- Email: [email protected]
Do nonprofits manipulate investment returns?
Accepted Manuscript Do nonprofits manipulate investment returns? Douglas Almond, Xing Xia
PII: DOI: Reference:
S0165-1765(17)30103-9 http://dx.doi.org/10.1016/j.econlet.2017.03.014 ECOLET 7549
To appear in:
Economics Letters
Received date : 26 January 2017 Revised date : 7 March 2017 Accepted date : 10 March 2017 Please cite this article as: Almond, D., Xia, X., Do nonprofits manipulate investment returns?. Economics Letters (2017), http://dx.doi.org/10.1016/j.econlet.2017.03.014 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*Highlights (for review)
Highlights • We find a sharp discontinuity around zero in cross-sectional distribution of the rates of return on investments for tax-exempt organizations. • Rates of return are significantly more likely to be slightly positive than slightly negative. • This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. • These findings suggest that a significant percentage of tax-exempt organizations with negative investment returns manipulate their reported investment returns upwards to avoid reporting negative returns.
1
Title Page
Economics Letters Manuscript Title: Do Nonprofits Manipulate Investment Returns? Authors: Douglas Almond Department of Economics and School of International and Public Affairs Columbia University 420 West 118th Street New York, NY 10027, USA.
Xing Xia (Corresponding Author) Division of Social Sciences Yale-NUS College 10 College Avenue West 01-101 Singapore 138609 E-mail address: [email protected]
Abstract We provide evidence that nonprofit organizations manipulate reported investment returns to avoid investment losses. We find a sharp discontinuity around zero in cross-sectional distribution of the rates of return on investments for tax-exempt organizations: rates of return are significantly more likely to be slightly positive than slightly negative. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. JEL classification: L31; K13; M40 Keywords: Nonprofit organization; Investment returns; Financial statement manipulation
1
*Manuscript Click here to view linked References
Do Nonprofits Manipulate Investment Returns? Douglas Almond⇤ and Xing Xia†‡ March 7, 2017
Abstract We provide evidence that nonprofit organizations manipulate reported investment returns to avoid investment losses. We find a sharp discontinuity around zero in cross-sectional distribution of the rates of return on investments for tax-exempt organizations: rates of return are significantly more likely to be slightly positive than slightly negative. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. JEL classification: L31; K13; M40 Keywords: Nonprofit organization; Investment returns; Financial statement manipulation
⇤
Department of Economics and SIPA, Columbia University and NBER. Division of Social Sciences, Yale-NUS College. ‡ Corresponding author. Mailing address: 10 College Avenue West 01-101, Singapore 138609. Telephone: +656601-5544. E-mail address: [email protected] †
1
Public charities frequently rely on the investment returns of their endowment and other internal investment funds to finance operations. Tax-exempt status releases charities from paying income tax on investment returns. Nevertheless, they must disclose their performance to the general public through IRS tax forms. The information disclosure per se carries no tax implications. However, potential donors may take investment returns as a signal of sound financial management. Furthermore, employees who oversee investments may view their job performance as tied to reported returns. In this paper, we provide evidence that public charities manipulate their investment returns information. To the extent that returns on risky investments are stochastic and are not manipulated, we expect the pdf of risky investment returns to be continuous in the rate of return (Lee and Lemieux, 2010). We find a sharp discontinuity of the rate of return on investments around zero for tax-exempt organizations: rates of return are significantly more likely to be slightly positive than slightly negative. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. These findings suggest that a significant percentage of tax-exempt organizations with negative investment returns manipulate their reported investment returns upwards. This paper relates to a larger literature on earnings management that considers how firms use accounting techniques to generate financial reports that depict a positive view of the firm’s financial position. Burgstahler and Dichev (1997) provide evidence that for-profit firms manipulate reported earnings to avoid earnings decreases and losses. Bergstresser et al. (2006) show that for-profit firms with large pension plans manipulate reported earnings through manipulating the assumed rates of return on pension assets. Bollen and Pool (2009) find discontinuities in the pooled distribution of monthly hedge fund returns but not in the bimonthly returns, which suggests that hedge funds manipulate reported returns but reconcile them with actual returns over the course of two months. In contrast, we find a discontinuity at zero even for the six-year average rate of return of nonprofit endowment funds. A separate strand of research studies financial disclosure management of nonprofits. Most research in this area uses accounting models to compare reported activities with predicted activities. There is evidence that some nonprofits manipulate the allocation of expenses across different categories to reduce tax liabilities on taxable activities (Weisbrod and Cordes, 1998; Yetman, 2001) or to increase reported program-spending ratios (Trussel, 2003). Hofmann and McSwain (2013) review this literature. To our knowledge, ours is the first paper to consider the manipulation of reported investment returns at nonprofits. More broadly, our regression discontinuity approach to evaluate the potential for manipulation in investment returns fits within the sleuthing literature of "forensic economics" (Zitzewitz, 2012).
1
Data
We analyze microdata from the IRS Statistics of Income (SOI) Form 990 sample files. Most taxexempt organizations with gross receipts over $200,000 or total assets greater than $500,000 are
2
required to file Form 990. The SOI sample files contain all financial variables on Form 990 for all filing organizations with more than $10 million in assets plus a random sample of approximately 4,000 filing organizations with assets between $25,000 and $10 million.1 All major universities and hospitals are included in the SOI files because of their size. However, because religious organizations of any size are not required to file Form 990, the SOI files do not contain a representative sample of religious organizations. Although the IRS SOI data files are available from 1988 onwards, full information on investment returns was not included in the IRS SOI files until 2012. Investment returns are reported through four separate components on Form 990: interest and dividends, bond proceeds, net realized gains from sales of securities, and net unrealized gains on investments. While the first three components have always been part of Form 990, the last component – net unrealized gains on investments – was not included in Form 990 until 2012.2 We make use of the two years of investment returns data that are currently available – 2012 and 2013. For graphical clarity, we restrict our analysis sample to institution-year observations with rates of return on investments between -0.2 and 0.4. Over 95% of institution-year observations with non-missing data for all four components of investment returns have rates of return in this range. Key for us, Form 990 also includes detailed balance sheet information. In particular, investments in interest-bearing checking or savings accounts and investments on securities are separately reported, which allows us to estimate the share of investments held in risky assets. Below, we focus on organizations with at least 95% of their investments in (risky) securities. To examine a longer time horizon, we utilize information on endowment funds reported in Schedule D of Form 990. Starting from 2008, organizations that have established endowment funds are also required to report the amount and returns of their endowment funds. We use this information to estimate the rate of return on endowment funds for years 2008 to 2013. Although endowment returns information is available for a longer time horizon, we do not observe the asset allocation of endowment funds. As a result, our endowment returns analysis sample may include organizations that invest a large percent of their funds in risk-less assets. Appendix Table A.1 lists the definitions of variables used in subsequent analysis and the locations of the corresponding data items on Form 990.
2
Evidence for the Manipulation of Investment Returns
We begin with graphs showing the distribution of the rate of return on investments. We use McCrary (2008)’s density test to test for a discontinuity in the distribution of the rate of return around zero. 1
The SOI data files can be found at https://www.irs.gov/uac/soi-tax-stats-charities-and-other-tax-exemptorganizations-statistics 2 Prior to 2012, “net unrealized gains on investments” was included in Schedule D of Form 990, part XI. However, this particular line in Schedule D was only completed by those institutions that have obtained separate, independent audited financial statements and whose audited statement employed a different reporting methodology from Form 990. Some institutions also report their unrealized investment returns through a separate attachment to Form 990. These data are not included in the IRS SOI files, although it is possible to manually extract information on unrealized gains and losses from these attachments.
3
McCrary’s density test entails estimating local linear regressions of the density function of the running variable (the rate of return, in this case) on either side of the cutoff (zero, in this case). The discontinuity at the cutoff is then estimated as the log difference in the value of the density functions on the two sides of the cutoff.3 Figure 1: Histogram for rate of return on investment
Note: This figure shows a histogram of the rate of return on investment for institution-year observations during 2012 and 2013. The sample includes tax-exempt organizations in the SOI study that held at least 95% of their total investments on securities (as opposed to risk-free, interest-bearing assets). The bin size is 0.00117, chosen to match the bin size that would be used in the McCrary’s density test.
Figure 1 presents a histogram of the rate of return on investment, with a bin width of 0.00117, chosen to match the bin size that would be used in the McCrary’s density test.4 A discontinuity is clearly discernible at zero. The first six bins to the right of zero together account for 546 organizations. This means that out of the 11,395 organizations in our restricted sample, 546 reported investment returns between zero and 0.007 (excluding zero). If we assume that the rates of return follow a continuous normal distribution with the same mean and variance as the sample mean and variance, there would be 1.4% of observations between zero and 0.007. In other words, if the rates of return were normally distributed with no discontinuity at zero, there would be approximately 160 organizations reporting rates of returns between zero and 0.007. Under these assumptions, approximately 386 out of the 11,395 tax-exempt organizations in our sample (3.4%) may have misreported their returns to avoid reporting negative returns. 3 Formally, McCrary (2008) defines the discontinuity estimate to be ✓ˆ = ln limr#c fˆ(r) ln limr"c fˆ(r), where r is the running variable and c is the cutoff. 4 Formally, McCrary (2008) defines the bin size as ˆb = 2ˆ n 1/2 , where ˆ is the sample standard deviation of the running variable.
4
Figure 2 presents the McCrary’s density test of a discontinuity around zero. The graph shows estimates of local linear regressions of the density function on either side of zero. The discontinuity estimate is 1.726, with a standard error of 0.10 (the t-stat is 17.2). Figure 2: McCrary’s density test for the rate of return on investment
Note: This figure shows results from a McCrary density test for discontinuity in the distribution at 0 for the rate of return on investment. The sample includes institution-year observations during 2012 and 2013. Institutions include tax-exempt organizations in the SOI study that held at least 95% of their total investments on securities (excluding organizations that hold more than 5% of investments in risk-free assets). The discontinuity estimate, which measures the log difference in density on the two sides of the cutoff, and the t-statistic for the significance of the discontinuity estimate is listed at the upper-right corner of the graph.
An alternative explanation for the discontinuity at zero is that some organizations invested a large share of their funds in risk-less assets that happened to yield close to zero returns. So far, we have restricted our sample to institutions that held at least 95% of their investment funds in securities at the beginning of the fiscal year. This sample restriction raises confidence that the true returns are stochastic. As a robustness check, we perform the test on organizations that held at least 95% in securities both at the beginning and at the end of the fiscal year. The results that are almost identical to those presented in Figure 2. The discontinuity estimate is 1.734, with a standard error of 0.11. We also perform the McCrary’s density test for alternative samples with different restrictions on the percentage of investments held in securities. Appendix Figure A.5 presents these results. The discontinuity at zero is discernible and significant regardless of the level of restriction placed on the percentage invested in securities. Notably, the magnitude of the discontinuity at zero is substantial even when we restrict the sample to institutions with at least 99% of their investment funds in securities. To the extent that asset allocations are correctly reported on Form 990, these 5
results suggest that we can rule out the possibility that our results are driven by the small positive returns of safety assets. However, we cannot rule out the possibility that our results are driven by misclassification of securities. If a significant share of organizations misreports investment in safety assets as investment in securities, we may also observe a discontinuity at zero. To further probe the robustness of our results, we perform the McCrary’s density test for alternative samples and on alternative running variables. Appendix Figure A.6 presents the McCrary’s density test for institutions whose financial statements were audited by an independent accountant. Appendix Figure A.7 presents results for institutions whose investment funds include donorrestricted or board-designated endowment funds. Appendix Figure A.8 presents the McCrary’s density test for an alternative running variable: the dollar amount of investment returns. In all cases, the discontinuity at zero is readily discerned and statistically significant. Additionally, excluding bond proceeds from investment returns yields results that are very similar to those presented in Appendix Figure A.8. Figure 3: McCrary’s density test for the rate of return on endowment funds
Note: These figures show results from McCrary density tests for discontinuity in the distribution at 0 for the rate of return on endowment funds. Sample includes tax-exempt organizations in the SOI study that reported endowment information on Schedule D of Form 990 and maintained a positive balance on endowment funds for all years 2008 to 2013. Left panel pools all institution-year observations from 2008 to 2013. Right panel includes the six-year average rate of return on endowment funds for each organization. The discontinuity estimate and the t-statistic for the significance of the discontinuity estimate is listed at the upper-right corner of the graph.
If organizations that manipulate reported investment returns reconcile reported returns with actual returns over time, we might expect little or no discontinuity in the pdf of the long-term average rate of return per organization. We extend our analysis to the rate of return on endowment funds to examine a longer time horizon - 2008 to 2013. Figure 3 presents the McCrary’s density test for pooled rates of return (left panel) and the six-year average rate of return per organization (right panel). Appendix Figure A.9 presents results separately for each fiscal year from 2008 to 2013. The discontinuity at zero is statistically significant for all six years. Interestingly, there is a substantial and statistically significant discontinuity at zero for the six-year average rate of return. 6
This suggests that some nonprofits repeatedly manipulate their investment returns. Next, we examine whether the tendency to manipulate investment returns differs for different types of tax-exempt organizations. We use the National Taxonomy of Exempt Entities (NTEE) classification system to categorize nonprofits organizations. Then we perform McCrary’s density test separately for each group to examine whether there is a discontinuity at zero within each group. Figure 4: McCrary’s density test for the rate of return, by type of tax-exempt organization
Note: These figures show results from McCrary density tests for discontinuity in the distribution at 0 for the rate of return on investments. Institutions are grouped according to the National Taxonomy of Exempt Entities (NTEE) classification system. The sample includes institution-year observations during 2012 and 2013. Institutions include tax-exempt organizations in the SOI study that held at least 95% of their total investments on securities. Only those industry groups that have more than 150 observations with a rate of return between -0.2 and 0.4 are shown here. The discontinuity estimate and the t-statistic for the significance of the discontinuity estimate is listed at the upper-right corner of each graph.
Figure 4 presents results for the nine major groups that have at least 150 observations within our sample restriction criterion: investing at least 95% in securities and having rates of return between -0.2 and 0.4. Among the nine groups of organizations, recreation and sports organizations have the highest discontinuity estimate (3.24), followed by religious organizations (2.40), community 7
improvement and capacity building organizations (2.10), education organizations (1.95), healthcare organizations (1.81), employment organizations (1.79), human services organizations (1.50), and mutual and membership benefit organizations (1.38). Philanthropy, voluntarism and grant making foundations (NTEE group T) have the lowest discontinuity estimate (1.08) and is the only group whose discontinuity estimate is not statistically significant at the 5% level.
3
Conclusion
Prima facie, nonprofits would appear to have little incentive to over-represent investment performance. “Under penalties of perjury,” the nonprofit’s responsible officer also avows to the IRS that “the [tax] return, including accompanying schedules and statements, and to the best of my knowledge and belief, it is true, correct, and complete.” Nevertheless, we observe nonprofits appear to behave, so that small negative reported returns are avoided and instead appear as gains. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. In this respect, nonprofits behave much like for-profit firms, which may have stronger incentives to manipulate returns and, moreover, have previously been found to do so. However, that some nonprofits report very small losses to the IRS implies that not all nonprofits engage in “window-dressing” behavior.
Acknowledgments This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The authors thank an anonymous referee for helpful comments.
References Bergstresser, Daniel, Mihir Desai, and Joshua Rauh, “Earnings Manipulation, Pension Assumptions, and Managerial Investment Decisions,” The Quarterly Journal of Economics, 2006, pp. 157–195. Bollen, Nicolas PB and Veronika K Pool, “Do Hedge Fund Managers Misreport Returns? Evidence from the Pooled Distribution,” The Journal of Finance, 2009, 64 (5), 2257–2288. Burgstahler, David and Ilia Dichev, “Earnings Management to Avoid Earnings Decreases and Losses,” Journal of Accounting and Economics, 1997, 24 (1), 99–126. Hofmann, Mary Ann and Dwayne McSwain, “Financial Disclosure Management in the Nonprofit Sector: A Framework for Past and Future Research,” Journal of Accounting Literature, 2013, 32 (1), 61–87. Lee, David S. and Thomas Lemieux, “Regression Discontinuity Designs in Economics,” Journal of Economic Literature, June 2010, 48 (2), 281–355.
8
McCrary, Justin, “Manipulation of the Running Variable in the Regression Discontinuity Design: A Density Test,” Journal of Econometrics, 2008, 142 (2), 698–714. Trussel, John, “Assessing potential accounting manipulation: The financial characteristics of charitable organizations with higher than expected program-spending ratios,” Nonprofit and Voluntary Sector Quarterly, 2003, 32 (4), 616–634. Weisbrod, Burton A and Joseph J Cordes, “Differential Taxation of Nonprofits and the Commercialization of Nonprofit Revenues,” Journal of Policy Analysis and Management, 1998, 17 (2), 195–214. Yetman, Robert J, “Tax-Motivated Expense Allocations by Nonprofit Organizations,” The Accounting Review, 2001, 76 (3), 297–311. Zitzewitz, Eric, “Forensic Economics,” Journal of Economic Literature, July 2012, 50 (3), 731–69.
9
PII: DOI: Reference:
S0165-1765(17)30103-9 http://dx.doi.org/10.1016/j.econlet.2017.03.014 ECOLET 7549
To appear in:
Economics Letters
Received date : 26 January 2017 Revised date : 7 March 2017 Accepted date : 10 March 2017 Please cite this article as: Almond, D., Xia, X., Do nonprofits manipulate investment returns?. Economics Letters (2017), http://dx.doi.org/10.1016/j.econlet.2017.03.014 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*Highlights (for review)
Highlights • We find a sharp discontinuity around zero in cross-sectional distribution of the rates of return on investments for tax-exempt organizations. • Rates of return are significantly more likely to be slightly positive than slightly negative. • This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. • These findings suggest that a significant percentage of tax-exempt organizations with negative investment returns manipulate their reported investment returns upwards to avoid reporting negative returns.
1
Title Page
Economics Letters Manuscript Title: Do Nonprofits Manipulate Investment Returns? Authors: Douglas Almond Department of Economics and School of International and Public Affairs Columbia University 420 West 118th Street New York, NY 10027, USA.
Xing Xia (Corresponding Author) Division of Social Sciences Yale-NUS College 10 College Avenue West 01-101 Singapore 138609 E-mail address: [email protected]
Abstract We provide evidence that nonprofit organizations manipulate reported investment returns to avoid investment losses. We find a sharp discontinuity around zero in cross-sectional distribution of the rates of return on investments for tax-exempt organizations: rates of return are significantly more likely to be slightly positive than slightly negative. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. JEL classification: L31; K13; M40 Keywords: Nonprofit organization; Investment returns; Financial statement manipulation
1
*Manuscript Click here to view linked References
Do Nonprofits Manipulate Investment Returns? Douglas Almond⇤ and Xing Xia†‡ March 7, 2017
Abstract We provide evidence that nonprofit organizations manipulate reported investment returns to avoid investment losses. We find a sharp discontinuity around zero in cross-sectional distribution of the rates of return on investments for tax-exempt organizations: rates of return are significantly more likely to be slightly positive than slightly negative. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. JEL classification: L31; K13; M40 Keywords: Nonprofit organization; Investment returns; Financial statement manipulation
⇤
Department of Economics and SIPA, Columbia University and NBER. Division of Social Sciences, Yale-NUS College. ‡ Corresponding author. Mailing address: 10 College Avenue West 01-101, Singapore 138609. Telephone: +656601-5544. E-mail address: [email protected] †
1
Public charities frequently rely on the investment returns of their endowment and other internal investment funds to finance operations. Tax-exempt status releases charities from paying income tax on investment returns. Nevertheless, they must disclose their performance to the general public through IRS tax forms. The information disclosure per se carries no tax implications. However, potential donors may take investment returns as a signal of sound financial management. Furthermore, employees who oversee investments may view their job performance as tied to reported returns. In this paper, we provide evidence that public charities manipulate their investment returns information. To the extent that returns on risky investments are stochastic and are not manipulated, we expect the pdf of risky investment returns to be continuous in the rate of return (Lee and Lemieux, 2010). We find a sharp discontinuity of the rate of return on investments around zero for tax-exempt organizations: rates of return are significantly more likely to be slightly positive than slightly negative. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. These findings suggest that a significant percentage of tax-exempt organizations with negative investment returns manipulate their reported investment returns upwards. This paper relates to a larger literature on earnings management that considers how firms use accounting techniques to generate financial reports that depict a positive view of the firm’s financial position. Burgstahler and Dichev (1997) provide evidence that for-profit firms manipulate reported earnings to avoid earnings decreases and losses. Bergstresser et al. (2006) show that for-profit firms with large pension plans manipulate reported earnings through manipulating the assumed rates of return on pension assets. Bollen and Pool (2009) find discontinuities in the pooled distribution of monthly hedge fund returns but not in the bimonthly returns, which suggests that hedge funds manipulate reported returns but reconcile them with actual returns over the course of two months. In contrast, we find a discontinuity at zero even for the six-year average rate of return of nonprofit endowment funds. A separate strand of research studies financial disclosure management of nonprofits. Most research in this area uses accounting models to compare reported activities with predicted activities. There is evidence that some nonprofits manipulate the allocation of expenses across different categories to reduce tax liabilities on taxable activities (Weisbrod and Cordes, 1998; Yetman, 2001) or to increase reported program-spending ratios (Trussel, 2003). Hofmann and McSwain (2013) review this literature. To our knowledge, ours is the first paper to consider the manipulation of reported investment returns at nonprofits. More broadly, our regression discontinuity approach to evaluate the potential for manipulation in investment returns fits within the sleuthing literature of "forensic economics" (Zitzewitz, 2012).
1
Data
We analyze microdata from the IRS Statistics of Income (SOI) Form 990 sample files. Most taxexempt organizations with gross receipts over $200,000 or total assets greater than $500,000 are
2
required to file Form 990. The SOI sample files contain all financial variables on Form 990 for all filing organizations with more than $10 million in assets plus a random sample of approximately 4,000 filing organizations with assets between $25,000 and $10 million.1 All major universities and hospitals are included in the SOI files because of their size. However, because religious organizations of any size are not required to file Form 990, the SOI files do not contain a representative sample of religious organizations. Although the IRS SOI data files are available from 1988 onwards, full information on investment returns was not included in the IRS SOI files until 2012. Investment returns are reported through four separate components on Form 990: interest and dividends, bond proceeds, net realized gains from sales of securities, and net unrealized gains on investments. While the first three components have always been part of Form 990, the last component – net unrealized gains on investments – was not included in Form 990 until 2012.2 We make use of the two years of investment returns data that are currently available – 2012 and 2013. For graphical clarity, we restrict our analysis sample to institution-year observations with rates of return on investments between -0.2 and 0.4. Over 95% of institution-year observations with non-missing data for all four components of investment returns have rates of return in this range. Key for us, Form 990 also includes detailed balance sheet information. In particular, investments in interest-bearing checking or savings accounts and investments on securities are separately reported, which allows us to estimate the share of investments held in risky assets. Below, we focus on organizations with at least 95% of their investments in (risky) securities. To examine a longer time horizon, we utilize information on endowment funds reported in Schedule D of Form 990. Starting from 2008, organizations that have established endowment funds are also required to report the amount and returns of their endowment funds. We use this information to estimate the rate of return on endowment funds for years 2008 to 2013. Although endowment returns information is available for a longer time horizon, we do not observe the asset allocation of endowment funds. As a result, our endowment returns analysis sample may include organizations that invest a large percent of their funds in risk-less assets. Appendix Table A.1 lists the definitions of variables used in subsequent analysis and the locations of the corresponding data items on Form 990.
2
Evidence for the Manipulation of Investment Returns
We begin with graphs showing the distribution of the rate of return on investments. We use McCrary (2008)’s density test to test for a discontinuity in the distribution of the rate of return around zero. 1
The SOI data files can be found at https://www.irs.gov/uac/soi-tax-stats-charities-and-other-tax-exemptorganizations-statistics 2 Prior to 2012, “net unrealized gains on investments” was included in Schedule D of Form 990, part XI. However, this particular line in Schedule D was only completed by those institutions that have obtained separate, independent audited financial statements and whose audited statement employed a different reporting methodology from Form 990. Some institutions also report their unrealized investment returns through a separate attachment to Form 990. These data are not included in the IRS SOI files, although it is possible to manually extract information on unrealized gains and losses from these attachments.
3
McCrary’s density test entails estimating local linear regressions of the density function of the running variable (the rate of return, in this case) on either side of the cutoff (zero, in this case). The discontinuity at the cutoff is then estimated as the log difference in the value of the density functions on the two sides of the cutoff.3 Figure 1: Histogram for rate of return on investment
Note: This figure shows a histogram of the rate of return on investment for institution-year observations during 2012 and 2013. The sample includes tax-exempt organizations in the SOI study that held at least 95% of their total investments on securities (as opposed to risk-free, interest-bearing assets). The bin size is 0.00117, chosen to match the bin size that would be used in the McCrary’s density test.
Figure 1 presents a histogram of the rate of return on investment, with a bin width of 0.00117, chosen to match the bin size that would be used in the McCrary’s density test.4 A discontinuity is clearly discernible at zero. The first six bins to the right of zero together account for 546 organizations. This means that out of the 11,395 organizations in our restricted sample, 546 reported investment returns between zero and 0.007 (excluding zero). If we assume that the rates of return follow a continuous normal distribution with the same mean and variance as the sample mean and variance, there would be 1.4% of observations between zero and 0.007. In other words, if the rates of return were normally distributed with no discontinuity at zero, there would be approximately 160 organizations reporting rates of returns between zero and 0.007. Under these assumptions, approximately 386 out of the 11,395 tax-exempt organizations in our sample (3.4%) may have misreported their returns to avoid reporting negative returns. 3 Formally, McCrary (2008) defines the discontinuity estimate to be ✓ˆ = ln limr#c fˆ(r) ln limr"c fˆ(r), where r is the running variable and c is the cutoff. 4 Formally, McCrary (2008) defines the bin size as ˆb = 2ˆ n 1/2 , where ˆ is the sample standard deviation of the running variable.
4
Figure 2 presents the McCrary’s density test of a discontinuity around zero. The graph shows estimates of local linear regressions of the density function on either side of zero. The discontinuity estimate is 1.726, with a standard error of 0.10 (the t-stat is 17.2). Figure 2: McCrary’s density test for the rate of return on investment
Note: This figure shows results from a McCrary density test for discontinuity in the distribution at 0 for the rate of return on investment. The sample includes institution-year observations during 2012 and 2013. Institutions include tax-exempt organizations in the SOI study that held at least 95% of their total investments on securities (excluding organizations that hold more than 5% of investments in risk-free assets). The discontinuity estimate, which measures the log difference in density on the two sides of the cutoff, and the t-statistic for the significance of the discontinuity estimate is listed at the upper-right corner of the graph.
An alternative explanation for the discontinuity at zero is that some organizations invested a large share of their funds in risk-less assets that happened to yield close to zero returns. So far, we have restricted our sample to institutions that held at least 95% of their investment funds in securities at the beginning of the fiscal year. This sample restriction raises confidence that the true returns are stochastic. As a robustness check, we perform the test on organizations that held at least 95% in securities both at the beginning and at the end of the fiscal year. The results that are almost identical to those presented in Figure 2. The discontinuity estimate is 1.734, with a standard error of 0.11. We also perform the McCrary’s density test for alternative samples with different restrictions on the percentage of investments held in securities. Appendix Figure A.5 presents these results. The discontinuity at zero is discernible and significant regardless of the level of restriction placed on the percentage invested in securities. Notably, the magnitude of the discontinuity at zero is substantial even when we restrict the sample to institutions with at least 99% of their investment funds in securities. To the extent that asset allocations are correctly reported on Form 990, these 5
results suggest that we can rule out the possibility that our results are driven by the small positive returns of safety assets. However, we cannot rule out the possibility that our results are driven by misclassification of securities. If a significant share of organizations misreports investment in safety assets as investment in securities, we may also observe a discontinuity at zero. To further probe the robustness of our results, we perform the McCrary’s density test for alternative samples and on alternative running variables. Appendix Figure A.6 presents the McCrary’s density test for institutions whose financial statements were audited by an independent accountant. Appendix Figure A.7 presents results for institutions whose investment funds include donorrestricted or board-designated endowment funds. Appendix Figure A.8 presents the McCrary’s density test for an alternative running variable: the dollar amount of investment returns. In all cases, the discontinuity at zero is readily discerned and statistically significant. Additionally, excluding bond proceeds from investment returns yields results that are very similar to those presented in Appendix Figure A.8. Figure 3: McCrary’s density test for the rate of return on endowment funds
Note: These figures show results from McCrary density tests for discontinuity in the distribution at 0 for the rate of return on endowment funds. Sample includes tax-exempt organizations in the SOI study that reported endowment information on Schedule D of Form 990 and maintained a positive balance on endowment funds for all years 2008 to 2013. Left panel pools all institution-year observations from 2008 to 2013. Right panel includes the six-year average rate of return on endowment funds for each organization. The discontinuity estimate and the t-statistic for the significance of the discontinuity estimate is listed at the upper-right corner of the graph.
If organizations that manipulate reported investment returns reconcile reported returns with actual returns over time, we might expect little or no discontinuity in the pdf of the long-term average rate of return per organization. We extend our analysis to the rate of return on endowment funds to examine a longer time horizon - 2008 to 2013. Figure 3 presents the McCrary’s density test for pooled rates of return (left panel) and the six-year average rate of return per organization (right panel). Appendix Figure A.9 presents results separately for each fiscal year from 2008 to 2013. The discontinuity at zero is statistically significant for all six years. Interestingly, there is a substantial and statistically significant discontinuity at zero for the six-year average rate of return. 6
This suggests that some nonprofits repeatedly manipulate their investment returns. Next, we examine whether the tendency to manipulate investment returns differs for different types of tax-exempt organizations. We use the National Taxonomy of Exempt Entities (NTEE) classification system to categorize nonprofits organizations. Then we perform McCrary’s density test separately for each group to examine whether there is a discontinuity at zero within each group. Figure 4: McCrary’s density test for the rate of return, by type of tax-exempt organization
Note: These figures show results from McCrary density tests for discontinuity in the distribution at 0 for the rate of return on investments. Institutions are grouped according to the National Taxonomy of Exempt Entities (NTEE) classification system. The sample includes institution-year observations during 2012 and 2013. Institutions include tax-exempt organizations in the SOI study that held at least 95% of their total investments on securities. Only those industry groups that have more than 150 observations with a rate of return between -0.2 and 0.4 are shown here. The discontinuity estimate and the t-statistic for the significance of the discontinuity estimate is listed at the upper-right corner of each graph.
Figure 4 presents results for the nine major groups that have at least 150 observations within our sample restriction criterion: investing at least 95% in securities and having rates of return between -0.2 and 0.4. Among the nine groups of organizations, recreation and sports organizations have the highest discontinuity estimate (3.24), followed by religious organizations (2.40), community 7
improvement and capacity building organizations (2.10), education organizations (1.95), healthcare organizations (1.81), employment organizations (1.79), human services organizations (1.50), and mutual and membership benefit organizations (1.38). Philanthropy, voluntarism and grant making foundations (NTEE group T) have the lowest discontinuity estimate (1.08) and is the only group whose discontinuity estimate is not statistically significant at the 5% level.
3
Conclusion
Prima facie, nonprofits would appear to have little incentive to over-represent investment performance. “Under penalties of perjury,” the nonprofit’s responsible officer also avows to the IRS that “the [tax] return, including accompanying schedules and statements, and to the best of my knowledge and belief, it is true, correct, and complete.” Nevertheless, we observe nonprofits appear to behave, so that small negative reported returns are avoided and instead appear as gains. This pattern is found for a wide range of nonprofit missions, including religious related charities and community improvement organizations. In this respect, nonprofits behave much like for-profit firms, which may have stronger incentives to manipulate returns and, moreover, have previously been found to do so. However, that some nonprofits report very small losses to the IRS implies that not all nonprofits engage in “window-dressing” behavior.
Acknowledgments This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The authors thank an anonymous referee for helpful comments.
References Bergstresser, Daniel, Mihir Desai, and Joshua Rauh, “Earnings Manipulation, Pension Assumptions, and Managerial Investment Decisions,” The Quarterly Journal of Economics, 2006, pp. 157–195. Bollen, Nicolas PB and Veronika K Pool, “Do Hedge Fund Managers Misreport Returns? Evidence from the Pooled Distribution,” The Journal of Finance, 2009, 64 (5), 2257–2288. Burgstahler, David and Ilia Dichev, “Earnings Management to Avoid Earnings Decreases and Losses,” Journal of Accounting and Economics, 1997, 24 (1), 99–126. Hofmann, Mary Ann and Dwayne McSwain, “Financial Disclosure Management in the Nonprofit Sector: A Framework for Past and Future Research,” Journal of Accounting Literature, 2013, 32 (1), 61–87. Lee, David S. and Thomas Lemieux, “Regression Discontinuity Designs in Economics,” Journal of Economic Literature, June 2010, 48 (2), 281–355.
8
McCrary, Justin, “Manipulation of the Running Variable in the Regression Discontinuity Design: A Density Test,” Journal of Econometrics, 2008, 142 (2), 698–714. Trussel, John, “Assessing potential accounting manipulation: The financial characteristics of charitable organizations with higher than expected program-spending ratios,” Nonprofit and Voluntary Sector Quarterly, 2003, 32 (4), 616–634. Weisbrod, Burton A and Joseph J Cordes, “Differential Taxation of Nonprofits and the Commercialization of Nonprofit Revenues,” Journal of Policy Analysis and Management, 1998, 17 (2), 195–214. Yetman, Robert J, “Tax-Motivated Expense Allocations by Nonprofit Organizations,” The Accounting Review, 2001, 76 (3), 297–311. Zitzewitz, Eric, “Forensic Economics,” Journal of Economic Literature, July 2012, 50 (3), 731–69.
9