Nonprofit Resource Allocation Decisions: A Study of Marginal versus Average Spending

We investigate how managers of charities respond to exogenous changes in budgeted costs. We focus, in particular, on how the response influences financial performance measures used by charity oversight groups. Our analysis is motivated by competing characterizations about how charities allocate incremental resources and how donors respond to such allocations.

One characterization, suggested by empirical evidence advanced in prior studies (Wing et al. 2004; Tinkelman 2009; Parsons et al. 2011), is that charities allocate incremental resources in deference to how allocations influence average financial performance metrics. We use the term donor pressure when referring to incentives to allocate incremental resources according to this characterization, where donor pressure is defined as the pressure, either real or perceived, that donors or oversight organizations place on charities to spend on program activities rather than on overhead.

A competing, although not necessarily mutually exclusive, perspective is that charities allocate incremental resources in ways that most effectively advance the long-run charitable objective, regardless of how such allocations influence performance measures in the short run (Steinberg 1986a, 1986b; Young and Steinberg 1995). This means that for a given budget change, charity managers choose the combination of program and overhead expenses that best serves the organization's charitable objective. We use the term “socially optimal” to describe allocations of incremental resources that are consistent with this perspective.

To evaluate these resource allocation decisions, we compare reported marginal program spending (the difference between current and prior period program spending) with a benchmark marginal program spending. Benchmark marginal program spending is computed program spending “as if” marginal resources are allocated consistent with the prior period's average program spending ratio. Comparing reported marginal program spending with the benchmark indicates whether donor pressure influences the way charities allocate resources to programs. In particular, we posit that program spending less than the benchmark indicates that charities do not consider perceptions conveyed to donors when allocating marginal resources. In contrast, marginal program spending greater than the benchmark may or may not indicate decisions influenced by concerns about impressions conveyed to donors. Thus, we interpret as evidence that resource allocations are not influenced by donor pressure when reported marginal program spending is less than the benchmark, as spending less than the benchmark results in a lower reported program ratio.¹

Using a sample of 38,499 IRS Form 990 charity-year observations (5,626 charities) between 1986 and 2007, we document that when budget increases are less than 15 percent, marginal program spending patterns are the same as average program spending patterns in the prior period. Hence, for the average charity, reported program ratios are unchanged. We document that when budgets decrease, however, charities cut program spending more substantially than they cut spending on overhead, such that current program ratios are lower than the prior period. This suggests that charity managers are more willing to report declining program ratios when budgets decrease than when budgets increase. While the degree of donor pressure does not likely change (in the short run), the results suggest that if donor pressure exists, charities do not succumb to it when budgets decrease. This may be because justifying declines in program spending that result in a decrease to the program ratio is less problematic when budgets are cut. It is more difficult, however, to justify a decline in the program ratio when budgets increase.

We find that some charities appear to be less concerned than others about how donors perceive program spending and, thus, are more willing to make spending decisions that decrease reported program ratios. This is the case for small charities, for those that rely little on contributions, for charities that are not funded by the government, and for organizations that do not fundraise. We also find that charities with high program ratios in the prior period make resource allocation decisions that decrease the program ratio, and charities with low program ratios in the prior period make resource allocation decisions that increase the program ratio. This suggests that donor pressure is associated with reported program ratios and also suggests that charities strive to attain an “acceptable” program ratio.

We acknowledge that the amounts reported in the program, administration, and fundraising categories on Form 990 may not reflect the true economics of the organization, but may be managed (Trussell 2003; Jones and Roberts 2006; Krishnan et al. 2006; Keating et al. 2008; Krishnan and Yetman 2011; Parsons et al. 2011).² We argue that this possibility does not impact the interpretation of our results, primarily because we suspect that the incentive to avoid reporting decreases in the program ratio is stronger than the incentive to avoid reporting increases in the program ratio. Thus, although the functional expense categories on some of the 990s included in our sample may be managed, our results indicate that when resources decline, charities report declining program ratios. This result contradicts expectations if the average charity systematically manages the program ratio.

Contributions of the study are important for several reasons. First, to our knowledge, we are the first to document that spending decisions are asymmetric, in that charities reduce program spending faster when budgets decrease than they increase program spending when budgets increase. Second, we contribute to the literature that examines how charity managers respond to institutional pressure, and provide evidence that at least some charity managers consider donor pressure when allocating incremental resources. Most studies to date examine the impact that institutional pressures have on the incentive to manage reported financial information. Third, we contribute to the limited research that uses financial information to gain insight into charity managers' operating strategies (Baber et al. 2001; Hughes and Luksetich 2004; Tinkelman 2004). Finally, this paper contributes to the ongoing discussion about the use of marginal spending as a tool to evaluate charities (Steinberg 1986b, 1989; Bowman 2006; Tinkelman 2006).

The next section develops the research question. The third section describes the research design. The results and additional testing are in the fourth section. The fifth section concludes.

Both economic theory and practice considerations can guide decisions about how to allocate resources between programs and overhead (administrative and fundraising). Economic theory predicts managers spend on fundraising up to the point where marginal fundraising costs equal marginal revenue, and spend on administration up to the point where the marginal outcome for spending a dollar on programs equals the marginal output for spending a dollar on administration (Young and Steinberg 1995). That is, for a given budget increase, charity managers should choose the combination of program and overhead expenses that best serves their organization. Young and Steinberg (1995) provide the example of a “Meals on Wheels” charity that might waste a budget increase if that charity uses the increase to purchase additional food (i.e., spend on programs) when the organization does not have the necessary administrative infrastructure to distribute additional food. Jones et al. (2012) suggest that it may be particularly difficult to increase program spending, especially in the short run, since doing so may require the charity to hire additional staff, identify worthwhile projects or recipients, or build the capacity to provide services. As such, economists advise nonprofit managers to “think at the margin” and make socially optimal choices when making resource allocation decisions (Steinberg 1986a, 1986b; Young and Steinberg 1995).

A few academics argue that donors should also “think at the margin” when evaluating charity spending. Steinberg (1986a, 1986b, 1989) considers this point as it relates to donors' evaluation of fundraising activities. He urges donors to evaluate marginal and not average fundraising ratios when making donation decisions, and maintains that average historic fundraising ratios are informative only if these predict future marginal spending decisions. In support of Steinberg (1986a, 1986b, 1989), Tinkelman (2006) finds that historic fundraising ratios are not an appropriate proxy for a charity's future marginal fundraising spending decisions. Bowman (2006) also argues that donors should examine the change in overhead ratios. Using a small sample of donations by federal employees to Chicago charities, he finds, however, that the change in overhead ratios explains relatively little variation in contributions received.

Despite theoretical support for the use of marginal program spending, more practical considerations that are not based on marginal analysis can guide donors' contribution decisions. To illustrate, nonprofit watchdog groups such as Charity Navigator and the Better Business Bureau's (BBB) Wise Giving Alliance evaluate and rate charities based on average (not marginal) program spending (the program ratio). Also, the business press urges donors to use the historical average program ratio when deciding whether to donate to a charity, because it indicates how much of your donation will go toward supporting the underlying mission rather than administrative costs (Chatzky 2006). Further, a report by Princeton Survey Research (2001, 20) finds that over three-fourths (79 percent) of Americans surveyed believe it is important for donors to know the percentage of a nonprofit's spending that goes toward supporting the underlying mission rather than to administrative costs, and most believe the appropriate amount should be at least 70 to 80 percent of total spending.³

In addition, some regulators and watchdogs view spending on fundraising and administrative activities as a poor use of donor funds. For instance, some states attempt to impose limits on fundraising costs (Strom 2003), and many states require professional solicitors to report the amounts raised on behalf of a charity as a way of regulating fundraising costs (Greenhouse 2003).⁴ Also, some donors increasingly support charities that spend only a small fraction of their budget on overhead expenses, while others impose restrictions on how gifts can be used (Silverman and Beatty 2006). In addition to what we observe anecdotally, research consistently indicates that donors contribute more when program ratios are high than when ratios are low (Weisbrod and Dominguez 1986; Harvey and McCrohan 1988; Posnett and Sandler 1989; Callen 1994; Tinkelman 1999, 2004; Greenlee and Brown 1999; Okten and Weisbrod 2000; Tinkelman and Mankaney 2007; Parsons 2007; Marudas 2004; Jacobs and Marudas 2009; Kitching 2009).⁵

As a result, donor pressure to maintain or improve ratios may affect resource allocation decisions.⁶ Research provides evidence that donor pressure impacts spending decisions, and that managers forgo spending on overhead in order to report favorable program ratios (Hager 2004; Pallota 2008; Tinkelman 2009; Parsons et al. 2011). In fact, Parsons et al. (2011) survey charity managers and find that charities with given characteristics forgo spending on overhead to avoid reporting program ratio declines. Pallota (2008), Tinkelman (2009), and Petrovits et al. (2011) document cases where revenues decrease as a result of the charity sacrificing spending on overhead. Thus, donor pressure may discourage investment in administrative infrastructure (Hager et al. 2004).

In sum, although economic theory suggests managers make socially optimal decisions by considering marginal analysis, managers may conform to donor pressure and consider average measures. This is especially likely when economically efficient spending decreases program ratios (or increases overhead ratios).

To illustrate, suppose a charity's budget in period t−1 equals $100, overhead spending equals $20, and program spending equals $80. Thus, the program ratio is 80 percent in period t−1. During period t, the charity's budget increases $5. Since program spending is currently at an optimal level, the charity uses the opportunity to improve infrastructure and spends the $5 on overhead. As a result, marginal program spending is zero and the program ratio decreases to 76 percent ($80/$105). Using this same scenario, if spending decisions conform to donor pressure to avoid a decline in the program ratio, then the charity might either spend the $5 on programs to increase the ratio (which might not be optimal) or increase program spending by $4 and administrative spending by $1 so that the program ratio remains at 80 percent ($84/$105).

In this paper, we ask the following research question:

We assume that if charity managers consider donor pressure when making resource allocation decisions, then managers make allocation decisions to avoid reporting decreases in the program ratio.

Note that we consider the organization's annual budget to be the spending resources available to management and use total expenses as a budget proxy. We do not consider total revenues because, in addition to the allocation decision about how to spend between programs and overhead, the organization must decide whether to spend or save marginal revenues. Jones et al. (2012), who examine this particular decision, document that when charities have the flexibility to do so, they save most marginal revenue. We examine the spending decision once budgets are set.

We use a subset of 501(c)(3) nonprofits where managers are most likely to make spending decisions based on how donors view spending on overhead. We focus on public charities where donors use efficiency ratios to make contribution decisions. We exclude religious institutions, art and cultural organizations, hospitals, or educational institutions (Gordon and Khumawala 1999).

We use Form 990 data from the Internal Revenue Service Statistics of Income (SOI) file made available by the National Center for Charitable Statistics for years 1985–2007.⁷ We eliminate observations where we believe ex ante that the degree of monitoring or donor oversight is reduced. In particular, we exclude observations where administrative, program, or fundraising expense equals (or is less than) zero, because these organizations are either new, almost defunct, or affiliated with a nonprofit that raises funds or incurs administrative costs on its behalf (Yetman and Yetman 2011).^8,9 Prior research documents potential problems with both the Form 990 and SOI data (Froelich and Knoepfle 1996; Froelich et al. 2000; Gordon et al. 1999; Marudas 2004; Tinkelman 2004). Thus, we eliminate observations where there is a strong possibility that Form 990 has not been properly completed (i.e., the sum of program, administrative, and fundraising expenses on Part II of Form 990 does not equal total expense). Also, because errors in the SOI data can impact the analysis (Tinkelman 2004), we remove the extreme 1 percent of the percent change in total spending. We also remove observations when the organization indicates that the entity experiences a termination, liquidation, dissolution, or major contraction (Form 990, Part VI, #79). Finally, we lose observations because the empirical analysis requires the use of first differences, and we eliminate observations identified as influential using procedures outlined in Belsley et al. (1980). Table 1, Panel A, presents the sample selection procedures.

The final sample includes 2,932 environmental, 7,728 health (non-hospital), 18,031 human service, 1,164 international, and 8,644 public benefit charity-year observations (38,499 in total). Percentages of charities in each of the five major NTEE classifications are similar to the percentages for the file of SOI observations.

Designing a comprehensive model that unequivocally confirms whether managers respond to donor pressure requires access to details about spending choices and operations on a charity-by-charity basis. These data are not available on any publicly available dataset. Thus, we construct a model to infer how resource allocation decisions are made. In particular, we use the program ratio reported in the prior period as a benchmark to assess whether spending patterns differ between periods:

where:

• ΔPS_t = the period t change in reported program expense (Part II, line 44b);

• ΔTS_t = the period t change in total expense (Part II, line 44a); and

• PR_t−1 = the period t−1 program ratio (program expense divided by total expense).

To address heteroscedasticity, ΔPS_t and ΔTS_t variables are deflated by last year's total expense.

The dependent variable, ΔPS_t, represents incremental program spending. The primary variable of interest is the interaction [ΔTS_t × PR_t−1], which is benchmark incremental program spending and is the period t change in program spending given that total spending is allocated to programs in the same proportion as in the prior period. If charity managers conform to donor pressure when making resource allocation decisions, then, when resources increase, managers either (1) increase the proportion of spending on programs to increase the program ratio, or (2) do not change the proportion of spending on programs to keep the program ratio consistent with the prior period. Although reported incremental program spending greater than the prior period's program ratio (i.e., the first option) may suggest concern about donor pressure, we realize that it may also suggest an efficient allocation of resources. For example, charities with a high proportion of fixed administrative costs may find it optimal to spend more on programs. Hence, it is only when incremental program spending is less than benchmark marginal program spending that we conclude management does not conform to donor pressure when making spending decisions. Hence, if the charity does not overly conform to donor pressure when making resource allocation decisions, then the coefficient on our main variable of interest is less than 1 (α₁ < 1).¹⁰

LargeChange is a dummy variable equal to 1 if the absolute percent change in total spending is greater than 15 percent, and 0 otherwise. Thus, the product [ΔTS_t × PR_t−1 × LargeChange] represents the incremental effect of marginal program spending decisions when resource increases are large.¹¹ We examine whether the relative magnitude of the resource change impacts resource allocation decisions, because managers likely have more flexibility in their spending choices when budget changes are moderate. We also suspect boards of directors are more involved in approving large budget increases for very specific purposes (e.g., adding a new program or purchasing a new building). Jones et al. (2012), in fact, assert that charities save in the short run, owing to costs of adjusting spending. These costs include hiring additional staff, identifying worthwhile projects or recipients, and building the capacity to provide services. We presume that the spending choice is predetermined when budget increases are large. Thus, managers' discretion over resource allocation decisions, irrespective of donor pressure, is likely impacted by the size of the budget change.

Decrease is a dummy variable equal to 1 if the change in total spending is negative, and 0 otherwise. Thus, the product [ΔTS_t × PR_t−1 × Decrease] represents the incremental effect of marginal program spending decisions when resources decrease. We suspect that managers have more discretion when resources increase than when resources decrease. This is because when resources increase, managers can choose to either increase program spending or keep it the same. When resources decrease, managers are more likely to be forced to cut program spending. Thus, because there is more discretion in spending choice when resources increase, we examine whether resource allocation decisions differ when budgets decline. The product [ΔTS_t × PR_t−1 × LargeChange × Decrease] represents the incremental effect of marginal spending when budget declines are large.

Note that intercepts are suppressed to focus on the total impact to changes in budgets and to assess whether there is an asymmetric response to changes in budgets across the partitions.¹²

Table 2 displays descriptive information. We inflation-adjust all amounts to 2007 dollars using the Consumer Price Index (CPI) published by the Bureau of Labor Statistics. Descriptive statistics indicate sample charities vary in size. Mean (median) total assets equal $60 million ($20 million), mean (median) total revenue equals $27 million ($8 million), and mean (median) total expense equals $24 million ($7 million). Regarding the functional spending allocations, the mean (median) program ratio (program expense to total expense) is 80.4 (83.4) percent, the administrative ratio is 13.5 (11.1) percent, and the fundraising ratio is 6.0 (3.5) percent. The mean (median) percent change in total expense (after adjusting for inflation) is 6.7 percent (3.6 percent). The untabulated mean (median) change in the program ratio is −0.2 percent (0.0 percent), and the mean (median) change in both the fundraising and administrative ratios is 0.1 percent (0.0 percent). Approximately 34 percent of all observations decrease their budget, and approximately 27 percent of the sample change their budget by at least 15 percent.

Table 3 reports selected specifications in Columns 1 through 4. The focus of the discussion is on Column 4, where results for the complete specification are reported. We focus on whether marginal program spending patterns differ from average program spending patterns reported in the prior period and, more specifically, whether α₁ differs from 1. The coefficient α₁, which represents incremental program spending when budget changes are positive and moderate, is 0.993 and is not statistically different from 1 (t = 0.908).¹³ Thus, on average, marginal program spending patterns do not differ statistically from program spending patterns in the prior period (i.e., program ratios do not change).

The coefficient α₂ is positive and significantly different from 0 (α₂ = 0.054, t = 5.905). Thus, when budget increases are large, incremental program spending is 1.047 (0.993 + 0.054, t = 9.431), and significantly different from 1, indicating charities increase program spending at a faster rate than they increase overhead spending. These results provide some evidence that large budget increases are used to advance the organization's mission.

The coefficient α₃ is positive and significant (α₃ = 0.101, t = 12.100). Thus, when budget decreases are moderate, marginal program spending is 1.094 (0.993 + 0.101), indicating charities cut program spending at a faster rate than they cut overhead spending, resulting in declines to the program ratio. This result persists when budget declines are large (α₁ + α₂ + α₃ + α₄ = 1.143, t = 24.085).

Collectively, the results indicate that managers behave differently when budgets increase than when budgets decrease. Although for a given charity, the degree of donor pressure does not likely change (in the short run), it appears that managers' decisions are based less on donor pressure when budgets decline. This may be because it is easier to justify a reduction in the average proportion of program spending to total spending, which would result in a decline to the program ratio, when budgets are cut. It is more difficult, however, to justify a decline in the program ratio when budgets increase.

A competing explanation is that fixed overhead costs are a greater proportion of total costs than fixed program costs. If so, then it may be difficult for charities to cut fixed overhead costs in the short run. If managers only base decisions on cost structures, however, then we expect an increase in program spending for these charities when resources increase, which we do not observe. Another possible explanation is that a large fraction of variable overhead includes salaries (including executive salaries), and cutting salaries is not an acceptable operating strategy. Similarly, the result may suggest that, consistent with findings in Anderson et al. (2003) in the for-profit context, variable administrative costs are sticky and difficult to reduce in the short run. Future research is needed to determine how cost structures impact resource allocation decisions.

Next, we explore whether organizational attributes are associated with resource allocation decisions. In particular, we examine the relationship between program ratios, funding source, and size on marginal program spending decisions. We report the results for the complete model, but focus the discussion on spending decisions when resource increases are moderate—the range where management presumably has the most flexibility over spending decisions. Unless specifically noted, however, results for the decreases and large increases are qualitatively similar to those reported in Table 3, Column 4.

We examine whether prevailing program ratios impact resource allocation decisions. Parsons et al. (2011), who examine this issue directly, provide evidence that charities with relatively low administrative ratios (high program ratios) feel the greatest pressure to report favorable ratios.¹⁴

In Table 4, Columns 1–3, we partition the sample based on the prior period program ratio, and classify program ratios greater than 88.8 percent (the fourth quartile) “high” and ratios less than 75.9 percent (the first quartile) “low.”¹⁵ The “medium” partition consists of the second and third quartiles, or 50 percent of the sample. When budget increases are moderate, the estimates on α₁ are 0.904 (t = 9.434), 0.984 (t = 1.998), and 1.193 (t = 9.025) for charities with high, medium, and low reported program ratios, respectively, and are statistically different from each other (f = 80.6, p ≤ 0.0001). The results indicate that marginal program spending is less than average program spending in the prior period for charities with high program ratios, suggesting charities with high program ratios do not consider donor pressure when making spending decisions (or, alternatively, their program ratios already far exceed donor expectations, so donor pressure is irrelevant in the current year). For charities with relatively low program ratios, marginal program spending is greater than average program spending in the prior period, suggesting that charities with low program ratios (1) are most sensitive to increasing program ratios, or (2) use budget increases as an opportunity to increase program spending. Interestingly, we observe that when budget decreases are moderate, charities with the most overhead (i.e., those with low program ratios) increase the program ratio modestly (α₁ + α₃ = 0.936, t = 2.445), but when budget decreases are large, these same charities cut program spending at a much greater rate than charities with less overhead (f = 120.13, p ≤ 0.0001).¹⁶

Resource dependency theory (Pfeffer and Salancik 1978) suggests that because funding sources are associated with different levels of external oversight, managers must recognize this and appropriately respond to the criteria for continued funding for each key provider.¹⁷ Thus, we explore whether two types of key providers—donors (direct and indirect) and the federal government—impact managers' resource allocation decisions. We examine the effect donors have on spending decisions because it is well documented that donors respond to reported ratios. We also examine the effect the federal government has on spending decisions because of the detailed rules on how grantees can spend federal grant money.¹⁸

In Table 4, Columns 4–6, we partition the sample according to the proportion of contributions to total revenue (both direct and indirect), and classify the proportion of contributions to total revenue as “high” when the percentage is 73.0 percent (the fourth quartile) and “low” when the percentage is 10.7 percent (the first quartile). When budget increases are moderate, the estimates on α₁ are 1.006 (t = 0.642), 0.996 (t = 0.383), and 0.978 (t = 1.897). For charities with high and low dependency on contributions, the estimates are statistically different from each other at the 10 percent confidence interval (f = 3.23, p = 0.072).¹⁹ Hence, charities that rely on contributions the most appear to be more concerned about donor pressure when making resource allocation decisions. In Columns 7–9, we partition the sample based on the proportion of government grants to total revenue. We classify the organizations into three partitions: (1) “zeros”—those that report zero grants (because the median level of government grants is zero), (2) “high”—those with a median level of grant revenue to total revenue greater than 12.3 percent, and (3) “low”—those with a median level of grant revenue to total revenue below 12.3 percent. When budget increases are moderate, the estimates on α₁ are 1.021 (t = 2.356), 1.000 (t = 0.007), and 0.978 (t = 1.989), and statistically different from each other (f = 2.77, p = 0.06) for charities with high, low, and no dependency on government funding. For charities with high dependency on government funding, marginal program spending is greater than average program spending in the prior period.²⁰ This supports a characterization where charities that receive government funding must use those funds on programs. For charities that receive no government funding, marginal program spending is less than average program spending in the prior period, suggesting these charities do consider donor pressure when making resource allocation decisions.

In this section, we explore whether organizational size is associated with spending decisions. We do this for two reasons. First, due to economies of scale, administrative spending as a fraction of total spending is smaller for larger nonprofits (Hager et al. 2001). This fact alone may impact resource allocation decisions. Second, reputational effects may impact resource allocation decisions. Assuming size is a reasonable proxy for reputation (Tinkelman 1999), large organizations may have the greatest incentive to respond to donor pressure when making resource allocation decisions. On the other hand, because large organizations have cultivated reputations, they may be less reliant on reported ratios to garner contributions (Kitching 2009). Thus, expectations about the role of size are not obvious ex ante.

Because the SOI data are biased toward large organizations, we use the digitized database maintained by the NCCS to examine whether size matters.²¹ This database includes the population of charities that file Form 990 between 1999 and 2003. Table 4 reports the results. We partition the digitized sample according to size, and classify organizations “large” when total revenue is greater than $2.6 million (the fourth quartile of the digitized database sample), and “small” when total revenue is less than $0.4 million (the first quartile of the digitized database sample). The “medium” partition consists of the second and third quartiles, or 50 percent of the sample. Note that charities in the fourth quartile of the digitized sample are included in the first quartile of the SOI sample. When budget increases are moderate, the estimates on α₁ are 0.995 (t = 0.518), 0.984 (t = 2.858), and 0.958 (t = 9.257) for the large, medium, and small partitions, respectively.²² For small charities (that are not in the SOI data), marginal program spending is less than average program spending in the prior period. By using the population of charities that file Form 990, we document that size matters. Based on these results, it appears that smaller charities conform less to donor pressure when making resource allocations. This is most likely because larger charities are monitored and evaluated by charity oversight agencies. Further, smaller charities may be monitored by donors who can observe quality and, thus, do not use the program ratio to infer quality. When using the SOI data, results indicate that size does not matter.²³

We conduct several robustness tests to determine whether potential issues with the data or alternative explanations impact the results. First, we examine whether extreme performing charities, such as those that are in financial distress or are rapidly expanding, drive the results. We restrict the sample to charities that have at least seven years of data, with at least two years of positive changes in total spending and at least two years of negative changes in total spending. Thus, no charity consistently reports positive or negative changes during the sample period. When extreme performing charities are excluded from the sample (the sample size is 25,656), the results are similar to those reported in Table 3, Column 4 (α₁ = 0.998, t = 0.243, and α₃ = 0.091, t = 7.697). This supports our initial result about the asymmetry of resource allocation decisions, but also suggests that a given charity behaves differently when budgets increase than when budgets decrease.

Second, we restrict the sample to include only 1999 to 2007 data. We do this because the 1996 Taxpayer Bill of Rights 2 (TBOR2) (U.S. House of Representatives 1996) was enacted during our sample period. TBOR2 significantly enhanced the mandate for nonprofit entities to make the Form 990 publicly available by requiring nonprofits to mail copies of their 990s to anyone upon written request. The objective of this mandate was to enhance the accuracy and accountability of financial disclosures (Smith and Shaver 2009). As a result of the law, Form 990 reporting quality may have improved, and charities may have become even more sensitive to how program ratios are reported. For these data, the estimates on α₁ are similar to that reported in Table 3, Column 4 (coefficient = 1.000, t = 0.012, n = 22,971). We also consider the earlier period of our main sample, as well, to investigate whether charities are influenced less by donor pressure when making resource allocation decisions during 1986 to 2006. We find that when we use pre-TBOR2 data, the coefficient α₁ is slightly less than 1 (coefficient = 0.9831, t = 1.786, n = 15,528), but the pre- and post- TBOR2 coefficients are not statistically significant from each other (f = 1.26, p = 0.26). Thus, considering TBOR2 does not influence interpretations of the primary results.

Third, we consider the possibilities that managers need more than one period to make adjustments to spending patterns or have a pent-up demand for administrative spending. We restrict the sample to charities that have a consecutive increase in total spending (n = 14,313). When the second year's increase is less than 15 percent, the estimate on α₁ is similar to that reported in Table 3, Column 4 (coefficient = 0.994, t = 0.703). These results are comparable to our primary results, even after charities experience consecutive increases in budget levels. We acknowledge, however, that program spending decisions may change given management has a horizon longer than two years to adjust resource allocations.

Fourth, we investigate whether there is an industry effect. We partition the sample into the five major NTEE classifications: environmental, health, international, public benefit, and human services. When budget increases are moderate, the coefficients in all five partitions are not statistically different from 1. Thus, we cannot conclude that donor pressure is more prevalent in certain types of charities.

Finally, our sample does not include observations with zero fundraising costs. Recall that we excluded these observations from our main sample, because we believe that charities that do not engage in fundraising do not have donors that monitor the organizations' program ratios. We estimate the specification using a sample of only zero fundraisers (48,223 observations).²⁴ When budget increases are moderate, the coefficient on α₁ is significantly less than 1.0 (coefficient = 0.958, t = 7.179). This result suggests that, on average, charities that have zero fundraising costs do not conform to donor pressure. This result supports our reasoning for excluding these observations from the sample.

In this paper, we examine whether managers' spending patterns reflect conformity to donor pressure when making spending decisions by comparing marginal and average spending patterns. We document that in many instances, spending patterns do not change. That is, when resource increases are moderate, charity managers consistently allocate resources in the same proportion as in the prior period and, thus, program ratios are maintained. Hence, we cannot dismiss the idea that charities conform to donor pressure when making resource allocation decisions. We find that only a select group of organizations appear not to be overly concerned with donor pressure, and make resource allocation decisions that result in declines to the program ratio. These organizations are those that have relatively high program ratios, small organizations, organizations that rely little on contributions or government funding, and those that report zero fundraising expense. We find organizations with low program ratios change spending patterns, but these organizations increase marginal program spending when provided the opportunity.

We document that in most instances when budgets decrease, program spending is reduced at faster rates than overhead spending. This finding not only indicates that resource allocation decisions are affected asymmetrically when resources increase than when resources decrease, but it also suggests that charity managers are less concerned about reporting decreases in the program ratio when budgets decline.

Finally, we observe that program ratios fall when resources decline, but do not generally observe program ratios rising when resource increases are moderate. That is, while program spending levels change, it is only up to the point where the program ratio does not change. Thus, although our tests focus on whether charities increase overhead spending and report falling program ratios, an increase in program spending might reflect a more efficient use of resources. One possible explanation for this general result is that charity managers lack incentives to increase program ratios and maximize program spending. This may be because donors and charity oversight agencies fix on an average measure to evaluate spending decisions by assessing whether the charity meets program ratio benchmarks and not program ratio changes. Further, if the program ratio is sufficiently high, managers may elect not to increase the ratio and ratchet up donor expectations in the future. When charity managers have discretion over spending decisions, actions are not directly observable, and no individual has a residual claim to profits or future performance, managers may increase overhead spending to reduce managerial effort or to increase perquisites when given the opportunity (Fama and Jensen 1983). Future research is required to explore these possibilities.