Economic Consequences of Expense Misreporting in Nonprofit Organizations: Are Donors Fooled?*

Economic Consequences of Expense Misreporting in Nonprofit Organizations: Are Donors Fooled?* Michelle H. Yetman** Associate Professor of Accounting The University of California at Davis July 31, 2009 Preliminary. Do not quote without permission. * I thank Chris Jones, Michael Maher, Carl Olson, Dan Tinkelman, Robert Yetman, workshop participants at the University of California at Davis, the 2009 Midyear Meeting of the American Accounting Association Government and Nonprofit Section, and the 2009 Western Meeting of the American Accounting Association for their valuable comments and advice. ** AOB IV, One Shields Avenue; Davis, CA 95616; phone (530) 754-7808; e-mail mhyetman@ucdavis.edu. Electronic copy available at: http://ssrn.com/abstract=1268582

Economic Consequences of Expense Misreporting in Nonprofit Organizations: Are Donors Misled? Abstract: Prior research finds that donors reward nonprofits that allocate larger proportions of their expenses to charitable purposes with more donations. Research also finds that managers overstate the amount of expenses reported as charitable, ostensibly to attract more donations. This paper examines the extent to which donors adjust their reliance on nonprofit financial reports that overstate charitable expenses when making donations. Results show that donors place significantly less weight on financial information for those organizations that overstate their charitable expenses. Results also show that as the ease of obtaining financial information has improved over time, so has donors disentanglement of low quality financial reporting. The findings suggest that donors are at least partially able to see through low quality financial reporting. JEL Classification: G1 G18 G3 G38 L3 L30 L31 M4 M41 M43 M48 Keywords: nonprofit organizations, program service ratios, financial reporting quality, regulation, cost shifting, agency problems 1 Electronic copy available at: http://ssrn.com/abstract=1268582

1. Introduction Prior research finds that donors consider nonprofits reported financial results when making donations allocation decisions. 1 Nonprofit organizations incur and report three broad categories of expenses on their primary publicly available IRS Form 990 report (IRS 990): program services, fundraising, and administrative. The program service ratio (PSR), which is the ratio of program service expenses to total expenses, measures the proportion of expenses that are directed towards the charitable mission. Prior research provides evidence that nonprofits that report higher PSRs are rewarded with more donations (Weisbrod and Dominguez 1986; Posnett and Sandler 1989; Tinkelman 1999; Okten and Weisbrod 2000; Tinkelman 2004; Parsons 2007). Donor reliance on reported nonprofit financial information provides nonprofit managers with an incentive to overstate charitable (and understate administrative and/or fundraising) expenses. Indeed, a growing body of research finds that nonprofits appear to manipulate their financial reports by increasing the proportion of expenses classified as charitable (Jones and Roberts 2006; Krishnan et al. 2006; Keating et al. 2008). This paper moves beyond documenting the existence of nonprofit financial statement management by investigating its economic consequences. More specifically, I examine whether donors disentangle nonprofit financial reporting management when making donations decisions. In addition, I explore whether donors disentanglement of financial reporting management has improved over time with the increased accessibility of financial information and/or is influenced by donor characteristics, such as the motivation for giving. The central research hypothesis I test is whether donors reduce their reliance on reported PSRs when they are of low quality when making donations decisions. To examine the question, I 1 Nonprofits receive three types of donations. Private donations are received from individuals and corporations. Indirect donations are received from collections agencies such as the United Way or from related organizations. Government grants are received from state or federal agencies. 2 Electronic copy available at: http://ssrn.com/abstract=1268582

rely on models used by prior research that show donors are sensitive to expense classification, rewarding firms that report spending proportionately more on charitable outputs (i.e., higher PSRs) with larger amounts of donations. I introduce into this model a reporting quality metric, and interact this quality metric with the ratio of charitable to total expenses. The basic premise underlying the model choice is that if donors can identify overstated charitable expenses as reflected in the reported quality metric, then they will reduce their reliance on the reported PSR when an organization has overstated their charitable expenses. My reporting quality metric is whether the nonprofit reports zero fundraising expenses despite earning substantial donations. Prior research finds that when a nonprofit receives substantial private donations (over $100,000 in my study) yet reports zero fundraising expenses, it is highly likely that the organization has understated its fundraising expenses and thus overstated its charitable expenses (Krishnan et al. 2006). The number of nonprofits that fall into this category is surprisingly large, approximately 35 percent. Results suggest that donors are able to see through low quality nonprofit financial information. In particular, when making donations decisions, donors place less weight on the reported ratio of charitable to total expenses for organizations that report zero fundraising expenses as compared to organizations that do report fundraising expenses. This result suggests donors do not naively respond to reported financial information, but are willing and able to at least partially disentangle some forms of nonprofit financial statement manipulation. I also find that as the ease of obtaining financial information has improved over time, so has donors disentanglement of low quality financial reporting. This study has important implications. From a public policy perspective, this study is important as donors inability to see through low quality financial reports can lead to improper or 3

unintended resource allocations. The amount of donation dollars at stake is large. In 2006 nonprofits received about $260 billion in donations, of which over 75 percent were from individual donors (Giving USA Foundation 2007). In addition, this study provides potentially useful information to government regulators. As in the for-profit setting, perceived deficiencies in nonprofit financial reporting have increased regulatory attention (Strom 2004). Concerns that donors are being misled by low quality financial information has led some state governments to impose some provisions of the Sarbanes-Oxley Act on nonprofits. 2 The U.S. Senate has held several hearings on how to enhance nonprofit reporting quality and also may impose some provisions of the Sarbanes-Oxley Act on nonprofit organizations (United States Senate Committee on Finance 2004, United States Senate Committee on Finance 2005). This study addresses this policy issue by investigating the extent to which donors are aware of poor financial reporting quality and consequently reduce their reliance on those reports in making donations decisions. The paper proceeds as follows. The next section presents background and theory, followed by the development of the research hypotheses. The following section describes my empirical model and its variables. Subsequent sections discuss the data and the empirical results, and the final section concludes. 2. Background and Hypothesis Development 2.1 The Role of, and Donor Response to, Nonprofit Financial Information Many organizations, regardless of their ownership structure, face agency problems because contracts are costly to be written and enforced (Jensen and Meckling 1976, Fama and Jensen 2 For example, California passed the Nonprofit Integrity Act of 2004, which imposes several of the provisions of the Sarbanes-Oxley laws on nonprofits. 4

1983). Agency theory recognizes that a manager s own objectives may not necessarily coincide with the owner s objective and typically presumes that managers will tend to maximize their own welfare, including benefits that are financial (wages) and non-financial (shirking or investing in pet projects) (Coase 1937). Although nonprofits are not owned in the traditional sense (i.e., they do not have residual claimants), they are accountable to several stakeholders including donors, lenders, customers, and regulators. Nonprofit stakeholders such as donors can reduce agency losses through monitoring (Hansmann 1996). Financial reports play an important role in monitoring managerial actions as they provide a means for donors and other stakeholders to evaluate whether the nonprofit is using donations towards the charitable mission in an efficient manner. Indeed, several studies show that donors reward nonprofits that report higher proportions of their total expenses as charitable (and correspondingly lower proportions as either administrative or fundraising) with more donations (Weisbrod and Dominguez 1986; Posnett and Sandler 1989; Tinkelman 1999; Okten and Weisbrod 2000; Tinkelman 2004; Parsons 2007). 3 2.2 Managerial Manipulation of Nonprofit Financial Information Given that donors rely on nonprofit financial reports for their donations allocation purposes, nonprofit managers have an incentive to manipulate those financial reports. Several studies find that nonprofit managers respond to this incentive by understating their administrative and/or fundraising expenses, which results in increased charitable expenses and improved reported efficiency. Jones and Roberts (2006) examine situations in which nonprofits engage in joint educational and fundraising campaigns (i.e., combining a plea for donations with information about their charitable purpose). Under these situations, the nonprofit is required to partition the costs of such campaigns into charitable (the education portion) and fundraising expenses. Jones 3 The exact empirical specification of the measure of charitable relative to total expenses varies across the studies, but all capture a similar underlying relationship. 5

and Roberts find that nonprofits use joint costs to mitigate changes in the program ratio. Krishnan et al. (2006) find that many nonprofits report zero fundraising expenses when they in fact are engaging in fundraising activities. By understating their fundraising expenses (all the way to zero), these nonprofits are increasing the proportion of expenses classified as charitable. More recent research by Keating et al. (2008) finds that nonprofits that engage in telemarketing campaigns systematically understate their fundraising costs, which has the effect of overstating their charitable costs. 2.3 Donors Use of Low Quality Financial Information To summarize the existing body of research, it is relatively well established that donors respond to reported nonprofit financial information, rewarding nonprofits with higher ratios of charitable to total expenses with more donations. A growing body of research shows that managers are apparently aware of donor sensitivity to this ratio and respond by manipulating the amounts of expenses reported as charitable. The question that prior research leaves largely unanswered is the extent to which donors disentangle managerial manipulation when making their donations allocation decisions. My primary research hypothesis (in null form) is that donors do not disentangle nonprofit financial statement management but rather naively use the information as reported regardless of its quality. Tinkelman (1996, 1999) examines whether the relation between donations received by a nonprofit and the PSR is affected by the reliability of the PSR (as measured by whether the nonprofit reported any fundraising expenses) using a small dataset of New York regulatory filings. There are several key differences between Tinkelman s work and this study. First, Tinkelman uses a limited sample of New York nonprofit organizations across a limited time period, whereas this study uses a broad sample of nonprofit organizations across a long time 6

period. His sample of New York nonprofits may not be generalizable. In addition, the available of nonprofit information has changed substantially since the early 1990s (his sample time frame) and it is possible that the results from that time period do not hold in more current time periods. Second, Tinkelman does not screen on reasons that a firm could plausibly report zero fundraising expenses (Krishnan et al. 2006), which introduces bias into his reliability measure. Third, Tinkelman uses New York regulatory filings to measure whether a firm reports fundraising expenses, whereas this study uses the publicly available IRS 990. It is key to use the publicly available IRS 990 to address the question of whether donors are fooled by expense misreporting because nonprofits have both a stronger ability and incentive to manage their IRS 990 relative to their regulatory reports and because empirical evidence has documented differences in the reporting of amounts in their expense categories across their publically available and regulatory reports (Krishnan and Yetman 2009). Fourth, Tinkelman uses reported fundraising to control for the effect of fundraising effort on donations. This is a biased proxy as many firms that report zero fundraising are actually underreporting it (Krishnan et al. 2006). This bias has the effect of potentially inducing his results. I improve upon the proxy for fundraising effort by estimating fundraising effort for organizations reporting zero fundraising. 2.3.1 Time-Series Variation I also examine whether the disentanglement of low quality financial reporting has improved as the ease of obtaining nonprofit financial information has increased. Both new IRS regulations and the development of Guidstar have made it easier and more cost effective to obtain IRS 990s over the past ten years. Before 1999, IRS regulations required nonprofits to allow the public to inspect the IRS 990 in person. Without having access to an IRS 990, donors would have to rely on published reports of nonprofit efficiency ratios, such as Wise Giving Alliance or donations 7

solicitation information received from the nonprofit. Thus, donors may have used the PSR when making donations decisions before 1999, but it was much more difficult and costly to obtain the details regarding expense reporting and therefore was more difficult to identify low quality reporting. In June 1999, the regulations changed and required organizations to respond to a written request for their IRS 990, making it much easier for the public to obtain IRS 990 (although a formal request was still necessary and the nonprofit was permitted to charge fees for coping and mailing). Guidestar was created in 1994, and by 1996 they collected information (e.g., data on revenues and expenses) on over 40,000 charities and launched their website (www.guidestar.org). In 1998 they expanded their database to include all 501(c)(3) public charities in the IRS Business Master File and added a digitized IRS 990 to charities' reports. By 1999, they report IRS 990s for more than 200,000 public charities. This leads to my second research hypothesis (in alternative form), which is that donors ability to disentangle nonprofit financial reporting management improved in the time periods in which the ease of obtaining financial reports increased (i.e., 1996, 1998, 1999). 3. Empirical Model and Variable Definitions To test my hypotheses, I build on prior empirical economic models on the determinants of donations (Weisbrod and Dominguez 1986, Okten and Weisbrod 2000). The model is as follows: Private Donations i,t = β 0 + β 1 PSR i,t-1 + β 2 Zero Fundraising Indicator i,t-1 + β 3 PSR i,t * Zero Fundraising Indicator i,t-1 + β 4 Fundraising Effort i,t-1 + β 5 Age i,t + β 6 Age i,t * Fundraising Effort i,t-1 + β 7 Government Grants i,t-1 + β 8 Feeder Donations i,t-1 + β 9 Sales Revenues i,t-1 + β 10 Assets i,t + ε. (1) 8

Consistent with prior research, the model is estimated using natural logs and expresses private donations as a function of lagged information (with the exception of age and assets), consistent with the idea that donors are expected to respond to information from the prior year. Private Donations (line 1a of the IRS 990) is the dollar amount of donations from individuals and corporations, as well as grants from foundations. PSR is the proportion of charitable expenses (line 13 on the IRS 990) as a fraction of total expenses (line 17 on the IRS 990); thus β 1 captures donor sensitivity to an organization s relative efficiency of operations. I expect the coefficient on the ratio to be positive, in line with prior research, suggesting that donors reward nonprofits that are more efficient. The primary variable of interest is the interaction of this ratio with the reporting quality variable, Zero Fundraising Indicator, which is equal to one if the nonprofit reported zero fundraising expenses (line 15 on the IRS 990), and zero otherwise. If donors at least partially see through low quality financial reports and adjust their sensitivity to the ratio, the interaction term will be negative and significantly different from zero. Alternatively, if donors are unable to see through (or see through but don t care about) low quality nonprofit financial reports, the interaction will not be statistically different from zero. There are several possible candidates for the reporting quality metric, including misreported telemarketing expenses (Keating et al. 2008), misreported joint costs (Jones and Roberts 2006), and misreported fundraising expenses (Krishnan et al. 2006). The reporting quality metric I use is from Krishnan et al., who find that when a nonprofit receives substantial private donations (over $10,000 in my study), yet reports zero fundraising expenses, it is highly likely that the organization has understated its fundraising expenses and, thus, overstated its charitable expenses. 4 I favor this reporting quality metric over other possible candidates for two reasons. 4 The central notion of this prior analysis is that earning donations is a production function and fundraising effort is a primary input. Although some donations can plausibly be raised with zero fundraising effort, in order to consistently 9

First, it is the only metric that applies to the majority of nonprofits. Information on telemarketing expenses as examined by Keating et al. is available for approximately one percent of all nonprofits, and the joint cost metric examined by Jones and Roberts applies to less than one percent of all nonprofits. Second, the Krishnan et al. (2006) quality metric is arguably the easiest for a donor to observe prior to making their donation allocation decisions. To observe misreported telemarketing expenses as examined by Keating et al. (2008), a donor would need to access state regulatory filings and compare them to the organization s IRS 990. To observe misreported joint costs as examined by Jones and Roberts (2006), a donor would need to employ a statistical expectations model to determine expected joint costs and compare that figure to actual joint costs. By contrast, the metric examined by Krishnan et al. is simply whether or not the nonprofit reports any fundraising expenses. To observe zero reported fundraising expenses, a donor needs to examine only a single line on the front page of the IRS 990 (i.e., line 15). 5 If it is zero and the nonprofit reports earning substantial donations and has no other plausible reason to be reporting zero fundraising expenses (e.g., it received donations through a feeder organization), the nonprofit is likely understating its fundraising expenses and correspondingly overstating its charitable expenses. Thus, a large benefit of using this metric is that it is relatively easy to disentangle. To the extent donors do not disentangle zero reported fundraising, it is unlikely that they would disentangle any of the other, arguably more complex, types of financial statement manipulation documented in the literature. In other words, the results in this study set a lower bound on donors ability to disentangle nonprofit financial statement management. earn larger amounts of donations, a nonprofit must undertake some fundraising effort, and this effort should reveal itself as an expense on the IRS 990. 5 There are many ways to obtain an organization s IRS 990. First, every organization must supply a copy if asked. Second, the IRS will supply a copy for a small fee if asked. Third, all IRS 990s of all organizations can be found at www.guidestar.org, a free Internet-based service. 10

Fundraising Effort is the reported dollar amount of fundraising expenses (line 15 of the IRS 990) for those that report non-zero fundraising, or estimated values of fundraising for those that report zero fundraising. This variable serves to control for the direct effects of fundraising effort, which is that of informing potential donors of the charitable need, much as advertising informs potential buyers of a product or service (Weisbrod and Dominguez 1986, Okten and Weisbrod 2000). For firms that report non-zero fundraising on their IRS 900, in my primary analyses I estimate their fundraising effort by using the statistical method of calibration (i.e., reverse regression). I begin by using the firms that report non-zero fundraising to observe the relationship between donations and fundraising by estimating industry specific regressions of fundraising on donations. Then, using the industry specific regression estimates, I estimate fundraising effort for the sample of firms that report zero fundraising. In robustness tests, I utilize alternative methods to estimate fundraising effort, as well as estimate the model without controlling for fundraising effort to ensure the results are not sensitive to my primary method for estimating fundraising effort. Age is the age of the nonprofit in years and is intended to be a proxy for reputation capital (Weisbrod and Dominguez 1986, Okten and Weisbrod 2000). The next three variables control for other sources of revenues that could crowd out (or crowd in) private donations (Posnett and Sandler 1989). Government Grants is the dollar amount of grants received from federal or state agencies (line 1b on the IRS 990), and Feeder Donations is the dollar amount of donations from federated fundraising organizations such as the United Way (line 1b on the IRS 990), Sales Revenues is the dollar amount of revenues received from the sales of products and services (line 2 of the IRS 990), which could also crowd out (or crowd in) private donations (need cite). Donors could view sales revenues favorably as a form of self-help or, unfavorably, as a 11

distraction from the primary exempt mission. Assets is year-end total assets and is included to control for size. All models include single digit industry and year indicator variables. 6 Because the analysis uses a pooled cross section, the possibility of non-independence of observations arises. In particular, it is very likely that an organization s financial reporting behavior and the amount of donations received is correlated across time, in which case ordinary least squares will produce downwardly biased standard errors (although the coefficient estimates will remain unbiased). A direct way to address this issue is to use the full panel of data, but to adjust the standard errors using the method of White (1980), with an additional adjustment for repeated firm observations (i.e., clustered" standard errors). 7 In addition, in all analyses, I screen for influential observations using Cooks D and Welsch Distance (Belsley et al. 1980). 4. Data 4.1 Sample The IRS 990 is the primary source of publicly available nonprofit financial information. All nonprofits with revenues over $25,000 must file the IRS 990 annually. Congressional reports suggest that the IRS 990 serves as the primary source of publicly available nonprofit financial information (Joint Committee on Taxation 2000). To ensure the wide dissemination of IRS 990 information, the IRS Statistics of Income division sponsors the Urban Institute to collect and make freely available IRS 990 data for virtually all nonprofits. This data can be found on the 6 The National Taxonomy of Exempt Entities was established by the Internal Revenue Service as a means of categorizing nonprofits into 26 broad categories. Information related to the nomenclature can be found at http://nccs.urban.org/classification/ntee.cfm. 7 Peterson (2009) provides an extensive review and analysis of the various methods used to address correlations across time and/or firms and provides the unambiguous recommendation that if a firm effect is suspected to be present (i.e., there is correlation across time within firms), the standard errors should be clustered by firm. Froot (1989), Rogers (1993), and Peterson (2009) show that clustered standard errors are unbiased and produce correctly sized confidence intervals in the presence of either temporary or permanent firm effects. Furthermore, clustered standard errors are robust to heteroskedasticity. The fixed effects estimator is not tractable as the zero fundraising indicator variable has little within-firm variation. 12

Internet at www.guidestar.org or can be obtained in computer readable form from the National Center for Charitable Statistics for a fee at www.nccs.urban.org. The IRS 990 contains typical financial statements, including a statement of revenues and expenses and a balance sheet, as well as a substantial amount of other information related to the nonprofit s charitable purpose and activities. 8 The data I use for the analysis is from the Internal Revenue Statistics of Income files. The sample includes only public charities (and excludes private foundations) exempt under Internal Revenue Code 501(c)(3). Although there are well over 10 types of tax-exempt organizations, 501(c)(3) organizations comprise the economic bulk and are differentiated from other types of nonprofits both because they are tax exempt and because their donors receive tax deductions for their donations. The IRS data is available from 1982 to 2005, with the exception of 1984. Since I need lagged values in my analysis, I begin with 1985. The original full sample from 1985 to 2005 is reduced by requiring a minimum of $10,000 in private donations as nonprofits that receive donations in excess of $10,000 are likely to have undertaken at least some fundraising effort, thus reducing possible measurement error in the zero fundraising indicator variable. 9 In addition, I remove observations that have alternative explanations for reporting zero fundraising as described Krishnan et al. (2006). Alternative explanations for reporting zero fundraising involve organizations with transactions between related parties (where one organization could reimburse the other for its administrative or fundraising expenses) or organizations with fundraising as their primary purpose (as these organizations could classify those expenses as charitable rather than fundraising). Thus, I remove 8 Keating and Frumkin (2003) provide an in-depth discussion of the IRS 990. 9 Consistent with Krishnan et al. (2006), I exclude the other two types of donations (i.e., government grants and feeder donations such as those from the United Way) as those can frequently be raised with little consistent fundraising effort. In addition, results are not sensitive to decreasing this minimum private donations threshold down to $1 or increasing it up to $1 million. 13

organizations that are members of affiliated groups as reported on lines H and J of the IRS 990, auxiliary and fundraising organizations, whose primary purpose is fundraising (denoted by two digit NTEE codes ending in a 11, 12 and 19), and philanthropic organizations for reasons similar to fundraising organizations (denoted by NTEE code T ). Finally, I require nonmissing values for all of the regression variables. The final sample includes 96,217 usable observations. 4.2 Descriptive Statistics Table 1 provides the descriptive statistics of the analysis variables partitioned across Fundraising. All variables, except for Age and PSR, are scaled by $1,000. Of this final sample of 96,217, approximately 39 percent report zero fundraising. Of those with zero fundraising, average private donations are approximately $1.2 million. For these zero-reporters, the average PSR is 83 percent. On average, observations that report fundraising expenses report more donations ($6.6 million) and a lower PSR (79 percent). This difference in PSRs is at least partially due to under-reported fundraising expenses for those that report zero. Table 2 presents the Pearson correlation statistics for some of the analysis variables. As expected, the amount of private donations received has a large economic and statistically significant positive correlation with the amount of fundraising expenses reported. Because of this, coupled with the fact that the PSR is negatively affected by the amount of fundraising effort, it is difficult to interpret the relationship between PSR and donations in a univariate sense. However, although the correlation between private donations and the PSR is negative, it is smaller for firms with lower quality reporting as measured by the zero fundraising indicator (- 0.4561), consistent with my hypothesis of donors being less sensitive to the program ratio when the nonprofit reports zero fundraising. This is consistent with donors at least partially disentangling nonprofit financial statement management. However, as evidenced by the negative 14

sign on the correlation between PSR and private donations, a multivariate analysis is required to control for variables correlated with both PSR and private donations. 5. Results The multivariate empirical results are contained in tables 3 through 6. Table 3 provides the full sample pooled analysis, whereas tables 4, 5, and 6 partition the sample across time, size, and industry. 5.1 Primary Results The model estimation reported in the first column in table 3 is based on the model used in prior research, which does not consider quality differences in the program service ratio and uses reported fundraising as a proxy for fundraising effort (Weisbrod and Dominguez 1986, Okten and Weisbrod 2000). The results show that donors respond positively both to the amount of fundraising expenses and to the level of the PSRs. Column 2 reports the primary full-sample results, which includes the zero fundraising indicator and its interactions and estimates fundraising effort for nonprofits that report zero fundraising. The coefficient on the PSR, which measures donor sensitivity to the PSR for those nonprofits that reported non-zero fundraising, remains statistically significant and more than doubles in magnitude. The coefficient on the PSR interacted with the zero fundraising indicator measures the incremental donor sensitivity to the PSR for those nonprofits that report zero fundraising and is statistically negative. This shows that donors are less sensitive to the PSR when a nonprofit reports zero fundraising expenses and suggests that donors at least partially disentangle nonprofit financial statement manipulation. To obtain the total donor sensitivity to the PSR for those nonprofits that report zero fundraising expenses, it is necessary to add the 15

coefficients of the PSR and the interaction of the PSR and zero fundraising indicator variable. The bottom of table 3 presents the result of this test and shows that the sum of the coefficient on the PSR plus the coefficient on the fundraising interaction variable is not statistically different from zero, suggesting that donors ignore the PSR when a nonprofit reports zero fundraising expenses. Finally, the magnitude of the coefficient on fundraising expenses increases from that reported in prior research (e.g., Okten and Weisbrod 2000, Tinkelman 1996) because I reduce the amount of bias in the measurement of this variable by estimating an amount of fundraising effort for those firms that report zero fundraising on their IRS 990s. Overall, the results suggest that for the full sample of nonprofits, the PSR has significant explanatory power for donations across firms that report fundraising expenses, but not for firms that report zero fundraising expenses. These results are robust to excluding the fundraising control (column 3) and to using reported fundraising as the fundraising control (column 4). 10 5.2 Time-Series Variation Table 4 reports the analysis by year. Consistent with financial reports providing a means for donors to evaluate whether the nonprofit is using donations towards the charitable mission in an efficient manner, the weight applied to the program service ratio has generally increased overtime. The increases are especially pronounced after 1999 (after increased regulations making it easier to obtain IRS 990s). Even more striking is the weight applied to the PSR for the low quality reporters. The decreases are especially pronounced around the 1996 1998 (around the time of Guidestar s 10 While it may seem odd that the coefficient on Zero Fundraising Indicator in column 3 is significantly positive (suggesting that firms that report zero fundraising obtain more donations), this is likely capturing the fact that there is no control for fundraising effort for the zero reporters, and thus the coefficient on Zero Fundraising Indicator is capturing part of that effect. This can be confirmed by excluding the Fundraising Effort control for all firms, as I do in the final column. Since there is no control for fundraising effort, its effect gets captured in the constant, and the Zero Fundraising Indicator returns to a negative sign. 16

reporting) and the 1999-2001 (after increased regulations making it easier to obtain IRS 990s). This is consistent with my conjecture that the ability to disentangle low quality financial reporting increased as the ease of obtaining nonprofit financial information increased. 5.3 Sensitivity Tests To assess the sensitivity of our results across size, we estimate the model across asset quartiles and report the results in table 5. Across the bottom three asset quartiles, consistent with the primary results, there is a significant positive relationship between the PSR and donations, which is significantly decreasing in accounting quality (i.e., PSR * Zero Fundraising Indicator). For the smallest size quartile, donors still reward nonprofits with higher PSRs even when they are of low quality, whereas for the middle quartiles, however, donors punish nonprofits with higher PSRs when the PSR is of low quality (see PSR + PSR * Zero Fundraising Indicator). For the largest size quartile, contrary to the primary results, the positive relationship between the PSR and donations is not significant. To assess the sensitivity of our results across industry, we provide an industry analysis in table 6. In general, the nonprofit sector is broken down into three broad categories of organizations: educational, medical, and charitable. The results in table 6 show that, with the exception of international and religious nonprofits, donations to nonprofits are associated with the PSR only for nonprofits reporting non zero fundraising (the t-test reveals that the coefficient on PSR is significantly positive, but the F-tests show that the sum of the PSR and the interaction are not different from zero). The results for religious charities need to be conditioned on the selfselected sample. Internal Revenue Service (IRS) rules do not require, but do permit, religious charities to file an IRS 990. Thus, the sample of religious organizations is not only a very small part of the total religious charity industry. 17

5.4 Summary To summarize my empirical results, consistent with prior research I find that donors reward nonprofits that devote larger proportions of their total expenses towards charitable purposes with more donations. However, when a nonprofit reports zero fundraising expenses (but ostensibly should be reporting some), donors generally appear to partially, and sometimes completely, discount the PSR. This is consistent with donors being able to detect disclosure management of the PSR through reporting zero fundraising expenses and suggests that nonprofits may not gain by reporting zero fundraising. 11 6. Conclusions Nonprofit organizations play an important role in society, and donations are an important nonprofit revenue source. By choosing which nonprofits are worthy of receiving their money, donors implicitly select the particular public goods they want supplied. One source of information relied upon by donors for making this decision is publicly available financial information. In particular, donors reward nonprofits that devote higher proportions of their expenses towards charitable purposes with more donations. Recent evidence suggests that nonprofit financial information is at best of varying quality and at worst is intentionally manipulated. If donors are unable to determine the quality of an organization s financial information, they could be misled into making donation decisions different from what they would otherwise make with high quality financial information. 11 It is unclear whether a nonprofit reporting zero fundraising receives more or less donations from its disclosure management. If donors allocate their donations to zero fundraising reporters in smaller amounts then they would have if they knew the actual charitable ratio, the nonprofit is worse off by managing its financial reports. On the other hand, if a donor gives larger amounts to zero fundraisers then they otherwise would have, the nonprofit is better off. My tests cannot shed light on this question. What is clear from my results is that donors are able to see through zero reported fundraising expenses and essentially ignore the reported amounts of expenses for those organizations that report zero fundraising. 18

This analysis has shown that donors are at least partially able to see through low quality nonprofit financial information (defined as those that report zero fundraising expenses when they should plausibly be reporting some amount). In particular, donors appear to discount the amounts of expenses reported as devoted towards the charitable mission for those organizations that report zero fundraising expenses. Whether or not nonprofits are worse off by reporting zero fundraising is an unanswered question that is worthy of future research. Additional areas of investigation include partitioning the results across different classes of donors. Although my results apply to the average donor, it is possible that some donors are fooled by nonprofit financial statement information even if the average donor is not. Identifying which donor types are more or less likely to be misled by low quality nonprofit financial information would be a valuable, but difficult, undertaking. 19

References Belsley, D., E. Kuh, and R. Welsch. 1980. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. New York, N.Y: Wiley. Coase, R. 1937. The nature of the firm. Economica 4: 386 405. Fama, E., and M. Jensen. 1983. Agency problems and residual claims. Journal of Law and Economics 26: 301 325. Froot, K. A. 1989. Consistent covariance matrix estimation with cross-sectional dependence and heteroskedasticity in financial data. Journal of Financial and Quantitative Analysis 24: 333 355. Giving USA Foundation, 2007. Giving USA. Glenview, IL. Hansmann. H. 1996. The Ownership of Enterprise. Page numbers? Cambridge, MA: Belknap Press. Jensen, M.C., and W.H. Meckling. 1976. Theory of the firm: Managerial behavior, agency costs, and ownership structure. Journal of Financial Economics 4: 469 506. Joint Committee on Taxation. 2000. Study of present-law taxpayer confidentiality and disclosure provisions as required by section 3802 of the Internal Revenue Service Restructuring and Reform Act of 1998. Volume I. (January). Jones, C., and A. Roberts. 2006. Management of financial information in charitable organizations: The case of joint cost allocations. The Accounting Review 81: 159 178. Keating, E., and P. Frumkin, P. 2003. Reengineering nonprofit financial accountability: Toward a more reliable foundation for regulation. Public Administration Review 63 (1) (January/February): 3 15. Keating, E., L.M. Parsons, and A.A. Roberts. 2008. Misreporting fundraising: How do nonprofit organizations account for telemarketing campaigns? The Accounting Review 83 (2): 417 446. Krishnan, R. and M.H. Yetman. 2009. Strategic cost shifting by nonprofit hospitals. Working paper, Michigan State University and University of California at Davis. Krishnan, R., M. Yetman, and R. Yetman. 2006. Expense Misreporting in Nonprofit Organizations. The Accounting Review 81: 399 420. National Center for Charitable Statistics. Internal Revenue Service Statistics of Income form 990 sample files. Center on nonprofits and Philanthropy/Urban Institute online database. Available at http://nccs.urban.org. 20

Okten, C., and B.A. Weisbrod. 2000. Determinants of donations in private nonprofit markets. Journal of Public Economics 75: 255 272. Parsons, L.M. 2007. The impact of financial information and voluntary disclosures on contributions to not-for-profit organizations. Behavioral Research in Accounting 19, 179 196. Peterson, M. 2009. Estimating standard errors in finance panel data sets: comparing approaches. Review of Financial Studies. 22(1): 435 480. Posnett, J., and T. Sandler, T. 1989. Demand for charity donations in private non-profit markets. Journal of Public Economics 40: 187 200. Rogers, W. 1993. Regression standard errors in clustered samples. Stata Technical Bulletin.13: 19 23. Strom, S. 2004. Charities face increase reviews by I.R.S. as Senate considers strengthening oversight. The New York Times, (June 23): page number. Tinkelman, D. 1996. An Empirical Study of the Effect of Accounting Disclosures upon Donations to Nonprofit Organizations. Dissertation, Stern School of Business. Tinkelman, D. 1999. Factors affecting the relation between donations to not-for-profit organizations and an efficiency ratio. Research in Government and Nonprofit Accounting 10: 135 161. Tinkelman, D. 2004. Using nonprofit organization-level financial data to infer managers fundraising strategies. Journal of Public Economics 88: 2181 2192. United States Senate Committee on Finance. 2004. Charity Oversight and Reform: Keeping Bad Things from Happening to Good Charities. June 22. United States Senate Committee on Finance. 2005. Charities and Charitable Giving: Proposals for Reform. April 5. Weisbrod, B., and N. Dominguez. 1986. Demand for collective goods in private nonprofit markets, Can fund-raising expenditures help overcome free-rider behavior?. Journal of Public Economics 30: 83 95. White, H. 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 56: 817 838. 21

Table 1 Descriptive Statistics Partitioned Across Zero Reported Fundraising Fundraising = 0: 37,564 observations Fundraising > 0: 58,653 observations t-test of difference Variable Mean Median Std. Dev. Mean Median Std. Dev. Private Donations 1,248 176 9,421 6,620 1,501 30,115 33.53 PSR 0.83 0.86 0.15 0.79 0.82 0.13-46.31 Fundraising Effort 359 202 1386 838 251 2,851 30.33 Age 31.19 30 18.68 37.95 39 19.28 53.69 Government Grants 1,638 0 11,661 4,260 59 29,926 3.36 Feeder Donations 312 0 30,160 446 0 7,345 16.21 Sales Revenues 48,775 7,419 264,712 33,386 5,025 117,646-12.31 Assets 66,307 18,755 242,185 121,301 24,831 731,240 14.09 Notes: The sample period includes 96,217 organization-year observations from 1985 to 2005. All variables except for Age and PSR are scaled by 1,000 for presentation purposes (logged values are used in the regression). All variables other than Private Donations, Age, and Assets are lagged by one year. Private Donations = the amount of donations from individuals and corporations (line 1a of the IRS 990), PSR = the amount of charitable expenses (line 13 on the IRS 990) scaled by total expenses (line 17 on the IRS 990), Fundraising Effort = the amount of fundraising expenses (line 15 of the IRS 990 for those that report non-zero fundraising, or estimated values of fundraising for those that report zero fundraising), Age = the age of the nonprofit in years, Government Grants = grants from federal or state agencies (line 1b on the IRS 990), Feeder Donations = donations from federated fundraising organizations such as the United Way (line 1b on the IRS 990), Sales Revenues = revenues from the sales of products and services (line 2 of the IRS 990), Assets = year end total assets (line 59b of the IRS 990). 22

Table 2 Pearson Correlation Statistics Private Donations PSR Zero Zero Fundraising Fundraising Indicator * Indicator PSR Fundraising Effort Age Government Grants Feeder Donations Sales Revenues PSR -0.0399* Zero Fundraising Indicator -0.4616* 0.1334* Zero Fundraising Indicator * PSR -0.4561* 0.2602* 0.9809* Fundraising Effort 0.5828* 0.0170* -0.0172* -0.0146* Age 0.2535* 0.0234* -0.1556* -0.1551* 0.2751* Government Grants 0.3989* 0.0233* -0.1733* -0.1715* 0.4878* 0.8719* Feeder Donations 0.3015* 0.0433* -0.1957* -0.1903* 0.2867* 0.1166* 0.1802* Sales Revenues 0.0211* 0.0592* -0.0418* -0.0383* 0.0233* 0.0895* 0.0700* 0.1416* Assets 0.1442* 0.1484* -0.0037 0.0055 0.3497* 0.3022* 0.3428* 0.0929* 0.0356* Notes: Variable definitions are in Table 1. Values marked with * are statistically significant at the 5 percent level. 23

Table 3 The Effects of Financial Reporting Quality on Donor Sensitivity Reported Fundraising as proxy for Fundraising Effort Reported Fundraising as proxy for Fundraising Effort No Fundraising Control Fitted Fundraising as proxy for Fundraising Effort CONSTANT 4.703 2.412 4.678 3.474 (34.17) (18.21) (32.45) (24.24) PSR 0.384 0.988 0.677 1.199 (3.33) (7.88) (4.44) (9.57) Zero Fundraising Indicator 4.204-0.503-0.405 (27.91) (4.01) (3.62) PSR * Zero Fundraising Indicator -1.181-0.853-1.445 (6.08) (4.07) (7.72) Fundraising Effort 0.089 0.396 0.409 (27.16) (43.23) (45.32) Age -0.046 0.010 0.026-0.165 (2.66) (0.58) (1.76) (7.74) Age * Fundraising Effort 0.000-0.000 0.001 (5.93) (1.35) (8.03) Government Grants 0.022 0.015 0.026 0.008 (12.86) (9.28) (14.67) (4.92) Feeder Donations 0.011 0.011 0.012 0.007 (5.30) (5.56) (5.63) (3.88) Sales Revenues -0.044-0.045-0.044-0.045 (15.19) (16.34) (14.63) (17.38) Total Assets 0.525 0.421 0.574 0.361 (61.08) (47.81) (65.12) (43.55) PSR + PSR * Zero Fundraising Indicator -0.193-0.176-0.246 Observations 96,217 96,217 96,217 96,217 Adjusted R-squared 0.60 0.63 0.57 0.64 Notes: Variables are defined in Table 1. Logged values are used in the regression. t-statistics are in parentheses under their respective coefficient estimates. All models correct the standard errors for clustering within an organization using the method of White (1980), with an adjustment for repeated organizations across years. All models include single digit industry controls as well as year controls. * indicates a statistically significant F test (at the five percent level) for whether the sum of the coefficient for PSR plus the coefficient for PSR * Zero Fundraising Indicator is equal to zero. 24

Table 4 The Effects of Financial Reporting Quality on Donor Sensitivity, by Year Years: 87-89 90-92 93-95 96-98 99-01 02-05 CONSTANT 3.528 3.544 3.064 3.348 3.367 3.656 (15.88) (12.92) (11.41) (14.50) (16.36) (19.33) PSR 0.858 1.201 0.987 1.173 1.565 1.451 (4.38) (5.43) (4.56) (5.74) (8.31) (7.89) Zero Fundraising Indicator -0.743-0.563-0.237-0.009-0.241-0.199 (4.53) (3.17) (1.18) (0.04) (1.18) (1.04) PSR * Zero Fundraising -0.734-1.048-1.572-2.105-1.825-1.895 Indicator (2.61) (3.50) (4.68) (6.05) (5.38) (5.93) Fundraising Effort 0.344 0.379 0.389 0.411 0.425 0.451 (22.24) (23.37) (25.21) (26.92) (31.80) (39.35) Age -0.170-0.169-0.073-0.140-0.202-0.226 (4.26) (3.54) (1.64) (3.73) (5.93) (7.30) Age * Fundraising 0.001 0.001 0.000 0.001 0.001 0.001 Effort (4.67) (4.59) (2.97) (5.75) (5.67) (7.45) Government Grants 0.012 0.011 0.008 0.005 0.007 0.004 (4.49) (4.13) (3.28) (1.96) (3.48) (2.02) Feeder Donations 0.005 0.011 0.012 0.006 0.005 0.005 (1.53) (3.37) (3.85) (2.15) (1.79) (2.19) Sales Revenues -0.041-0.047-0.048-0.044-0.041-0.045 (10.08) (10.30) (10.63) (11.41) (12.32) (14.94) Assets 0.409 0.377 0.397 0.367 0.350 0.317 (29.01) (22.52) (26.96) (27.34) (29.96) (32.23) PSR + PSR * Zero Fundraising Indicator 0.124 0.153-0.585* -0.932* -0.260-0.444 Observations 10,746 12,930 13,504 14,440 18,374 24,679 Adjusted R 2 0.65 0.62 0.62 0.64 0.64 0.66 Notes: Variables are defined in Table 1. Logged values are used in the regression. t-statistics are in parentheses under their respective coefficient estimates. All models correct the standard errors for clustering within an organization using the method of White (1980), with an adjustment for repeated organizations across years. All models include single digit industry controls as well as year controls. * indicates a statistically significant F test (at the five percent level) for whether the sum of the coefficient for PSR plus the coefficient for PSR * Zero Fundraising Indicator is equal to zero. 25

Table 5 The Effects of Financial Reporting Quality on Donor Sensitivity by Asset Quartiles Quartile 1 < $6.7 million Quartile 2 $6.7 $22.3 million Quartile 3 $23.4 $68.4 million Quartile 4 > $68.4 million CONSTANT 5.112 3.447 2.904-1.392 (23.92) (4.90) (3.96) (2.71) PSR 1.177 0.921 0.788 0.384 (6.58) (4.83) (3.45) (1.19) Zero Fundraising Indicator -0.645-0.244-0.271-0.387 (4.82) (1.38) (1.14) (1.02) PSR * Zero Fundraising Indicator -0.860-1.534-1.537-1.471 (3.73) (5.13) (3.87) (2.40) Fundraising Effort 0.362 0.352 0.398 0.474 (30.94) (24.48) (22.33) (18.36) Age -0.116 0.004 0.003 0.005 (3.76) (0.09) (0.07) (0.09) Age * Fundraising Effort 0.000 0.000 0.000 0.000 (2.10) (0.31) (2.57) (0.88) Government Grants -0.006-0.003 0.003 0.015 (2.44) (1.41) (1.23) (4.42) Feeder Donations -0.000 0.019 0.014 0.001 (0.13) (6.18) (4.23) (0.29) Sales Revenues -0.029-0.043-0.046-0.056 (9.38) (10.43) (8.19) (7.41) Assets 0.238 0.387 0.408 0.618 (20.89) (9.07) (9.73) (21.74) PSR + PSR * Zero Fundraising Indicator 0.317* -0.613* -0.749* -1.087* Observations 24,054 24,053 24,053 24,054 Adjusted R-squared 0.39 0.45 0.58 0.67 Notes: Variables are defined in Table 1. Logged values are used in the regression. t-statistics are in parentheses under their respective coefficient estimates. All models correct the standard errors for clustering within an organization using the method of White (1980), with an adjustment for repeated organizations across years. All models include single digit industry controls as well as year controls. * indicates a statistically significant F test (at the five percent level) for whether the sum of the coefficient for PSR plus the coefficient for PSR * Zero Fundraising Indicator is equal to zero. 26