in specific circumstances that charitable donations can substitute public funding. Across the board, charitable donations and gov-ernment support are not substitutes and should not be treated like that by policy makers and nonprofit professionals.

184

Samenvatting

Er is veel onderzoek naar het argument dat toenemende overheidsuitgaven donaties aan goede doelen “verdrijven”, terwijl afnemende overheidsuit-gaven juist zouden leiden tot meer donaties. Het is maar de vraag of deze verdrijvingshypothese klopt. Het meeste bewijs komt van laboratoriumex-perimenten in de gedragseconomie, terwijl sociologische studies meestal geen verband vinden. Deze dissertatie probeert het gat tussen de gedragsec-onomie en de sociologie te overbruggen. Er zijn allerlei mediërende en mo-dererende factoren die de uiteenlopende resultaten kunnen verklaren. Het onderzoek in deze dissertatie bekijkt een aantal van deze factoren met ex-perimentele en niet-experimentele methoden. In de literatuur naar het ver-drijvingseffect van overheidsbestedingen is dit de eerste studie die gebruik maakt van internationale gegevens over donaties aan goede doelen op indi-vidueel niveau, de eerste die longitudinale enquêtes analyseert en de eerste die een inhoudsanalyse doet op verslaggeving in de media.

Over het algemeen is er weinig bewijs is voor de verdrijvingshypothese. Analyses met landenvergelijkende data, een longitudinale enquête en sce-nariovragen laten onder de streep ofwel geen, ofwel een positief verband zien.

In de Nederlandse cijfers zijn geen aanwijzingen te vinden dat organisa-ties meer investeren in fondsenwerving als ze minder overheidssubsidies ontvangen. Wel laten de analyses zien dat overheidssteun in sommige geval-len kan fungeren als een signaal dat een goed doel het waard is om gesteund te worden.

Natuurbescherming is het terrein waar substitutie tussen overheidsbeste-dingen en donaties het meest waarschijnlijk is. Het sterkste bewijs voor een positief verband is te zien op de terreinen milieu, onderwijs en onderzoek, en internationale hulp. De bevindingen voor gezondheid en sociale doelen zijn niet eenduidig. De resultaten laten zien dat er substitutie tussen sectoren is. Het lijkt erop dat overheidsbestedingen op het gebied van gezondheid en so-ciale voorzieningen donateurs naar meer “expressieve” doelen drijven, zoals internationale hulp of kunst en cultuur.

Mensen moeten wel weten dat overheidsbestedingen veranderen voordat ze hun donatie daarop kunnen aanpassen. Dit lijkt een open deur, maar de rol van informatie is grotendeels onbenoemd gebleven in het academische debat. Veranderingen in overheidsbestedingen worden vaak niet genoemd

185

in de media, waardoor het onwaarschijnlijk is dat ze een direct effect heb-ben op geefgedrag. Als mensen op de hoogte worden gebracht van bestaande bezuinigingen kan dat leiden tot een toename van het aantal donateurs met zo’n twintig procent.

Verder blijken niet alle sociale groepen op dezelfde manier te reageren op veranderingen in overheidsbestedingen. Mensen die een hogere opleiding hebben afgerond, die sterkere prosociale waarden hebben, die meer ver-trouwen hebben in goededoelenorganisaties en die vaker gevraagd worden om te geven, zijn eerder geneigd om bezuinigingen te compenseren. Mensen die meer empathisch zijn en meer vertrouwen hebben in de overheid zijn juist minder geneigd hun donaties te verhogen na overheidsbezuinigingen. Organisaties die geconfronteerd worden met afnemende overheidssubsidies kunnen dit soort donorprofielen gebruiken in de fondsenwerving.

Het advies voor beleidsmakers is duidelijk: wees voorzichtig met bezuin-igingen als je de particuliere financiering van een gezonde filantropische sector wilt bevorderen. Zelfs als particuliere giften toenemen kunnen ze de afnemende overheidsbijdragen nooit volledig compenseren. In specifieke gevallen, echter, kunnen burgers geactiveerd worden om te geven. Door te overleggen met organisaties in de filantropische sector, de beleidscontext zorgvuldig te bestuderen en nota te nemen van het beschikbare onderzoek, kan duidelijk worden wat de mogelijke gevolgen zijn van verschillende bud-gettaire keuzes.

Al met al kunnen we stellen dat het bewijs voor de verdrijvingshypothese weinig overtuigend is. Alleen in specifieke gevallen kunnen donateurs in het gat springen dat de overheid achterlaat. Over het algemeen zijn filantropie en overheidsbestedingen geen substituten, en dat is een belangrijk inzicht voor beleidsmakers en professionals in de filantropische sector.

186

Acknowledgements

I could not have wished for a better supervising team than the one I had for this dissertation. René Bekkers inspired me with his intelligent views on giving and volunteering in the Netherlands and abroad. His views on respon-sible research and open science guided this dissertation research. Marjolein Broese van Groenou was always there to comment on pieces of research, even if she was not involved as a co-author. Her comments on theoretical angles, methodological choices and writing styles were always sharp and constructive. Beate Völker immediately decided to help me out by creating a collaboration between her university and ours, which guaranteed the con-tinuations of the project. With great enthusiasm and useful comments she provided daily supervision from February to July 2016.

Besides me and my supervising team, many colleagues contributed to the work in this dissertation.

I thank René Bekkers, Evelien Boonstoppel, Floor de Nooij, Suzanne Felix, Saskia Franssen, Barbara Gouwenberg, Barry Hoolwerf, Danique Karamat Ali, Elly Mariani, Tjeerd Piersma, Brigitte Schouten, Theo Schuyt, Claire van Teunenbroek, Dave Verkaik and other (former) colleagues at Vrije Universi-teit (VU) Amsterdam for all the conversations, outings and great work atmo-sphere.

Colleagues at the Amsterdam Institute for Social Science Research at the University of Amsterdam (UvA) provided an inspiring place to work during the first half of 2016. Despite the unusual situation, the two directors of op-erations Hanneke Reuling (VU) and José Komen (UvA) did a great job in mak-ing the collaboration between the two universities work.

Although the results did not end up in this dissertation, Jochem Miggel-brink, Mireille van der Meij and Christina Ceulemans at the Amsterdam University Fund generously helped us with developing a field experiment among alumni.

Years earlier, Brian Burgoon and Tom van der Meer pushed me in the right direction when supervising my Bachelor and Master thesis, respectively.

I am grateful to Femida Handy and her colleagues at the School of Social Policy & Practice at the University of Pennsylvania for their great hospitality from September to November 2016. A special thanks to Andrea Nurse for helping me out with the practical issues around my research visit.

Pamala Wiepking not only provided great substantive and personal sup-

187

port, I also benefited from her important work on the Individual Interna-tional Philanthropy Database (IIPD) which is used in Chapter 1. With this project, Pamala and Femida made an important contribution to the field and were awarded ARNOVA’s Virginia A. Hodgkinson Research Book Prize. The IIPD was created with the help of René Bekkers, Steffen Bethman, Oonagh Breen, Beth Breeze, Chris Einolf, Chulhee Kang, Hagai Katz, Michael D. Lay-ton, Kuang-Ta Lo, Irina Mersianova, Michaela Neumayr, Una Osili, Anne Bir-gitta Pessi, Karl-Henrik Sivesind, Wendy Scaife, and Naoto Yamauchi. A large part of the data coding and synchronizing was done by Sohyun Park.

Michaela Neumayr contributed a lot to Chapter 1, and was a great col-league to work with. I consider Michaela and myself both as first authors on this article.

I benefited a lot from conversations at different occasions with Rich Stein-berg, Mark Ottoni-Wilhelm and other colleagues from the Lilly School of Philanthropy at Indiana University–Purdue University Indianapolis.

I thank Peter Frumkin for the great pieces of advice on my work and ca-reer, and the other 2014 Penn Social Impact Doctoral Fellows, Sabith Khan, Mirae Kim, Bethany Slater, Amanda Stewart and Rachel Wright, for their ex-tensive and useful comments.

Roza Meuleman, Gerbert Kraaykamp, Marion Wittenberg and other par-ticipants of the 5th Dutch ESS Workshop provided a great platform to think about cross-national studies like the one in Chapter 1.

Michael Berbaum and other teachers in the ICPSR Summer Program in Quantitative Methods of Social Research considerably contributed to im-proving my skills and knowledge.

Mark Koetse and Boris van Zanten helped us with developing the me-ta-analysis in Chapter 2.

Ad Graaman from the Dutch Central Bureau on Fundraising (CBF) provid-ed the organizational data for Chapter 3.

Other colleagues who generously devoted time to comment on my work include Jeff Brudney, Maria Radyati, Adalbert Evers, Judith van der Veer and Hans Keman.

Last but not least, a special word for Theo Schuyt. Although he was not part of my team of supervisors, his views on the relationship between the state and the nonprofit sector inspired the context of this dissertation. As founder, former director and part-time professor, his never-lasting enthusi-asm makes the Center for Philanthropic Studies work.

188

Funding

Arjen de Wit was partly supported by different travel grants from the Gradu-ate School of Social Sciences at Vrije Universiteit (VU) Amsterdam.

Femida Handy received partial support for this research from the School of Social Policy and Practice, University of Pennsylvania.

Pamala Wiepking was partly supported by the Netherlands Organization for Scientific Research (NWO).

René Bekkers received support from the Van der Gaag Foundation of the Royal Netherlands Academy of Sciences.

Funding for the Giving in the Netherlands Panel Survey was provided by the Ministry of Justice, the Ministry of Health, Wellbeing and Sports, and various other Ministries of the Netherlands.

189

Appendix

Table A1: Possible moderators of the crowding-out effect

Tested in meta-analysis Not tested in meta-analysisOften distinguished in empirical studies

• Data source (experimental/non-experimental)

• Sample country• Beneficiary of

government support (subsidies to organizations/direct expenditures)

• Regression specification (FE/FD/other)

• Use of instrumental variables

Sometimes distinguished in empirical studies

• Level of government (central/lower)

• Nonprofit sectors• Non-linear effect of

government support• Private donor

(individual/company/foundation/other)

Often not distinguished in empirical studies

• Types of government support (lump-sum grants/matching grants/contracts/purchase of services/vouchers)

• Tax salience• Number of other

donors• Different types of

public goods• Linearity of public good

cost function• Number of initial non-

donors• Substitution between

organizations

190

Tabl

e A

2: S

tudi

es in

the

met

a-an

alys

is

Refe

renc

e#

Data

Coun

-tr

ySe

ctor

(s)

Gove

rnm

ent

supp

ort

Leve

l of g

over

n-m

ent

Find

ing(

s)M

ean

effe

ct

size

est

.Ab

ram

s & S

chm

itz, 1

978

3Ar

chiv

alUS

ACo

mbi

ned

Expe

nditu

res

Fede

ral/

low

erCr

owdi

ng-o

ut-0

.300

Abra

ms &

Sch

mitz

, 198

43

Arch

ival

USA

Com

bine

dEx

pend

iture

sLo

wer

Crow

ding

-out

-0.2

80An

dreo

ni, 1

993

21La

b ex

p.US

An/

aTa

xn/

aCr

owdi

ng-o

ut-0

.716

Andr

eoni

& P

ayne

, 201

114

Arch

ival

USA

Com

bine

dSu

bsid

ies

Com

bine

dM

ixed

-0.7

60Be

cker

& L

inds

ay, 1

994

3Ar

chiv

alUS

AEd

ucat

ion

Subs

idie

sLo

wer

Crow

ding

-out

-0.8

70Bl

anco

et a

l., 2

012

6La

b ex

p.Sp

ain

Envi

ron-

men

tTa

xn/

aM

ixed

-0.4

70

Bolto

n &

Kat

ok, 1

998

2La

b ex

p.US

An/

aTa

xn/

aCr

owdi

ng-o

utn/

aBö

nke,

Mas

sarr

at-M

ashh

adi,

&

Siel

aff,

2013

2Ar

chiv

alGe

rma-

nyCo

mbi

ned

Expe

nditu

res

Com

bine

dCr

owdi

ng-o

utn/

a

Borg

onov

i, 20

0611

Arch

ival

USA

Cultu

reSu

bsid

ies

Fede

ral/

low

er/

com

bine

dCr

owdi

ng-in

n/a

Broo

ks, 1

999

1Ar

chiv

alUS

ACu

lture

Subs

idie

sCo

mbi

ned

Crow

ding

-in0.

075

Broo

ks, 2

000a

5Ar

chiv

alUS

ACu

lture

Subs

idie

sCo

mbi

ned

Crow

ding

-inn/

aBr

ooks

, 200

0b7

Arch

ival

USA

Mix

edEx

pend

iture

sFe

dera

l/lo

wer

Mix

ed0.

089

Broo

ks, 2

003a

2Ar

chiv

alUS

ACo

mbi

ned

Subs

idie

sCo

mbi

ned

Mix

ed1.

535

Broo

ks, 2

003b

3Ar

chiv

alUS

ACu

lture

Subs

idie

sCo

mbi

ned

Crow

ding

-in0.

433

Brun

ner,

1997

1Ar

chiv

alUS

ACu

lture

Subs

idie

sCo

mbi

ned

Crow

ding

-out

-0.0

75Br

unne

r & S

onst

elie

, 200

36

Arch

ival

USA

Educ

atio

nSu

bsid

ies

Com

bine

dM

ixed

-0.1

35Ca

llen,

199

43

Arch

ival

Cana

daH

ealth

Subs

idie

sCo

mbi

ned

Crow

ding

-out

n/a

Chan

et a

l., 1

996

16La

b ex

p.US

An/

aTa

xn/

aCr

owdi

ng-o

ut-0

.574

Chan

et a

l., 2

002

5La

b ex

p.Ca

nada

n/a

Tax

n/a

Crow

ding

-out

-0.7

15

191

Tabl

e A

2 (c

ontin

ued)

Refe

renc

e#

Data

Coun

-tr

ySe

ctor

(s)

Gove

rnm

ent

supp

ort

Leve

l of g

over

n-m

ent

Find

ing(

s)M

ean

effe

ct

size

est

.Du

ncan

, 199

91

Surv

eyUS

ACo

mbi

ned

Expe

nditu

res

Low

erCr

owdi

ng-o

ut-0

.234

Ecke

l et a

l., 2

005

2La

b ex

p.US

ACo

mbi

ned

Mix

edn/

aCr

owdi

ng-o

ut-0

.678

Ferr

is &

Wes

t, 20

033

Arch

ival

USA

Soci

alEx

pend

iture

sCo

mbi

ned

Crow

ding

-out

n/a

Galb

iati

& V

erto

va, 2

008

16La

b ex

p.Ita

lyn/

aTa

xn/

aCr

owdi

ng-o

ut-0

.628

Galb

iati

& V

erto

va, 2

014

2La

b ex

p.Ita

lyn/

aTa

xn/

aCr

owdi

ng-o

ut-0

.791

Garr

ett &

Rhi

ne, 2

010

41Ar

chiv

alUS

AH

ealth

/ed

ucat

ion/

soci

al/

com

bine

d

Expe

nditu

res

Fede

ral/

low

er/

com

bine

dCr

owdi

ng-o

utn/

a

Gron

berg

et a

l., 2

012

1La

b ex

p.US

An/

aTa

xn/

aCr

owdi

ng-o

ut-0

.900

Güth

et a

l., 2

006

3La

b ex

p.Au

stri

aSo

cial

Tax

n/a

Crow

ding

-out

-0.3

00H

erze

r & N

unne

nkam

p, 2

013

2A

rchi

val

USA

Inte

rna-

tiona

lSu

bsid

ies

Com

bine

dC

row

ding

-in0.

128

Hsu

, 200

86

Lab

exp.

Taiw

ann/

aTa

xn/

aC

row

ding

-out

-2.1

88H

ughe

s & L

ukse

tich,

199

98

Arc

hiva

lU

SACu

lture

Subs

idie

sFe

dera

l/low

erM

ixed

3.01

1H

ughe

s et a

l., 2

014

4A

rchi

val

USA

Cultu

reSu

bsid

ies

Com

bine

dM

ixed

1.59

0H

unge

rman

, 200

52

Arc

hiva

lU

SAC

ombi

ned

Expe

nditu

res

Low

erC

row

ding

-out

-0.0

36Is

aac &

Nor

ton,

201

32

Lab

exp.

USA

n/a

Tax

n/a

Mix

ed-0

.906

Kha

nna

& S

andl

er, 2

000

8A

rchi

val

UK

Mix

edSu

bsid

ies

Com

bine

dM

ixed

0.95

6K

hann

a et

al.,

199

55

Arc

hiva

lU

KC

ombi

ned

Subs

idie

sC

ombi

ned

Mix

ed0.

070

Kim

& V

an R

yzin

, 201

44

Surv

ey

exp.

USA

Cultu

reSu

bsid

ies

Fede

ral/c

om-

bine

dM

ixed

n/a

192

Tabl

e A

2 (c

ontin

ued)

Refe

renc

e#

Data

Coun

-tr

ySe

ctor

(s)

Gove

rnm

ent

supp

ort

Leve

l of g

over

n-m

ent

Find

ing(

s)M

ean

effe

ct

size

est

.Ki

ngm

a, 1

989

10Su

rvey

USA

Cultu

reSu

bsid

ies

Com

bine

dM

ixed

-0.0

17Ki

ngm

a &

McC

lella

nd, 1

995

3Su

rvey

USA

Cultu

reSu

bsid

ies

Com

bine

dCr

owdi

ng-o

ut-0

.000

Kono

w, 2

010

1La

b ex

p.

USA

n/a

Expe

nditu

res

n/a

Crow

ding

-out

-0.1

75Ko

reno

k et

al.,

201

22

Lab

exp.

US

An/

aEx

pend

iture

sn/

aCr

owdi

ng-o

ut-0

.223

Kore

nok

et a

l., 2

014

3La

b ex

p.US

An/

aTa

xn/

aCr

owdi

ng-o

ut-0

.261

Krop

f & K

nack

, 200

31

Arch

ival

USA

Cultu

reSu

bsid

ies

Com

bine

dCr

owdi

ng-in

0.04

0Li

lley

& S

loni

m, 2

014

2La

b ex

p.Au

stra

-lia

Inte

rna-

tiona

lTa

xn/

aM

ixed

-0.2

37

Lucc

asen

, 201

21

Lab

exp.

US

An/

aTa

xn/

aCr

owdi

ng-o

ut-0

.972

Luks

etic

h &

Lan

ge, 1

995

6Ar

chiv

alUS

ACu

lture

Subs

idie

sCo

mbi

ned

Mix

ed-0

.138

Man

zoor

& S

trau

b, 2

005

3Su

rvey

USA

Cultu

reSu

bsid

ies

Com

bine

dM

ixed

0.05

9M

arcu

ello

& S

alas

, 200

02

Arch

ival

Spai

nIn

tern

a-tio

nal

Subs

idie

sFe

dera

l/lo

wer

Crow

ding

-in0.

039

Mar

cuel

lo &

Sal

as, 2

001

8Ar

chiv

alSp

ain

Inte

rna-

tiona

lSu

bsid

ies

Fede

ral/

low

erM

ixed

n/a

Nel

son

& G

azle

y, 20

146

Arch

ival

USA

Educ

atio

nEx

pend

iture

sFe

dera

l/lo

wer

Mix

edn/

aN

unne

nkam

p &

Öhl

er, 2

012

7Ar

chiv

alUS

AIn

tern

a-tio

nal

Subs

idie

sCo

mbi

ned

Crow

ding

-inn/

a

Okte

n &

Wei

sbro

d, 2

000

10Ar

chiv

alUS

ACu

lture

/so

cial

/he

alth

/ed-

ucat

ion

Subs

idie

sCo

mbi

ned

Mix

edn/

a

O'Re

gan

& O

ster

, 200

22

Arch

ival

USA

Com

bine

dSu

bsid

ies

Com

bine

dCr

owdi

ng-o

utn/

a

193

Tabl

e A

2 (c

ontin

ued)

Refe

renc

e#

Data

Coun

-tr

ySe

ctor

(s)

Gove

rnm

ent

supp

ort

Leve

l of g

over

n-m

ent

Find

ing(

s)M

ean

effe

ct

size

est

.Pa

qué,

198

64

Arch

ival

Germ

a-ny

Com

bine

dEx

pend

iture

sCo

mbi

ned

Mix

edn/

a

Payn

e, 1

998

10Ar

chiv

alUS

ASo

cial

Subs

idie

sCo

mbi

ned

Mix

ed-0

.408

Payn

e, 2

001

29Ar

chiv

alUS

AEd

ucat

ion

Subs

idie

sFe

dera

l/co

m-

bine

dM

ixed

0.34

6

Posn

ett &

San

dler

, 198

911

Arch

ival

UKRe

ligio

n/he

alth

/in

tern

a-tio

nal/

com

bine

d

Subs

idie

sFe

dera

l/lo

wer

Mix

edn/

a

Reec

e, 1

979

7Su

rvey

USA

Relig

ion/

educ

atio

n/so

cial

/co

mbi

ned

Expe

nditu

res

Com

bine

dM

ixed

-0.0

09

Rees

on &

Tis

dell,

200

84

Lab

exp.

Aust

ra-

lian/

aTa

xn/

aM

ixed

-0.8

10

Riba

r & W

ilhel

m, 2

002

8Ar

chiv

alUS

AIn

t. ai

dSu

bsid

ies

Com

bine

dM

ixed

-0.0

79Sa

v, 20

1215

Arch

ival

USA

Educ

atio

nSu

bsid

ies

Fede

ral/

low

erM

ixed

-0.1

54Sc

hiff,

198

55

Surv

eyUS

ASo

cial

/co

mbi

ned

Subs

idie

s/ta

xLo

wer

Mix

ed-0

.060

Smith

, 200

717

Arch

ival

USA

Cultu

reSu

bsid

ies

Com

bine

dM

ixed

0.73

6So

kolo

wsk

i, 20

133

Arch

ival

Cros

sCo

mbi

ned

Subs

idie

sFe

dera

lCr

owdi

ng-in

n/a

Song

& Y

i, 20

112

Arch

ival

USA

Cultu

reSu

bsid

ies

Com

bine

dCr

owdi

ng-o

ut-0

.500

Stei

nber

g, 1

985

3Su

rvey

UKCo

mbi

ned

Expe

nditu

res

Com

bine

dCr

owdi

ng-o

ut-0

.006

194

Tabl

e A

2 (c

ontin

ued)

Refe

renc

e#

Data

Coun

-tr

ySe

ctor

(s)

Gove

rnm

ent

supp

ort

Leve

l of g

over

n-m

ent

Find

ing(

s)M

ean

effe

ct

size

est

.Su

tter

& W

eck-

Han

nem

ann,

200

415

Lab

exp.

Aust

ria

n/a

Tax

n/a

Crow

ding

-out

-0.7

69Ti

nkel

man

& N

eely

, 201

17

Arch

ival

USA

Com

bine

dSu

bsid

ies

Com

bine

dM

ixed

-0.0

53W

einb

latt

, 199

22

Arch

ival

Isra

elCo

mbi

ned

Subs

idie

sCo

mbi

ned

Crow

ding

-out

-0.4

00Ye

tman

& Y

etm

an, 2

003

24Ar

chiv

alUS

AM

ixed

Subs

idie

sCo

mbi

ned

Mix

ed-0

.150

Yetm

an &

Yet

man

, 201

33

Arch

ival

USA

Com

bine

dSu

bsid

ies

Com

bine

dCr

owdi

ng-in

0.01

9

195

References

A

Abraham, K.G., Helms, S., & Presser, S. (2009). How Social Processes Distort Measurement: The Impact of Survey Nonresponse on Estimates of Volun-teer Work in the United States 1. American Journal of Sociology, 114(4), 1129-1165.

Abrams, B.A., & Schmitz, M.D. (1978). The ‘crowding-out’ effect of govern-mental transfers on private charitable contributions. Public Choice, 33(1), 29-39.

Abrams, B. A., & Schmitz, M.D. (1984). The Crowding-out Effect of Govern-mental Transfers on Private Charitable Contributions: Cross-Section Evi-dence. National Tax Journal, 37(4), 563-568.

Andreoni, J. (1988). Privately Provided Public Goods in a Large Economy: The Limits of Altruism. Journal of Public Economics, 35(1), 57-73.

Andreoni, J. (1989). Giving with Impure Altruism: Applications to Charity and Ricardian Equivalence. Journal of Political Economy, 97(6), 1447–1458.

Andreoni, J. (1990). Impure Altruism and Donations to Public Goods: A The-ory of Warm-Glow Giving. Economic Journal, 100(401), 464–477.

Andreoni, J. (1993). An Experimental Test of the Public-Goods Crowding-Out Hypothesis. American Economic Review, 83, 1317-1327.

Andreoni, J, & Payne, A.A. (2003). Do Government Grants to Private Charities Crowd Out Giving or Fund-raising? American Economic Review, 93, 792-812.

Andreoni, J., & Payne, A.A. (2011). Is crowding out due entirely to fundrais-ing? Evidence from a panel of charities. Journal of Public Economics, 95, 334-343.

Andress, H.J., & Heien, T. (2001). Four Worlds of Welfare State Attitudes? A Comparison of Germany, Norway and the United States. European Socio-logical Review, 17(4), 337–356.

Angrist, J.D., & Pischke, J.S. (2009). Mostly Harmless Econometrics: An Empir-icist’s Companion. Princeton: Princeton University Press.

Anheier, H.K., & Toepler, S. (1999). Philanthropic Foundations: An Interna-tional Perspective. In: H.K. Anheier & S. Toepler (Eds.) Private Funds, Pub-lic Purpose: Philanthropic Foundations in International Perspective. New York: Kluver Academic/Plenum Publishers, pp. 3-23.

196

Ansell, C., & Gash, A. (2008). Collaborative Governance in Theory and Prac-tice. Journal of Public Administration Research and Theory, 18(4), 543-571.

Arts, W., & Gelissen, J. (2001). Welfare States, Solidarity and Justice Princi-ples: Does the Type Really Matter? Acta Sociologica, 44(4), 283-299.

B

Barth, A. (2011). Meer marktwerking maakt culturele sector ondernemend-er. Reformatorisch Dagblad, December 29.

Batson, C.D. (2010). Altruism in Humans. New York: Oxford University Press.Baumol, W.J. (1996). Children of Performing Arts, the Economic Dilemma:

The Climbing Costs of Health Care and Education. Journal of Cultural Eco-nomics, 20(3), 183-206.

Becker, E., & Lindsay, C.M. (1994). Does the Government Free Ride? Journal of Law & Economics, 37(1), 277-296.

Beerbohm, E. (2016). The Free-Provider Problem: Private Provision of Pub-lic Responsibilities. In: R. Reich, C. Coredelli & L. Bernholz (Eds.) Philan-thropy in Democratic Societies: History, Institutions, Values. Chicago: The University of Chicago Press, pp. 207-225.

Bekkers, R. (2003). Trust, accreditation, and philanthropy in the Nether-lands. Nonprofit and Voluntary Sector Quarterly, 32(4), 596-615.

Bekkers, R. (2005a). Words and Deeds of Generosity: Are Decisions About Real and Hypothetical Money Really Different? ICS/Department of Sociology. Utrecht: Utrecht University.

Bekkers, R. (2005b). It’s Not All in the Ask: Effects and Effectiveness of Recruit-ment Strategies Used by Nonprofits in The Netherlands. Paper presented at the 34th Arnova Annual Conference, Washington DC.

Bekkers, R. (2007). Measuring Altruistic Behavior in Surveys: The All-Or-Nothing Dictator Game. Survey Research Methods, 1(3), 139-144.

Bekkers, R. (2008). Straight From the Heart. In: S. Chambré and M. Goldner (Eds.) Advances in Medical Sociology, Volume 10: Patients, Consumers and Civil Society: US and International Perspectives. Emerald Group Publishing, pp. 197-221.

Bekkers, R. (2013a). De maatschappelijke betekenis van filantropie (inaugu-ral lecture). Amsterdam: Vrije Universiteit (VU) Amsterdam.

Bekkers, R. (2013b). De vermogende gever. In: T. Schuyt, B. Gouwenberg, & R. Bekkers (Eds.) Geven in Nederland 2013: Giften, Nalatenschappen, Spon-

197

soring en Vrijwilligerswerk. Amsterdam: Reed Business, pp. 205-226.Bekkers, R. (2015). When and Why Matches are More Effective Subsidies

Than Rebates. In: C. Deck, E. Fatas & T. Rosenblat (Eds.) Research in Ex-perimental Economics, Volume 18: Replication in Economic Experiments. Emerald Group Publishing.

Bekkers, R. (2016). Regional Differences in Philanthropy. In: J. Harrow, T. Jung, & S. Phillips (Eds.) Routledge Companion to Philanthropy. London: Routledge, pp. 124-138.

Bekkers, R., Boonstoppel, E., & De Wit, A. (2016). Giving in the Netherlands Panel Survey - User Manual (version 2.5). Amsterdam: Center for Philan-thropic Studies, Vrije Universiteit (VU) Amsterdam.

Bekkers, R. & De Wit, A. (2014). Look who’s crowding-out! Paper presented at the ISTR Conference, Muenster, Germany.

Bekkers, R., & Schuyt, T. (2008). And who is your neighbor? Explaining de-nominational differences in charitable giving and volunteering in the Netherlands. Review of Religious Research, 74-96.

Bekkers, R., Schuyt, T., & Gouwenberg, B. (2017) Geven in Nederland 2017. Amsterdam: Lenthe.

Bekkers, R., & Wiepking, P. (2006). To Give or Not to Give, That Is the Ques-tion: How Methodology Is Destiny in Dutch Giving Data. Nonprofit and Vol-untary Sector Quarterly, 35(3), 533-540.

Bekkers, R., & Wiepking, P. (2011a). Who gives? A literature review of pre-dictors of charitable giving part one: religion, education, age and socialisa-tion. Voluntary Sector Review, 2(3), 337-365.

Bekkers, R., & Wiepking, P. (2011b). A Literature Review of Empirical Studies of Philanthropy: Eight Mechanisms that Drive Charitable Giving. Nonprofit and Voluntary Sector Quarterly, 40(5), 924-973.

Bekkers, R. & Wiepking, P. (2011c). Accuracy of Self-reports on Donations to Charitable Organizations. Quality & Quantity, 45(6): 1369-1383.

Bekkers, R. & Wilhelm, M.O. (2016). Principle of Care and Giving to Help Peo-ple in Need. European Journal of Personality, 30(3), 240-257.

Bennett, R. (2003). Factors underlying the inclination to donate to particular types of charity. International Journal of Nonprofit and Voluntary Sector Marketing, 8(1), 12–29.

Blanco, E., Lopez, M.C., & Coleman, E.A. (2012). Voting for Environmental Do-nations: Experimental Evidence from Majorca, Spain. Ecological Econom-ics, 75, 52-60.

Bode, I. (2006). Disorganized welfare mixes: Voluntary agencies and new

198

governance regimes in Western Europe. Journal of European Social Policy, 16, 346–59.

Body, A., & Breeze, B. (2016). What are ‘unpopular causes’ and how can they achieve fundraising success? International Journal of Nonprofit and Volun-tary Sector Marketing, 21(1), 57–70.

Bolton, G.E., & Katok, E. (1998). An experimental test of the crowding out hy-pothesis: The nature of beneficent behavior. Journal of Economic Behavior & Organization, 37(3), 315-331.

Bönke, T., Massarrat-Mashhadi, N., & Sielaff, C. (2013). Charitable giving in the German welfare state: fiscal incentives and crowding out. Public Choice, 154(1-2), 39-58.

Bonoli, G., George, V., & Taylor-Gooby, P. (2000). European Welfare Futures. Cambridge: Blackwell Publishers.

Borenstein, M., Hedges, L.V., Higgins, J.P.T., & Rothstein, H.R. (2009). Introduc-tion to Meta-Analysis. Chichester: Wiley.

Borgonovi, F. (2006). Do public grants to American theatres crowd-out pri-vate donations? Public Choice, 126(3-4), 429-451.

Bredtmann, J. (2016). Does Government Spending Crowd Out Voluntary Labor and Donations? IZA World of Labor Working Paper. Bonn: IZA.

Brooks, A.C. (1999). Do Public Subsidies Leverage Private Philanthropy for the Arts? Empirical Evidence on Symphony Orchestras. Nonprofit and Vol-untary Sector Quarterly, 28(1), 32-45.

Brooks, A.C. (2000a). Public Subsidies and Charitable Giving: Crowding Out, Crowding In, or Both? Journal of Policy Analysis and Management, 19(3), 451–464.

Brooks, A.C. (2000b). Is There a Dark Side to Government Support for Non-profits? Public Administration Review, 60(3), 211-218.

Brooks, A.C. (2003a) Do Government Subsidies To Nonprofits Crowd Out Do-nations or Donors? Public Finance Review, 31(2), 166-179.

Brooks, A.C. (2003b). Taxes, Subsidies, and Listeners Like You: Public Policy and Contributions to Public Radio. Public Administration Review, 63, 554-561.

Brooks, A.C. (2004). The Effects of Public Policy on Private Charity. Adminis-tration & Society, 36(2), 166-185.

Brown, E. (2005). College, Social Capital, and Charitable Giving. In: A.C. Brooks (Ed.) Gifts of Time and Money: The Role of Charity in America’s Com-munities. Lanham: Rowman & Littlefield Publishers, pp. 185-204.

Brown, E., & Ferris, J. M. (2007). Social capital and philanthropy: An analysis

199

of the impact of social capital on individual giving and volunteering. Non-profit and Voluntary Sector Quarterly, 36(1), 85-99.

Brown, P. H., & Minty, J. H. (2008). Media coverage and charitable giving after the 2004 Tsunami. Southern Economic Journal, 75(1), 9-25.

Brunner, E.J. (1997). An empirical test of neutrality and the crowding-out hypothesis. Public Choice, 92, 261–279.

Brunner, E., & Sonstelie, J. (2003). School finance reform and voluntary fiscal federalism. Journal of Public Economics, 87(9-10), 2157-2185.

Bryan, M.L., & Jenkins, S.P. (2016). Multilevel Modelling of Country Effects: A Cautionary Tale. European Sociological Review, 32(1), 3-22.

Burger, A., Dekker, P., Toepler, S., Anheier, H. & Salamon, L. (1999). The Neth-erlands: Key Features of the Dutch Nonprofit Sector. In L. Salamon, H. An-heier, R. List, S. Toepler & S. W. Sokolowski (Eds.) Global Civil Society: Di-mensions of the Nonprofit Sector. Balimore, MD: The John Hopkins Centre for Civil Society, pp. 145-162.

C

Callen, J.L. (1994). Money Donations, Volunteering and Organizational Effi-ciency. Journal of Productivity Analysis, 5(3), 215-228.

Camerer, C.F. (2015). The Promise and Success of Lab-Field Generalizability in Experimental Economics: A Reply to Levitt and List. In: G. Frechette, & A. Schotter (Eds.) Handbook of Experimental Economic Methodology. Ox-ford: Oxford University Press, pp. 249–95.

Carpenter, J., Connolly, C., & Myers, C.K. (2008). Altruistic Behavior in a Repre-sentative Dictator Experiment. Experimental Economics, 11(3), 282–298.

Carroll, D.A., & Stater, K.J. (2009). Revenue diversification in nonprofit orga-nizations: Does it lead to financial stability? Journal of Public Administra-tion Research and Theory, 19(4), 947-966.

Cats, R. (2010). Grootscheeps banenverlies op stapel bij hulpclubs; Ontwik-kelingssubsidies gekort. Het Financieele Dagblad, December 17.

Central Bureau on Fundraising. (2014). Neem een kijkje in de financiën van KWF Kankerbestrijding, St. http://www.cbf.nl/Instelling-financien/4024/KWF-Kankerbestrijding-St (Retrieved May 6, 2014).

Center on Philanthropy. (2011). Review of Literature on Giving and High Net Worth Individuals. Indianapolis: Center on Philanthropy.

Chan, K.S., Mestelman, S., Moir, R., & Muller, R.A. (1996). The Voluntary Provi-

200

sion of Public Goods Under Varying Income Distributions. Canadian Jour-nal of Economics - Revue Canadienne D Economique, 29(1), 54-69.

Chan, K.S., Godbyb, R., Mestelman, S., & Muller, R.A. (2002). Crowding-out voluntary contributions to public goods. Journal of Economic Behavior & Organization, 48, 305–317.

Civil Exchange (2015). Whose Society? The Final Big Society Audit. http://www.civilexchange.org.uk/whose-society-the-final-big-society-audit (Re-trieved May 17, 2017).

Cucciniello, M., Porumbescu, G. A., & Grimmelikhuijsen, S. (2017). 25 Years of Transparency Research: Evidence and Future Directions. Public Admin-istration Review, 77(1), 32-44.

D

Day, K.M., & Devlin, R.A. (1996). Volunteerism and Crowding out: Canadian Econometric Evidence. The Canadian Journal of Economics - Revue cana-dienne d’Economique, 29(1): 37-53.

De Tocqueville, A. (1970[1840]). Democracy in America – Part the Second: The Social Influence of Democracy. New York: Knopf.

De Swaan, A. (1988). In Care of the State: Health Care, Education and Welfare in Europe and the USA in the Modern Era. Oxford: Polity Press.

De Wit, A. (2016). Mechanisms of crowding-out and crowding-in: Private con-tributions in 20 European welfare states. In: R. Meuleman, G. Kraaykamp, & M. Wittenberg (Eds.) Nederland in context: verschillen en overeenkom-sten. Proceedings vijfde Nederlandse Workshop European Social Survey. Den Haag: DANS.

De Wit, A., & Bekkers, R. (2017). Government Support And Charitable Dona-tions: A Meta-Analysis of the Crowding-out Hypothesis. Journal of Public Administration Research and Theory, 27(2), 301-319.

De Wit, A., Bekkers, R., & Broese Van Groenou, M. (2017). Heterogeneity in Crowding-out: When Are Charitable Donations Responsive To Govern-ment Support? European Sociological Review, 33(1), 59-71.

Dokko, J.K. (2009). Does the NEA Crowd Out Private Charitable Contribu-tions to the Arts? National Tax Journal, 62(1), 57-75.

Duncan, B. (1999). Modeling charitable contributions of time and money. Journal of Public Economics, 72, 213–242.

201

E

Eckel, C.C., & Grossman, P.J. (2003). Rebate versus matching: does how we subsidize charitable contributions matter? Journal of Public Economics, 87(3–4), 681–701.

Eckel, C.C., & Grossman, P.J. (2008). Subsidizing charitable contributions: a natural field experiment comparing matching and rebate subsidies. Exper-imental Economics, 11(3), 234-252.

Eckel, C.C., Grossman, P.J., & Johnston, R.M. (2005). An Experimental Test of the Crowding Out Hypothesis. Journal of Public Economics, 89, 1543–60.

Einolf, C. (2016). The Social Origins of the Nonprofit Sector and Charitable Giving. In: F. Handy & P. Wiepking (Eds.) The Palgrave Handbook of Global Philanthropy. Basingstoke: Palgrave Macmillan, pp. 509-529.

Ek, C. (2017). Some Causes Are More Equal than Others? The Effect of Simi-larity on Substitution in Charitable Giving. Journal of Economic Behavior & Organization, 136, 45-62.

Elmelund-Præstekær, C. & Emmenegger, P. (2013). Strategic Re-framing as a Vote Winner: Why Vote-seeking Governments Pursue Unpopular Reforms. Scandinavian Political Studies, 36(1), 23-42.

Esping-Andersen, G. (1990). The Three Worlds of Welfare Capitalism. Prince-ton: Princeton University Press.

Evans, C.A., Evans, G.R. & Mayo, L. (2017). Charitable Giving as a Luxury Good and the Philanthropic Sphere of Influence. Voluntas, 28: 556.

European Commission. (2013). Communication from the commission to the european parliament, the council, the european economic and social com-mittee and the committee of the regions: Towards social investment for growth and cohesion including implementing the european social fund 2014-2020. Brussels: COM.

F

Falk, A., Meier, S., & Zehnder, C. (2013). Do Lab Experiments Misrepresent Social Preferences? The Case of Self-selected Student Samples. Journal of the European Economic Association, 11(4), 839–852.

Ferragina, E. (2017). The welfare state and social capital in Europe: Reas-sessing a complex relationship. International Journal of Comparative So-ciology, 58(1), 55-90.

202

Ferris, J.S., & West, E.G. (2003). Private versus public charity: Reassessing crowding out from the supply side. Public Choice, 116(3-4), 399-417.

Foster, W., & Fine, G. (2007). How Nonprofits Get Really Big. Stanford Social Innovation Review, Spring 2007.

Francis, G. (2012). Too good to be true: Publication bias in two prominent studies from experimental psychology. Psychonomic Bulletin & Review, 19, 151–156.

Franssen, S.E. & Bekkers, R. (2016). Culturele instellingen in Nederland: Ve-randeringen in geefgedrag, giften, fondsenwerving en inkomsten tussen 2011 en 2014. Amsterdam: Center for Philanthropic Studies, Vrije Univer-siteit (VU) Amsterdam.

Froelich, K.A. (1999). Diversification of Revenue Strategies: Evolving Re-source Dependence in Nonprofit Organizations. Nonprofit and Voluntary Sector Quarterly, 28(3), 246-268.

Froelich, K.A., Knoepfle, T.W., & Pollak, T.H. (2000). Financial Measures in Nonprofit Organization Research: Comparing IRS 990 Return and Audited Financial Statement Data. Nonprofit and Voluntary Sector Quarterly, 29(2), 232-254.

Frumkin, P. (2002). On being nonprofit: A conceptual and policy primer. Cam-bridge, MA: Harvard University.

G

Galbiati, R., & Vertova, P. (2008). Obligations and cooperative behaviour in public good games. Games and Economic Behavior, 64(1), 146-170.

Galbiati, R., & Vertova, P. (2014). How laws affect behavior: Obligations, in-centives and cooperative behavior. International Review of Law and Eco-nomics, 38, 48-57.

Galizzi, M.M., Navarro-Martinez, D. (2015). On the External Validity of So-cial-Preference Games: A Systematic Lab-Field Study. Working Paper.

Garrett, T., & Rhine, R. (2010). Government growth and private contributions to charity. Public Choice, 143(1-2), 103-120.

Gerber, A.S., Karlan, D., & Bergan, D. (2009). Does the Media Matter? A Field Experiment Measuring the Effect of Newspapers on Voting Behavior and Political Opinions, American Economic Journal: Applied Economics, 1(2), 35-52.

Gesthuizen, M., Scheepers, P., van der Veld, W., & Völker, B. (2013). Structural

203

aspects of social capital: tests for cross-national equivalence in Europe. Quality & Quantity, 47(2), 909-922.

Gesthuizen, M., Van der Meer, T.W.G., & Scheepers, P. (2008). Education and Dimensions of Social Capital: Do Educational Effects Differ Due to Educa-tional Expansion and Social Security Expenditure? European Sociological Review, 24, 617-632.

Geys, B., & Sørensen, R.J. (2016). Revenue Scarcity and Government Out-sourcing: Empirical Evidence from Norwegian Local Governments. Public Administration, 94(3), 769-788.

Glass, G.V. (1976). Primary, Secondary, and Meta-Analysis of Research. Edu-cational Researcher, 5(10), 3-8.

Gordon, C. W., & Babchuk, N. (1959). A typology of voluntary associations. American Sociological Review, 22-29.

Grimmelikhuijsen, S., & Klijn, A. (2015). The effects of judicial transparency on public trust: evidence from a field experiment, Public Administration, 93(4), 995-1011.

Grimmelikhuijsen, S.G., & Meijer, A.J. (2014). The Effects of Transparency on the Perceived Trustworthiness of a Government Organization: Evidence from an Online Experiment, Journal of Public Administration Research and Theory, 24(1), 137-157.

Gronberg, T.J.., Luccasen III, R.A., Turocy, T.L., & Van Huyck, J.B. (2012). Are Tax-Financed Contributions to a Public Good Completely Crowded-out? Experimental Evidence. Journal of Public Economics, 96, 596–603.

Güth, W., Sutter, M., & Verbon, H. (2006). Voluntary versus Compulsory Sol-idarity: Theory and Experiment. Journal of Institutional and Theoretical Economics-Zeitschrift Fur Die Gesamte Staatswissenschaft, 162, 347–363.

H

Handy, F. (2000). How We Beg: The Analysis of Direct Mail Appeals. Nonprofit and Voluntary Sector Quarterly, 29(3), 439-454.

Handy, F., Seto, S., Wakaruk, A., Mersey, B., Mejia, A., & Copeland, L. (2010). The Discerning Consumer: Is Nonprofit Status a Factor? Nonprofit and Vol-untary Sector Quarterly, 39(5), 866-883.

Harbaugh, W.T., Mayr, U., & Burghart, D.R. (2007). Neural Response to Tax-ation and Voluntary Giving Reveal Motives for Charitable Donations. Sci-ence, 316, 1622–1625.

204

Heald, D. (2006). Varieties of transparency. In: C. Hood and D. Heald (Eds.) Transparency: the Key to Better Governance? New York: Oxford University Press, pp. 25-43.

Hedström, P., & Bearman, P. (2009). The Oxford Handbook of Analytical So-ciology. Oxford: Oxford University Press.

Heerma van Voss, L.H., & Van Leeuwen, M. (2012). Charity in the Dutch Re-public: an introduction. Continuity and Change, 27(2), 175-197.

Henrich, J., Heine, S.J., & Norenzayan, A. (2010). The Weirdest People in the World? Behavioral and Brain Sciences, 33, 61-135.

Herzer, D., & Nunnenkamp, P. (2013). Private Donations, Government Grants, Commercial Activities, and Fundraising: Cointegration and Causality for NGOs in International Development Cooperation. World Development, 46, 234-251.

Heutel, G. (2014). Crowding Out and Crowding In of Private Donations and Government Grants. Public Finance Review, 42(2), 143-175.

Horne, C.S., Johnson, J.L., & Van Slyke, D.M. (2005). Do Charitable Donors Know Enough - And Care Enough - About Government Subsidies to Affect Private Giving to Nonprofit Organizations? Nonprofit and Voluntary Sector Quarterly, 34, 136-149.

Horne, C.S., Van Slyke, D.M., & Johnson, J.L. (2006). Charitable choice imple-mentation: what public managers should know about public opinion and the potential impact of government funding on private giving. Internation-al Journal of Public Administration, 29(10-11), 819-836.

Hsu, L.C. (2008). Experimental evidence on tax compliance and voluntary public good provision. National Tax Journal, 61(2), 205-223.

Hughes, P.N.H., & Luksetich, W.A. (1999). The Relationship Among Funding Sources for Art and History Museums. Nonprofit Management and Leader-ship, 10(1), 21–37.

Hughes, P., Luksetich, W., & Rooney, P. (2014). Crowding-Out and Fundraising Efforts: The Impact of Government Grants on Symphony Orchestras. Non-profit Management and Leadership, 24(4), 445-464.

Hungerman, D.M. (2005). Are church and state substitutes? Evidence from the 1996 welfare reform. Journal of Public Economics, 89, 2245–2267.

I

IIPD (2016). Individual International Philanthropy Database [Machine-read-

205

able data file]. P. Wiepking & F. Handy [principle investigators]. Rotter-dam, the Netherlands: Erasmus University Rotterdam [distributor].

Isaac, R.M., & Norton, D.A. (2013). Endogenous Institutions and the Possibil-ity of Reverse Crowding Out. Public Choice, 156(1-2), 253-284.

Inglehart, R. (1997). Modernization and Postmodernization. Princeton: Princ-eton University Press.

Inglehart, R. (2000). Globalization and Postmodern Values. Washington Quarterly, 23(1), 215-228.

Ingram, P., & Clay, K. (2000). The Choice-Within-Constraints New Institution-alism and Implications for Sociology. Annual Reviews of Sociology, 26, 525-546.

J

Jæger, M.M. (2006). Welfare Regimes and Attitudes towards Redistribution: The Regime Hypothesis Revisited. European Sociological Review, 22(2), 157-170.

Jones, D.B. (2015). Education’s Gambling Problem: Earmarked Lottery Rev-enues and Charitable Donations to Education. Economic Inquiry, 53(2), 906-921.

K

Kääriäinen J., & Lehtonen, H. (2006). The Variety of Social Capital in Welfare State Regimes – A Comparative Study of 21 Countries. European Societies, 8, 27-57.

Khanna, J, & Sandler, T. (2000). Partners in giving: The crowding-in effects of UK government grants. European Economic Review, 44, 1543-1556.

Khanna, J., Posnett, J., & Sandler, T. (1995). Charity Donations in the UK: New Evidence Based on Panel Data. Journal of Public Economics, 56, 257-272.

Kim, S., & Lee, J. (2012). E-participation, transparency, and trust in local gov-ernment, Public Administration Review, 72(6), 819-828.

Kim, M., & Van Ryzin, G.G. (2014). Impact of Government Funding on Dona-tions to Arts Organizations: A Survey Experiment. Nonprofit and Voluntary Sector Quarterly, 43(5), 910-925.

Kingma, B.R. (1989). An Accurate Measurement of the Crowdout Effect, In-

206

come Effect, and Price Effect for Charitable Contributions. Journal of Polit-ical Economy, 97, 1197–1207.

Kingma, B.R., & McClelland, R. (1995). Public radio stations are really, really not public goods: Charitable contributions and impure altruism. Annals of Public and Cooperative Economics, 66(1), 65-76.

Klamer, A. (2004). Art as a common good. Paper presented at the 13th confer-ence of the Association of Cultural Economics International.

Klijn, E.H. (2008). Governance and Governance Networks in Europe. Public Management Review, 10(4), 505-525.

Knowles, S., & Sullivan, T. (2017). Does Charity Begin at Home or Overseas? Nonprofit and Voluntary Sector Quarterly. Online first, April 11.

Konow, J. (2010). Mixed feelings: Theories of and evidence on giving. Journal of Public Economics, 94(3-4), 279-297.

Korenok, O., Millner, E.L., & Razzolini, L. (2012). Are Dictators Averse to In-equality? Journal of Economic Behavior & Organization, 82(2-3), 543-547.

Korenok, O., Millner, E.L., & Razzolini, L. (2014). Taking, Giving, and Impure Altruism in Dictator Games. Experimental Economics, 17(3), 488-500.

Koster, F. (2007). Globalization, social structure, and the willingness to help others: A multilevel analysis across 26 countries. European Sociological Review, 23(4), 537-551.

Kropf, M., & Knack, S. (2003). Viewers like You: Community Norms and Con-tributions to Public Broadcasting. Political Research Quarterly, 56(2), 187-197.

Künemund, H., & Rein, M. (1999). There Is More to Receiving than Needing: Theoretical Arguments and Empirical Explorations of Crowding In and Crowding Out. Ageing and Society, 19(01), 93-121.

L

Lassen, D.D. (2005). The Effect of Information on Voter Turnout: Evidence From a Natural Experiment, American Journal of Political Science, 49(1), 103-118.

Lecy, J.D., & Van Slyke, D.M. (2013). Nonprofit Sector Growth and Density: Testing Theories of Government Support. Journal of Public Administration Research and Theory, 23(1): 189-214.

Lee, B., Fraser, I., Fillis, I. (2017). Nudging Art Lovers to Donate. Nonprofit and Voluntary Sector Quarterly. Online first, April 11.

207

Lergetporer, P., Schwerdt, G., Werner, K., & Woessmann, L. (2016). Informa-tion and Preferences for Public Spending: Evidence from Representative Sur-vey Experiments. Bonn: IZA Discussion Paper No. 9968.

Levitt, S.D., & List, J.A. (2007). What Do Laboratory Experiments Measuring Social Preferences Reveal about the Real World? The Journal of Economic Perspectives, 21(2), 153-174.

Li, H., & McDougle, L. (2017). Information Source Reliance and Charitable Giving Decisions. Nonprofit Management and Leadership. Published online February 17.

Lilley, A., & Slonim, R. (2014). The price of warm glow. Journal of Public Eco-nomics, 114, 58-74.

Lindsey, L.B., & Steinberg, R. (1990). Joint crowdout: An empirical study of the impact of federal grants on state government expenditures and charita-ble donations. National Bureau of Economic Research, Working Paper No. 3226.

Lobb, A., Mock, N., & Hutchinson, P. L. (2012). Traditional and Social Media Coverage and Charitable Giving Following the 2010 Earthquake in Haiti. Prehospital and Disaster Medicine, 27(4), 319-324.

Lu, J. (2016). The Philanthropic Consequence of Government Grants to Non-profit Organizations: A Meta-Analysis. Nonprofit Management and Leader-ship, 26(4), 381-400.

Luccasen, R.A. (2012). Individual Differences in Contributions and Crowd-ing-Out of a Public Good. Scottish Journal of Political Economy, 59, 419–441.

Luksetich, W.A., & Lange, M.D. (1995). A simultaneous model of nonprofit symphony orchestra behavior. Journal of Cultural Economics, 19(1), 49-68.

M

Manzoor, S.H., & Straub, J.D. (2005). The robustness of Kingma’s crowd-out estimate: Evidence from new data on contributions to public radio. Public Choice, 123, 463–476.

Marcuello, C., & Salas, V. (2000). Money and time donations to Spanish non-governmental organizations for development aid. Investigaciones Económicas, 24(1), 51-73.

Marcuello, C., & Salas, V. (2001). Nonprofit organizations, monopolistic com-petition, and private donations: Evidence from Spain. Public Finance Re-

208

view, 29(3), 183-207.Maslow, A.H. (1954). Motivation and personality. New York: Harper and Row.McDougle, L., & Handy, F. (2014). The Influence of Information Costs on Do-

nor Decision Making. Nonprofit Management and Leadership, 24(4), 465–485.

Meier, S. (2006). A survey of economic theories and field evidence on pro-social behavior. Working paper series // Federal Reserve Bank of Boston, No. 06-6.

Milward, H.B., & Provan, K.G. (2003). Managing the Hollow State: Collabora-tion and Contracting. Public Management Review, 5(1), 1-18.

Morgan, S.L., & Winship, C. (2007). Counterfactuals and Causal Inference: Methods and Principles for Social Research. New York: Cambridge Univer-sity Press.

Museumvereniging (2015). Museumcijfers 2014. Amsterdam: Stichting Mu-seana.

N

Nelson, A.A., & Gazley, B. (2014). The Rise of School-Supporting Nonprofits. Education Finance and Policy, 9(4), 541-566.

Nguyen, P.A. (2015). The Influence of Government Support for the Nonprofit Sector on Philanthropy across Nations. In: F. Handy & P. Wiepking (Eds.) The Palgrave Handbook of Global Philanthropy. Basingstoke: Palgrave Macmillan, pp. 530-539.

Nikolova, M. (2015). Government Funding of Private Voluntary Organiza-tions: Is There a Crowding-Out Effect? Nonprofit and Voluntary Sector Quarterly, 44(3), 487-509.

Nisbet, R.A. 1962[1953]. Community and power [originally The quest for com-munity]. New York: Oxford University Press.

Nunnenkamp, P., & Öhler, H. (2012). How to attract donations: the case of US NGOs in international development. The Journal of Development Studies, 48(10), 1522-1535.

O

Okten, C., & Weisbrod, B.A. (2000). Determinants of Donations in Private

209

Nonprofit Markets. Journal of Public Economics, 75(2), 255–272.O’Regan, K., & Oster, S. (2002). Does Government Funding Alter Nonprofit

Governance? Evidence from New York City Nonprofit Contractors. Journal of Policy Analysis and Management, 21(3), 359-379.

P

Paqué, K.H. (1986). The efficiency of tax incentives to private charitable giving–Some econometric evidence for the Federal Republic of Germany. Weltwirtschaftliches Archiv, 122(4), 690-712.

Payne, A.A. (1998). Does the Government Crowd-out Private Donations? New Evidence from a Sample of Non-profit Firms. Journal of Public Eco-nomics, 69, 323–345.

Payne, A.A. (2001). Measuring the Effect of Federal Research Funding on Private Donations at Research Universities: Is Federal Research Funding More than a Substitute for Private Donations? International Tax and Public Finance, 8(5-6), 731-751.

Payne, A.A. (2009). Does Government Funding Change Behavior? An Empir-ical Analysis of Crowd Out. Tax Policy and the Economy, 23(1), 159-184.

Peloza, J., & Steel, P. (2005). The price elasticities of charitable contributions: a meta-analysis. Journal of Public Policy & Marketing, 24(2), 260-272.

Pennerstorfer, A., & Neumayr, M. (2017). Examining the Association of Wel-fare State Expenditure, Non-profit Regimes and Charitable Giving. Volun-tas, 28(2), 532–555.

Petrovski, E. (2017). Whether and How Much to Give: Uncovering the Con-trasting Determinants of the Decisions of Whether and How Much to Give to Charity with Two-Stage Alternatives to the Prevailing Tobit Model. Vol-untas, 28(2), 594–620.

Petticrew, M., & Roberts, H. (2006). Systematic Reviews in the Social Sciences: A Practical Guide. Oxford: Blackwell Publishing.

Pfeffer, J., & Salancik, G.R. (1978). The External Control of Organizations: A Resource Dependence Perspective. New York: Harper & Row, Publishers.

Piantadosi, S., Byar, D.P., & Green, S.B. (1988). The Ecological Fallacy. Ameri-can Journal of Epidemiology, 127(5), 893-904.

Piotrowski, S.J., & Van Ryzin, G.G. (2007). Citizen attitudes toward transpar-ency in local government. The American Review of Public Administration, 37(3), 306-323.

210

Pontzen, R. (2011). Zet open die ramen! Beeldende kunst; Terugblik op 2011 Kunstwereld gedwongen tot zelfreflectie. De Volkskrant, December 30.

Posnett, J., & Sandler, T. (1989). Demand for charity donations in private non-profit markets: the case of the UK. Journal of Public Economics, 40(2), 187-200.

R

Randall, R., & Wilson, C. (1989). The Impact of Federally Imposed Stress Upon Local-Government and Nonprofit Organizations. Administration & Society, 21(1), 3-19.

Reece, W.S. (1979). Charitable contributions: New evidence on household behavior. The American Economic Review, 142-151.

Reeson, A.F., & Tisdell, J.G. (2008). Institutions, Motivations and Public Goods: An Experimental Test of Motivational Crowding. Journal of Economic Be-havior and Organization, 68, 273–281.

Reinstein, D. (2006). Does One Contribution Come at the Expense of Another? Empirical Evidence on Substitution Between Charitable Donations. Univer-sity of Essex Discussion Paper Series No. 618.

Reinstein, D. (2007). Substitution Between (and Motivations for) Charitable Contributions: An Experimental Study. University of Essex Discussion Pa-per Series No. 648.

Ribar, D.C., & Wilhelm, M.O. (2002). Altruistic and Joy-of-Giving Motivations in Charitable Behavior. Journal of Political Economy, 110(2), 425-457.

Rijghard, R. (2010). Wereld van de kunst is onthutst; Er komen protesten aan. NRC.NEXT, October 8.

Rijksoverheid (2011). Memorie van toelichting Geefwet. https://www.rijksoverheid.nl/documenten/kamerstukken/2011/09/20/memo-rie-van-toelichting-geefwet (Retrieved April 12, 2017).

Roberts, R.D. (1984). A Positive Model of Private Charity and Public Trans-fers. Journal of Political Economy, 92, 136-48.

Rodriguez, H.P., Laugesen, M.J., & Watts, C.A. (2010). A randomized experi-ment of issue framing and voter support of tax increases for health insur-ance expansion. Health Policy, 98(2), 245-255.

Rooney, P., Osili, U., Thayer, A., Baranwoski, G., Hayat, A., Davis Kalugyer, A., & Hyatte, C. (2014). The 2014 U.S. Trust ® Study of High Net Worth Philan-thropy. Boston/Indianapolis: U.S. Trust/Indiana University Lilly Family

211

School of Philanthropy.Rooney, P.M., Steinberg, K.S., & Schervish, P.G. (2001). A Methodological Com-

parison of Giving Surveys: Indiana as a Test Case. Nonprofit and Voluntary Sector Quarterly, 30(3), 551-568.

Rooney, P.M., Steinberg, K.S., & Schervish, P.G. (2004). Methodology is Desti-ny: The Effect of Survey Prompts on Reported Levels of Giving and Volun-teering. Nonprofit and Voluntary Sector Quarterly, 33(4), 628-654.

Rose-Ackerman, S. (1981). Do Government Grants to Charity Reduce Private Donations? In: M. White (Ed.) Nonprofit Firms in a Three Sector Economy. Washington: The Urban Institute, pp. 95-114.

Rosenthal, R. (1984). Meta-Analytical Procedures for Social Research. Beverly Hills: Sage Publications.

Rothstein, B. (1998). Just Institutions Matter – The Moral and Political Logic of the Universal Welfare State. Cambridge: Cambridge University Press.

S

Salamon, L.M., & Anheier, H.K. (1998). Social Origins of Civil Society: Explain-ing the Nonprofit Sector Cross-Nationally. Voluntas, 9, 213-248.

Salamon, L.M., & Sokolowski, W. (2001). Volunteering in Cross-National Per-spective: Evidence From 24 Countries. Working Papers of the Johns Hop-kins Comparative Nonprofit Sector Project, no. 40. Baltimore: The Johns Hopkins Center for Civil Society Studies.

Salamon, L.M., Sokolowski, S.W., & Anheier, H.K. (2000). Social Origins of Civil Society: An Overview. Working Papers of the Johns Hopkins Comparative Nonprofit Sector Project, no. 38. Baltimore: The Johns Hopkins Center for Civil Society Studies.

Sargeant, A., & Lee, S. (2004). Donor trust and relationship commitment in the UK charity sector: The impact on behavior. Nonprofit and Voluntary Sector Quarterly, 33(2), 185-202.

Sav, G.T. (2012). Government free riding in the public provision of higher ed-ucation: panel data estimates of possible crowding out. Applied Economics, 44(9), 1133-1141.

Scheepers, P., & Grotenhuis, M. T. (2005). Who cares for the poor in Europe? Micro and macro determinants for alleviating poverty in 15 European countries. European Sociological Review, 21(5), 453-465.

Schervish, P.G., & Havens, J.J. (1995). Wherewithal and beneficence: Charita-

212

ble giving by income and wealth. New Directions for Philanthropic Fund-raising, 8, 81-109.

Schiff, J. (1985). Does government spending crowd out charitable contribu-tions? National Tax Journal, 535-546.

Schiff, J. (1990). Charitable Giving and Government Policy: An Economic Anal-ysis. Westport, Connecticut: Greenwood Press.

Schuyt, T. (2010). Philanthropy in European welfare states: a challenging promise? International Review of Administrative Sciences, 76(4), 774-789.

Schuyt, T. (2014). Nederland is en blijft een filantropisch land. De Volkskrant, 19 november.

Shah, K.K., Sussex, J., & Hernandez-Villafuerte, K. (2015). Government and charity funding of cancer research: public preferences and choices. Health Research Policy and Systems, 13(38), 1-14.

Simmons, W., & Emanuele, R. (2004). Does Government Spending Crowd Out Donations of Time and Money? Public Finance Review, 32(5): 498–511.

Slothuus, R. (2007). Framing deservingness to win support for welfare state retrenchment. Scandinavian Political Studies, 30(3), 323-344.

Smith, T.M. (2007). The Impact of Government Funding on Private Contri-butions to Nonprofit Performing Arts Organizations. Annals of Public and Cooperative Economics, 78(1), 137-160.

Smith, S.R., & Lipsky, M. (1993). Nonprofits for hire: The welfare state in the age of contracting. Cambridge: Harvard University Press.

Sokolowski, S.W. (2013). Effects of Government Support of Nonprofit Insti-tutions on Aggregate Private Philanthropy: Evidence from 40 Countries. Voluntas, 24(2), 359-381.

Song, S., & Yi, D.T. (2011). The fundraising efficiency in U.S. non-profit art organizations: an application of a Bayesian estimation approach using the stochastic frontier production model. Journal of Productivity Analysis, 35(2), 171-180.

Stadelmann-Steffen, I. (2011). Social Volunteering in Welfare States: Where Crowding Out Should Occur. Political Studies, 59(1): 135-155.

Stanley, T.D. (2005). Beyond Publication Bias. Journal of Economic Surveys, 19(3), 309–345.

Steinberg, R. (1985). Empirical Relations Between Government Spending and Charitable Donations. Nonprofit and Voluntary Sector Quarterly, 14(2-3), 54-64.

Steinberg, R. (1989). The Theory of Crowding Out: Donations, Local Govern-ment Spending, and the “New Federalism” In: R. Magat (Ed.) Philanthropic

213

Giving: Studies in Varieties and Goals. New York/Oxford: Oxford University Press.

Steinberg, R. (1991). Does Government Spending Crowd Out Donations? An-nals of Public and Cooperative Economics, 62(4), 591-612.

Steinberg, R. (1997). Overall Evaluation of Economic Theories. Voluntas, 8(2), 179-204.

Suanet, B., Broese Van Groenou, M., & Van Tilburg, T. (2012). Informal and Formal Home-Care Use among Older Adults in Europe: Can Cross-nation-al Differences be Explained by Societal Context and Composition? Ageing and Society, 32(3), 491-515.

Sutter, M., & Weck-Hannemann, H. (2004). An experimental test of the pub-lic goods crowding out hypothesis when taxation is endogenous. Finan-zArchiv: Public Finance Analysis, 60(1), 94-110.

Svallfors, S. (1997). Worlds of Welfare and Attitudes to Redistribution: A Comparison of Eight Western Nations. European Sociological Review, 13(3), 283-304.

T

Teeuwen, D. (2014). Generating Generosity: Financing poor relief through charitable collections in Dutch towns, c. 1600-1800 (Doctoral thesis, Utrecht University). https://dspace.library.uu.nl/handle/1874/294327 (Retrieved May 15, 2017).

Thaler, R.H. (1999). Mental Accounting Matters. Journal of Behavioral Deci-sion Making, 12(3): 183-206.

Thomas, W.I., & Thomas, D.S. (1928). The Child in America: Behavior Prob-lems and Programs. New York: Knopf.

Tinkelman, D. (2010). Revenue Interactions: Crowding Out, Crowding In, or Neither? In: B.A. Seaman & D.R. Young (Eds.) Handbook of Research on Nonprofit Economics and Management. Cheltenham: Edward Elgar Pub-lishing Limited.

Tinkelman, D., & Neely, D.G. (2011). Some Econometric Issues in Studying Nonprofit Revenue Interactions Using NCCS Data. Nonprofit and Voluntary Sector Quarterly, 40(4), 751-761.

Tolbert, C.J., & Mossberger, K. (2006). The effects of e-government on trust and confidence in government. Public Administration Review, 66(3), 354-369.

214

Tukey, J.W. (1962). The Future of Data Analysis. The Annals of Mathematical Statistics, 33, 1-67.

Tummers, L., Weske, U., Bouwman, R., & Grimmelikhuijsen, S. (2015). The Impact of Red Tape on Citizen Satisfaction: An Experimental Study. Inter-national Public Management Journal, 19(3), 320-341.

U

Uslaner, E.M. (2003) Trust and civic engagement in East and West’. In: G. Ba-descu & E. Uslaner (Eds.) Social Capital and the Transition to Democracy. London and New York: Routledge, pp. 81-94.

V

Vamstad, J., & Von Essen, J. (2013). Charitable Giving in a Universal Welfare State: Charity and Social Rights in Sweden. Nonprofit and Voluntary Sector Quarterly, 42(2), 285-301.

Van der Meer, T., Te Grotenhuis, M., & Pelzer, B. (2010). Influential Cases in Multilevel Modeling: A Methodological Comment. American Sociological Review, 75(1), 173-178.

Van Ingen, E., & Van der Meer, T. (2011). Welfare State Expenditure and In-equalities in Voluntary Association Participation. Journal of European So-cial Policy, 21(4), 302-322.

Van Leeuwen, M.H.D. (1999). Armenzorg en charitas, ca. 1800-2000. Een historische erfenis. In: R. van der Bie & P. Dehning (Eds.) Nationaal Goed. Feiten en cijfers over onze samenleving ca. 1800-1999. Amsterdam: CBS, pp. 159-178.

Van Oorschot, W., & Arts, W. (2005). The Social Capital of European Welfare States: The Crowding Out Hypothesis Revisited. Journal of European Social Policy, 15(1), 5-26.

Verba, S., Schlozman, K.L. & Brady, H.E. (1995). Voice and Equality: Civic Vol-untarism in American Politics. Cambridge: Harvard University Press.

Verschuere, B., & De Corte, J. (2014). The Impact of Public Resource Depen-dence on the Autonomy of NPOs in Their Strategic Decision Making. Non-profit and Voluntary Sector Quarterly, 43(2), 293-313.

Vos, N. (2011). Kleine kunstinstellingen kunnen wel inpakken; Zijlstra laat

215

Concertgebouworkest, Nederlandse Opera en beste musea en festivals ongemoeid. Het Parool, June 11.

W

Warr, P.G. (1982). Pareto Optimal Redistribution and Private Charity. Journal of Public Economics, 19(1), 131-138.

Weinblatt, J. (1992). Do Government Transfers Crowd out Private Transfers to Nonprofit Organizations? The Israeli Experience. International Journal of Social Economics, 19(2), 60-66.

Weber, M. (1922[1987]). Economy and Society: An Outline of Interpretative Sociology. Los Angeles: University of California Press.

Weisbrod, B.A. (1977). The Voluntary Nonprofit Sector: An Economic Analysis. Lexington: Lexington Books.

Wiepking, P. (2010). Democrats support international relief and the upper class donates to art? How opportunity, incentives and confidence affect donations to different types of charitable organizations. Social Science Re-search, 39(6), 1073-1087.

Wiepking, P., & Bekkers, R. (2012). Who gives? A literature review of pre-dictors of charitable giving. Part Two: Gender, family composition and in-come. Voluntary Sector Review, 3(2), 217-245.

Wiepking, P., & Handy, F. (2015). Introduction. In: F. Handy & P. Wiepking (Eds.) The Palgrave Handbook of Global Philanthropy. Basingstoke: Pal-grave Macmillan, pp. 3-8.

Wiepking, P., & Handy, F. (2016). Documentation Individual International Philanthropy Database (IIPD). A Comparative Study of Global Giving. Rot-terdam: Erasmus University Rotterdam.

Wiepking, P., Handy, F., Park, S., Bekkers, R., Bethmann, S., Breen, O., Breeze, B., Einolf, C.J., Kang, C., Katz, H.A., Krasnopolskaya, I., Layton, M.D., Lo, V.K.T., Neumayr, M., Osili, U., Pessi, A.B., Sivesind, K.H., Scaife, W., De Wit, A., Xiulan, Z., Yamauchi, N. (2016). The Matthew Effect in Philanthropy: How Philanthropic Structure Enables Philanthropic Giving. Paper presented at the 45th ARNOVA Annual Conference, Washington, DC.

Wiepking, P., & Maas, I. (2009). Resources that make you generous: Effects of social and human resources on charitable giving. Social Forces, 87(4), 1973-1995.

Wilhelm, M.O., & Bekkers, R. (2010). Helping Behavior, Dispositional Em-

216

pathic Concern, and the Principle of Care. Social Psychology Quarterly, 73(1), 11-32.

Wolf, F.M. (1986). Meta-Analysis: Quantitative Methods for Research Synthe-sis. Beverly Hills: Sage Publications.

Worthy, B. (2010). More open but not more trusted? The effect of the Free-dom of Information Act 2000 on the United Kingdom central government. Governance, 23(4), 561-582.

Worthy, B. (2015). The Impact of Open Data in the UK: Complex, Unpredict-able, and Political. Public Administration, 93(3), 788-805.

Wortmann, M. (1997). Het ontstaan van het Centraal Bureau Fondsenwerv-ing: Lessen voor het heden? In: V. Kingma & M.H.D. Van Leeuwen (Eds.) Filantropie in Nederland: Voorbeelden uit de periode 1770-2020. Amster-dam: Aksant, pp. 110-119.

Y

Yetman, M.H., & Yetman, R.J. (2003). The Effect of Nonprofits’ Taxable Ac-tivities on the Supply of Private Donations. National Tax Journal, 56(1), 243-258.

Yetman, M.H., & Yetman, R.J. (2013). How Does the Incentive Effect of the Charitable Deduction Vary across Charities? Accounting Review, 88(3), 1069-1094.

Yoruk, B.K. (2012). The Effect of Media on Charitable Giving and Volunteer-ing: Evidence from the “Give Five” Campaign. Journal of Policy Analysis and Management, 31(4), 813-836.

Young, D.R. (2000). Alternative Models of Government-Nonprofit Sector Re-lations: Theoretical and International Perspectives. Nonprofit and Volun-tary Sector Quarterly, 29(1), 149-172.

philanthropy in the welfare stated r o o w r o o v 7 introduction: philanthropy in the welfare state...

Documents