Near zero taxable income reporting by nonprofit organizations.
This paper investigates the propensity of nonprofit organizations to report near zero taxable income profitability to avoid or mitigate the impact of unrelated business income taxes (UBIT) and then examines the organization-specific characteristics associated with this propensity. Although most profits earned by nonprofit organizations are free of income tax, profits from activities that are unrelated to the primary exempt purpose are subject to federal and state UBIT. The UBIT is functionally similar to the corporate income tax in terms of tax rate schedules, filing requirements, and applicable penalties. The UBIT, one of four income tax systems in the U.S., was enacted in 1950 after Congress determined that nonprofits' tax-exemption should not apply to activities that would otherwise unfairly compete with fully taxable for-profit entities (U.S. House 1950).(1)
Using a sample of more than 1,300 hand-collected confidential nonprofit tax returns we begin our analysis by examining the distribution of reported taxable profitability (i.e., taxable profits scaled by taxable revenues) relying on an approach similar to that used in prior research to identify deviations from smoothness in the distribution (Mills and Newberry 2001; Degeorge et al. 1999; Burgstahler and Dichev 1997). If nonprofits are managing their taxable income to near zero to avoid the UBIT, then there would be an abnormally large number of nonprofits that report very close to zero profitability. The setting is somewhat similar to that in the multinational tax literature where foreign-controlled U.S. corporations (FCDCs) are thought to manage taxable income reported to U.S. taxing authorities toward a target of zero by transferring profits from the U.S. corporations to the foreign parents (Grubert et al. 1993; Collins et al. 1997). The nonprofit setting creates similar incentives for organizations to shift profits from their taxable activities within a single entity. Although taxable entities also have differentially taxed assets, such as municipal vs. corporate bonds, the nonprofit setting is unique in that the identical asset may simultaneously produce both taxable and tax-exempt income. A finding that an abnormally large number of nonprofits report near zero taxable income is consistent with nonprofits shifting profits from their taxable to their tax-exempt activities to reduce their tax liabilities. In the second part of our analysis, we examine the various tax-planning frictions and restrictions that introduce cross-sectional variation in near zero taxable income reporting behavior by these organizations.
By examining nonprofit taxable income management, we hope to provide information of interest to policymakers as well as academics interested in earnings management and the effects of taxes on business decisions. To the extent that nonprofits are effectively able to manage their taxable income to near zero, the UBIT fails to fully accomplish its joint purposes of preventing unfair competition and raising revenues. From a policy standpoint, documenting variations in tax reporting can shed light on the efficiency of a tax. To the extent that there is cross-sectional variation in tax burdens borne by similar taxpayers, due to either aggressive positions or investments in tax-planning technologies, the tax is inefficient and potentially inequitable. Our study also contributes to the accounting tax literature by extending the Scholes-Wolfson income-shifting paradigm to the nonprofit setting.
Although the taxable revenues earned by nonprofits are relatively small compared to individual or corporate income taxes, they are growing at almost 30 percent per year. The increasing commercialism of nonprofit organizations has caused concern among firms that compete with nonprofits (Brady 2000; Berss 1994; Gomes and Owens 1988). Congress continues to be concerned over the rapid expansion of nonprofits' commercial activities and the potential for "unfair competition" arising from the differential taxation of nonprofits and for-profit firms (Manzullo 2001; Orban 1999; Ramsey 1986).
We begin our analysis by grouping observations into taxable income profitability intervals and testing for abnormal deviations from a smoothly changing distribution. Notwithstanding the general tendency of a profitability distribution to be somewhat "bell" shaped, we test whether an abnormal number of nonprofits report taxable profitability's in the near zero range. Our statistical test assumes a smoothly changing distribution (rather than an absolute smooth distribution) because we anticipate an abnormality in the distribution around the expected peak, rather than off-peak. We find that a statistically abnormal number of nonprofits report taxable income in an interval around zero. No other interval on the distribution contains an abnormal number of observations. We then conduct a similar analysis on nonprofits' tax-exempt net income and find no abnormal clumping of near zero tax-exempt net income. We interpret these results as consistent with the hypothesis that nonprofit organizations manage their taxable income to near zero by shifting income out of their taxable activities.
In the second part of our analysis, we classify observations contained in the abnormal interval as potential taxable income managers and conduct an analysis of the various frictions and restrictions that can impede a nonprofit's ability to manage its taxable income to an interval around zero. The purpose of this analysis is to examine why many nonprofits do not report taxable income in the interval around zero. We consider this portion of our analysis to be a significant contribution because, as noted by Maydew (2001), it is not surprising to find that organizations wish to pay fewer taxes. It is more interesting to examine why there is cross-sectional variation in the behavior observed.
We examine whether the probability that a nonprofit reports near zero taxable income is related to donations, net operating loss carryforwards (NOLs), the use of a paid CPA preparer, nonprofit type, and whether the taxable activity has a tax-exempt counterpart. By tax-exempt counterpart, we mean an activity that produces products that can be both taxable and non-taxable, depending on to whom the product is sold. (2) We find that the likelihood that a nonprofit reports near zero taxable income is decreasing in size and when the taxable activity has a tax-exempt counterpart. We also find that charitable nonprofits are less likely to report near zero taxable income than are hospitals. We do not find any relationship between donations or NOLs and the probability of reporting near zero taxable income. Finally, we find that the use of a paid CPA preparer is associated with a higher probability of reporting near zero taxable income, regardless of the size of the CPA firm used (i.e., Big 5 or non-Big 5 CPA firm). This result is consistent with paid CPA preparers assisting their nonprofit clients in managing their taxable income to near zero.
We make two distinct contributions to the tax accounting literature. First, we document that many nonprofits are adept at managing their taxable income to near zero, providing additional evidence that the UBIT as currently structured at least partially fails to achieve its goals of preventing unfair competition and raising revenue. Second, we provide evidence on the frictions and restrictions that affect the cross-sectional incidence of near zero taxable income reporting by nonprofits. This evidence is important because it helps to explain why all nonprofits do not report near zero taxable profits. In addition to contributing to the general academic interest in income tax systems, our results provide information of potential interest to policymakers concerned with the efficiency and equity of the UBIT.
In what follows, the second section provides a background on the UBIT and prior research, while the third section describes our methodology. In the fourth section we discuss our data and then our empirical analysis and results in the fifth section. The final section concludes.
BACKGROUND AND PRIOR LITERATURE
Background on Nonprofit Taxation and Unfair Competition
Due to a zero marginal tax rate on all "related" income (which includes all forms of passive income such as interest, dividends, rents, and royalties), nonprofits provide a unique opportunity for tax planning. By shifting profits from taxable to tax-exempt activities (via revenue or expense shifting), nonprofits can reduce their tax burdens from a maximum combined federal and state marginal rate of over 40 percent to zero percent. This compares to the multinational setting where corporations are hypothesized to shift income between countries whose marginal tax rates vary by only a few percentage points. Anecdotally, nonprofit tax planning appears to be effective. In the aggregate, nonprofits report losses in excess of 30 percent of revenues (Meckstroth and Amsberger 1998). The rapid growth of nonprofits' taxable activities in conjunction with the low rate of tax paid has increased concerns over unfair competition, which in turn has led to continued interest by Congress (Brady 2000; Manzullo 2001).
Financial and Tax Earnings Management Literatures
Our analysis builds on the body of literature that examines similar issues such as transfer pricing, or the shifting of profits between related multinational corporations to avoid taxes, and earnings management by shifting profits over time to avoid losses. In a seminal transfer-pricing study, Grubert et al. (1993) find a persistent concentration of foreign-controlled U.S. corporations with taxable income around zero. After accounting for various potential causes of near zero taxable income (i.e., start-up effects, debt costs, exchange rate fluctuations, and capital costs), they find that a substantial portion (i.e., 50 percent) of near zero taxable income persists, which they attribute to transfer pricing.
Building on the Grubert et al. (1993) study by using confidential tax return data, Collins et al. (1997) further examine the concentration of near zero reported taxable income by FCDCs and report results contrary to those of prior studies. Collins et al. (1997) document evidence of concentrations of taxable income around zero at the firm-specific level for a sample of FCDCs as well as a control sample (although the distribution was shifted to the right for the control sample). However, Collins et al. (1997) go on to show that the ratio of gross profit changes to sales changes is not different for FCDCs (which empirically report near zero taxable income) and a control group of U.S. only firms (which ostensibly do not report near zero taxable income) as the firms' net incomes approach zero. They interpret their results as evidence that FCDCs do not shift income out of the United States to their foreign parents.
In the earnings management literature, Hayn (1995) finds suggestive evidence that managers shift only enough income or expense to move pre-managed earnings to the targeted earnings level. Burgstahler and Dichev (1997) were the first to show that firms tend to avoid reporting small losses and negative earnings changes on their publicly available financial statements. The primary analysis tool used by Burgstahler and Dichev (1997) is examination of deviations from a smooth distribution in reported income levels. Mills and Newberry (2001) use a similar technique to examine the effect of income management on book-tax reporting differences between public and private firms. Degeorge et al. (1999) consider three earnings management thresholds by looking at discontinuities off-peak, at the peak, and at an interval that includes the peak of the distribution.
Prior studies have found evidence that nonprofits reallocate expenses between their taxable and tax-exempt activities to reduce their tax liabilities (Cordes and Weisbrod 1998; Sansing 1998; Yetman 2001; Omer and Yetman 2002). The results in this paper differ from and expand upon those prior studies in four important ways. First, by examining the distribution of taxable income we consider overall income management, which implicitly includes both revenue and expense shifting, whereas prior literature considered only expense shifting. Second, we identify taxable income management based on numbers as reported by the nonprofits, rather than relying on expense reallocation models. Third, although prior nonprofit tax research suggests that nonprofits shift expenses to their taxable activities, there is no evidence that they target any particular goal amount, such as near zero. In the limit, nonprofits could use these income-shifting techniques to systematically reduce their taxable income to zero, above zero, or below zero, although there is no evidence of where the shifting places them on the distribution. Fourth and perhaps most importantly, we extend prior literature by examining what organization-specific characteristics are associated with potential taxable income management, while prior studies have not.
Abnormal Frequency of Near Zero Taxable Income Reporting
A common assumption made about any earnings distribution is that it is continuous (i.e., the first and second derivative exists). Burgstahler and Dichev (1997) suggest that researchers can detect earnings management using a pooled cross-sectional distribution approach, although they note that the method is most effective when the earnings goal can be precisely defined. We rely on prior research in multinational taxation to assist us in precisely identifying nonprofits' taxable income management goal. Grubert et al. (1993) find that foreign-controlled U.S. corporations report persistent near zero taxable income. "If shifting were costless, all companies would be at exactly zero taxable income at all times. In the general case, bookkeeping costs, potential penalties, and legal scruples prevent perfect shifting, but a concentration near zero would still be-the expected pattern." (See Grubert et al. 1993, 259.)
The operation of differentially taxed activities within a single organization provides nonprofits with a similar incentive to shift profits from their taxable activities to their tax-exempt activities to reduce their tax liabilities. We therefore expect that nonprofits are more likely to manage their taxable profits to near zero when possible and/or feasible. A limitation of our study is that we are not able to determine what the distributional location of a nonprofit was before it shifted its taxable income to near zero. Although it is possible that nonprofits with the highest marginal tax rates had more incentive to shift to near zero, any nonprofit with a positive tax rate has an incentive to shift, and we observe only reported, or ex post taxable income figures. Because of this, we are unable to examine for corresponding decreases in any particular interval on the distribution.
To determine if an abnormally large number of nonprofits are reporting near zero taxable profits, we group all sample firm-years into 2 percent-wide profitability intervals and plot the frequency that a nonprofit reports taxable income in these intervals. We assume a smoothly changing distribution, and thus expect that the number of observations in any interval of the taxable income distribution will equal the number of observations in the immediately preceding interval.(3) The test statistic is the difference between the actual and expected number of observations in a given interval, divided by the estimated standard deviation of the difference. (4) It is important to note that this methodology embodies the notion that the distribution is likely to grow as it approaches some mean value, but that the growth should be smoothly changing. The assumption of a smooth (or continuously differentiable) density function is common and is substantially weaker than assuming any particular form of the function (Mills and Newberry 2001; Degeorge et al. 1999; Burgstahler and Dichev 1997). We estimate the test statistic by starting with the left-most interval (i.e., most negative taxable profit interval) and move to the right. (5) As an additional distributional test, we conduct a similar analysis (assuming a smooth distribution) on nonprofits' tax-exempt income. Given that nonprofits are not operated primarily to generate and accumulate profits, but rather to expend their resources on charitable purposes, it is possible that any significant clumping of near zero taxable income is an artifact of the nonprofits' overall financial results, which may also clump around zero if they attempt to maximize expenditures on their charitable outputs. (6)
Factors Associated with Near Zero Taxable Income Reporting
In this analysis we seek to examine why some nonprofits report near zero taxable income while others do not. Our examination is in the spirit of the Scholes-Wolfson paradigm in that we test the effects of potential frictions and restrictions on nonprofit tax planning via income shifting (Scholes et al. 2002). To determine what factors are associated with the probability of reporting near zero taxable income, we estimate the following logistic model:
(1) Near [Zero.sub.it] = [alpha] + [[beta].sub.1] [Size.sub.it] + [[beta].sub. 2] [Donations.sub.it] + [[beta].sub.3][NOL.sub.it] + [[beta].sub.4] [CPA.sub.it] (7) + [[beta].sub.5] [Similar Activities.sub.it] + [[beta].sub.6] [Type.sup.it] + [[epsilon].sub.it].
The following sections discuss our dependent and independent variables.
Our dependent variable is a firm-year specific indicator variable equal to 1 if the nonprofit observation is in the interval [-0.01, 0.01). This variable undoubtedly contains measurement error in that it likely misclassifies both those nonprofits that were attempting to manage their taxable profits to near zero, but missed, as well as those nonprofits that were not attempting to manage their taxable income to near zero, but wound up there by chance. In either of these cases, the measurement error in the dependent variable will not bias our coefficient estimates to the extent that it is not correlated with any of the included independent variables and will be included in the model's error term (Greene 2000). It is important to note that if the measurement error were correlated with our covariates, it would introduce bias into the coefficient estimates. We attempted to mitigate these misclassification effects by widening our analysis interval by 0.01 profitability increments. Results were qualitatively unchanged by increasing the "near zero" interval to include those nonprofits that reported taxable income from -0.03 to 0.03. (8)
Our independent variables are intended to capture potential tensions (i.e., frictions and restrictions) that could introduce cross-sectional variation in a nonprofits' propensity to report near zero taxable income. We include the amount of total public donations divided by total donations received by the nonprofit as our proxy for reputation as prior research finds that the proportion of public donations received by a nonprofit is increasing in reputation (Okten and Weisbrod 2000). (9) It is possible that our donations variable captures additional constructs beyond reputation, such as organizational mission or funding sources, which could introduce other tax-planning frictions and make it difficult to ex ante predict a relationship between donations and the probability of reporting near zero profitability.
Our next independent variable, NOL, is the accumulated net operating loss carryforwards. The presence of a net operating loss carryforward reduces the need for a nonprofit to manage its taxable profits toward zero. We estimate NOL for each nonprofit using the amount of net operating loss carryforward reported on 990T not including current year's results. (10)
We include the variables Big-Five CPAs and Other CPAs to measure the effects of using paid preparers on the probability of reporting near zero profitability. Paid tax return preparers could provide a friction to near zero reporting (by encouraging their clients to report true amounts rather than manipulated amounts), or they could increase the propensity to report near zero by assisting their clients in aggressive tax reporting behavior. To create our tax return preparer variable we divide all tax returns into those prepared by a CPA firm and those that are self-prepared by the nonprofit. We then subdivide CPA prepared returns into those prepared by Big-Five CPAs and Other CPAs. Prior literature suggests that Big 5 CPAs have more resources and are more likely to have in-house expertise, and this expertise could affect the ability or propensity to report near zero profitability (Becker et al. 1998; Francis et al. 1999). Prior analytical research in the individual setting generally suggests that taxpayers seek out tax preparers to resolve complex tax issues and that using paid tax preparers increases compliance (Beck and Jung 1989; Klepper et al. 1991; Phillips and Sansing 1998). Experimental evidence is generally consistent with the analytical predictions and suggests that taxpayers do not demand risky or aggressive tax advice (Hite and McGill 1992). In contrast to the analytical and experimental evidence, archival empirical evidence finds that tax preparers are associated with less individual income tax compliance (Erard 1993). It is not ex ante clear how these findings in the individual income tax setting will extrapolate to the nonprofit setting.
We include the variable Similar Activities because it could introduce a tax-planning friction. Nonprofits often jointly produce taxable and tax-exempt products using the same facilities. For example, a nonprofit hospital can use its pharmacy to jointly produce both taxable sales (to non-admission patients) and tax-exempt sales (to currently admitted patients). If the hospital reported higher profit margins on its tax-exempt pharmacy sales relative to its taxable pharmacy sales, it is more likely that the IRS would discover and likely prove intentional profit understatement. However, if the same hospital engaged in a taxable activity for which there were no tax-exempt counterpart, this tension is less likely to impede near zero taxable income reporting because there is no similar tax-exempt activity against which to "benchmark." The tension is similar to that imposed by book-tax conformity for fully taxable corporations, although in the nonprofit setting the tension is across the book and tax profitability differences of identical production activities.
To measure the effects of Similar Activities, we include an indicator variable equal to 1 if the nonprofit carries out an activity that has both a taxable and tax-exempt counterpart, and 0 otherwise. We constructed this variable by examining the nonprofits' description of their taxable activity (i.e., line H of the IRS 990-T). For each nonprofit, we attempted to determine if the taxable activity had a tax-exempt activity counterpart in several ways. First, we examined Part VII of the IRS 990 where a nonprofit reports its revenues broken out into taxable and tax-exempt portions by revenue type. For example, if a nonprofit reported both taxable revenues (column A) and tax-exempt revenues (columns D or E) of a particular revenue category, then we coded Similar Activities as 1. Second, we read through the description of the nonprofits operations as described in Parts III and VIII of the IRS 990 and compared these descriptions to those of the taxable activity from line H on the IRS 990-T. If the descriptions contained similar activities, then we coded Similar Activities for that nonprofit as 1. Given the method used to construct this variable it likely contains measurement error, which will bias the coefficient estimate toward zero. We hypothesize that the probability a nonprofit will report near zero taxable income is decreasing when the taxable activity has a tax-exempt counterpart.
We include an indicator variable for nonprofit Type (medical versus other charitable organizations) and hypothesize that other charitable nonprofits will report near zero taxable income less often than will hospitals because prior research finds that hospitals behave more opportunistically with respect to tax laws and are, in general, more "profit-oriented" than are charitable nonprofits (Steinberg 1986). We include Size, which is equal to total assets, primarily as a control variable.
Nonprofit organizations must annually file a form 990 with the IRS. The IRS 990 is the nonprofit's publicly available information return and includes an income statement, balance sheet, and other information related to nonprofits' overall activities. It is not possible to examine nonprofits' taxable activities using the IRS 990 because it aggregates both taxable and non-taxable activities. Nonprofits with taxable activities must also file a form 990-T. The IRS 990-T, which is not publicly available, is the nonprofit's income tax return and includes only taxable information. The IRS 990-T is similar in appearance to the IRS 1120 for corporations. The database of confidential IRS form 990-Ts used for the analysis is a subsample of the National Center for Charitable Statistics database (NCCS 1999), which itself is a subsample of all nonprofits. The NCCS database includes all nonprofits with total assets of $10 million or more, plus a stratified random sample of smaller organizations, for a total annual sample of approximately 12,000 nonprofits.
In 1995, there were 2,316 nonprofits in the NCCS database that reported earning taxable revenues. We sent a written request to all 2,316 nonprofits that reported earning taxable revenues in 1995. In response to our request, 703 nonprofits voluntarily supplied matching sets of their forms 990 and 990-T. (11) Although we requested three consecutive years of data, nonprofits supplied us with an average of 2.6 returns leaving us with a pooled sample of 1,824 observations. The sample contains a relatively small 8 percent of all nonprofits that earned taxable revenues by number, but captures an average of 33 percent of the total taxable revenues earned for each of 1995, 1996, and 1997. Untabulated sample representativeness tests suggest that the sample is not jointly different from the population across total assets, total revenues, taxable revenues, total expenses, and total donations.
We examine the details of nonprofit tax law for provisions that may mechanically limit taxable income to zero and find special provisions for periodical advertising, and "exploited" activities, which restrict nonprofits from reporting losses). (12) Mechanically, tax law for advertising and exploited activities limits the expense deduction to the amount of revenues generated by the activities, which would cause nonprofits with excess expenses to report exactly zero taxable income. (13) After examining our data, we find 70 observations where nonprofits report exactly zero taxable income due to the loss limitation provisions. We exclude these 70 observations from the distribution analysis because, given the nature of the statistical tests, their inclusion would possibly lead to the spurious conclusion that an abnormal number of nonprofits report near zero taxable profits. Removing these observations reduces our sample size from 1,824 to 1,754 observations, although including these 70 observations does not qualitatively alter results. We further remove 387 observations where the type of tax return preparer could not be observed due to redaction or being left blank, leaving us with a sample of 1,367 observations. (14) Results (exclusive of the CPA variable) are not sensitive to including these additional 387 observations.
Abnormal Frequency of Near Zero Taxable Income Reporting
Figure 1 presents the distribution of the pooled cross-section of nonprofit taxable income as reported on the IRS form 990-T scaled by taxable revenues. The distribution is somewhat "bell-shaped" and, as expected, grows around some mean value. There is a clear and notable discontinuity at the near zero interval. The apparently irregular interval contains 138 observations that report taxable profitability in the interval [-0.01, 0.01). Applying our statistical test, we find that the standardized difference for the number of nonprofits reporting in the range [-0.01, 0.01) is 7.68, which represents a p-value of < 0.00001, suggesting that nonprofits report taxable income in that range more frequently than would be expected under the assumption of a smoothly changing distribution. Of the 138 observations in this interval, 19 reported exactly zero taxable income.
[FIGURE 1 OMITTED]
The distributional results indicate that a statistically abnormal number of nonprofits report taxable income near zero, suggesting that those nonprofits intentionally manage the profitability of their taxable activities. For our logit analysis of the factors associated with this reporting behavior, we code these 138 observations as 1 and all other observations as 0.
We also conduct a similar distributional analysis for nonprofits' tax-exempt net income scaled by total exempt revenues. We define tax-exempt net income as total net income of the nonprofit, minus taxable income. Because nonprofits' primary purpose is to make expenditures on their exempt purpose and not to earn and accumulate profits, it is possible that there would be a similarly significant amount of nonprofits that report near zero tax-exempt net income (i.e., on average they spend what they earn). If true, then our results for taxable income could simply be an artifact of nonprofits' overall objective functions. Figure 2 presents the distribution of nonprofits' tax-exempt net income. We find no statistically significant deviations from smoothness for the distribution of nonprofits' tax-exempt net income. This result mitigates concerns that the nonprofits' taxable income distribution is simply an artifact of their tax-exempt income distribution.
[FIGURE 2 OMITTED]
As an untabulated robustness test, we re-constructed Figure 1, partitioning our sample on the median amount of taxable revenues. We conduct this test because it is possible that the near zero profitability group is driven by the set of nonprofits that report near zero taxable revenues (i.e., it is possible that small revenues translate into small profits). Results of this robustness test show that both distributions of taxable income (i.e., one using hightaxable revenue earners and the other using low-taxable revenue earners) have statistically abnormal spikes in the distributions at the zero point. Interestingly, more nonprofits in the larger revenue sample report near zero taxable income (i.e., 86 observations) than do nonprofits in the smaller revenue sample (i.e., 52 observations). This result suggests that Figure 1 is not driven by nonprofits that earn relatively small amounts of taxable revenues. As an additional robustness test, we used various alternative scalars (as well as no scalar) with no change in inferences.
Factors Associated with Near Zero Taxable Income Reporting
We report summary statistics for the analysis variables in Table 1. (15) The descriptive statistics suggest that, on average, nonprofits in our sample receive a little over 65 percent of their donations from public sources and that about two-thirds of the sample hire CPAs to complete their tax return. Of those hiring CPAs, the distribution is nearly split between Big 5 CPAs and Other CPAs. The average nonprofit has $167 million in total assets and accumulated NOLs of slightly over $300 thousand on its taxable activities. Note that, on average, the net loss for nonprofits in our sample on taxable activities is just over $9,000 and net profit overall is slightly over $12 million.
A Pearson correlation analysis (untabulated) of all primary test variables indicates that many of the variables are statistically correlated, but only four correlations exceed 15 percent. Not surprisingly, Big-Five CPA preparer and Other CPA preparer are highly negatively correlated (49 percent). Type and Donations are also negatively correlated (21 percent) suggesting that hospitals receive substantially fewer public donations. The remaining correlations are positive between Big-Five CPA and Type, (16 percent) and negative between Other CPA and Type (17 percent).
Table 2 presents the results of estimating Equation (1) using a standard logit technique. The model is signifcant with a pseudo [R.sup.2] of 0.0674. The model prediction success is slightly better for nonprofit organizations reporting near zero taxable income (97 out of 138 observations) than all other nonprofits (751 out of 1,229 observations). Our reported prediction rate uses the observed sample rate of near zero reporting nonprofits. To determine whether the model is reasonably able to distinguish between near zero and non-near zero nonprofits, we estimate a Receiver Operating Characteristic (ROC) curve. ROC curves are commonly used to determine how well a model can predict outcomes when a naive model predicts a disproportionate number of zero occurrences. The area under the ROC curve for our model (untabulated) is 69.14 percent, which places the model in the "adequate" category (Hosmer and Lemeshow 2000) for detecting near zero versus non-near zero nonprofits. Below we discuss the coefficients for the variables in the model. We interpret these results cautiously because we cannot draw a direct causal link between the variables we have selected and reporting in the near zero group. Thus, we suggest that a significant coefficient is representative of an association between the variable in question and being in the near zero taxable income group.
The coefficient for NOL is significantly negative, suggesting that nonprofits with large prior accumulated net operating losses are less likely to report near zero taxable income in the current year. Because it is possible that nonprofits that report large losses in the current year (and thus are far away from zero taxable income) show similar large losses year after year (thus producing a large NOL carryforward), any results with respect to the NOL variable could be partially driven by these organizations (i.e., big NOL carryforwards are negatively associated with being at the near zero point). To investigate this concern, we reestimated our models using a multinomial logit method in which we partition the sample across observations with greater than zero and less than zero taxable income. Untabulated results find that the NOL result is driven entirely by the negative taxable income nonprofits. Based on this result, we cannot conclude that NOLs are associated with the propensity to report near zero taxable income. Results of all other independent variables are robust to the multinomial logit analysis.
The coefficient for Donations is insignificant. Because Donations and Size may have an interaction effect, we conduct additional (untabulated) analysis that shows that, when we interact the two variables, the main effect of Donations, as well as the interaction effect, is not statistically different from zero. Results with respect to Big-Five CPA and Other CPA paid preparers show that the odds of reporting taxable income in the near zero group go up by 122 and 175 percent if the nonprofit hires a Big 5 CPA or other CPA, respectively. This result is consistent with the hypothesis that CPAs assist their nonprofit clients in managing their taxable income toward zero. In light of the discussion, regarding the difference in expertise between Big 5 CPAs and other CPAs, tests of the difference in coefficients suggest that both CPA types are equally effective for their clients. (16)
We find that the probability that a nonprofit reports near zero taxable income is increasing in Type, supporting the conjecture of prior research that hospitals tend to be more aggressive tax reporters. Finally, we find that the probability that a nonprofit reports near zero taxable income is decreasing in Similar Activities, suggesting that nonprofits are concerned about potential "benchmarking" of their taxable and tax-exempt activities when conducted in the same environment. The odds of being in the near zero taxable income group go down by 47 percent if the nonprofit engages in activities that have both taxable and tax-exempt counterparts, and the odds of being the near zero group go up by over 50 percent if the nonprofit is a medical facility. A Pearson Chi-square test suggests that hospitals end up in the near zero taxable income group significantly more often than other charities based on our sample.
Nonprofit organizations are subject to an unrelated business income tax on net profits from activities not closely related to their tax-exempt purpose. The simultaneous operation of differentially taxed activities provides nonprofits with an incentive to shift profits from taxable to tax-exempt activities to minimize overall tax liabilities. In the limit, nonprofits could use these income-shifting techniques to reduce their taxable income to zero. We use a distributional analysis based on prior literature, which assumes that the underlying distribution is smoothly changing, to determine if an abnormally large number of nonprofits report near zero taxable income. Based on our distributional test we find evidence that an abnormal number of nonprofits report near zero taxable income, which we interpret as consistent with intentional managerial manipulation.
Using our findings from the distributional analysis, we create an indicator variable equal to 1 if the nonprofit reported near zero taxable income, and 0 otherwise. We then examine the factors that are associated with a nonprofit's propensity to report near zero taxable income. We find that the likelihood that a nonprofit reports near zero taxable income is decreasing in size and when the taxable activity has a tax-exempt counterpart. We also find that charitable nonprofits are less likely to report near zero taxable income than are hospitals. Finally, we find that the use of a paid CPA preparer is associated with a higher probability of reporting near zero taxable income.
Our results add to the growing body of research that attempts to understand the effect of the unrelated business income tax on nonprofits' behavior. We document for the first time the empirical regularity that nonprofits report near zero taxable income. Our results suggest that some nonprofits are particularly adept at managing taxable income toward zero and trade off the benefits with various tax-planning frictions and restrictions. Of particular interest is the finding that CPA-prepared tax returns are more likely to report near zero taxable income, suggesting that CPA firms are either more aggressive or more adept at managing their clients' taxable income to near zero.
Summaries of Papers in this Issue
Near Zero Taxable Income Reporting by Nonprofit Organizations
Thomas C. Omer and Robert J. Yetman
In this paper, we investigate the propensity of nonprofit organizations to report near zero taxable income profitability to avoid or mitigate the impact of unrelated business income taxes (UBIT) and then we examine which organization-specific characteristics are associated with this propensity. Although most profits earned by nonprofit organizations are free of income tax, profits from activities that are unrelated to the primary exempt purpose are subject to federal and state UBIT.
The approach we use is to presume that, if nonprofits are managing their taxable income to near zero to avoid the UBIT, there will be an abnormally large number of nonprofits that report very close to zero profitability. The setting is somewhat similar to that in the multinational tax literature where foreign-controlled U.S. corporations are thought to manage taxable income reported to U.S. taxing authorities toward a target of zero by transferring profits from the U.S. corporations to the foreign parents. The nonprofit setting creates similar incentives for organizations to shift profits from their taxable activities within a single entity. In the second part of our analysis, we examine the various tax-planning frictions and restrictions that introduce cross-sectional variation in near zero taxable income reporting behavior by these organizations.
To the extent that nonprofits are effectively able to manage their taxable income to near zero, the UBIT fails to fully accomplish its joint purposes of preventing unfair competition and raising revenues. From a policy standpoint, documenting variations in tax reporting can shed light on the efficiency of a tax. To the extent that there is cross-sectional variation in tax burdens borne by similar taxpayers, due to either aggressive positions or investments in tax-planning technologies, the tax is inefficient and potentially inequitable. Our study also contributes to the accounting tax literature by extending the Scholes-Wolfson income-shifting paradigm to the nonprofit setting.
We find that a statistically abnormal number of nonprofits report taxable income in an interval around zero. No other interval on the distribution contains an abnormal number of observations. We then conduct a similar analysis on nonprofits' tax-exempt net income and find no abnormal clumping of near zero tax-exempt net income. We interpret these results as consistent with the hypothesis that nonprofit organizations manage their taxable income to near zero by shifting net profits out of their taxable activities.
In the second part of our analysis, we find that the likelihood that a nonprofit reports near zero taxable income is decreasing in size and when the taxable activity has a tax-exempt counterpart. We also find that charitable nonprofits are less likely to report near zero taxable income than are hospitals. We do not find any relationship between donations or NOLs and the probability of reporting near zero taxable income. Finally, we find that the use of a paid CPA preparer is associated with a higher probability of reporting near zero taxable income regardless of the size of the CPA firm used (i.e., Big 5 or non-Big 5 CPA firm). This result is consistent with paid CPA preparers assisting their nonprofit clients in managing their taxable income to near zero.
TABLE 1 Summary Statistics for a Sample of 1,367 Matched Sets of IRS Forms 990 and 990-T Description Mean Std. Dev. Continuous variables Donations 0.66 0.40 NOL 304,862 955,293 Size 167,017,643 1,117,261,369 n = Categorical Variables Similar Activities 587 Big-Five CPA 463 Other CPA 438 Type 600 Other variables Taxable Revenues 527,575 1,263,927 Taxable Expenses 538,740 1,211,973 Total Revenues 76,522,482 185,752,694 Total Expenses 64,185,381 134,822,042 Variable definitions: Donations = total public donations/total donations (Form 990); NOL = total accumulated net operating losses; Size = total assets; Similar Activities = 1 if taxable activity has a tax-exempt counterpart within the same nonprofit, 0 otherwise; Big-Five CPA = 1 if tax return prepared by Big 5 CPA, 0 otherwise; Other CPA = 1 if tax return prepared by non-Big 5 CPA, 0 otherwise; Type = l if nonprofit is a medical facility, 0 otherwise; Total (Taxable) Revenues = the sum of taxable and tax-exempt revenues on the IRS Form 990; and Total (Taxable) Expenses = the sum of taxable and tax-exempt expenses on the IRS Form 990. TABLE 2 Logit Analysis of Nonprofit Characteristic Associatd with Reporting Near-Zero Taxable Income RIG * = [alpha] + [[beta].sub.1] + [[beta].sub.2] Donations + [[beta].sub.3] NOL + [[beta].sub.4] Other CPA + [[beta].sub.5] Big-Five CPA + [[beta].sub.6] Similar Activities + [[beta].sub.7] Type + [micro] (a) Percent Change in Variable Estimate Robust-Z Odds Intercept Size -2.74e-09 -2.66 -2 (b) Donations 0.18 0.80 0 NOL -7.06e-07 -1.96 -7 (c) Other CPA 1.01 3.87 175 Big-Five CPA 0.81 3.11 122 Similar activities -0.63 -3.03 -47 Type 0.43 2.19 53 LR [x.sup.2] Log-Likelihood Ratio Pseudo [R.sup.2] 60.10 -417.11 0.067 (a) We tested for the necessity of including fixed effects in the model and result indicate that controlling for the year the observation was from was not necessary. (b) The change in odds for Size is calculated based on a size change of 10 million dollars in total assets. 10 million dollars is approximately 6 percent of the asset value of the average nonprofit in our sample. [c] The change in odds for NOL are calculated based on a change in accumulated NOLs of 100,000 about one third of the average accumulated NOLs in our sample. Variable definitions: Donations = total public donations/total donations (Form 990): NOL = total accumulated NOLs; Size = total assets; Similar Activities = 1 if nonprofit activity has taxable and tax-exempt counterpart, 0 otherwise. Big-Five CPA = 1 if tax return prepared by Big 5 CPA, 0 otherwise; Other CPA = 1 if tax return prepared by non-Big 5 CPA, 0 otherwise. Type = 1 if nonprofit is a medical facility, 0 otherwise.
We thank Rick Hatfield, Michelle Yetman, Sonja Olhoft, workshop participants at The University of Arizona and The University of Iowa, participants at the 2002 American Accounting Association Annual Meeting, and two anonymous referees for helpful and constructive comments.
(1) The other income tax systems are those levied on corporations (1120 series of tax forms), individuals (1040 series of tax forms), and trusts and estates (700 series of tax forms).
(2.)For example, sales of pharmacy items by a nonprofit hospital to currently admitted patients is tax-exempt, while sales of identical items to non-patients is taxable.
(3.) Burgstahler and Dichev (1997) assume that the amount of observations within any particular income interval should be the average of the two adjacent intervals. We are not able to use this procedure for the distribution of taxable profitability because we examine deviations from a smooth distribution that includes the expected peak of the distribution (i.e., near zero), although our measure captures a similar concept (i.e., smoothness).
(4.) Burgstahler and Dichev (1997, 102-103 and footnote 6) derive a similar statistic. The variance of the difference is approximately the sum of the variances of the components of the difference. If the variance of a binomial variable is equal to npq (where n is the number of observations, p is the probability of the interval in question, and q is p--1), then the variance of the difference can be approximated as: [Np.sub.i] (1 - [p.sub.i]) + [Np.sub.i-1](1 - [P.sub.i-1]), where N is the total number of observations and [p.sub.i] is the probability that an observation will fall within a particular interval calculated as the expected number of observations in the interval divided by N.
(5.) Results of this analysis are robust to various alternatives including beginning with the right-most interval and moving to the left, adding a drift term to the expected value equal to the change in the preceding interval, and adding a variety of higher-order drift terms.
(6.) For the analysis of the tax-exempt income profitability, we use the same methodology and test statistic as Burgstahler and Dichev (1997) in which the expected value of an interval is the average of the two surrounding intervals. We use the same test statistic for the distribution of tax-exempt profitability because the expected abnormality is off-peak. Results are not sensitive to using the same test statistic as was used for the taxable distribution (which assumed a smoothly changing distribution).
(7.) The model is a reduced form for presentation. CPA is partitioned into Big 5 CPAs and Other CPAs in the estimated model.
(8.) Note that this particular robustness test lessens one type of error (excluding taxable income managers) at the expense of potentially increasing the other type of error (including taxable income non-managers).
(9.) Nonprofits receive several forms of donations, some of which are more sensitive to donor tastes. The three types are public donations from individuals and corporations, government grants, and feeder donations (from other charitable organizations). We prefer public donations rather than total donations (the sum of the three) because they may be more susceptible to public opinion.
(10.) The amount of the NOL carryforwards is from line 31 of the IRS 990-T. As an integrity check, we searched through the IRS 990-Ts for the NOL carryforward schedules and matched the amounts reported on line 31 with those schedules, reconciling any difference in favor of the carryforward schedule.
(11.) At the time of the data request, 1995 was the most recently available year of data.
(12.) Exploited activities are those that profit from the use of a nonprofit's intangible "good name." An example would be a radio advertisement for a sports beverage, which is endorsed by a collegiate football program.
(13.) Treas. Reg. [sections]l.512(a)-l(d)(l).
(14.) The signature portion of a tax return kept in client files is often blank because the file copy is made before the signature block is filled out. Rather than assume who the tax preparer was, we excluded these observations.
(15.) Because the preparer identified sample (i.e., 1,367 observations) is smaller than the full sample (i.e., 1,824 observations), we tested to see whether the smaller subset is significantly different across our analysis variables. Untabulated results show that the only significant difference between the full and reduced samples is the level of donations with the reduced sample receiving significantly fewer donations when compared to the full sample (p < .000). The remaining variables are not significantly different between the full and reduced sample.
(16) We also estimated the model with a signal indicator variable for CPA-filed returns and find a modest decrease in predictive power and area under the ROC curve, although inferences with respect to the preparer variable were robust.
Beck, P. J., and W-O. Jung. 1989. Taxpayers' reporting decisions and auditing under information asymmetry. The Accounting Review 64 (July): 468-487.
Becker, C., M. DeFond, J. Jiambalvo, and K. R. Subramanyam. 1998. The effects of audit quality on earnings management. Contemporary Accounting Research 15: 1-24.
Berss, M. 1994. Taxation by other means. Forbes (April 1): 64-68.
Brady, D. 2000. When nonprofits go after profits. Business Week (June 26): 173-178.
Burgstahler, D., and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses. Journal of Accounting and Economics 24:99-126.
Collins, J., D. Kemsley, and D. Shackelford. 1997. Transfer pricing and the persistent zero taxable income of foreign-controlled U.S. corporations. The Journal of the American Taxation Association 19 (Supplement): 68-83.
Cordes, J. J., and B. A. Weisbrod. 1998. Differential taxation of nonprofits and the commercialization of nonprofit revenues. Journal of Policy Analysis and Management 17:195-214.
Degeorge, G., J. Patel, and R. Zeckhauser. 1999. Earnings management to exceed thresholds. The Journal of Business 72 (January): 1-33.
Erard, B. 1993. Taxation with representation: An analysis of the role of tax practitioners in tax compliance. Journal of Public Economics 52 (September): 163-197.
Francis, J., E. L. Maydew, and H. C. Sparks. 1999. The role of Big 6 auditors in the credible reporting of accruals. Auditing: A Journal of Practice & Theory 18: 17-34.
Gomes, G., and J. Owens. 1988. Commercial nonprofits, untaxed entrepreneurialism, and "unfair competition." Journal of Small Business Management 26 (April): 8-17.
Greene, W. H. 2000. Econometric Analysis. New York, NY: Macmillan Publishing Company.
Grubert, H., T. Goodspeed, and D. Swenson. 1993. Explaining the low taxable income of foreigncontrolled companies in the U.S. In Studies in International Taxation, edited by A. Giovannini, R. G. Hubbard, and J. Slemrod. Chicago, IL: University of Chicago Press.
Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20: 125-153.
Hite, P., and G. McGill. 1992. An examination of taxpayer preference for aggressive tax advice. National Tax Journal 45 (December): 389-404.
Hosmer, D. W., and S. Lemeshow. 2000. Applied Logistic Regression. 2nd edition. Wiley Series in Probability and Statistics. New York, NY: John Wiley & Sons, Inc.
Klepper, S., M. Mazur, and D. Nagin. 1991. Expert intermediaries and legal compliance: The case of tax preparers. Journal of Law and Economics 34 (April): 205-229.
Manzullo, D. 2001. Manzullo asks Treasury to report on unfair competition between exempt organizations and small businesses. Tax Notes Today 66 (April 5): 32.
Maydew, E. L. 2001 Empirical tax research in accounting: A discussion. Journal of Accounting and Economics 31 (September): 389-403.
Meckstroth, A., and P. Arnsberger. 1998. A twenty-year review of the nonprofit sector, 1975-1995. Internal Revenue Service Statistics of Income Bulletin 18: 149-169.
Mills, L., and K. Newberry. 2001. The influence of tax and non-tax costs on book-tax reporting differences: Public and private firms. The Journal of the American Taxation Association 23 (Spring): 1-19.
National Center for Charitable Statistics (NCCS). 1999. 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.
Okten, C., and B. Weisbrod. 2000. Determinants of donations in private nonprofit markets. Journal of Public Economics 75 (February): 255-272.
Omer, T. C., and R. Yetman. 2002. Taxable income management by nonprofit organizations. Working paper, University of Illinois and The University of Iowa.
Orban, R. 1999. Testimony before the Internal Revenue Service Panel on Unrelated Business Income Tax, Office of Advocacy, U.S. Small Business Administration. February 10. Washington, D.C.: Government Printing Office.
Phillips, J., and R. Sansing. 1998. Contingent fees and tax compliance. The Accounting Review 73 (January): 1-18
Ramsey, D. 1986. The 1986 White House conference on small businesses. Agency Sales 16: 36-39.
Sansing, R. 1998. The unrelated business income tax, cost allocation, and productive efficiency. National Tax Journal 51: 291-302.
Scholes, M., M. Wolfson, M. Erickson, E. Maydew, and T. Shevlin. 2002. Taxes and Business Strategy: A Planning Approach. 2nd edition. Englewood Cliffs, NJ: Prentice Hall.
Steinberg, R. 1986. The revealed objective functions of nonprofit firms. Rand Journal of Economics 17 (Winter): 508-526.
U.S. House, Committee on Ways and Means. 1950. Revenue Revision of 1950: Hearings before the House Committee on Ways and Means, 579-580. 81st Cong., 2d sess. Washington, D.C.: Government Printing Office.
Yetman, R. 2001. Tax-motivated expense allocations by nonprofit organizations. The Accounting Review 76 (July): 297-311.
Thomas C. Omer is an Associate Professor at the University of Illinois at Chicago and Robert J. Yetman is an Assistant Professor at The University of Iowa.
|Printer friendly Cite/link Email Feedback|
|Author:||Omer, Thomas C.; Yetman, Robert J.|
|Publication:||Journal of the American Taxation Association|
|Date:||Sep 22, 2003|
|Previous Article:||How well do Compustat NOL data identify firms with U.S. tax return loss carryovers?|
|Next Article:||An international investigation of the influence of dividend taxation on research and development tax credits.|