Do Donors Penalize Nonprofits With Higher Non-Program Costs? A Meta-Analysis of Donor Overhead Aversion

Abstract

Disparate research on overhead aversion and nonprofit starvation can benefit from a conceptual model that explains their relationships. Following resurrection of such a model, we focus on one important piece: the relationship between overhead spending and nonprofit donations. Studies on this topic have produced inconclusive results. Our meta-analysis clarifies the relationship by synthesizing a sample of 30 original studies with 244 effect sizes. We uncover a negative association suggesting that donors penalize nonprofits with higher overhead costs. Moreover, our meta-regression models reveal that experimental designs detect higher donor aversion than studies that use other research designs and that amateur donors have more intense overhead aversion than professional donors. However, studies that measure administrative costs do not report more negative effects than studies that measure both administrative and fundraising costs. The overall contribution of the meta-analysis solidifies the conceptual link between reported capacity costs and funders’ giving decisions, a key arc in the nonprofit starvation cycle.

Keywords

overhead aversion overhead costs meta-analysis nonprofit organizations administrative costs donations fundraising management

When an international relief organization appeals for my donation to help a starving child, I may well expect that my $100 will be spent directly on starving children. The fact that the organization also needs my donation to support executive and human resource management, information technology, and future fundraising efforts might not be clear in the appeal. If it was, I might make different choices: I might consider these costs to be a diversion of funds from program services, and I might thereby re-evaluate my gift. “Overhead aversion” refers to donor’s distaste for non-program spending, including fundraising and other administrative costs like human resource, financial, and technology management.

Donors are not always averse to non-program spending. Scholars report that overhead spending is indeed sometimes negatively related to donations (e.g., Greenlee & Brown, 1999; Marudas et al., 2014), but sometimes positively related (e.g., Ashley & Van Slyke, 2012; Frumkin & Kim, 2001), and sometimes unrelated. These inconsistencies call for a need to place the puzzle in a conceptual framework and evaluate what prevailing scholarship tells us about donor reactions to fundraising and other administrative spending.

The following article proceeds in three sections. First, we situate overhead spending in a model of organizational behavior. Second, we incorporate findings from previous studies of overhead spending into a meta-analysis that summarizes scholarship on a key part of the model: the relationship between overhead costs and the amount of contributions an organization receives. Third, we investigate how characteristics of different studies (or their statistical models) influence the “effects” reported in those studies. Variation across studies raises the question of which forces (some methodological, some substantive) explain differences in effect sizes and direction. An example is whether individual donors view overhead differently from those who are prone to scrutinize organizational finances and outcomes more closely, such as foundations or a nonprofit’s own trustees. Our goal is to establish both a conceptual and empirical marking point for future scholarship regarding the relationship between overhead spending and fundraising success.

Overhead Aversion and the Starvation Cycle

Nearly two decades ago, researchers at the Urban Institute and Indiana University (Wing, Hager, et al., 2004) conducted surveys, case studies, and literature reviews under the auspices of their Nonprofit Overhead Cost Project. The project focused on field research and education, consequently influencing how the BBB Wise Giving Alliance and Charity Navigator defined their spending ratios, how the Internal Revenue Service (IRS) asked public charities to account for expenses associated with fundraising contractors, and how industry watchdogs in 2013 advanced a campaign to raise awareness of “the overhead myth.” Perhaps most influential, however, is the project’s articulation of the dynamics of donor decisions and nonprofit spending that was later branded by Gregory and Howard (2009) as the nonprofit starvation cycle. Gregory and Howard cite the Nonprofit Overhead Cost Project as the progenitor of the cycle. However, since the cycle was never presented in an academic article or book, it did not enter into the literature beyond Gregory and Howard’s popular field article.

The original (unpublished) model can provide a much-needed map for research on the value of and popular aversion to overhead. The Nonprofit Overhead Cost Project’s conceptual offerings arguably hit its pinnacle in May of 2004, when Ken Wing presented the project’s culminating ideas at the 3rd International Conference on Systems Thinking in Management (Wing, Pollak, & Rooney, 2004a; see also Wing, Pollak, & Rooney, 2004b). Later in November, Wing and Hager (2004) presented a simplified version of the May model, and it was this simplified version that Gregory and Howard branded as the starvation cycle and it became viral. The more sophisticated Wing Model was buried. In Figure 1, we adopt a revival of the original Wing Model as a means for placing overhead spending in the broader context of organizational behavior. The model provides a more thorough and nuanced treatment of the relationships between nonprofit behavior, donor choices, and various outcomes. The model also helps sort and organize a recent spate of starvation cycle research that sheds light on various portions of the model.

Figure 1.

Wing Model of Limited Nonprofit Organizational Effectiveness.

Although only the final model is presented here, Wing presented it in developmental stages. He described the left one third of the model as a virtuous cycle of effectiveness, wherein effective nonprofits attract resources that reinforce further effectiveness. The positive sign represents the constructive relationships among these forces. We place an A in the model to call attention to research that considers how overhead spending might positively influence program effectiveness. Marwell and Calabrese (2015) and Kim (2017) investigate this relationship, but Altamimi and Liu (2022) theorize it specifically in the context of overhead aversion and the starvation cycle. Altamimi and Liu specify and empirically validate a curvilinear influence of administrative costs on effectiveness of arts nonprofits, with non-program spending improving outcomes up to a point, but then harming outcomes when this spending rises above a certain level. This squares with the Wing Model to the extent that reasonable and sufficient overhead spending contributes to the virtuous (left) side of the model. Altamimi and Liu’s work suggests that overhead aversion might arise when administrative spending exceeds a threshold. The middle and right-hand thirds of the Wing Model then become relevant.

A first complication arises in the middle portion of the model when capacity spending is reported to funders. Although some funders might be attracted to such capacity spending, Wing asserted that aversions to these costs might cause some funders to re-think their relationship with these organizations. The negative cycle, what we might call the overhead aversion loop, limits the ability of nonprofits to continue investing in capacity beyond the level of funder preferences.

The right side of the model houses the starvation cycle loop. The key dynamic here is underreporting of capacity costs (Wing et al., 2005). This underreporting might be unintentional, a consequence of inadequate investment in financial tracking systems. It might also be intentional, an effort to manage the public face of how the nonprofit is spending its money. If funders have unrealistic expectations about how much capacity spending is consistent with organizational effectiveness, nonprofits face a dilemma. If they spend what is necessary to be effective, they risk penalization from their funders. If they limit capacity spending in accordance with funder preferences, they risk reducing their effectiveness. In the short term, intentional underreporting of capacity costs appears to be a way out. They can spend on capacity what they need to be effective while appearing to be good financial stewards to their funders. The difficulty with this strategy is that, in the long term, funders are socialized to see artificially low-capacity spending as normal. To the extent that funders reduce their preferred level of capacity spending, pressures on nonprofits to keep capacity spending low only increase.

Point B demarcates a recently popular point of empirical tests related to the starvation cycle. Both Lecy and Searing (2015) and Schubert and Boenigk (2019) focus on changes in the reporting of overhead costs over time. The under-reporting of capacity costs (labeled model Point C) is not considered in those tests but can be found in Trussel (2003), Krishnan et al. (2006), and others who consider accounting manipulation as a likely response to overhead aversion among donors. Different studies shed light on particular pieces.

The model is multi-faceted, and no one empirical article will test all its dimensions. Accordingly, the study you are reading now focuses on a separate portion of the Wing Model, which we label D in Figure 1. The question we consider is whether overhead costs influence the contributions that nonprofits attract. Consistent with the Wing Model, Bowman (2006) argued that donors hold competing perspectives on the costs. One perspective is that the costs are wasteful. A nonprofit with high overhead costs features “excessive salaries, numerous perquisites, and unnecessary staff” (p. 292). So, such costs might signal that a nonprofit does not use its resources productively. Overhead aversion reigns: Instead of spending money on administrative and fundraising matters, nonprofits are expected to allocate more of their expenses into program services (Bowman, 2006; Buchheit & Parsons, 2006; Hager, 2003; Keating & Frumkin, 2003). Donors might only support nonprofits with low or no overhead costs (Curran & Bonilla, 2010; Gneezy et al., 2014). Donor expectation of low or no overhead creates the cycle that threatens to compromise the effectiveness of nonprofit organizations and leads to “an underdeveloped nonprofit sector and a loss of community trust and confidence in philanthropy” (Hager & Flack, 2004, p. 4).

The other perspective regards administrative costs as a necessary capacity expense, key for an effective nonprofit (Altamimi & Liu, 2022; Hager et al., 2004). As illustrated in the virtuous (left) side of the Wing Model, nonprofits are theoretically most effective when adequate staffing, salaries, and other supports are available in sufficient measure. The case is perhaps clearer when fundraising costs are considered: Reasonable donors do not begrudge a nonprofit spending money to advertise its needs, and organizations that spend money on fundraising generally raise more money.

So, non-program spending can be either a boon or a drag on fundraising, depending on the amount spent and the attitudes of target donors. Point D on Figure 1 poses the question, do donors penalize or reward nonprofits with high overhead costs? The Wing Model asserts a negative relationship, with giving slowed as non-program spending increases. However, empirical studies on this question have produced inconsistent results. Different researchers study different empirical situations, but the sources of variation in research findings might be related to the use of different measures, data, or research methods in original studies. Such inconsistency in results on a central conceptual question is ripe for meta-analysis. To shine a bright light on Point D in the Wing Model, we turn our attention to the question of whether scholarship supports or refutes the overhead aversion thesis.

Method

Meta-analysis is appropriate when inconsistent and contradictory findings in the literature constitute a barrier to offer clear and coherent knowledge for both research and practice. The method offers a set of techniques to summarize mixed findings. Magnitudes of relationships between variables reported in original study models, or what meta-analysis calls effect sizes, are the inputs for meta-analysis. Comparison of these effects leads to conclusions about why the relationship varies across studies. Meta-analysis is an analysis of analyses. The method was defined by Glass (1976) as a “statistical analysis of a large collection of analysis results from individual studies for the purpose of integrating the findings” (p. 3). Over the past decades, meta-analysis has become a major research method in the fields of education, psychology, and medicine (Shadish & Lecy, 2015). The method has penetrated into the field of public and nonprofit management in recent years. Meta-analyses that consider a broad or holistic concept (such as “performance”) might include hundreds of studies (Gurevitch et al., 2018), while meta-analyses that focus on specific relationships between variables (such as government grants and donations) are somewhat smaller, limited to those studies that have examined the association. A meta-analysis of overhead spending and donations only became possible in recent years, since a substantial majority of research on this topic has been completed over the past decade.

Sample Selection

The first step in conducting meta-analysis is the selection of relevant quantitative studies. The first conceptual choice for our study was whether to include a related literature on price of giving, which also considers donor response to non-program costs. Since overhead ratio (non-program spending/total spending) is calculated differently than price (total spending/program expenses), model effects are not directly comparable. Calabrese (2011) summarizes the conceptual difference between these measures, noting that price “measures how much a donor would have to give. . . to generate a single dollar of nonprofit output,” whereas the overhead ratio “is an efficiency measure used to compare similar nonprofit organizations” (p. 862). Calabrese calculates and uses both measures in his models and reports differential effects. We follow the field’s lead in differentiating overhead costs and price, despite their conceptual overlap. We leave price aside and focus on the literature on overhead costs. However, since program spending is a simple inverse of non-program spending, we include studies that consider the relationship between program spending and donations.

To select studies, we employed three literature searches and two reality checks suggested by Lipsey and Wilson (2001) and Ringquist (2013). First, we used three academic databases (EBSCO, Web of Science, and ProQuest) to search for relevant studies. Keywords combined with the Boolean operators “AND” or “OR” for searching were as follows: (nonprofit OR not-for-profit OR nonprofit OR charitable organization OR voluntary organization) AND (overhead OR administrative cost OR administrative expense OR management cost OR management expense OR general cost OR general expense OR efficiency OR program spending OR program ratio OR program expense) AND (donation OR contribution OR giving OR grant OR funding). The searches of the academic databases yielded 2,899 candidate studies.

This initial pool of candidate studies was source data; few would ultimately include tests of the relationship that is the subject of our meta-analysis of overhead aversion. We reviewed the empirical details of each manuscript and decided which candidate studies to include based on the following criteria. First, we define overhead costs as non-program costs, defined as administrative costs and fundraising costs or inversely as program costs. Administrative expenses are incurred in the provision of social and public services in the nonprofit organizations such as human resource management, general legal and accounting services, office management, and management of investments. Fundraising expenses are incurred to generate financial or in-kind contributions, including grants. Accounting standards and IRS Form 990 ask nonprofits to divide all nonprofit costs across program, administrative, and fundraising expenses.

Many consider fundraising expenses to be a dimension of overhead. Early quantitative scholarship on overhead aversion gives more attention to administrative costs and tends to leave fundraising costs aside. More recently, experimental studies have shifted the attention back to both administrative and fundraising costs. Therefore, we include early quantitative scholarships that focus on administrative expenses and recent experimental studies that used “overhead” scenarios that refer generically to non-program spending, including both administrative and fundraising expenses (Gneezy et al., 2014; Kinsbergen & Tolsma, 2013).

As a second criterion, our subject studies include the estimations of the relationship between spending and donations. Donations refer to financial contributions received by subject nonprofits from individuals, foundations, corporations, federated campaigns, and other sources that are neither earned (commercial) nor investment income. Third, we included studies on any country, but, fourth, we restricted our sample to those manuscripts written in English. Fifth, qualitative studies were excluded, since they do not quantify the relationship that is the unit of analysis in the meta-analysis. Sixth, unpublished studies were included to guard against overestimation of effects due to publication bias. Finally, studies with insufficient information to compute effect sizes were excluded. Following these criteria, the searches of the academic databases yielded 16 eligible studies.

We then performed backward and forward searches on the eligible studies to identify literature that was not flagged in the initial database searches. Moreover, we scanned programs of relevant academic conferences (e.g., ISTR, ARNOVA, and ASPA) and archives of working papers (e.g., SSRN) for relevant unpublished studies. The reality checks were a scan of online tables of contents of three leading journals (Nonprofit and Voluntary Sector Quarterly, Nonprofit Management & Leadership, and Voluntas) and searches via Google Scholar. After the searches and reality checks, we included in the meta-analysis 27 studies that reported a statistical relationship between overhead costs and donations. The last search date was July 27, 2022. In addition, we included three more studies identified by NVSQ reviewers, bringing the tally to 30 studies. Figure 2 summarizes the study sample selection process.

Figure 2.

Original Study Collection and Selection Process.

Measure Specification

Following Ringquist (2013), we coded the subject studies for effect size calculations and meta-regression analyses. In the language of meta-analysis, an “effect” is the magnitude and direction of the relationship between the forces under study (in our case, proportion of overhead spending going out and donations coming in). Some studies yield more effects than others because their estimates are produced by models using different samples, different sets of controls, or different assumptions about the error term. Different studies represent the statistical relationship in different ways. For those studies that reported t-statistics (e.g., Tinkelman & Mankaney, 2007), we used t-statistics to calculate the effect sizes. We used corresponding t-statistics to estimate the effect sizes when studies (e.g., Marudas et al., 2014) only report a level of statistical significance, but not standard errors.¹ For those studies that reported parameter estimates that were not statistically significant (e.g., Lee et al., 2018), we recorded effect sizes as zero (see Note 1). We used z-statistics to calculate effect sizes when studies estimated models using maximum likelihood (e.g., Krawczyk et al., 2017). We flipped the signs when studies used program expenses to predict donations to make them comparable with non-program expenses. This process yielded 244 effects reported in the 30 studies (see Table 1). Of the 244 effects, 70 indicated positive association between overhead spending (ratio to total spending) and donations, 17 indicated no association, and 157 indicated negative association.

Table 1.

Selected Studies, Effects, and Moderator Characteristics (n = 30 Studies, 244 Effects).

Study	Number of effects	Average effect size	Weight (%)	Experimental design	Advanced models	Administrative costs only	Amateur donors	Form 990 data	Longitudinal data	Subsectors
Aryee (2015)	1	−0.58	0.12	No	No	Yes	No	No	Yes	Various
Ashley & Faulk (2010)	2	0.00	0.36	No	Yes	Yes	No	Yes	No	Various
Balsam & Harris (2013)	15	−0.01	12.87	No	Yes	No	No	Yes	Yes	Various
Blouin et al. (2018)	1	−0.04	0.89	No	No	No	No	Yes	No	Various
Calabrese (2011)	18	0.02	14.34	No	Yes and No	Yes	No	Yes	Yes	Various
Charles (2018)	4	−0.05	0.72	No	Yes	Yes	No	No	Yes	Arts
Charles et al. (2020)	2	−0.19	0.34	Yes	N/A	No	Yes	No	No	Social services
Frumkin & Kim (2001)	6	0.01	1.11	No	Yes	Yes	No	Yes	Yes	Various
Gneezy et al. (2014)	8	−0.10	1.28	Yes	N/A	No	Yes	No	No	International development
Greenlee & Brown (1999)	3	−0.14	0.69	No	No	Yes	No	No	No	Various
Grizzle (2015)	1	−0.03	0.18	No	No	Yes	No	No	Yes	Arts
Harris et al. (2015)	30	−0.01	29.81	No	Yes	No	No	Yes	Yes	Various
Harris et al. (2019)	2	−0.02	1.95	No	Yes	No	No	Yes	Yes	Various
Jacobs & Marudas (2009)	12	−0.05	2.08	No	No	Yes	No	Yes	No	Various
Kinsbergen & Tolsma (2013)	4	−0.22	0.75	Yes	N/A	No	Yes	No	No	International development
Krawczyk et al. (2017)	5	−0.16	0.96	No	Yes	Yes	Yes and No	No	Yes	Arts
Lee et al. (2018)	2	0.01	0.39	No	Yes	Yes	No	No	Yes	Arts
Marudas et al. (2014)	6	−0.18	1.04	No	Yes	Yes	No	No	Yes	Various
Nicholson-Crotty (2011)	10	0.01	9.41	No	No	Yes	No	Yes	Yes	Various
Nikolova (2015)	18	−0.04	3.36	No	Yes	Yes	No	No	Yes	International development
Peng et al. (2019)	1	−0.90	0.15	Yes	N/A	No	Yes	No	No	Arts, culture, and education
Portillo & Stinn (2018)	8	−0.17	1.34	Yes	N/A	No	Yes	No	No	International development
Qu & Daniel (2021)	1	−0.27	0.18	Yes	N/A	No	Yes	No	No	Disaster relief
Quosigk & Forgione (2018)	1	0.08	0.47	No	No	No	No	Yes	Yes	Various
Steinberg (1986)	49	0.11	4.90	No	Yes and No	Yes	No	Yes	Yes	Various
Surysekar et al. (2015)	5	−0.03	5.08	No	Yes	No	No	Yes	Yes	Various
Tian et al. (2020)	2	−0.12	0.24	Yes	N/A	No	Yes	No	No	Environment
Tinkelman (1999)	6	0.07	1.13	No	No	Yes	No	No	Yes	Various
Tinkelman & Mankaney (2007)	18	−0.01	3.41	No	No	Yes	No	Yes	Yes and No	Various
Xie et al. (2016)	3	−0.10	0.46	No	No	Yes	No	No	Yes	Public benefits
Total (overall)	244	−.03	100	7	13	17	8	13	19	N/A

We combined the 244 effects to estimate an average effect size across studies. The summary effect magnitude and direction is −.03 (z = −9.40, p ≤ .01), with a 95% confidence interval of [−.04, −.03], when both experimental and non-experimental studies are included in the analysis. The effect size and significance level attenuate when experimental studies are not included: −.02 (z = −5.07, p ≤ .01), with a 95% confidence interval of [−.02, −.01]. Both average effect sizes are small. So, when all available studies are considered together, the overall pattern indicates a negative and statistically significant association between overhead spending and donations. Donations decrease as overhead spending increases.

Since the average effect sizes are small, we further examined why, though, studies report conflicting evidence on this question? The third aim of this article is to provide guidance for future scholarship on the methodological and substantive conditions that might drive the detection of overhead aversion. Toward this end, we conducted a moderator analysis that has become common in meta-analyses published in the fields of nonprofit and public management (Hung & Lu, in press; Ringquist, 2013). In the next section, we present our method, define the moderators, and describe findings that help us understand the diversity of findings in the empirical literature.

Analysis of Moderators

Table 3 presents random-effects meta-regression models that explore the variation in effect sizes due to assorted potential moderators.² We prefer the general estimating equations (GEE) modeling approach since it allows us to address effect size heteroscedasticity and non-independence of observations issues (Liang & Zeger, 1986; White, 1980). Moreover, GEE models are a better fit for our situation since the approach places less weight on original studies that report a large number of effect sizes, which in turn produces fewer biased estimates. In our case, Steinberg (1986) contributes around 20% of the effects to this meta-analysis. We prefer the technique that attenuates the weight of Steinberg’s (1986) study on the meta-analysis; therefore, we consider GEE models to be most appropriate in our case. The dependent variable in all Table 3 models is effect size; the independent variables in the models are moderators selected to explain the variation in effect sizes.

Although the measures of overhead are consistent, original studies in the meta-analysis exhibit differences in theoretical argument, measurement, sample, design, analysis choices, and study quality. These differences might be the factors that cause the variation in effect sizes or direction across studies. To test the sources of the variation we consider six moderators (see Table 1). To facilitate analysis and explanation, we divide the moderators into categories of research design, substantive factors, and data structure.

Research Design Moderators

Experimental Design Versus Non-Experimental Designs

One complication is that our analysis includes studies that use both experimental and non-experimental research designs. The differences in research design compromise their comparability. Experimental designs have become increasingly popular in nonprofit research. In general, a well-designed experiment is conducted in a controlled environment where the causal relationships between overhead spending and donations can be identified. Non-experimental studies are prone to endogeneity and omitted variable issues that might bias analysis (Stock & Watson, 2012). We, therefore, created a moderator variable to test whether measurement of overhead aversion varies according to research design choice. That is, we coded all effects according to whether they are derived from an experimental design (1) or a non-experimental design (0), and then test whether this distinction represents a statistically significant difference in effect (i.e., relationship between overhead costs and donations).

Advanced Regression Models Versus Basic Regression Models

The second research design moderator distinguishes effects derived from advanced models and effects derived from basic ones. Meta-analyses have been criticized for throwing bad studies together with good ones. How one defines quality is subjective, but recent meta-analyses have considered whether research techniques in subject studies seek to reduce estimation bias (de Wit & Bekkers, 2017; Lu, 2016). Studies that use fixed-effects, instrumental variable, or other advanced regression models to handle omitted variable bias and endogeneity issues seek to produce less biased estimates (Stock & Watson, 2012). We, therefore, test whether the association between overhead spending and donations varies according to whether the estimates are generated from advanced regression models (1) or basic regression models such as ordinary least squares, logistic, or probit (0).

Substantive Factor Moderators

Administrative Costs Versus Administrative and Fundraising Costs

Our subject studies differ in what non-program costs are included in the overhead costs numerator. Many studies expressly consider only administrative costs, while others add in fundraising costs. Although we consider both to be dimensions of overhead, each plays a different role in organizations and may be seen differently by donors. Lecy and Searing (2015) report that nonprofit administrative expenses have fallen substantially since 1985 while fundraising expenses have almost doubled. This suggests that donors might have more intense aversion toward administrative costs than fundraising costs (Portillo & Stinn, 2018). We, therefore, assume that subject studies that measure overhead costs as administrative costs (1) might produce less positive effects than subject studies that measure overhead costs as both administrative costs and fundraising costs, or inversely as program costs (0).

Amateur Versus Professional Donors

Another potential influential source of variation across studies concerns the nature of the donor. Different types of donors may have different sensibilities regarding overhead costs. We label donors as amateur when they respond to appeals rather than study financial statements or outcome measurements. Individual donors with distance from the operations of organizations are more likely to be based on instinct (Bowman, 2006). In contrast, professional donors such as foundations, businesses, or even a nonprofit’s own trustees may be more likely to incorporate nonprofits’ financial and program performance or reporting into their giving decisions (Balsam & Harris, 2013; Kitching, 2009; Parsons, 2003; Tinkelman, 1998). The defining characteristic of the professional donor is their additional investment in time in understanding the operations of the subject nonprofit. We code effects accordingly, between models that draw on amateur (1) versus professional donor (0). However, we were not always able to distinguish since some studies aggregate contributions from all sources; these effects are coded as (2) for this moderator.

Data Structure Moderators

IRS Form 990 Data Versus Other Data

The IRS Form 990s are readily accessible to the public and are a common data source on nonprofits in the United States. The section on functional expenses readily separates self-reported program, administrative, and fundraising spending. Some donors consult the form when making giving decisions, and this visibility has given the public the opportunity to scrutinize nonprofits. American nonprofits find themselves under pressure of the public to manage overhead numbers on financial statements, as suggested by both the Wing Model and a quarter-century of research (Froelich et al., 2000; Froelich & Knoepfle, 1996; Hager, 2003; Krishnan & Yetman, 2011; Lecy & Searing, 2015). If costs are under-reported on public reports (as illustrated at point C of the Wing Model), then IRS Form 990 would be susceptible to this error. In our first data structure moderator, we test whether the effect of overhead varies according to the data that original studies use. This moderator is coded 1 for studies that use IRS Form 990 data and 0 for studies that use other data.

Longitudinal Data Versus Cross-Sectional Data

A second data structure moderator investigates the potential value of longitudinal studies in detecting the association between overhead spending and donations. Unlike cross-sectional data, longitudinal data track the dynamic of the relationship between overhead and donations over time. Studies that use longitudinal or panel data to estimate parameters are equipped with different variance components, which is considered superior to studies that use cross-sectional data. Thus, we included a moderator to test whether the association between overhead spending and donations varies according to the ability of analysts to measure change over time. We assigned a value of 1 for studies that used longitudinal or panel data and 0 for studies that used cross-sectional data.

Results

Following de Wit and Bekkers (2017), we take experimental versus non-experimental design as our first moderator and test it separately from other moderators. We then examine non-experimental studies to test whether the effect of overhead varies according to regression techniques, overhead measures, donor types, data sources, and data structure. Of the moderators examined, the highest correlation is 0.55, between models using advanced regression models and models using longitudinal data (see Table 2). The mean variance inflation factor is 1.47, which suggests negligible collinearity between the variables. Model results are presented in Table 3. The results from the moderator analyses follow.

Table 2.

Correlations Among the Moderators in Non-Experimental Studies.

No.	Variables	1	2	3	4	5
1	Advanced models	—
2	Administrative costs	−.41	—
3	Amateur donors	.08	.07	—
4	IRS Form 990	.09	−.33	−.10	—
5	Longitudinal data	.55	−.09	.05	.25	—

Note. N = 218 observations; experimental versus non-experimental design is not included in this table because it is tested separately from other moderators. IRS = Internal Revenue Service.

Table 3.

GEE Model Results (n= 30 Studies, 244 Effects).

Moderator	Model 1	Model 2	Model 3	Model 4	Model 5
Research design
Experimental designs	−.21** (.08)
Advanced models		−.01(.01)			−.01(.02)
Substantive factors
Administrative costs			−.03(.02)		−.01(.02)
Amateur donors			−.05*** (.01)		−.05*** (.01)
Data structure
IRS Form 990				.05^† (.03)	.04(.03)
Longitudinal data				−.01(.02)	.01(.02)
Constant	−.03** (.01)	−.03^† (.02)	−.02** (.01)	−.06** (.02)	−.05^† (.03)
N of effect size	244	218	218	218	218

Note. Robust standard errors in parentheses. IRS = Internal Revenue Service.

†

p ≤ .10. *p ≤ .05. **p ≤ .01. ***p ≤ .001.

Research Design Moderators

Experimental Design Versus Non-Experimental Designs

This moderator tests for the difference in the effect size between studies that used experimental designs versus studies that used non-experimental designs. Notably, the difference is statistically significant in our model. The effect sizes are, on average, significantly more negative from the studies that used experimental designs to examine the relationship, β = −.21, z = −2.81, p ≤ .01. This difference is substantial given the average effect size is −.03. The use of experimental designs makes a big difference. In sum, studies that used experimental designs are more likely to detect negative associations.

Advanced Regression Models Versus Basic Regression Models

This moderator tests whether the effect of overhead varies according to regression techniques employed in the original studies. The results of our model indicate effect sizes in analyses of the relationship do not differ according to modeling choice, β = −.01, z = −.72, p = .47.

Substantive Factor Moderators

Administrative Costs Versus Administrative and Fundraising Costs

Our moderator that flags effects that include fundraising costs in its measure of overhead is negative and but is not statistically significant. This result suggests that original studies that used administrative costs to measure overhead do not report more negative effects than original studies that used both administrative and fundraising costs (or only program costs) to measure overhead, β = −.01, z = −.33, p ≤ .74.

Amateur Versus Professional Donors

Our moderator for donor types examines differences between effect sizes in studies focused on amateur (individual, general public) donors versus studies of professional donors who might scrutinize organizations more closely. The coefficient indicates that amateur donors return more negative effects than professional donors, β = −.05, z = −4.61, p ≤ .001. The difference is large given the average effect size of −.03.

Data Structure Moderators

IRS Form 990 Data Versus Other Data

Our moderator analysis indicates that effects calculated from data drawn from Form 990s are not significantly different from effects calculated from other sources, β = .04, z = 1.56, p = .12.

Longitudinal Data Versus Cross-Sectional Data

As described above, we test whether the association between overhead spending and donations varies according to how much the study considers the influence of time. Our moderator analysis indicates that effects calculated from studies that used longitudinal data are not significantly different from effects calculated from studies that used cross-sectional data, β = .01, z = .16, p = .87.

Discussion

This article seeks to make three contributions to the literature on nonprofit spending and their ability to attract grants and other contributions. First, we believe that publication of the Wing Model will provide a useful corollary to the simpler starvation cycle model. Whereas the starvation cycle concentrates solely on the negative feedback from overhead aversion and cutbacks in capacity spending, the Wing Model describes both the value of capacity spending and the possible negative reactions by donors. This model allows us to make sense of the variety of research projects that refer to overhead, the starvation cycle, or outcomes, but are focused on different parts of a complex puzzle. The Wing Model helps the field organize the complex influences of overhead spending. It gives us a map for sorting the disparate literature on overhead and shows what each piece contributes to the whole.

The current study highlights one complex pathway in the Wing Model, or the starvation cycle generally, namely, the troublesome relationship between reported spending allocations and the reactions of donors. Indeed, our analysis of studies of the relationship between overhead ratios and contributions confirms practitioner concerns that their transparency will harm their standing with stakeholders. If they spend less on capacity, their operations may suffer. However, nonprofits may choose to mis-report administrative and fundraising as program costs, compromising values to skirt overhead aversion. Other studies will investigate these specific pathways, validating or re-writing key portions of the Wing Model, including the virtuous relationship between capacity investments and program outcomes.

Second, the meta-analysis provides strong empirical validation for that one pathway in the Wing Model, namely, the relationship between reported capacity costs and funders’ giving decisions (Point D in Figure 1). The moniker overhead aversion asserts a negative relationship on this pathway. Our investigation demonstrates a small, negative, and statistically significant association between overhead costs and donations. The results are consistent with the observation that many donors view nonprofit overhead costs as wasteful expenses or an inappropriate usage of nonprofit resources, at least when the costs become high enough to warrant attention. This finding generally validates the overhead aversion thesis, that nonprofits are expected by donors to allocate most of their expenses into program services. When capacity costs exceed donor expectations, spending is chilled.

Third, the moderator analysis considers the generalizability of this claim across studies. We found that the disparate findings on the overhead aversion issue lie in research design and donor types examined in the original studies. For one, our analysis detects a significant difference in effect sizes between experimental and non-experimental designs. Studies that use experimental designs detect higher overhead aversion than studies that use other research designs. Perhaps this detection of aversion is real, a result of the famous ability of experiments to uncover causation. However, the difference may be methodological: Experiments are better able to handle endogeneity and omitted variable issues that potentially bias the association between overhead costs and donations. The difference may be an artifact of research design: Experimental and non-experimental designs target different sample populations. Although experimental designs center around individual donors who are provided full information about donation scenarios in controlled settings, non-experimental designs use aggregate data sets where both individual and institutional (professional) donors are combined. Finally, we should not overlook the concern that experiments are artificial: Subjects are fed context that may be obscure in real donor situations. Experiments provide some advances in measuring key relationships, but they have limitations that compromise their value.

We also find that original studies that use administrative costs to measure overhead do not report more negative effects than studies that use both administrative and fundraising costs (or simply program costs) to measure overhead. This finding is surprising given that a previous study suggests that fundraising costs create less aversion than pure administrative costs. Portillo and Stinn (2018) attribute this to donor perceptions of fundraising endeavors as investments that will lead to more contributions. Fundraising directly raises contributions, whereas administrative costs do not. However, our finding suggests that contributions reduced by fundraising aversion may outweigh contributions raised by fundraising efforts. That is, donors’ fundraising aversion may be greater than their administrative aversion. Thus, readers of this literature should consider that donors may have different reactions to administrative versus fundraising costs, even if both produce the similar effect on nonprofit donations. Administrative aversion, fundraising aversion, and a combined overhead aversion may have different effects on the starvation cycle.

Another substantial difference regards studies that focus on professional donors. In our study, some of our professionals are individuals working in institutional settings: The key is that they are professionals with access to more information and greater socialization to the value of capacity building. Previous research has observed that institutional donors are more likely than individual donors to comprehensively review nonprofits’ portfolios when making charitable giving decisions (Balsam & Harris, 2013; Cnaan et al., 2011; Hibbert & Horne, 1996; Kitching, 2009; Parsons, 2003; Tinkelman, 1998). Their decisions to give are less likely to be solely based on a cost ratio. That is, although an increase in overhead might result in a decrease in contributions from some professional donors, the negative relationship is weaker when compared with the overhead aversion of amateur donors. Our findings are consistent with this observation. Professional donors exhibit a comparatively low level of overhead aversion. This might be because, compared with amateur donors, professional donors have more resources and time to evaluate nonprofits. They may have a better understanding of the value that capacity spending brings to nonprofit operations.

The Wing Model considers the fact that different funders will have different tolerances or preferences for capacity (non-program) spending. On one hand, these preferences theoretically influence their giving decisions. On the other hand, nonprofit perceptions of donor preferences influence nonprofit spending or reporting decisions. These are common precepts but repeated without strict empirical tests or shared understanding of how arcs of both virtuous and starvation cycles dictate the behaviors and fates of nonprofit organizations. With the Wing Model in hand, the field can test, hone, and develop common understanding of these complex and contested relationships. For our part, we find that overhead aversion is a real and consequential consideration for organizations that rely on public goodwill.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

ChiaKo Hung

Mark Hager

Yuan Tian

Notes

Author Biographies

ChiaKo Hung, PhD, is an assistant professor at the University of Hawaiʻi at Mānoa. His areas of interest are nonprofit finance, nonprofit-government collaboration, social entrepreneurship, and meta-analysis. His research has been featured in Nonprofit and Voluntary Sector Quarterly, VOLUNTAS, Public Performance & Management Review, Nonprofit Management and Leadership, and the Journal of Public Administration Research and Theory.

Mark A. Hager, PhD, is an emeritus professor at Arizona State University. His research includes studies of the scope, dimensions, administration, and financial operations of and reporting by nonprofit organizations. From 2015 to 2020, he served as editor-in-chief of Nonprofit Management & Leadership. He is gratified to finally be able to bring Ken Wing’s model back to the field. The Wing Model promises to organize our future thinking about how nonprofits react to funder priorities.

Yuan Tian, PhD, is an associate professor at the School of International and Public Affairs, Shanghai Jiao Tong University. Her research specialized in philanthropic studies, experimental studies, and cross-sector collaboration.

References

Altamimi

Liu

(2022). The nonprofit starvation cycle: Does overhead spending really impact program outcomes? Nonprofit and Voluntary Sector Quarterly, 51(6), 1324–1348.

Aryee

T. E.

(2015). Corporate governance, fundraising and operational efficiency of non-governmental organizations in Ghana [Doctoral dissertation, University of Ghana].

Ashley

S. R.

Faulk

(2010). Nonprofit competition in the grants marketplace. Nonprofit Management and Leadership, 21(1), 43–57.

Ashley

S. R.

Van Slyke

D. M.

(2012). The influence of administrative cost ratios on state government grant allocations to nonprofits. Public Administration Review, 72(s1), S47–S56.

Balsam

Harris

E. E.

(2013). The impact of CEO compensation on nonprofit donations. The Accounting Review, 89(2), 425–450.

Blouin

M. C.

Lee

R. L.

Erickson

G. S.

(2018). The impact of online financial disclosure and donations in nonprofits. Journal of Nonprofit & Public Sector Marketing, 30(3), 251–266.

Bowman

(2006). Should donors care about overhead costs? Do they care? Nonprofit and Voluntary Sector Quarterly, 35(2), 288–310.

Buchheit

Parsons

L. M.

(2006). An experimental investigation of accounting information’s influence on the individual giving process. Journal of Accounting and Public Policy, 25(6), 666–686.

Calabrese

T. D.

(2011). Do donors penalize nonprofit organizations with accumulated wealth? Public Administration Review, 71(6), 859–869.

10.

Charles

(2018). Nonprofit arts organizations: Debt ratio does not influence donations—Interest expense ratio does. The American Review of Public Administration, 48(7), 659–667.

11.

Charles

Sloan

M. F.

Schubert

(2020). If someone else pays for overhead, do donors still care? The American Review of Public Administration, 50(4–5), 415–427.

12.

Cnaan

R. A.

Jones

Dickin

Salomon

(2011). Nonprofit watchdogs: Do they serve the average donor? Nonprofit Management & Leadership, 21(4), 381–397.

13.

Curran

C. J.

Bonilla

(2010). Taking OD to the bank: Practical tools for nonprofit managers and consultants. Journal for Nonprofit Management, 22, 1–7.

14.

de Wit

Bekkers

. (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.

15.

Froelich

K. A.

Knoepfle

T. W.

(1996). Internal revenue service 990 data: Fact or fiction? Nonprofit and Voluntary Sector Quarterly, 25(1), 40–52.

16.

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.

17.

Frumkin

Kim

M. T.

(2001). Strategic positioning and the financing of nonprofit organizations: Is efficiency rewarded in the contributions marketplace? Public Administration Review, 61(3), 266–275.

18.

Glass

G. V.

(1976). Primary, secondary, and meta-analysis of research. Educational Researcher, 5(10), 3–8.

19.

Gneezy

Keenan

E. A.

Gneezy

(2014). Avoiding overhead aversion in charity. Science, 346(6209), 632–635.

20.

Greenlee

J. S.

Brown

K. L.

(1999). The impact of accounting information on contributions to charitable organizations. Research in Accounting Regulation, 13, 111–126.

21.

Gregory

A. G.

Howard

(2009). The nonprofit starvation cycle. Stanford Social Innovation Review, 7(4), 49–53.

22.

Grizzle

(2015). Efficiency, stability and the decision to give to nonprofit arts and cultural organizations in the United States. International Journal of Nonprofit and Voluntary Sector Marketing, 20(3), 226–237.

23.

Gurevitch

Koricheva

Nakagawa

Stewart

(2018). Meta-analysis and the science of research synthesis. Nature, 555(7695), 175–182.

24.

Hager

M. A.

(2003). Current practices in allocation of fundraising expenditures. New Directions for Philanthropic Fundraising, 41, 39–52.

25.

Hager

M. A.

Flack

(2004). The pros and cons of financial efficiency standards (Nonprofit Overhead Cost Project, Brief No. 5). Center on Nonprofits and Philanthropy, Urban Institute and the Center of Philanthropy, Indiana University.

26.

Hager

M. A.

Pollak

Wing

Rooney

P. M.

(2004). Getting what we pay for: Low overhead limits nonprofit effectiveness (Nonprofit Overhead Cost Project, Brief No. 3). Center on Nonprofits and Philanthropy, Urban Institute and the Center of Philanthropy, Indiana University.

27.

Harris

E. E.

Petrovits

C. M.

Yetman

M. H.

(2015). The effect of nonprofit governance on donations: Evidence from the revised form 990. The Accounting Review, 90(2), 579–610.

28.

Harris

E. E.

Tate

S. L.

Zimmerman

A. B.

(2019). Does hiring a local industry specialist auditor matter to nonprofit organizations? Nonprofit and Voluntary Sector Quarterly, 48(3), 633–664.

29.

Hibbert

Horne

(1996). Giving to charity: Questioning the donor decision process. Journal of Consumer Marketing, 13(2), 4–13.

30.

Hung

(2022). Meta-analysis for nonprofit research: Synthesizing quantitative evidence for knowledge advancement. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations. Advance online publication. https://doi.org/10.1007/s11266-022-00505-3

31.

Jacobs

F. A.

Marudas

N. P.

(2009). The combined effect of donation price and administrative inefficiency on donations to US nonprofit organizations. Financial Accountability & Management, 25(1), 33–53.

32.

Keating

E. K.

Frumkin

(2003). Reengineering nonprofit financial accountability: Toward a more reliable foundation for regulation. Public Administration Review, 63(1), 3–15.

33.

Kim

(2017). The relationship of nonprofits’ financial health to program outcomes: Empirical evidence from nonprofit arts organizations. Nonprofit and Voluntary Sector Quarterly, 46(3), 525–548.

34.

Kinsbergen

Tolsma

(2013). Explaining monetary donations to international development organisations: A factorial survey approach. Social Science Research, 42(6), 1571–1586.

35.

Kitching

(2009). Audit value and charitable organizations. Journal of Accounting and Public Policy, 28(6), 510–524.

36.

Krawczyk

Wooddell

Dias

(2017). Charitable giving in arts and culture nonprofits: The impact of organizational characteristics. Nonprofit and Voluntary Sector Quarterly, 46(4), 817–836.

37.

Krishnan

Yetman

M. H.

(2011). Institutional drivers of reporting decisions in nonprofit hospitals. Journal of Accounting Research, 49(4), 1001–1039.

38.

Krishnan

Yetman

M. H.

Yetman

R. J.

(2006). Expense misreporting in nonprofit organizations. The Accounting Review, 81(2), 399–420.

39.

Lecy

J. D.

Searing

E. A.

(2015). Anatomy of the nonprofit starvation cycle: An analysis of falling overhead ratios in the nonprofit sector. Nonprofit and Voluntary Sector Quarterly, 44(3), 539–563.

40.

Lee

K. C.

Kim

(2018). Marketing expense and financial performance in arts and cultural organizations. International Journal of Nonprofit and Voluntary Sector Marketing, 23(3), Article e1588.

41.

Liang

K. Y.

Zeger

S. L.

(1986). Longitudinal data analysis using generalized linear models. Biometrika, 73(1), 13–22.

42.

Lipsey

M. W.

Wilson

D. B.

(2001). Practical meta-analysis. SAGE.

43.

(2016). The philanthropic consequence of government grants to nonprofit organizations: A meta-analysis. Nonprofit Management and Leadership, 26(4), 381–400.

44.

Marudas

N. P.

Hahn

Jacobs

F. A.

(2014). An improved model of effects of accounting measures of inefficiency on donations. Journal of Finance and Accountancy, 21, 1–15.

45.

Marwell

N. P.

Calabrese

(2015). A deficit model of collaborative governance: Government–nonprofit fiscal relations in the provision of child welfare services. Journal of Public Administration Research and Theory, 25(4), 1031–1058.

46.

Nicholson-Crotty

(2011). Does reported policy activity reduce contributions to nonprofit service providers? Policy Studies Journal, 39(4), 591–607.

47.

Nikolova

(2015). Government funding of private voluntary organizations: Is there a crowding-out effect? Nonprofit and Voluntary Sector Quarterly, 44(3), 487–509.

48.

Parsons

L. M.

(2003). Is accounting information from nonprofit organizations useful to donors? A review of charitable giving and value-relevance. Journal of Accounting Literature, 22, 104–129.

49.

Peng

Kim

Deat

(2019). The effects of nonprofit reputation on charitable giving: A survey experiment. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations, 30(4), 811–827.

50.

Portillo

J. E.

Stinn

(2018). Overhead aversion: Do some types of overhead matter more than others? Journal of Behavioral and Experimental Economics, 72, 40–50.

51.

Daniel

J. L.

(2021). Is “overhead” A tainted word? A survey experiment exploring framing effects of nonprofit overhead on donor decision. Nonprofit and Voluntary Sector Quarterly, 50(2), 397–419.

52.

Quosigk

Forgione

D. A.

(2018). Do donors respond to discretionary accounting information consolidation? Journal of Public Budgeting, Accounting & Financial Management, 30(1), 35–58.

53.

Ringquist

(2013). Meta-analysis for public management and policy. John Wiley.

54.

Schubert

Boenigk

(2019). The nonprofit starvation cycle: Empirical evidence from a German context. Nonprofit and Voluntary Sector Quarterly, 48(3), 467–491.

55.

Shadish

W. R.

Lecy

J. D.

(2015). The meta-analytic big bang. Research Synthesis Methods, 6(3), 246–264.

56.

Steinberg

(1986). The revealed objective functions of nonprofit firms. The RAND Journal of Economics, 17, 508–526.

57.

Stock

J. H.

Watson

M. W.

(2012). Introduction to econometrics: Global edition. Pearson Education.

58.

Surysekar

Turner

E. H.

Wheatley

C. M.

(2015). On the association between donor-imposed financial inflexibility and future donations to charitable organizations. Journal of Management Accounting Research, 27(1), 63–79.

59.

Tian

Hung

C. K.

Frumkin

(2020). Breaking the nonprofit starvation cycle: An experimental test. Journal of Behavioral Public Administration, 3(1), 1–19.

60.

Tinkelman

(1998). Differences in sensitivity of financial statement users to joint cost allocations: The case of nonprofit organizations. Journal of Accounting, Auditing & Finance, 13(4), 377–393.

61.

Tinkelman

(1999). Factors affecting the relation between donations to not-for-profit organizations and an efficiency ratio. Research in Government and Nonprofit Accounting, 10(1), 135–161.

62.

Tinkelman

Mankaney

(2007). When is administrative efficiency associated with charitable donations? Nonprofit and Voluntary Sector Quarterly, 36(1), 41–64.

63.

Trussel

(2003). Assessing potential accounting manipulation: The financial characteristics of charitable organizations with higher than expected program-spending ratios. Nonprofit and Voluntary Sector Quarterly, 32(4), 616–634.

64.

White

(1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48(4), 817–838.

65.

Wing

Hager

(2004, November). Who feels pressure to contain overhead costs? Results from a national survey [Conference session]. Presentation at the annual conference of the Association for Research on Nonprofit Organizations and Voluntary Action, Los Angeles, CA, United States.

66.

Wing

Hager

Rooney

Pollak

(2004). Donating to charity: A guide. Nonprofit Overhead Cost Study.

67.

Wing

Hager

Rooney

Pollak

(2005, May-June, 32–37). Paying for not paying for overhead. Foundation News and Commentary.

68.

Wing

Pollak

Rooney

(2004a, May). Toward a theory of limited nonprofit organizational effectiveness [Conference session]. Presentation at the 3rd International Conference on Systems Thinking in Management, Philadelphia, PA, United States. http://ssrn.com/abstract=4256936

69.

Wing

Pollak

Rooney

(2004b). Toward a theory of organizational fragility in the nonprofit sector [Working paper]. http://ssrn.com/abstract=4256921

70.

Xie

Huang

C. C.

Wang

Pei

(2016). Administrative efficiency and donation of foundations in China. Human Service Organizations, 40(4), 410–420.