In the absence of owners, how effective are the constraints imposed by the state in promoting effective organization governance? This paper develops state-level indices of the governance environment facing not-for-profits and examines the effects of these rules on not-for-profit behavior. Stronger provisions aimed at detecting and punishing managerial misbehavior are associated with significantly greater charitable expenditures, increased foundation payouts, and lower managerial compensation. The paper also examines how governance influences an alternative metric of not-for-profit performance—the provision of social insurance. Stronger governance measures are associated with intertemporal smoothing of resources in response to economic shocks.
JEL Classifications: L30; G30; H40; K20.
The not-for-profit sector is sizable, with total assets of over $3.5 trillion across more than 270,000 organizations (IRS 2010). The not-for-profit sector is characterized by the absence of owners but the presence of legal and reporting rules. In the U.S., the primary governing responsibility over the activities of not-for-profit organizations resides with the various states (Fremont-Smith 2004; Brakman-Reiser 2004, 2005). States not only approve the incorporation of all not-for-profit organizations (and retain sole authority to dissolve them as well), but also establish a variety of laws and reporting requirements intended to monitor and oversee the activities of not-for-profits operating within their borders. These governance laws and reporting requirements vary significantly by state within the U.S., creating the opportunity to analyze the efficacy of legal rules in a setting without owners and without the complicating features of cross-country studies. The extent to which these state laws are effective in governing not-for-profit behavior is largely unknown, and is the purpose of our analysis.
We examine both public charities and private foundations. Public charities and private foundations are similar in that both can receive tax-deductible donations, and both operate free of income taxes. Perhaps the most significant difference is that, in contrast to private foundations, public charities receive their support from a broad set of donors and directly engage in charitable activities. Private foundations receive their support from a very small set of donors (frequently a single individual or family), and do not directly engage in charitable activities but rather make grants to public charities from their endowments. Another important difference is that private foundations must annually spend (commonly referred to as “payout”) at least 5 percent of their investment assets (IRC 4942(e)(1)(A)).
If these entities were motivated purely by altruistic motives, or if the laws themselves were ineffective, then governance constraints would not affect not-for-profit outcomes. Alternatively, governance laws might improve organization performance. The results of our paper should interest state-level lawmakers, as the results inform the extent to which state-level governance and reporting laws have a measurable effect on not-for-profit behaviors.
We use two measures of state-level governance. The first is a linear aggregation of eight state governance laws and nine reporting requirements. There is significant variation in how many of these 17 laws and reporting requirements a state has adopted, ranging from a low of three (in Hawaii) to a high of 16 (in Tennessee). Our second governance measure is the ratio of fully adjudicated (i.e., completed trial) cases brought against not-for-profits for breach of fiduciary duty in a state, scaled by the number of not-for-profits in that state.
Assessing the efficacy of these governance measures requires observable metrics of organization behavior. In order to identify specific behaviors that would be reasonably expected to respond to state-level oversight and monitoring, we rely on the legislative intent of these state laws. In general, state governance laws have two primary purposes: to ensure that not-for-profits employ their assets in the accomplishment of a charitable purpose, and to prevent the transfer of assets to those in positions of influence, such as managers and directors (Fremont-Smith 2004; Brakman-Reiser 2004, 2005).
We rely on prior research to identify empirically tractable measures of these two theoretical constructs for both public charities and private foundations. With respect to public charities, we use two measures of the extent to which an organization uses its assets to accomplish its charitable mission. The first is the Program Service Ratio (the ratio of charitable expenses to total expenses) and the second is the Administrative Expense Ratio (ratio of administrative expenses to total expenses). We use the ratio of chief executive officer (CEO) compensation to total expenses as our measure of the extent to which a public charity is transferring assets to those in positions of influence.
For private foundations, we use three measures of the extent to which a foundation uses assets to accomplish the charitable mission, all of which are related to the amount and timeliness of foundation payouts. The first is the Payout Ratio, which is the ratio of actual payouts to the required legal minimum of 5 percent. The second is Delayed Payouts, which is an indicator of how long a foundation delays its legally required payout. Our third measure is the Grants Ratio, which is the proportion of payouts made as grants to public charities rather than being consumed by internal administrative expenses. As with public charities, we use the ratio of chief executive officer (CEO) compensation to total expenses as our measure of the extent to which a private foundation is transferring assets to those in positions of influence.
In the second part of our analysis, we examine the extent to which governance affects not-for-profits' willingness to provide social insurance against economic shocks.1 When a not-for-profit receives an unexpected negative shock to its revenues, there will be a natural tendency to reduce overall charitable spending, with the result that charitable beneficiaries bear the cost of the negative shock. Not-for-profits can attenuate these reductions in charitable outputs by drawing down accumulated savings so as to smooth their charitable outputs. We test the extent to which stronger state-level governance is associated with not-for-profits' propensity to smooth their charitable outputs in both good and bad times in an effort to alleviate the negative consequences of economic cycles, at least partially shielding their charitable beneficiaries from shocks to organizational revenues.
Our results show that, for public charities, both state laws and the enforcement of those laws are positively associated with larger proportions of expenses being directed toward the charitable mission, and negatively associated with CEO compensation. With respect to private foundations, we find both state laws and the enforcement of those laws to be positively associated with larger and faster payouts. State laws (but not enforcement) are also positively associated with larger proportions of foundation payouts that are given to public charities as grants (as opposed to being consumed by administrative activities). We do not find state laws or reporting requirements to be associated with foundation CEO compensation.
Overall, these are consistent with state-level laws and reporting requirements, as well as state-level enforcement of those laws and requirements, being associated with not-for-profit organizations devoting more of their resources toward accomplishing their charitable missions and reducing private benefit transfers to persons of influence.
Finally, we find that public charities smooth their charitable spending in the face of revenue shocks. This finding is consistent with a social insurance behavior in that charities in more well-governed states decrease their charitable spending less as resources fall, shielding their charitable beneficiaries from bearing the full costs of negative economic shocks. These results can be viewed as supporting the view that good governance rules encourage not-for-profit entities to provide social insurance.
Our paper is the first to link a comprehensive set of contemporary state-level laws and reporting requirements to specific public charity and private foundation output behaviors. Our work is closest to that of Fisman and Hubbard (2003, 2005), who examine the role of endowments and how state-level governance affects the relationship between endowments and donation flows. Our paper also parallels La Porta, Lopes de Silanes, Shleifer, and Vishny (1997, 1998) and subsequent work in the law and finance literatures. Empirically our paper is similar to Gompers, Ishii, and Metrick (2003), who create country-level governance indices and link these indices to firm performance.
Our paper differs from Fisman and Hubbard (2003, 2005) in several ways. First, the measure of state laws used by Fisman and Hubbard (2003, 2005) include eight laws and reporting requirements from 1973, while our measure includes 17 laws and reporting requirements based on data that are 30 years more contemporary.2 Second, we also use a measure of how frequently a state enforces its not-for-profit laws. Third, we examine both private foundations and public charities. Finally, we directly examine specific not-for-profit outcome measures found by prior research to be indicators of how well a charity or foundation is accomplishing its charitable purpose and the extent to which not-for-profits provide any form of social insurance by smoothing their charitable outputs across economic shocks.
Several other papers have examined not-for-profit governance, but do not examine the effects of state laws on not-for-profit behaviors. Pauly and Redisch (1973) were among the first to empirically examine not-for-profit governance within the context of not-for-profit hospitals, while Glaeser (2003) extended the analysis to other settings. Core, Guay and Verdi (2006), building on Fisman and Hubbard (2005), show that not-for-profits with large “excess” endowments pay higher managerial compensation and also spend less on charitable outputs. Sansing and Yetman (2006) examine private foundations and find that tax laws play a role in incentivizing foundations to distribute (or “pay out”) additional funds to public charities. Finally, M. Yetman and R. Yetman (2011) find that several sources of governance are effective in reducing financial statement manipulations by not-for-profit organizations.
In addition to the relevant literature on not-for-profit organizations, the analysis also frames not-for-profit organizations within a more comprehensive literature on social insurance mechanisms. This literature typically emphasizes programs that are explicitly designed to provide insurance, such as unemployment insurance as in Hamermesh (1982) or Gruber (1997), and their effects in allowing recipients to smooth consumption. The intuition of social insurance, in which beneficiaries are shielded from some portion of unexpected economic shocks, has been extended to the mechanisms that are operative between and within families (Hayashi, Altonji, and Kotlikoff 1996) or through the progressivity of the tax code (Auerbach and Feenberg 2000; Kniesner and Ziliak 2004). While Cochrane (1991) alludes to the role of not-for-profits in smoothing consumption, there do not appear to be any empirical studies that conceptualize not-for-profits in this way. Our paper is the first attempt to empirically document whether not-for-profit organizations smooth charitable spending in the presence of revenue shocks.
STATE-LEVEL NOT-FOR-PROFIT GOVERNANCE
The primary contribution of our paper is investigating the efficacy of state-level not-for-profit governance laws and reporting requirements, as well as enforcement of those laws. As such, we next provide a discussion of the origin of these laws. The primary locus for legal authority over not-for-profit organizations has resided at the state level since the creation of the first state in 1776 (Fremont-Smith 2004; Brakman-Reiser 2004, 2005). Not-for-profits are incorporated at the state level and therefore agree to be bound by state laws. Only the state has the legal power to create and dissolve not-for-profit organizations. Other than the state, no other person or entity has the inviolable legal right to monitor and oversee the operations of not-for-profits, unless contractually granted.3 Typically states empower their attorneys general (either directly or indirectly) to enforce their laws and reporting requirements over not-for-profit organizations conducting business or raising donations within their borders. As each state entered the Union, it was free to adopt the set of not-for-profit laws and reporting requirements it saw fit and, unsurprisingly, the combination of laws and reporting requirements adopted by each state varies widely.
We gather these laws from Fremont-Smith (2004), and the reporting requirements from the Multi-State Filer Project of the National Association of State Charities Officials and the National Association of Attorneys General. These two sources identify 17 not-for-profit governance laws and reporting requirements. Table 1 provides a brief discussion of these specific laws.
Our empirical measure of state-level governance and reporting laws (State Laws) is simply the sum of the number of governance laws and reporting requirements that a state has adopted. Table 2 reports a mean of 11.5 and a median of 13. The number of laws adopted by a state varies from a low of three (Hawaii) to a high of 16 (Tennessee). We avoid arbitrary assumptions about which laws are more or less effective and use a parsimonious additive index, consistent with research in the for-profit setting (La Porta et al. 1998; Gompers et al. 2003). One obvious question that naturally arises is why any of these laws are expected to influence not-for-profit resource allocation decisions. Perhaps the best response to that query is that the various states must believe that these laws have some effect, lest they would not incur the costs to pass and enforce them. In the end, the extent to which these laws and reporting requirements affect not-for-profit behavior is ultimately an empirical question.
In addition to having laws on the books, another important component of state-level not-for-profit governance is the extent to which those laws are actually enforced. The effects of enforcement on behaviors are a long-established and well-researched area (Beccaria 1764; Moberly 1968). In general, that research shows that enforcement of laws on one party reduces illegal acts by other parties, primarily due to observation of punishment leading to a desire to avoid similar punishment. In our setting, this suggests that more frequent visible prosecution via fully adjudicated cases (i.e., went to trial and received a public verdict) of not-for-profits in a state would be plausibly associated with other not-for-profits in that same state avoiding behaviors that would lead to their own prosecution.
A study by Fremont-Smith and Kosaras (2003) documents the number of fully adjudicated cases brought against not-for-profits in the various states over the period 1995 to 2002, a period that overlaps with our sample period (i.e., 1995 to 2007). We use the ratio of the total number of cases for breach of fiduciary duty across all years in a state, divided by the average number of not-for-profits in a state over the period 1995 to 2002 (State Enforcement) as our enforcement variable.4 Descriptive statistics in Table 2 report a mean ratio of 0.68, which is actually 0.00068, as we multiply by 1,000 for presentational purposes. Thus, on average, one of every 11,765 not-for-profits located in the U.S. was prosecuted in a fully adjudicated case over the period 1995 to 2002 (i.e., 1/0.00068 × 8 years = 11,765).
We seek to determine the extent to which not-for-profit behaviors are influenced by not-for-profit governance laws and reporting requirements, as well as enforcement of those laws. To guide us in our selection of specific behaviors that these laws, requirements, and enforcement actions plausibly affect, we turn to the historical intent of those laws. The essence of these state laws dates back to 1601 as included in the Statute of Charitable Uses Act enacted by Queen Elizabeth I. The 1601 Statute forms the basis for most state-level not-for-profit governance laws that exist today (Fremont-Smith 2004). These laws have two primary goals.5 First, they are intended to ensure that not-for-profits devote their resources toward accomplishing their charitable purpose. States recognize that some not-for-profits might be tempted to use their resources for ancillary or noncharitable purposes.6
Second, the laws are intended to prevent the transfer of private benefits to disqualified persons. Most state statutes (as well as Internal Revenue Code Section 4975(e)(2)) define disqualified persons as donors, managers, and directors. States recognize that persons in these positions of influence within a not-for-profit might be tempted to use the organization's assets for their own purposes. To prevent this behavior, state laws prohibit the transfer of assets to disqualified persons above amounts commensurate with the services provided to the organization.
Based on the intent of these laws, our first hypothesis is that governance laws affect the extent to which a not-for-profit accomplishes its charitable mission, stated in the null:
State-level governance laws and enforcement have no effect on the extent to which not-for-profit organizations accomplish their charitable mission.
Our second hypothesis, which also follows the intent of state laws, is that governance laws affect the extent to which a not-for-profit transfers private benefits to disqualified persons, or stated in the null:
State-level governance laws and enforcement have no effect on the extent to which not-for-profit organizations transfer private benefits to disqualified persons.
Our third hypothesis considers the extent to which governance laws affect not-for-profits' willingness to provide social insurance, by which we mean that the organization at least partially insulates their charitable beneficiaries from negative shocks to revenues. We test this hypothesis on public charities only (and not on private foundations), as private foundations do not directly conduct charitable activities and, thus, cannot directly insure charitable recipients.
Adapting logic from other economic literatures, we suggest that the degree to which not-for-profits provide social insurance—by building stocks of reserves and responding to local negative income shocks—represents a plausible measure of positive performance. Much as the efficiency of private firms is measured by their responsiveness to investment opportunities, this test capitalizes on the idea that not-for-profits should expand activity at times when their activity is most warranted—that is, do the organizations help when helping helps the most? To envision how this insurance effect could work, consider a not-for-profit that experiences a negative shock to its revenue streams. This organization can either reduce its charitable spending by the full amount of the shock or, instead, “smooth” its charitable spending by reducing it less than by the entire amount of the negative shock. Ostensibly, resources to make up the difference would come from financial “reserves,” perhaps from saving some portion of past positive shocks to revenues. An insurance reaction to both positive and negative shocks would result in “smoothing” charitable spending in response to those shocks. Based on this reasoning our third hypothesis, stated in the null, is:
State-level governance laws and enforcement have no effect on the extent to which not-for-profit organizations insure their charitable beneficiaries by smoothing their charitable outputs in response to revenue shocks.
Empirically testing these three hypotheses requires valid and tractable measures of our behavioral constructs, an issue that we address in the following sections.
While approximately 30 different types of not-for-profit organizations are recognized by the IRS, those exempt under Section 501(c)(3) of the Internal Revenue Code (i.e., public charities and private foundations) account for over 90 percent of the total assets and revenues of the not-for-profit sector. In addition to receiving relief from federal and state income, sales, and property taxes, contributions to these two types of not-for-profits are tax deductible. Since churches are not required to report their information to the IRS, they are not included in the analysis. Similarly, schools and hospitals are frequently exempted from many state-level not-for-profit laws and are subject to other specific laws and regulations. Accordingly, schools and hospitals are also excluded from the analysis.
All public charities and private foundations with revenues over $25,000 must file an IRS Form 990 or 990-PF annually that are available to the public.7 The public charity dataset we use is the 1985 to 2007 IRS Statistics of Income sample files. Although the number of observations in the IRS SOI samples is considerably fewer than the full population, samples are constructed in such a way to as include the vast majority of economic activity, and still include a representative sample of smaller organizations.
Our public charity sample starts with 135,981 total observations across the years 1985 to 2007. Consistent with Core et al. (2006), we exclude 17,618 grant-making foundations, mutual organizations, and organizations whose industry is “unknown” (classified as “T,” “Y,” and “Z,” respectively by the National Taxonomy of Exempt Entities [NTEE] industry classification system).
Because our sample is a panel across several years, each not-for-profit appears in the sample multiple times. However, our state-level governance variables do not vary across observations within a state. To control for repeated values of our state governance variables at the not-for-profit level, we use the average value of our variables across our sample time period for each observation. Thus, each not-for-profit appears in our analysis sample only once, with variables equal to across-time averages. This reduces the sample by 104,102 observations leaving us with 14,261 unique observations. We reduce the sample further by imposing three additional screens. First, we remove 1,464 observations that report average donations of zero. Most state governance efforts are intended to protect donors, and thus not-for-profits that receive zero donations are less likely to be subject to state governance. Second, we remove 4,900 observations that report zero administrative expenses or zero fundraising expenses, which prior research suggests are incorrectly reported (Krishnan, M. Yetman, and R. Yetman 2006; Keating, Parsons, and Roberts 2008). Our final public charity analysis sample contains 7,804 unique observations.
The foundation dataset contains 120,208 observations across the years 1993 through 2007 (IRS data for private foundations only begin in 1993). As with the public charity sample, each foundation appears only once in our analysis sample, with variables set as means for values. This reduces the sample to 17,977 observations. Finally, we reduce the sample to 14,003 observations as some of our variables require one-year lagged values and many observations either occur only once in the sample (and thus do not have a lagged value), or occur more than once but not in consecutive years.
Not-for-Profit Behavior Empirical Model
Our first two hypotheses test whether state not-for-profit laws affect the extent to which organizations accomplish their charitable missions and transfer private benefits to disqualified persons. To test H1 and H2, we use the following empirical model:
In all of our analyses the primary variables of interest are State Laws and State Enforcement, both of which have been previously defined and discussed. We use a variety of dependent variables measuring non-for-profit behaviors as empirical measures of the theoretical constructs of mission accomplishment and the transfer of private benefits to disqualified persons. Specific dependent variables depend on whether the observation is a public charity or a private foundation. In cases where our measure of Behavior is continuous (dichotomous), we use an OLS (Logit) model. Consistent with Core et al. (2006), we winsorize all variables (before calculating their organization-specific average) by year at the 1 percent and 99 percent levels (i.e., for each variable we reassign its value if it is less (greater) than the 1st (99th) percentile to the value of the 1st (99th) percentile in a given year) to mitigate the influence of outliers. We also remove observations with a studentized residual greater than 4.
In addition to our test variables, we also include Total Revenues (line 12 of the IRS 990, and line 12 of the IRS 990-PF), and Total Assets (line 59b of the IRS 990, and line I of the IRS 990-PF). We include these scale-related controls as not-for-profits significantly vary in size and scale, and prior not-for-profit research (as cited throughout this paper) routinely includes them. Finally, we also include single-digit NTEE industry indicator variables to control for cross-industry effects, again consistent with research in the literature. Results are unchanged when a finer industry partition (i.e., two-digit) NTEE classification is used.
Empirical Measures of Public Charity Mission Accomplishment
Recall that the first intent of state not-for-profit governance laws is to ensure that not-for-profits devote their resources toward accomplishing their charitable purpose. We create two empirical measures intended to capture the construct of how well a public charity accomplishes its charitable purpose and avoids the pursuit of ancillary activities. Existing measures of this construct are well established in the literature, and are based on various expense ratios. On their IRS 990, not-for-profit organizations are required to partition all of their expenses into one of three categories: charitable, administrative, or fundraising, with charitable expenses being those devoted to accomplishing the primary charitable mission, and administrative expenses being those devoted to either the operation of the organization or to ancillary activities not directly related to the charitable purpose. Based on these definitions, our first empirical measure is the Program Service Ratio, or the ratio of charitable expenses (line 13 of the IRS 990) divided by total expenses (line 17 of the IRS 990). Research shows that the program service ratio is a reasonable measure of how well a not-for-profit is accomplishing its charitable mission (Weisbrod and Dominguez 1986; Posnett and Sandler 1989; Tinkelman 1999, 2004; Greenlee and Brown 1999; Okten and Weisbrod 2000; Parsons 2003, 2007; Tinkelman and Mankaney 2007).
One empirical issue with using the reported program service ratio is that a growing body of research shows that not-for-profits systematically under report their fundraising expenses and over report their charitable expenses, thereby inflating their Program Service Ratio (Tinkelman 1998; Wing, Gordon, Hager, Pollak, and Rooney 2006; Roberts 2005; Khumawala, Parsons, and Gordon 2005; Jones and Roberts 2006; Krishnan et al. 2006; Keating et al. 2008). The upshot of this literature is that not-for-profits that earn over $10,000 in donations yet report exactly zero fundraising expenses have likely shifted their fundraising expenses into their charitable expenses, and thus should be removed from any analysis that relies on the reported program service ratio (Krishnan et al. 2006). As previously discussed, we follow this advice and remove all such observations.
Our second empirical measure is the Administrative Expense Ratio, or the ratio of administrative expenses (line 14 of the IRS 990) divided by total expenses (line 17 of the IRS 990). Prior research suggests that larger administrative expense ratios are consistent with spending on noncharitable ancillary activities (Wing et al. 2006; Tinkelman 1999; Yetman and Yetman 2011).8
Empirical Measures of Public Charity Private Benefit Transfers
Recall that the second intent of state not-for-profit governance laws is to prevent the transfer of private benefits to disqualified persons. As discussed by Fremont-Smith and Kosaras (2003), the three most common forms of private benefit transfers are misuse of assets, self-dealing, and excessive compensation. Misuse of assets occurs when a disqualified person uses the charitable assets for their own personal benefit, such as using the organization's automobiles for personal vacations. Self-dealing occurs when a disqualified person directs the business of the not-for-profit toward entities in which the disqualified person has a financial interest, such as giving favorable contracts to companies owned by the disqualified person. Finally, excessive compensation is defined by Treasury Regulation 1.162-7(b)(3) as compensation above “the amount that would ordinarily be paid for like services by like organizations in like circumstances.”
Identification of asset misuse and self-dealing is a difficult, if not impossible, task as financial disclosures do not provide a basis for constructing more than crude variables. Therefore, the most commonly used empirical measure of private benefit transfers to disqualified persons is excessive CEO compensation. Not-for-profit financial information on the IRS 990 and 990-PF discloses the CEO's compensation, and the CEO is a disqualified person. However, neither the IRS 990 nor the 990-PF identifies whether the compensation is excessive. Prior research uses the level of CEO compensation or the ratio of CEO compensation to total expenses as measures of “excessive” compensation, with higher values being more likely to reflect “excessive” amounts (Fisman and Hubbard 2005; Core et al. 2006).
Based on this, we use the ratio of chief executive officer (CEO) compensation to total expenses (CEO Compensation Ratio) as our empirical measure. By scaling compensation by total expenses, our measure reflects the level of compensation relative to what would be expected based on the organizational size, consistent with the intent of Treasury Regulation 1.162-7(b)(3). We presume that higher ratios are more likely to indicate excessive compensation.
We obtain officer's compensation from Part V of the IRS 990, which lists compensation by person, but without an identifying title. Although we do not know which compensation line item relates to the CEO, we assume the highest compensation number is that of the CEO. This procedure is now common practice in not-for-profit compensation research, and has been shown to be reasonably valid and reliable.9
Empirical Measures of Private Foundation Mission Accomplishment
We construct three empirical measures intended to capture the construct of how well a private foundation accomplishes its charitable purpose. Private foundations primarily exist for one legal purpose, to provide grants to public charities.10 Therefore, measures of how well a private foundation is accomplishing its charitable purpose are related to how much, and how soon, foundations give to public charities.
Prior research has investigated the extent to which specific tax laws are effective in governing private foundation behaviors; we rely on that research to identify our empirical measures of foundation output behaviors (Sansing and Yetman 2006; Core and Donaldson 2010; Yoder, Addy, and McAllister 2011). Consistent with this research, our first empirical measure is the Payout Ratio, which is the ratio of actual payouts (line 6 of Part XII of the IRS 990-PF) to required payouts (line 7 of Part XI of the IRS 990-PF). Foundations are required to payout at least 5 percent of their prior-year's investment assets each year under IRC 4942, although they can payout additional amounts if they choose. IRS data reveal that the majority of foundations adhere closely to this 5 percent minimum (Belmonte 2009). Minimum (i.e., 5 percent) payouts have historically led to more rapid endowment growth, and research suggests that large foundation endowments have smaller social value as compared to the costs of forgone current-period charitable outputs (Hansmann 1990; Craig 1999; Mehrling 1999). Some scholars challenge the concept of an endowment altogether (Brody 1997), but the consensus is that optimal payouts likely vary across foundations and the charities they support, and adhering to minimum payouts, on average, is evidence of nonoptimal (from a societal viewpoint) behavior (Deep and Frumkin 2001). Thus, we presume that larger payouts are consistent with foundations better accomplishing their charitable missions.
Our second empirical measure is Delayed Payouts, or how much a foundation delays its payout obligations. Tax law permits foundations to delay their required payouts by up to one full year. The historical basis of this law was to provide foundations with sufficient time to ascertain the fair market value of their prior-year's ending assets. Although this valuation was not an easy task 50 years ago, given today's computerized technology, asset valuation is comparatively simple and thus most foundations are aware of their necessary payout amounts and no longer need the one-year lag. Therefore, a foundation's choice to delay their payout by a full year is not likely a choice based on asset valuation constraints, but rather based on their desire to retain their assets as long as possible (and thus delay grants to public charities as long as possible). Consistent with Sansing and Yetman (2006), we consider this delay to be less consistent with a foundation accomplishing its charitable purpose. We measure Delayed Payouts as an indicator variable set equal to 1 if the foundation delayed at least 90 percent of its required payouts for one full year, and 0 otherwise.
Our third empirical measure of how well a private foundation is accomplishing its charitable mission is the Grants Ratio, or the ratio of payouts in the form of grants to public charities (line 25 of the IRS 990-PF) divided by the amount of payouts consumed by foundation administrative expenses (line 24 of the IRS 990-PF). Tax law does not require that all (or any) foundation payouts be given to public charities in the form of grants. If a foundation chooses, then it could spend all of its payouts on itself in the form of administrative expenses and salaries.11 Higher grant ratios are consistent with the foundation spending more of its required payouts on grants to public charities, rather than on internal administration, which we consider to be more consistent with accomplishing its charitable purpose.
Empirical Measures of Private Foundation Private Benefit Transfers
As with public charities, we use one measure of private benefit transfers, CEO Compensation Ratio, which is the ratio of CEO compensation (line 13 of the IRS 990-PF) to total expenses (line 11 of the IRS 990-PF). We presume that higher ratios are more likely to indicate private benefit transfers, although no prior research used this variable in the foundation setting.
Not-for-Profit Social Insurance Empirical Model
Our third hypotheses tests the extent to which state not-for-profit laws affect an organization's willingness to provide social insurance by limiting the costs borne by charitable recipients when the organization experiences a negative income shock. To investigate this effect, we predict a negative change in charitable spending when there is a negative shock to revenues, but that this effect is smaller in the presence of stronger governance (and vice versa). The model we use to test H3 is:
The dependent variable is the one-year percentage change in total charitable spending from line 13 of the IRS 990. The ideal revenue source for our independent variable is one that is economically important (i.e., large) to the not-for-profit, and that is also subject to unexpected positive and negative shocks over time. We consider Program Revenues (line 2 of the IRS 990) to be the ideal revenue candidate. Program Revenues are from sales of products and services, which are subject to market conditions often outside the control of the organization.12 Furthermore, Program Revenues are a significant revenue source, accounting for roughly half of total revenues on average.13 The primary variables of interest are the interactions of State Laws and State Enforcement with %Change in Program Revenues. As an additional test, we also examine “unexpected” shocks to program revenues, where “unexpected” is the organization's average change in Program Revenues over the sample period (recall that variables are the organization-level average across all years so that each organization appears only once in our sample) less the average change in Program Revenues average across all years for the organization's single-digit NTEE industry.
Table 2 contains the descriptive statistics for our sample of 7,804 public charities. State Laws and State Enforcement have been previously discussed. The average Program Service Ratio is 0.77, suggesting that the average public charity spends 77 percent of its total expenses on its charitable output. The average Administrative Expense Ratio is 0.18. On average, public charity CEO compensation comprises 5 percent of total expenses, with a median of 3 percent of total expenses. Mean annual CEO compensation for charities is $116 thousand, with a median of $90 thousand.14 In our empirical analysis of public charity CEO compensation, we require a minimum compensation of $10,000. This requirement reduces the sample from 7,804 to 6,740 observations, a reduction of approximately 14 percent.
Turning to private foundations, we find an average Payout Ratio of 1.43, suggesting that the average foundation pays out 43 percent more each year than is legally required, although the sample median of 1.1 suggests that the median foundation pays out just above the minimum requirement. There is significant variation in the Payout Ratio, with some foundations paying out more than 500 percent of their minimum requirement (i.e., a Payout Ratio larger than 5.0), an issue we later address in our empirical analysis. Results show that 23 percent of the sample delays at least 90 percent of their required payouts as long as possible. On average, 90 percent of private foundation payouts are made as grants to public charities (and thus administrative expenses consume only 10 percent of total expenses on average). The ratio of foundation CEO compensation to total expenses is 0.09 (slightly higher than the value of 0.05 for public charities). Mean CEO compensation for private foundations is $100 thousand, with a median of $44 thousand.15 As with public charities, we require a minimum compensation of $10,000, which reduces the sample from 14,003 to 5,317 observations, a rather large reduction in the sample of approximately 62 percent. This suggests that the majority of private foundations pay relatively small amounts of CEO compensation.
Public Charity Results
H1 predicts that state governance laws and reporting requirements will affect the extent to which not-for-profits accomplish their charitable missions, while H2 predicts that those laws and reporting requirements reduce private benefit transfers to disqualified persons. Table 3 contains the results for H1, while Table 4 contains the results for H2.
Results in Table 3 show that, consistent with H1, State Laws and State Enforcement are statistically (at the 1 percent level) positively associated with the Program Service Ratio, and negatively associated with the Administrative Expense Ratio. Several prior papers that examine the Program Service Ratio use a log transformation, and as a robustness test we replicate our results using the log of the ratio with no change in the inferences of our results. Results in Table 4 show that, consistent with H2, State Laws and State Enforcement are statistically (at the 1 or 5 percent level) positively associated with the ratio of CEO compensation to total expenses. Again, we transform our dependent variable using its logged value as a robustness test and find similar results.
In summary, our public charity results support the notion that state-level laws and reporting requirements, as well as state-level enforcement of those laws and requirements, are associated with public charities that devote more of their resources toward accomplishing their charitable missions and reduce private benefit transfers to disqualified persons.
Private Foundation Results
Private foundation results for H1 are in Table 3, while results for H2 are in Table 4. Results in Table 3 show that, consistent with H1, State Laws and State Enforcement are statistically (at the 1 percent or 5 percent level) positively associated with the Payout Ratio, and negatively associated with the Delayed Payouts indicator variable. Results for Grants Ratio is mixed, in that State Laws (but not State Enforcement) is statistically (at the 1 percent level) positively associated with Grants Ratio. With respect to CEO Compensation, results in Table 4 show no statistical association between State Laws and State Enforcement and CEO Compensation. One plausible explanation for this null result is that, as previously discussed, the majority of private foundations in our sample pay relatively small amounts of CEO compensation. Recall that we limited our analysis to foundations that pay their CEOs at least $10,000, and that screen removed over 60 percent of the sample. Given that most foundations pay little CEO compensation, it seems plausible that our null result is due to state attorneys general not focusing their attention on foundation CEO compensation.
In addition, as previously discussed, many private foundations pay out over 500 percent more than is required. To ensure that our results are not being driven by these extreme observations, we conduct two additional robustness tests. First, we remove the 3,050 observations with a Payout Ratio above 2.0. Second, we substitute our continuous measure of Payout Ratio with an indicator variable set equal to 1 if Payout Ratio is greater than its median value of 1.1, and 0 otherwise. Results are robust to both of these additional tests.
In summary, our private foundation results support the notion that state-level laws and reporting requirements, as well as state-level enforcement of those laws and requirements, are associated with private foundations that do not minimize or delay their payouts. We find some support that these laws and requirements are associated with foundations giving a larger proportion of their required payouts as grants to charities, rather than consuming them on internal administrative expenses.
Social Insurance Results
Results for the social insurance tests are found in Table 5. The first column uses the nominal changes in spending and resource variables, while the second column uses the unexpected change in resources. Results from both models consistently suggest that when resources increase, charitable spending also increases (and vice versa). However, consistent with H3, we find that the presence of stronger governance alters the change in charitable outputs as resources from programs rise or fall. This finding is consistent with a social insurance effect in that charities in well-governed states decrease their charitable spending less as resources fall, protecting their charitable beneficiaries from bearing the full costs of negative economic shocks. The opposite is also true in that charities in well-governed states increase their charitable spending less as resources rise, accumulating additional assets to be used in the future, perhaps as a means of protecting their charitable beneficiaries from bearing the full costs of future negative economic shocks. These results can be viewed as supporting the notion that good governance rules help not-for-profit entities fulfill a social insurance function. Further analysis might usefully examine the other factors that allow not-for-profit organizations to fulfill a social insurance function.
Our results suggest that the legal and reporting requirements facing not-for-profit organizations appear to shape their emphasis on charitable activities, their compensation patterns, and their willingness to smooth and time their activities most effectively. These findings are consistent with the notion that state-level laws and regulations constitute an effective governance environment in the absence of owners.
By “social insurance” we mean the extent to which a not-for-profit bears some of the costs of negative revenue shocks, thereby ensuring that less of those costs are borne by its charitable recipients.
The correlation between our state-level governance measure and the one used by Fisman and Hubbard (2003, 2005) is less than 40 percent. The measures used by Fisman and Hubbard (2003, 2005) are based on 1973 data from the Ohio Attorney General, while ours are based on 2003 data from Fremont-Smith (2004).
One exception is that the Internal Revenue Service has the legal right to oversee federal income tax matters of all not-for-profit organizations that have received federal tax-exempt status.
Breach of fiduciary duty is defined in Fremont-Smith and Kosaras (2003) as “breaches of the duties of loyalty and prudence: self-dealing, failing to carry out the mission of the charity, and negligent management of assets.” These activities reflect precisely on the two behavioral constructs that the laws are intended to govern, which are the accomplishment of a charitable mission and prevention of asset transfers to persons of influence.
The Statute's stated purpose is twofold. First, it is “to redresse the misemployment of landes goodes and stockes of money heretofore given to charitable uses,” and second it requires that all assets of the charity be “imployed accordinge to the charitable intente of the givers and founders thereof” (Austin Wakeman Scott, Select Cases and Other Authorities on the Law of Trusts ). See Fremont-Smith (2004) for a comprehensive history of charitable trusts.
One notable example of this is TIAA-CREF, a not-for-profit investment management company that provides retirement vehicles to academics. TIAA-CREF's primary charitable mission is to secure the retirement future of not-for-profit teachers. In 1997 the federal government revoked TIAA-CREF's tax-exempt status because the organization was making substantial profits offering its retirement services to employees of nonacademic for-profit businesses.
To ensure the wide dissemination of Form 990 information, the IRS Statistics of Income (SOI) division provides data to the Urban Institute, which, in turn, makes the data freely available on the Internet at: http://www.guidestar.org. Computer readable data are available from the National Center for Charitable Statistics directly at: http://www.nccs.urban.org.
The Administrative Expense Ratio is not equal to 1 minus the Program Service Ratio, as neither includes fundraising expenses. Although it is certainly true to state that observations with higher Administrative Expense Ratios tend to have lower Program Service Ratios, each has its own unique variation separate and apart from the other.
Studies that use this method include Fisman and Hubbard (2005) and Core et al. (2006). Sedatole, Swaney, M. Yetman, and R. Yetman (2013) use officer's compensation from the IRS “Digitized” database (which has a very limited time series) and compare the CEO's compensation from that database to the numbers we use, and find the two compensation values to be the same 93 percent of the time.
Although a small number of foundations directly conduct charitable activities (i.e., operating foundations), we exclude them from our analysis.
This rule is not well understood or known by the general public, many of whom believe that the 5 percent payout rule requires all the payouts to be grants to charities, which is not the case. However, persistently low (or zero) Grants Ratios would most certainly attract state-level and IRS scrutiny (Boris, Renz, Barve, Hager, and Hobor 2006).
Common sources of Program Revenues include ticket sales at aquariums or theaters, or membership dues at clubs.
We do not use donations, as donors frequently put “restrictions” on their donations and thus additional donations are not necessarily available for spending in the current period. In addition, donations are less subject to market driven shocks, but rather tend to be fairly stable over time (average donations scaled by their standard deviation is 2.4, while average program revenues scaled by their standard deviation is 4.05).
In terms of inter-measure reliability, we find the Program Service Ratio to be statistically (at the 1 percent level) negatively correlated with both the Administrative Expense Ratio and the CEO Compensation Ratio. Similarly, the Administrative Expense Ratio and the CEO Compensation Ratio are positively correlated. These correlations, which are all in the expected direction, provide some degree of comfort that our measures of public charity behavior are coincident with each other.
With respect to inter-measure reliability, we find Payout Ratio to be negatively correlated with the Delayed Payouts as expected, and negatively correlated with CEO Compensation Ratio, also as expected, but not correlated with Grants Ratio. Grants Ratio is negatively correlated with CEO Compensation Ratio, as expected.
We thank Bill Allen, Heitor Almeida, Lucian Bebchuk, Dana Brakman-Reiser, Marion Fremont-Smith, Allan Grossman, Doug Guthrie, Karen Horn, Louis Kaplow, Andrew Metrick, Ken Prewitt, Ross Watts, Rosalie Wolf, David Yermack, William R. Baber (editor), an anonymous referee, and various seminar participants (particularly those at the 2004 Federal Reserve Bank of New York's Conference on the Governance of Not-for-Profit Organizations) for helpful discussions and comments. Professor Desai thanks the Division of Research at Harvard Business School for generous financial support.