Abstract
Many nonprofits rely on private donations and government grants, but it is still unclear how these sources of funding may interact or even influence each other. To examine the behavioral aspect of the crowding-out hypothesis, we conducted an online survey experiment (n = 562) to test if government funding of a hypothetical nonprofit would influence donations. Our results show that a nonprofit with government funding, compared to an identical hypothetical organization without government funding, received 25% less in average donations (US$35 vs. US$47) and was about half as likely (21% vs. 38%) to receive all the money in a forced-choice scenario. However, the crowding-out effect of government funding appears much weaker for those who are arts patrons or who have previously contributed to the arts. Interestingly, this crowding-out effect seems insensitive to the amount of government funding and to labeling the government funding as coming from a prestigious source (e.g., National Endowment for the Arts [NEA]).
Context
The number of registered nonprofits in the United States was about 1.6 million as of 2010, according to the Nonprofit Sector in Brief (Blackwood, Roeger, & Pettijohn, 2012). The same report shows that private donations represent 13.3% of revenues in the nonprofit sector overall while government grants account for 8.3%, with the remaining revenue coming largely from fees for services and goods. For nonprofit arts organizations, the focus of this study, private donations represent fully 31% of total income, with 24% coming from individual donors and the rest from foundations and corporations; government support represents approximately 9% of income for arts organizations (Americans for the Arts, 2012). This dynamic funding structure of the nonprofit sector has generated much debate on the question of whether government grants may crowd-out, or crowd-in, private giving to nonprofit organizations. The term crowding-out refers to a situation in which the presence of government funding displaces or discourages private giving. On the other hand, crowding-in refers to the possibility that government funding may leverage or encourage private donations to nonprofit organizations. The results of previous empirical studies on institutional crowding phenomena in the arts have been somewhat mixed.
A number of studies have explored both the crowding-out and crowding-in hypotheses in the particular context of nonprofit arts organizations (Borgonovi, 2006; Brooks, 2000; Dokko, 2009; Kingma, 1989; Okten & Weisbrod, 2000; Smith, 2003, 2007), either by focusing on one type of arts organizations or as a part of larger sample of various types of nonprofit organizations. A study by Kingma (1989), for example, found some evidence of crowding-out in the case of giving to public radio stations. Dokko (2009) estimated that private charitable contributions to arts organizations increased after a major funding cut to the National Endowment for the Arts (NEA) between fiscal years 1995 and 1996, in part due to corresponding increases in nonprofits’ fundraising expenditures. In a study of symphony orchestras, Brooks (2000) found evidence of a nonlinear relationship between public subsidies and private contributions. Specifically, his study found a crowding-out effect occurred at low levels of public funding, but a crowding-in effect appeared at higher levels of public funding. Borgonovi (2006) supports Brooks’ inverted U-shape crowding effect with his analysis of a panel of American nonprofit theaters. Meanwhile, Okten and Weisbrod (2000) found no significant relationship between private and public funding sources for nonprofit organizations across various nonprofit sectors, including arts galleries. Examining the crowding-out hypothesis with a panel of nonprofit performing arts organizations, Smith (2007) concludes that there is a lack of evidence of crowding-out for the nonprofit performing arts group.
Earlier studies suggested that private donors may perceive government dollars as substitutes for their donations and thus give less, resulting in a direct or complete crowding-out effect (Abrams & Schmitz, 1984; Roberts, 1984; Warr, 1982). Yet, Andreoni (1990) and Steinberg (1991) claim that private donors regard government funding as an imperfect substitute, and therefore a partial crowding-out effect occurs instead of a dollar-for-dollar displacement. In the line of inquiry on crowding-out, Andreoni and Payne (2003, 2011) examined how nonprofits adjust their fundraising efforts after receiving government support and found that increased public funding was associated with decreased fundraising budget. This reduced fundraising effort, in turn, would likely lead to a decrease in private donations. Another empirical study in support of a crowding-out effect (Dokko, 2009) found that fundraising expenditures increased when nonprofits experienced a drop in government grants. As such, the mechanism of the crowding-out effect involves organizations’ fundraising efforts, but it is important to also test the crowding-out effect from the perspective of individual decision-making behavior rather than relying on institutional outcomes. In other words, it remains to be seen how government funding influences private giving decision directly, independent of the efforts made by nonprofits to encourage private contributions. This is important because the extent to which an individual gives to a nonprofit may be directly influenced by the information available about the presence or level of government support received by the nonprofit. Previous studies of the crowding-out effect have not addressed this cognitive or behavioral dimension of private giving.
Another research strand argues that government support for nonprofits may encourage private giving, in other words, a crowding-in effect (Schiff, 1990; Smith, 2003). For instance, NEA applicant organizations have to go through a difficult and highly competitive application process that, according to Wyszomirski and Mulcahy (1995), serves as a signal of the quality of the successful recipients’ programs. Potential donors who would like to reward good arts organizations, but without enough time and information to discern differences, may rely on the acknowledgment of merit signified by government funding (Rose-Ackerman, 1986). Therefore, receiving a grant from a prestigious source such as the NEA, or even a state arts council, can be perceived as a stamp of approval signaling artistic excellence. To the extent such funding enhances the organization’s reputation for excellence, it may lead to more rather than less funding from private donors. Smith (2003)’s analysis of financial data showed the larger leverage effect that NEA grants make compared to non-NEA public funding on nonprofit dance companies. Through interviews, Borgonovi and O’Hare (2004) found a majority of arts managers and NEA panelists believe that NEA funding influences private giving with its signaling function and thus receiving a NEA grant can be beneficial to appeal to donors. Interviewed program officers at foundations, however, seem to regard NEA funding not much differently from other merit-based foundation grants even though they generally agree to the positive connotation of having a NEA grant. Nevertheless, the crowding-in effect of government funding remains uncertain and so does the signaling effect of prestigious merit-based funding opposed to other general public supports.
Still, the crowding-out or crowding-in studies introduced above are based on the implicit assumption that individuals possess fairly good knowledge about the level of government funding received by the nonprofits to which they donate. A survey by Horne, Johnson, and Van Slyke (2005), however, challenges the validity of this assumption. Using data from 675 survey respondents, they compared people’s estimates of government funding levels to the actual government funding levels of nonprofits to which the people had donated in the previous year. The survey results found that donors had little knowledge and awareness of the amount of government funding received by the organizations they had supported. Furthermore, over 80% of donors responded that they would give about the same even if a charity they supported were to get an increase in government funding. Yet, it is unclear how these self-reports would predict actual giving behavior. Moreover, there is some evidence that individual donors often make their giving decision to nonprofit organizations for a variety of reasons, such as family tradition or their personal connection to the organization or its mission (Prince & File, 2001). To the extent donors make their giving decisions in ways that do not take account of government funding, even if they have access to this information, we might expect to see neither the crowding-out effect nor the crowding-in effect.
Objectives
Our aim in this study, therefore, is to examine the crowding-out and crowding-in effects on giving to nonprofit arts organizations using an experimental design. In particular, we seek to test how information about the amount of government funding received by a nonprofit arts organization influences the willingness of individuals to donate to the organization. We are interested in the crowding-out effect of the presence of government funding, as opposed to no government funding at all, or what might be called a categorical crowding-out hypothesis. And we are interested in the crowding-out effect of a larger share of government funding, as opposed to a smaller share of the nonprofit’s budget, or what might be called a continuous crowding-out hypothesis. In addition, we aim to test how labeling the government funding as coming from a prestigious source might encourage private donations by signaling the legitimacy of the organization or the quality of its programs—the crowding-in hypothesis. To put things more formally, our study aims to test the following three research hypotheses:
Hypothesis 1 (H1): Individual donors will donate less when they know that a nonprofit receives some amount of government funding. (Categorical crowding-out hypothesis)
Hypothesis 2 (H2): Individual donors will give less if a nonprofit receives a greater share of its funding from government. (Continuous crowding-out hypothesis)
Hypothesis 3 (H3): Knowing that government funding comes from a competitive merit-based program will increase individuals’ willingness to donate. (Crowding-in hypothesis)
Experimental research designs have the important advantage of increasing the internal validity of a study and thus are especially strong studies for demonstrating causal relationships (Remler & Van Ryzin, 2011; Shadish, Cook, & Campbell, 2002). In certain disciplines of social science, such as psychology or political science, the use of experiments and especially survey experiments has become more common in recent years (Morton & Williams 2010; Van de Walle & Van Ryzin, 2011). In the nonprofit sector, only a few studies to date have utilized experimental methods. For example, Hager, Wilson, Pollak, and Rooney(2003) examined the influence of survey research design on the response rate using an experimental setup to find that questionnaire complexity and the use of monetary incentives make no difference but the use of Federal Express mail does. Also, Lin and Van Ryzin (2012) used an experimental design to compare the quality of data from an online and mail survey of nonprofit organizations. But no prior studies that we know of in the nonprofit literature have employed the experimental method to examine the willingness of individuals to donate to the nonprofit sector.
Experimental Design and Participants
To test the hypotheses stated above, we created an online survey experiment in which participants were randomly assigned to one of four arms (or treatment groups), which will be described in detail shortly. The survey instrument was composed of 14 mostly close-ended questions divided into a beginning section with a few general questions about the arts and charitable giving, a main experimental section (with the four experimentally varied arms or treatment groups), and a final section of sociodemographic questions. In the main experimental section, participants were first instructed that they were going to be presented with descriptions of two arts organizations and then asked to divide a US$100 budget between the two organizations. As Figure 1 shows, all respondents were then presented with information on the first nonprofit organization, the Performing Arts Centers (PAC), which was the same across the four experimentally varied groups. The PAC thus served as a point of initial comparison so that respondents had a meaningful alternative to allocate their US$100 budget. Next, respondents were shown a description of the second organization, the Cultural Trust (Trust), but in this case the last sentence of the description of the Trust, referring to its funding, was experimentally varied across the four arms (or groups) as follows: (a) The Trust receives no direct government funding; (b) The Trust receives direct government funding for 33% of its budget; (c) The Trust receives direct government funding for 33% of its budget from the National Endowment for the Arts. This varied description is the experimental treatment, and the aim was to see how it influenced respondents when dividing their US$100 budget between the Trust and PAC. In this survey experiment, the NEA funding in treatment group d represents a case of well-known labeled government grants. Table 1 presents the instructions, images, and full descriptions of the organizations as presented to participants in the experiment. Admittedly, the specified levels of government funding are substantially above the real average level of government funding of arts organizations in the United States. However, the aim of the experiment was to make the levels and variation in government funding salient to respondents in order to clearly test the main hypotheses.
Experimental design.
The Two Nonprofits Shown to Respondents, With Experimental Variations.
After reading about both organizations, respondents were then asked: From your $100 budget, how much would you give to each organization? (You can divide your budget anyway you choose, provided the total adds up to $100.) From the answers to this question, we calculated a mean dollar amount given to the Trust under the varied funding descriptions (treatments a, b, c, and d). To provide another measure of giving preferences, the survey next asked respondents: If you had to give all of your $100 budget to only one organization, which one would it be? In other words, the respondents were forced to choose one organization over the other. The question thus permits us to calculate the proportion of respondents who choose the Trust over the PAC, again under the varied funding descriptions for the Trust that represented the experimental condition (treatments a, b, c, and d).
Participants in the experiment were members of the CivicPanel project, an opt-in email panel of approximately 12,000 active panelists at the time of this study. CivicPanel is a university-affiliated online research project created to provide a general population of volunteers to participate in surveys and online studies about public and civic affairs. Voluntary panelists have been continuously recruited from various online postings including Google advertising, Craislist.org, and the open directory Dmoz.org. A single email invitation was sent to the entire panel in early December 2011, and a total of 719 agreed to participate in the study online. Given the substantive focus on the U.S. nonprofit sector, 133 responses from non-U.S. residents were dropped from the analysis. An additional 24 of those with partially completed responses were also removed, resulting in a final analytical sample n = 562.
Table 2 displays the number of participants in each of the four arms or treatment groups as well as their demographic characteristics and political attitudes. Compared to the U.S. population, there are relatively high proportions of females and Whites in the panel, a feature that Berrens, Bohara, Jenkins-Smith, Sliva, and Weimer (2003) find to be common in other voluntary online panels in the United States. Although this limits the external validity of studies using online panels, there is some evidence to suggest that, at least for some topics, this type of voluntary panel can approximate survey results based on probability sampling of the U.S. population (Van Ryzin, 2008). Even viewed simply as a group of volunteers for an experimental study, which focuses more on internal validity, the participants in our study are still diverse in terms of ages, incomes, education levels, and regions of the country. In a typical full randomized setup, treatment, and control groups should have the same characteristics except for the treatment they are given (Remler & Van Ryzin, 2011). The lack of statistically significant differences across groups randomly assigned to each condition in Table 2 confirms the statistical equivalence of the experimental groups.
Demographic Comparison Across Groups.
Results of the opening questions, 1 which took exact wordings from previous national surveys about the attitudes toward the arts and donating, provide more insight into the representativeness of the study participants. About a third (32%) of study participants said they attended any classical music performances in the previous year, much higher than the proportion (10%) in a national survey by the NEA in 2008. Also, the percentage of study participants who said they donated to the arts in the previous year (37.5%) was more than double the rate of donating (16.9%), found in the Giving and Volunteering in the United States (Independent Sector, 2001). Although not representative of the general population, the participants in our study clearly appear interested and supportive of the arts and thus perhaps more closely resemble the potential donor community for many U.S. arts nonprofits. Given the relatively high percentage of the sample who express interest or support of the arts, we will analyze the results separately for arts patrons (defined as those who reported having been to a live classical music performance in the last 12 months) and for arts donors (defined as those who reported having contributed money or property to the arts, culture, and humanities in the last 12 months). Both arts patrons and arts donors have higher levels of income and educational background, on average, compared to the nonpatron and nondonor groups. Also, patrons and donors are more likely to express support for government spending on arts, culture, and humanities.
Results
Table 3 reports comparisons for the average amount of giving to the Trust made by individuals in treatment groups a through d, as well as comparisons of the percentages of participants who chose to give all of their US$100 to the Trust over the PAC. Compared to the average of US$47 given to the Trust by those in treatment group a (no government funding), respondents in treatment groups b and c (33% and 67% government funding, respectively) both gave about 25% less, a substantively and statistically significant decline in giving. This comparison of group a with groups b and c effectively confirms the categorical crowding-out effect (Hypothesis 1). There was no significant difference, however, between treatment group b (33% government funding) and treatment group c (67% government funding), with group b giving an average of US$35 and group c a nearly identical average of US$36. This finding clearly suggests respondents were sensitive only to the presence of government funding, not the amount of government funding as a share of the organization’s budget. In short, there is no evidence for the continuous crowding-out effect (Hypothesis 2). In addition, the results show that the NEA branding of the 33% government funding in treatment group d, compared to treatment group b (33% government funding, with no mention of NEA), increases donations slightly to US$39 from US$35. This difference, however, is small and not significant statistically. Thus, our results provide little support for the crowding-in effect of a prestigious government grant (Hypothesis 3).
Results of the Experiment.
Note: p-values based on a two-sample t-test (two-tailed) for dollars given (which are means) and a two-sample z-test (two-tailed) for percent given (which are proportions). Tests were done only on theoretically meaningful differences (see text).
Table 3 also reports comparisons of the percentage choosing the Trust over the PAC. Overall, the comparisons closely parallel the findings for the mean amount of giving. While 38% of respondents in treatment group a (no government funding) chose to give all their money to the Trust, only 21% of respondents in treatment group b (33% government funding) and 22% of respondents in treatment group c (67% government funding) made this same choice. This finding again clearly provides support for a categorical crowding-out effect (Hypothesis 1). And the decline by almost half in the proportion willing to give all their money to the Trust is statistically and substantively significant. There was again no significant difference, however, in percentages between those in group b and group c (33% and 67% government funding, respectively), again providing no evidence to support a continuous crowding-out effect (Hypothesis 2). Finally, the NEA branding represented by treatment group d, compared to group b (the same funding amount, but without the NEA branding), shows that slightly more people were attracted to giving to the Trust when the NEA was mentioned as the government funding source. However, the 5-point difference is not statistically significant, thus again failing to provide clear evidence for a crowding-in effect (Hypothesis 3).
As mentioned, we were able to analyze the results of the experiment separately for arts patrons (defined as those who attended a live classical music performance in the last 12 months) and for arts donors (defined as those who contributed to the arts, culture, and humanities in the last 12 months). As Table 3 shows, the crowding-out effects of government funding in both the patron group and the donor group appear much weaker. Indeed, the mean amounts given to the Trust by both patrons and donors are not statistically different across treatment groups, although the means do trend slightly lower when government funding is mentioned. And the proportion of patrons and donors giving all of their money to the Trust was significantly lower (in the statistical sense) only in the most extreme case when the Trust was said to have 67% of its funding from government (and even then the difference was only marginally significant statistically). In contrast, the crowding-out effect remains quite strong for nonpatrons and nondonors, both in terms of the mean dollar amount given to the Trust and in the percentage choosing to give all of their money to the Trust. Thus, it seems that most of the crowding-out effect evident in the sample as a whole comes from participants who are neither patrons of the arts nor recent donors to arts organizations.
Discussion and Implications
Findings from this experimental study of giving to nonprofit arts organizations provide clear evidence for a categorical crowding-out effect, with the presence of government funding leading to a decline of 25% in the dollar amount given and to only about half as many respondents willing to donate all of their money to the organization. But there is no evidence that the level or share of government funding in a nonprofit’s budget influences the decision to donate, suggesting no continuous crowding-out effect. There was also no support found for the crowding-in effect of prestigious government giving as a marker of the organization’s quality or performance. However, the subgroup analysis of arts patrons and arts donors suggest that the crowding-out effect may be much weaker in these two key subpopulations; in other words, most of the decline in the willingness to give attributable to government funding appears to come from people less likely to give to the arts to begin with. Given the experimental design of this study, these results are robust with respect to the strength of causal evidence and thus have important theoretical as well as practical implications that we will discuss shortly.
But first, we should point out that the findings of this study need to be interpreted with caution for several reasons. In the first place, our findings are based on a simulated setting that may not generalize to actual giving behavior in the real world. Although the somewhat unrealistically high level of government funding presented in our survey was intended to make the experimental variation salient to respondents, it does limit the realism of the study. Also, participants knew they were involved in a hypothetical exercise and that they were not allocating their own (real) money despite our best efforts to make the hypothetical organizations and the budget task realistic. In future studies, it might be interesting to do an experiment in which participants are given real cash and stamped envelopes addressed to the organizations in order to provide more realism to their decision-making task.
It should be also noted that this study focuses only on arts organizations, thus the results may not be directly relevant to other types of nonprofits with different funding structures. For example, while arts organizations depend a great deal on private donations, many health and human service organizations rely much more on payments from government through contracts or reimbursement programs (Young, 2010). Nevertheless, it is clear that most nonprofits depend on a mix of both private donations and government support, regardless of the different funding mechanisms. As a consequence, we would argue that the conclusions of this research still may have implications for the nonprofit sector in general, although of course it is important to replicate the finding of a crowding-out effect using examples of other types of nonprofit organizations.
Although our experiment tested a crowding-out effect from government funding, it is not known whether a similar disincentive to individual giving would happen also in the case of corporate or foundation funding. Individual donors may decide to give less, or not to give at all, whenever they know that another large donor has already given a substantial amount to the organization. In other words, the crowding-out effect may not be limited to public grants per se but could be relevant to any large sum of money coming from well-known philanthropists, big corporations, or foundations. Certainly, future experiments along these lines could test this hypothesis as well. Another issue concerns individual income elasticity as a possible factor in determining the level of individual giving. In other words, people may change the amount of giving depending on how large a budget they have to allocate. Thus, replicating the study using either a larger or smaller total charitable budget may provide additional insights.
Despite these limitations, the findings from this study are especially meaningful because they add to those of previous empirical investigations that remained ambiguous with respect to the direction of the causal relationship, mostly because of hidden factors not easily controlled in nonexperimental research. In particular, the experimental design of this study effectively removes the complicating influence of an organization’s fundraising efforts on private donors. In nonexperimental studies based on existing data, the correlation may reflect the fact that organizations engage in a stronger effort to raise private donations after experiencing cuts in government funding. So far, most studies on the crowding issue have not incorporated the fundraising effect on charitable giving into their empirical estimations, with the exception of a few studies such as Andreoni and Payne (2003, 2011) and Dokko (2009). Taking advantage of the experimental setup employed in our study, we can say with some certainty that the presence of government funding can causally influence individual giving behavior.
Moreover, prior studies have often been based on the implicit assumption that all individuals obtain information about the level of government funding and act on that knowledge. Nonetheless, not every private donor has equal access to information on the level of government funding or the same amount of information regarding the budgets of organizations to which they contribute. Thus, it has not been possible to accurately measure intentions about the amount of giving when the level of government funding varies. The experimental approach used in this research made it possible for all donors to have the same information regarding the levels of government funding, since this was a controlled element of the manipulated scenario. Thus, this research is the first to remove both the influence of fundraising and varying levels of information regarding the amount of government funding in its empirical estimation of the crowding-out effect. The recent growth of online tools such as Charity Navigator or Guidestar help donors to become better informed about organizations they would like to support (Hager & Greenlee, 2004). Although nonprofits are required to allow public inspection of their Form 990, and various institutions mentioned above make it available online, the information available to donors is still far from perfect. This study clearly shows, however, that such information does have the potential to influence people’s willingness to donate.
Understanding the influence of government grants on charitable giving to the arts is especially critical at a time of fiscal crisis that typically affects arts funding over other government spending. If government aims in part to enhance the decision making powers of individual donors, subsidizing the arts through grants may require further conditions on grantees with respect to sharing information about their receipt of public money.
The lack of evidence for a continuous crowding-out effect may raise a concern. If the mere existence of government support crowds out private giving, regardless of the level of support, the consequence of minimal public funding could even be a reduced total amount of charitable contributions. Nevertheless, we should note that the crowding-out effects were clearly weaker among those much more likely to give to nonprofit arts organizations. Therefore, there may not be much of a crowding-out effect in the real world of nonprofit arts organizations and their donor communities. It is also noteworthy to recall Brooks’ (2000) detection of a nonlinear inverted U-shape relationship between government funding and private giving. That may explain the finding of no difference identified in this research with respect to the magnitude of government funding, which was admittedly somewhat high with respect to actual average funding levels (as discussed earlier). In particular, his research asserted that public subsidies first leverage or crowd-in more individual giving but then starts to displace or crowd-out donations as the level of government grants increases. Therefore, it would be worthwhile to run another experiment that tests for crowding-out effects at much lower (and thus more realistic) levels of government funding, for example at 5% and say 20% as the two extremes. Such follow-up experiments would be important to better and more thoroughly test the continuous crowding-out hypothesis, as well as the Brooks’ U-shape relationship hypothesis.
Finally, the lack of evidence for the prestige effect of an NEA grant raises questions about the signaling function of merit-based government funding. The survey results suggest that having an NEA grant itself does not seem to work as an indicator of unobservable qualities or capacities of an organization to leverage private giving. Conversely, earlier empirical study showed a leveraging effect of NEA grants on other charitable contributions (Smith, 2003). It could have been positively related to successful fundraising through the strategic promotion of an organization’s status as an NEA grant recipient. Perhaps nonprofit arts organizations emphasize their receipt of NEA grants as a testament to their worthiness when they appeal to donors, resulting in more private contributions. One way of investigating this idea would be to survey key fundraising or budgeting personnel in nonprofit organizations. An experimental approach could also be used to test if a fundraising campaign that highlights or emphasizes NEA funding would influence donations.
In summary, our survey experiment about charitable giving to nonprofit arts organizations provides some support for a crowding-out effect in the nonprofit arts sector. Nonetheless, the effect seems strongest among those least likely to give in the first place. Future studies may focus on extensions of this experiment to investigate the influence of varied levels and types of grants coming from both public and private institutions on individual charitable giving.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors received partial support for their research from the Alfred P. Sloan Foundation.