Explaining unrestricted giving by charitable foundations: A transaction cost approach

Explaining unrestricted giving by charitable foundations: A transaction cost approach

International Journal of Industrial Organization 28 (2010) 44–53 Contents lists available at ScienceDirect International Journal of Industrial Organ...

234KB Sizes 0 Downloads 28 Views

International Journal of Industrial Organization 28 (2010) 44–53

Contents lists available at ScienceDirect

International Journal of Industrial Organization j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / i j i o

Explaining unrestricted giving by charitable foundations: A transaction cost approach☆ Jeremy Thornton ⁎ Brock School of Business, Samford University, 800 Lakeshore Drive, Birmingham, AL 35080, United States

a r t i c l e

i n f o

Article history: Received 20 February 2008 Received in revised form 12 March 2009 Accepted 7 June 2009 Available online 23 June 2009 JEL classification: L2 L3 D8

a b s t r a c t This paper extends a transaction costs framework to the nonprofit sector where information asymmetries are typically acute. I explore the decision of charitable foundations to place material restrictions on grants to nonprofits. Foundations often place constraints on grant use to limit cross-subsidy of projects or ex-post opportunism by nonprofit managers. In contrast, nonprofits prefer fewer restrictions to smooth income across revenue streams or compensate for shifts in demand. The paper utilizes a pseudo panel of 6000 grant contracts to examine the relationship between the various costs of those restrictions and the observed characteristics of grant contracts. Empirical results confirm received theory that high contracting costs will reduce the probability of grant restrictions. © 2009 Elsevier B.V. All rights reserved.

Keywords: Nonprofit Grant Incomplete contracts Transaction costs

1. Introduction In the U.S., nonprofits provide key services such as healthcare, education, the arts, and human services. Arrow (1963) noted that nonprofit organizations flourished in industries where the quality or quantity of output was particularly difficult to observe. In such industries, stakeholders face a hidden-action problem where profitmaximizing firms have an incentive to shirk on non-contractible elements of the production process. Hansmann (1980) later characterized the nonprofit organizational form as a mechanism for stakeholders to mitigate hidden-action or hidden-information problems, particularly when complete contracts are difficult to write. Presumably, the legal commitment by nonprofit managers to a non-distribution constraint softens the incentives toward opportunistic behavior. Contract failure remains a central theoretical rationale for nonprofit formation (Glaeser & Shleifer, 2001). Yet, empirical evidence supporting the theory has been more difficult to obtain. This paper contributes to understanding the role of incomplete contracts in nonprofit formation by analyzing the particular issue of a charitable foundation's choice in offering restricted grants to nonprofit organizations. Institutional characteristics of the grant-making process allow

☆ Special thanks to Marco Castaneda, Stefan Norrbin, and two anonymous referees for their helpful input into early versions of this paper. ⁎ Tel.: +1 205 726 2128; fax: +1 205 726 2464. E-mail address: [email protected]. 0167-7187/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ijindorg.2009.06.001

us to observe some important components of incomplete contracts among nonprofits that are more difficult to observe elsewhere. This paper presents a unique dataset of grant contracts between private foundations and recipient charities. To interpret variation in contract structure, a transaction cost economics (TCE) framework is applied to generate testable hypotheses about conditions where foundations will likely place restrictions on the use of grant funds. The paper extends empirical analysis of incomplete contracts to the nonprofit sector, which is largely unexplored. Consistent with TCE, empirical results indicate that foundations are more likely to place restrictions on grants as the costs of monitoring/enforcing grant restrictions decline or as the propensity for opportunistic behaviors rise. An empirical examination of contractibility within the nonprofit sector opens the door for an improved understanding of nonprofit formation, capitalization, and behavior. 2. Motivation Reliance on grant restrictions by foundations has been a consistent debate within the nonprofit sector (Krehely & House, 2005). Throughout the 1990 s, the fraction of grant monies classified as unrestricted fell by more than 5 percentage points. By the year 2000, only 11.5% of total grant monies were classified as unrestricted. Unrestricted grants have rebounded in recent years, accounting for roughly 18% of foundation grant dollars in 2005. The underlying determinants of grant restrictions have a strong connection to the TCE literature, but this connection has not yet been explored.

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

The paper treats grants as a production contract between charitable foundations and recipient nonprofit firms. As with for-profit firms, complex contracts among nonprofits are typically incomplete. The key insight of TCE is that economic transactions vary in their cost to design, negotiate and monitor. Consequently, governance mechanisms (contract design) will economize on those costs (Williamson, 2002). 2.1. Previous research The development of the TCE framework is primarily attributed to Oliver Williamson (Williamson, 1975, 1985, 1979, 2002).1 The analytic power from Williamson's work comes from his characterization of three key dimensions of transactions. He proposes that contracts matter when assets are specific to the production process. Relationship-specific investments are then subject to ex-post exploitation. The cost of mitigating this potential for exploitation depends on the uncertainty surrounding the contracting environment and the frequency of the transaction. Contracts will generally become more complete when the costs associated with creating those contracts are reduced. Indeed, three decades of research has demonstrated that nearly any contract can be explored along these dimensions (Williamson, 2002). While there have been significant theoretical advances characterizing incomplete contracts, empirical evidence has lagged. Empirical progress has been hampered by difficulty in measuring the relevant concepts of observability, verifiability, and contract completeness. Even more difficult is the persistent problem of endogenous matching of agents to contracts. Ackerberg & Botticini (2002) document the consequences of ignoring heterogeneity among contract participants and offer some practical empirical solutions. Despite these challenges, there have been some recent advances. Kaplan & Stromberg (2004) examine a set of 213 investment contracts by 14 venture capitalist (VC) firms. They find that VC firms typically construct agreements in order to minimize the costs of expost opportunism as well as the monitoring burden on the VC firm. The tradeoff between agency and monitoring costs appears to play a key role in VC contract design. In a similar work, Baker & Hubbard (2003) exploit an exogenous change in monitoring technology among trucking firms to examine innovation in contract provisions. They find that lower monitoring costs, through new technology, encouraged the integration of shipping firms into trucking. Aggarwal (2007) takes up the more familiar issue of agricultural contracts. Using survey data, she examines groundwater contracts in rural India. Contracts among growers are typically incomplete due to a series of incentive and monitoring costs. Her empirical findings favor a transaction costs explanation of these contracts over traditional principle-agent models of risk sharing. A common theme among these papers is the use of production contract as the unit of observation. The governance form of the contract is regressed on specific characteristics of the contracting environment. This paper will follow that general approach while exploiting some unique institutional characteristics of the nonprofit sector to gain new insights. The paper makes two distinct contributions. A general contribution is made to the TCE literature by offering new empirical evidence on incomplete contracts within the nonprofit sector. While there has been steady progress in understanding contracts among for-profit firms, I am aware of no other empirical study examining similar issues among nonprofits.2 Second, nonprofit capital markets are under1 Williamson's work drew from previous research including: the cost of obtaining information (Stigler, 1961), the coordination of investments (Alchian & Demsetz, 1972), and the measurement of output (Barzel, 1982). Jensen and Meckling (1976) also demonstrated how incentive alignment and the possibility of ex-post opportunism will increase contracting costs. 2 Krashinsky (1986) offers a compact summary of the underlying theory of TCE as it bears on the nonprofit sector.

45

studied in their own right. Foundations play a significant role in financing and monitoring the activities of nonprofits. A more complete understanding of the mechanism by which capital is allocated within the sector should lead to greater efficiency and effectiveness of those agencies which receive funding. The paper is organized as follows. Section 3 motivates the paper by discussing the distinct institutional characteristics embedded within the grant-making process. Section 4 describes the sample data of charitable grants by foundations and its linkages to financial information provided by the recipient nonprofit organizations. In this section a set of proxies for both contractibility and the potential for ex-post opportunism by the nonprofit are introduced. Section 5 develops an empirical model which relates those proxies to the probability of formal grant restrictions by foundation. The empirical findings demonstrate that foundations behave much like VC firms or agricultural contractors. Foundations appear more likely to impose grant restrictions when recipient nonprofits have a greater opportunity to pursue sub-goals or the costs associated with creating, monitoring, and enforcing those restrictions are low. Section 6 addresses legitimate concerns regarding unobserved heterogeneity, particularly among foundations, which would encourage endogenous matching of grant contracts. The original crosssection is expanded to include two additional years of grant observations. Using repeated observations of foundations operating across years, this section attempts to control for unobserved heterogeneity among foundations. The final section explores the limitations and potential extensions of the work. 3. Institutional characteristics of the nonprofit sector Grants made by charitable foundations to recipient nonprofits can be viewed similarly to production contracts made by for-profit firms. Donors make financial gifts to nonprofit organizations with the expectation that those gifts are an input into the charitable organization's production process. Yet, donors face a common agency problem where the preferences guiding the nonprofit will not perfectly match those of the donor. 3.1. The nonprofit organizational form The nonprofit sector includes a vast array of organizations, ranging from hospitals to homeless shelters. Often the term nonprofit is equated with charity. However, the term charity more accurately describes a particular sub-set of nonprofits which receive preferential tax treatment under section 501(c)(3) of the U.S. tax code.3 These privileges include the ability to offer tax deductions to donors as well as exemption from federal income tax. To receive tax-exempt status the organization must pursue a charitable objective, as defined by the IRS (Hopkins, 2003). This paper restricts its attention to tax-exempt charitable organizations. The key economic distinction of the nonprofit organization, however, is its voluntary, public and permanent prohibition against private inurement.4 While a manager has operational control of firm assets, ownership rights are attenuated because there can be no legal claimant to residual earnings. The consequences of this separation of ownership from control were formalized by Jensen and Meckling (1976). They demonstrate how the cost of slack to the manager will decline in line with their ownership rights. The application of

3 Examples of nonprofits which are not charities would be political advocacy groups, mutual benefit organizations such as business trade groups, or fraternal societies. 4 The private inurement doctrine prohibits 1) the transfer of income or assets away from the organization, and 2) the use of organizational income by a person closely associated with the organization for inappropriate purposes (Hopkins, 2003).

46

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

managerial slack in Jensen and Meckling (1976) to the nonprofit organizational form should, however, be treated with caution. While attenuated residual rights should diminish the incentive to profit maximize, this circumstance may be characterized as a virtue. Profit maximization may be replaced by alternative objectives (such as output maximization) that are more favorable to the organization's constituents. Nonprofit firms are an extreme case of the condition described by Jensen and Meckling (1976). Nonprofit entrepreneurs have permanently and publicly given themselves a zero percent ownership stake in the firm. While low-powered incentives may not provide optimal effort in maximizing the objectives of the firm, this disadvantage is offset by a diminished incentive to shirk on non-contractible margins, such as quality (Glaeser & Shleifer, 2001). Low-powered incentives are particularly important in the production of trust goods such as healthcare, education, or social services where output is difficult to observe and contracting is costly. The wide array of nonprofit types implies significant variation in the contractibility of their activities across organizations. It is easy to consider some organizations, such as art museums or universities, where operations are observable and contracts are relatively low cost to construct. One can also conceive of organizations, such as international relief or medical research, where the opportunities for sub-goal pursuit and the cost of constructing complete contracts for their services are much higher. This paper follows previous empirical work on incomplete contracts by proposing that the variation in contractibility leads to substantive differences in the conditions placed on grants or private donations. 3.2. Foundation objectives Foundations are typically organized around a set of stated goals. For example, the Bill and Melinda Gates Foundation restricts the majority of its resources toward global health and development, while the Lilly Foundation emphasizes religion, education, and community development. In this sense, foundations have expressed a public set of preferences for activities they are willing to support. Within the boundaries of these preferences, foundations entertain proposals from organizations which vary in their ability to produce the desired services efficiently. Beyond the decision to finance production, a foundation must decide whether to place material restrictions on the use of the grant funds. It is likely that the recipient nonprofit has better information regarding its own capability/initiative to provide the desired services. Once the gift is given, the foundation has only a limited ability to monitor the allocation of the grant. Consequently, foundations face the following tradeoff. If nonprofit organizations never behaved opportunistically, then unrestricted grants would be efficient. Nonprofits “on the ground” would allocate grant resources to their highest valued use. Consequently, foundations would always wish to give nonprofits full discretion to allocate grant resources optimally. Unfortunately, the pursuit of sub-goals by nonprofit firms has been well documented (Freemont-Smith & Kosaras, 2003). Though the non-distribution constraint prevents the taking of profits as cash, other research indicates that nonprofit managers may consume residual earnings as perquisites (Castaneda et al., 2008; Fisman & Hubbard, 2005). Also common are cases of project cross-subsidy against the wishes of donors (Dreazen, 2005). The most public cases are instances of blatant fraud or embezzlement, such as the recent case of the United Way in Washington D.C. (Whoriskey, 2002). To mitigate the potential for sub-goal pursuit, foundations commonly revert to constraints on the use of the grant. Grant restrictions most commonly bind expenditures of donated funds to a particular project for a specified period of time. Yet, there are costs to imposing grant restrictions. First, there are explicit costs for identify-

Table 1 Summary of foundation grants sample vs. population, year 2000. Number of unique foundations Number of unique grants Number of unique recipients

603 3000 2531 Number

Sample General support/unrestricted Capital support Emergency funds Program support Student aid Research Others Population General support Capital support Emergency funds Program support Student aid Research Others Not specified

% of total

Grant total⁎

% of total

824 441 2 1404 167 144 18

27% 15% 0% 47% 6% 5% 1%

$63,100 $113,000 $21 $206,000 $29,900 $25,700 $3157

14.3% 25.6% 0.1% 46.7% 6.8% 5.8% 0.7%

23,293 14,216 1321 44,941 4965 6498 1521 36,023

19.4% 11.9% 0.0% 37.5% 4.1% 5.4% 1.3% 30.1%

$2,100,341 $3,396,964 $1,321 $6,870,895 $1,065,565 $1,588,386 $336,146 $2,503,148

14% 22.6% 0.0% 45.8% 7.1% 10.6% 2.2% 16.7%

⁎In thousands of dollars. Source: Foundation Center Statistical Services and Author's Calculations.

ing, monitoring, and credibly enforcing the terms of the grant restriction. Second, restrictions may impose implicit costs by forcing the recipient nonprofit to allocate resources sub-optimally. Demand for social services may change over time, the foundation may have insufficient information to gauge the return over various programs, or new technologies for service provision may emerge. Given significant variation in contractibility among nonprofits, foundations must weigh the tradeoff between contracting costs and the possibility of sub-goal pursuits which reduce the impact of the grant. 3.3. Nonprofit objectives Nonprofit recipients will (weakly) prefer unrestricted funds to those with restrictions. If an organization receives an unrestricted grant, it can allocate the donation according to its own preferences. If, instead, the nonprofit receives a grant with restrictions, it may be forced to allocate sub-optimally based on the constraints defined by the restriction. If demand conditions change, a charity may be stuck with resources allocated to a sub-optimal use. Examples abound: museums may be forced to display artwork for which there is little interest; universities may be constrained to provide academic programs for which there are few students; or, relief agencies may be forced to supply aid when there are relatively few victims. Like foundations, nonprofits are typically organized around a relatively narrow set of objectives (AIDS research, homeless shelter, literacy, etc.).5 Nonprofits will often seek funding from foundations with similar objectives to their own. If those nonprofits could offer a credible commitment to allocate grant resources according to the preferences of the foundation, there would be little need for grant restrictions. But this commitment is difficult. The nature of most nonprofit activities makes monitoring their activities costly. In summary, both foundations and recipient nonprofits seek to optimally allocate resources in the pursuit of specific objectives. Ideally, this would imply that all grants would be made without exante restrictions to allow the full flexibility of resources. Yet, foundations face a costly monitoring problem. Exploring the 5 To receive tax-exempt status, an organization must indicate in its bylaws the general category of charitable activities it will pursue. The National Center of Charitable Statistics collects information and categorizes nonprofit organizations according to the National Taxonomy of Exempt Entities (NTEE).

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

47

Table 2 Unrestricted grant tabulations, by NTEE sector. By number of grants

By grant total (in $)

NTEE code

Unrestricted

Restricted

Total

Unrestricted

Restricted

Total

Education

111 17.5% 128 32.2% 70 23.5% 45 24.2% 49 42.2% 29 25.2% 17 16.0% 27 31.8% 16 24.2% 17 30.4% 9 16.4% 12 22.2% 19 38.0% 8 19.0% 7 20.0% 8 23.5% 4 11.8% 6 19.4% 6 23.1% 5 20.0% 7 30.4% 5 22.7% 2 16.7% 1 33.3% 1 33.3% 0 0.0% 601

523 82.5% 269 67.8% 228 76.5% 141 75.8% 67 57.8% 86 74.8% 89 84.0% 58 68.2% 50 75.8% 39 69.6% 46 83.6% 42 77.8% 31 62.0% 34 81.0% 28 80.0% 26 76.5% 30 88.2% 25 80.6% 20 76.9% 20 80.0% 16 69.6% 17 77.3% 10 83.3% 2 66.7% 2 66.7% 2 100.0% 1909

634

128,000,000 91.4% 32,300,000 74.3% 23,000,000 89.4% 27,100,000 87.8% 9,191,704 76.4% 5,973,651 82.3% 12,000,000 88.6% 5,342,385 68.1% 2,660,859 87.7% 31,200,000 94.4% 4,947,751 81.3% 3,025,506 83.0% 1,285,978 53.9% 13,900,000 95.0% 5,208,750 94.0% 1,551,291 85.3% 3,818,425 95.5% 2,084,295 93.5% 1,558,888 93.5% 1,178,000 94.2% 614,352 80.4% 2,689,262 86.0% 7,997,557 99.5% 78,850 61.2% 50,000 76.9% 31,500

140,000,000

2

12,000,000 9% 11,200,000 26% 2,727,260 11% 3,755,199 12% 2,842,423 24% 1,286,000 18% 1,546,037 11% 2,503,040 32% 372,500 12% 1,837,000 6% 1,135,000 19% 619,734 17% 1,100,000 46% 731,666 5% 335,000 6% 266,741 15% 180,000 5% 145,244 7% 107,500 6% 73,000 6% 149,850 20% 439,000 14% 43,333 1% 50,000 39% 15,000 23% 0

31,500

2510

45,460,527

326,789,004

372,249,531

Arts & culture Human services Health care Philanthropy Community imp. Environment Youth development Housing & shelter International & fn. aff. Crime & legal Mental health Civil rights Social science Public benefit Medical research Recreation & sports Employment Animal related Food & agriculture Religion-related Medical research Science & tech Unknown Public safety Mutual benefit Total

397 298 186 116 115 106 85 66 56 55 54 50 42 35 34 34 31 26 25 23 22 12 3 3

43,500,000 25,727,260 30,855,199 12,034,127 7,259,651 13,546,037 7,845,425 3,033,359 33,037,000 6,082,751 3,645,240 2,385,978 14,631,666 5,543,750 1,818,032 3,998,425 2,229,539 1,666,388 1,251,000 764,202 3,128,262 8,040,890 128,850 65,000

Ranked by total grant dollars.

conditions under which foundations choose to offer restricted or unrestricted grants is the subject of the next section. 4. The data The Foundation Center maintains a proprietary database which includes all grants of $10,000 or more awarded by non-operating public and private foundations to nonprofit organizations.6 Grants to 6 For community foundations, only discretionary grants are included. Typically these are grants where community foundation staff has a role in defining the grants. It does exclude grants from restricted funds, typically where grants are designated by the donor for the same organization for the same purpose year after year. In the circa 2000 sample, community foundations represent only 5.9% of foundations and 5.4% of total grant dollars.

individuals are not included in the file. An initial random sample of grants was drawn from the Foundation Center's Grants Sample Database.7 Table 1 summarizes the sample drawn from the Foundation Center database. The initial sample includes 3000 unique grants (circa year 2000) distributed by 603 foundations to 2531 recipient nonprofit organizations. These grants are classified by the Foundation Center according to their

7 Similar to other empirical studies on the topic, the basic unit of analysis is the contract (or grant in this case). In the Foundation Center data, the grant can only be observed at the time of origination. Information on subsequent renegotiations of the grant stipulations is not observable.

48

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

funding type.8 From this classification it is possible to identify whether the grant was restricted to a particular program or allowed to be spent on general operating funds. Approximately 14% of grant monies and 28% of the total number of grants within the sample went to unrestricted operating support. Summary statistics from within the sample are consistent with descriptive statistics for the population as a whole. Each grant was subsequently linked with financial data on the recipient nonprofit organization as well as the donor foundation. The Form 990 (990PF for foundations) is the primary financial reporting device for charitable nonprofits. For this portion of the study, nonprofit financials were drawn from the National Center for Charitable Statistics.9 During the data merge, 298 nonprofit organizations (353 grant contracts) were dropped from the sample because they could not be matched to NCCS Form 990 data. It is likely that these grants were either made to organizations not required to file a public Form 990 (such is the case with religious congregations or very small nonprofits) or that the codes were incorrectly documented in the grant files.10 An additional 42 organizations (137 grant contracts) were case-wise deleted because of either missing or implausible values in their financial data.11 These omissions reduced the sample size by 17%. In the resulting sample of 2112 nonprofit organizations and their associated financial information were matched with 2510 grants received from 556 charitable foundations. The remaining organizations represent a wide cross-section of nonprofit types, from art museums to foreign aid. By identifying organizations based on their charitable purpose, it is possible to investigate variation in foundation giving across nonprofit type. Table 2 summarizes the types of nonprofit organizations within the sample as described by the National Taxonomy of Exempt Entities (NTEE).12 The table includes a tabulation of organizations that have received unrestricted grants versus those whose grants were restricted. NTEE sectors are ranked by total dollar value of grants received. Ignoring those sectors with only a few grant observations, sectors with relatively high unrestricted proportions include: Philanthropy (42%), Civil Rights (38%), Arts & Culture (32%), Youth Development (32%), Religion-related (30%), as well as International and Foreign Affairs (30%). In contrast, those sectors with the lowest unrestricted proportions of grants are: Recreation and Sports (12%), Environmental (16%), and Crime & Legal (16%). Examining the fraction of grant dollars which remain unrestricted reveals a slightly different rank-ordering. Ignoring the sectors which 8 The Foundation Center uses a detailed classification system to characterize the nature of the grant. This paper adopts this system which separates unrestricted grants from those that place a variety of different restrictions on the use of the grant. The Foundation Center assigns a two-digit code which designates the type of funding provided by the foundation. For this study, a grant is classified as unrestricted if it receives a “General/Operating support” classification by the foundation center. The Foundation Center defines General/Operating Support as, “grants for the day-to-day operating costs of an existing program or organization or to further the general purpose or work of an organization; also called unrestricted grants.” A detailed description of how the classifications are made is available at http://foundationcenter. org/gainknowledge/grantsclass/how.html. 9 GuideStar-NCCS National Nonprofit Research Database, ver. 1 [2000], http:// nccsdataweb.urban.org (2002). 10 With two exceptions, all tax-exempt nonprofit organizations are required to file IRS Form 990. Nonprofit organizations with less than $25,000 in revenue as well as religious congregations are not required to file Form 990. The IRS keeps no data on these firms. 11 Form 990 financial data is self-reported, consequently, it is prone to certain types of reporting errors. For more information on this issue, see (Krishnan et al., 2006; Khumawala et al., 2007). The following measures were taken to clean the data. Values less than or greater than 1 were considered reporting errors removed for TASSET and GRANT_INTENSITY. Similarly negative or (non-zero) missing values for TOTREV were deleted from the sample. Missing observations for HHI_PROG and AGE were deleted. In all, 42 observations were removed. 12 The National Center for Charitable Statistics has classified all nonprofit organizations according to the National Taxonomy of Exempt Entities (NTEE). The NTEE operates similarly to NAICS codes in that groups of firms can be categorized according to the desired level of aggregation. The broadest category within the NTEE breaks the population of nonprofit firms into twenty-six broad sub-sectors.

Table 3 Summary of variables. Description

Variable

Expected sign

Restricted grant

RESTRICT

Dependent

Proxies for contractibility Objective match Tangible asset MSA match Grant intensity

OBJ_MATCH TASSET MSA_MATCH GRANT_INT

+ + + +

Proxies for opportunism % Unrestricted fund balance Program concentration State legal index

UNRES HHI_PROG LEGAL

+ − −

Firm controls Age of the firm Total revenue

AGE TOTREV

N/A N/A

only have a limited number of grants, the sectors with high unrestricted proportions are: Civil Rights (46%), Youth Development (32%), and Arts & Culture (26%). Sectors with the low unrestricted proportions include: Recreation & Sports (5%), Employment (7%), Animal Related (6%), and International & Foreign Affairs (6%). 4.1. Description of the variables The form (governance structure) of the grant is the dependent variable of interest. The Foundation Center classifies each grant according to a detailed taxonomy of purposes, ranging from general operating support to program specific grants.13 The range of contractual options is simplified into a binomial variable. Using Foundation Center classifications, the variable (RESRICT) equals unity if the grant has been allocated with any material restrictions; zero otherwise. Similar to previous empirical work, contract form is regressed on a set of variables describing the principal, the agent, and the contracting environment. When considering grant restrictions, foundations must weigh both the full cost of the grant restriction and the propensity for the nonprofit to pursue its own sub-goals. Within this context, contract completeness should increase with factors that promote contractibility (by reducing monitoring & enforcement costs). Similarly, contract completeness should rise with opportunities for recipient ex-post opportunism. This section proposes a series of empirical proxies for contractibility and ex-post opportunism between charitable foundations and recipient nonprofits. The proxies, along with their expected signs, are displayed in Table 3. 4.1.1. Proxies for contractibility Nonprofits typically have good information about the preferences of foundations because they express those preferences in their own publications, websites, and in grant databases. Likewise, foundations have access to information about the objectives of nonprofits from similar sources. By focusing on a narrow set of objectives (i.e., global health, education, or medical research) foundations can gain expertise which lowers the cost of evaluating and monitoring grant applicants. If foundations share common objectives with the recipient nonprofit, then they will be able to monitor outcomes with less cost. The activities of charitable foundations are classified by the Foundation Center according to the National Taxonomy of Exempt 13 A more expansive definition of unrestricted grants was considered. Including all grants that carried a (1) as a first Grant Code digit added an additional 116 grants to the unrestricted category. This inclusion made little difference in the resulting estimates yet created a discrepancy between our definition of unrestricted grants and that of the Foundation Center. Consequently, the more restrictive definition was used.

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

Entities (NTEE), a classification system analogous to for-profit NAICS/ SIC codes. Nonprofit organizations are also classified according to the NTEE by the IRS. The variable OBJ_MATCH equals unity if the objectives (as classified under the NTEE) of the foundation match that of the recipient nonprofit.14 Under a TCE framework, foundations should be more likely to impose grant restrictions when its objectives match those of the charity, because transaction costs (due to common specialization) are lower. Next, while researchers have recognized that the nonprofit organizational form tends to dominate when information asymmetries are acute, they also note that nonprofits vary in the degree to which their services can be monitored by constituents (Steinberg, 2006). Previous research has attempted to characterize this variation in several ways. Weisbrod & Schlesinger (1986) characterize nursing homes by those traits which are easily observed (type 1) versus those which are not (type 2). Type 1 characteristics may include physical traits such as room size, while type 2 characteristics represent less tangible characteristics such as compassionate care or competence. Castaneda et al. (2008) instead examine the characteristics of nonprofit inputs. They use variation in the observability of firm assets to assess the influence contract failure on donor competition. This analysis more closely follows the latter by proposing that foundations typically place grant restrictions related to firm expenditures on inputs. Nonprofit firms vary in the degree to which they use inputs which are observable and verifiable. When those inputs are physically tangible assets such as buildings, equipment, or collections, they will be more observable (and verifiable) relative to “soft” inputs such as labor or financial capital. For example, an incremental donation to purchase an art collection, build a building, or purchase medical equipment is readily observable, making contracting on the specific characteristics of these expenditures relatively simple. In contrast, an incremental donation to expand the operations of a policy center or international aid organization is less observable and relatively costly to monitor. The variable TASSET represents the fraction of total assets which are physically tangible, indicating the relative importance of physically observable assets in the production process.15 If an increase in tangible assets improves contractibility, thereby lowering contracting costs from the perspective of the foundation, grant restrictions should become more likely. Site proximity should also be an important contributor to low cost contracting. Some foundations have a propensity to give locally. Grantors typically have local knowledge that makes oversight less costly relative to those given nationally or internationally. The dummy variable (MSA_MATCH) indicates that the foundation and the recipient nonprofit share a common Metropolitan Statistical Area. It is expected that if the foundation and the recipient nonprofit share the same MSA, then the probability of restrictions on the grant should increase. Finally, the size of the grant, relative to other donation streams, will influence the firm's ability to redirect funds away from the objectives of the grant. A grant which represents only a trivial fraction of donation revenues is typically more fungible with other revenue streams. However, large grants are more difficult to redirect because they must be replaced with significant revenues from other sources. To capture this effect, grant intensity (GRANT_INT) represents the

14 Nonprofit organizations are classified into the NTEE by “IRS determination Specialists” using descriptive information provided by the entity. The Foundation Center also classifies foundation objectives according to the NTEE, but allow up to three classifications for each foundation, beginning with their primary area of expertise. The Foundation Center uses the most broad “major groups” classification of the NTEE which breaks the sector down into 26 sub-categories, ranging from arts & culture to religion-related. In our case OBJ_MATCH = 1 if the primary area of expertise (the first grant code) matches the classification of the recipient nonprofit. For detailed information on the classification system see http://nccs.urban.org/classification/NTEE. cfm. 15 TASSET is constructed by dividing depreciated physical assets (Line 57b) by total assets (Line 59) from the Form 990.

49

fraction of total donation revenues accounted for by the observed grant. Greater grant intensities will lower the cost of monitoring, making restrictions more likely. 4.1.2. Proxies for moral hazard It is unlikely that the objectives of the nonprofit manager will precisely align with those of the foundation. While a basic premise of the nonprofit organizational form is to mitigate moral hazard, ample opportunities remain. Managers may engage in self-dealing through perquisite consumption, or simply employ resources to projects considered sub-optimal by the foundation. If foundations believe that ex-post opportunism is likely, formal grant restrictions will allow recourse. In the most extreme case nonprofits may bring legal action against the nonprofit if funds are not allocated according to the grant conditions.16 More commonly, nonprofit managers are careful to follow the spirit of grant agreements so as to not jeopardize future funding requests or tarnish their reputation with other funders. This section describes a set of parameters which proxy for an increased probability, not necessarily the existence, of manager opportunism. First, many organizations, such as homeless shelters, pursue a single activity. By contrast, the Salvation Army will engage in hundreds of relatively dissimilar programs. As the number of program activities increase, monitoring and preventing cross-subsidization of programs becomes increasingly difficult. PROGRAM_HHI is an authorconstructed Herfindahl–Hirschman Index (HHI) of program activities.17 Nonprofits which report only a single program activity will have a PROGRAM_HHI equal to one, while firms with diffuse programs will have HHI values approaching zero. Consequently, PROGRAM_HHI will increase as the firm becomes less complex (fewer programs). An inverse relationship is expected between the probability of grant restrictions and the relative simplicity of the nonprofit's activities (high PROGRAM_HHI). Second, previous donations which were given with material restrictions are reported as restricted fund balances on the Form 990. These restrictions may be programmatic or endowment. As restricted fund balances increase relative to the total funds available, nonprofit managers have diminished ability to reallocate resources. In this sense the direct impact of a grant restriction is strategic, where the influence of a particular restriction is contingent upon the actions of other donors. Bilodeau (1992) explores this type of situation theoretically in the context of united charities. For the nonprofit manager, a grant restriction reduces the pool of discretionary resources which can be diverted to uses outside the objectives of the donor. As this pool shrinks so does the range of sub-goals which could be pursued with relative ease by the nonprofit manager. The fraction of total fund balances which are unrestricted are calculated and presented as UNRES.18 A high proportion of unrestricted fund balances should increase budget fungibility, thereby 16 Grant restrictions typically fall under common contract law. However, legal enforcement against nonprofit organizations has, historically, been the domain of the state attorney general (Freemont-Smith, Governing nonprofit organizations: federal and state law and regulation, 2004). More recently, donors have been increasingly willing to litigate directly against charities if nonprofits do not adhere to the original intent of their gifts (Blum, 2002). 17 Part III of the Form 990 requests a description of each program service accomplishment by the nonprofit over the past year. The form also requests the budget allocation to that program. The program lists, along with its associated budget, are used to create a program concentration ratio which captures both the number and budget concentration in a single metric. Recall that the typical way to calculate the Herfindahl– Hirschman Index is the sum of the squared market shares (PROGRAM_HHI = Σα2) where α is the fraction of total expenditures allocated to a particular program). In this case “market share” represents the fraction of total program expenditures allocated to a particular program, as defined in Part III of the Form 990. A single program would have a value of 1. An equally distributed duopoly would have a HHI of .5, and so on as the project allocations become more diffuse. 18 Part IV of the Form 990 classifies fund balances as restricted, temporarily restricted, or unrestricted. We calculate UNRES as the fraction of total fund balances which are unrestricted.

50

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

Table 4 Sample summary statistics. Description

Variable

Restricted grant RESTRICT Proxies for contractibility Objective OBJ_MATCH match Tangible asset TASSET MSA match MSA_MATCH Grant intensity GRANT_INT Proxies for opportunistic behavior State legal LEGAL index HHI_PROG Program concentration index % unrestricted UNRES fund balance Age of the firm AGE Total revenue TOTREV

N

Std. dev.

Min

2510 0.76

Mean

0.42

0.00 1.00

Max

2510 0.41

0.49

0.00 1.00

2510 0.37 2510 0.31 2510 0.30

0.87 0.46 0.40

0.00 1.00 0.00 1.00 0.00 1.00

2510 6.08

1.31

3.00 8.00

2510 0.71

0.30

0.00 1.00

whose objectives match the donor foundation. Thirty-seven percent of assets within sample firms are considered “tangible”. Thirty-one percent of firms receive funds from foundations in the same MSA. The average grant represents approximately 30% of revenues. The typical firm in the sample has two-thirds of its financial revenues at its disposal as unrestricted funds and reports relatively few (less than two) programs on its Form 990. Firms in there sample average 34 years of age and generate $35 million in total revenues. 5. The cross-sectional model The base model is estimated for grant i given to a nonprofit in NTEE sector j. Gij = α1 Χij + α2 Ζij + α3 Vi + α4 Ιj + εij

2510 0.67

0.41

0.00 1.00

2510 34 20 6 2510 35,100,000 106,000,000 0

87 994,000,000

For the original sample of grant contracts, 298 nonprofit organizations (353 grant contracts) were dropped from the sample because they could not be matched to NCCS Form 990 data. It is likely that these grants were made to organizations not required to file a public Form 990 (such is the case with religious congregations or very small nonprofits) or that the codes were incorrectly documented in the grant files. An additional 42 organizations (137 grant contracts) were case-wise deleted because of either missing or implausible values in their financial data. These omissions reduced the sample size by 17%.

increasing the probability manager opportunism. Consequently, it is expected that a foundation will be more likely to place material restrictions on its own grant when high levels of unrestricted funds are observed. Finally, even if opportunistic behaviors can be detected, the legal environment constraining the behaviors of nonprofit managers varies substantially across states. States such as New York maintain relatively robust regulatory frameworks where charities are required to file regular reports with the state's attorney general. Foundations may perceive state oversight as a complement to their own monitoring efforts, reducing the propensity for a manager to pursue opportunistic behaviors. Yetman & Yetman (2004) developed an index of state governance measures which characterize the legal environment regarding nonprofit activities. The index is a simple linear combination of dummy variables such as whether the state legally distinguishes nonprofit firms, or whether the State Attorney General must be notified of lawsuits.19 This index (LEGAL) is applied in this paper to control for variation in the legal environment across states. It is expected that the probability of observing a grant restriction will decline as the legal environment within the nonprofit's state becomes more stringent. To round-out the model, the age of the firm (AGE) and the size of the firm, measured by total revenues (TOTREV), were included as basic firm controls.

Where Gij is a binary variable indicating whether the grant is unrestricted (G = 0) or restricted (G = 1). Xij represents a vector of proxies for contractibility, including: OBJ_MATCH, MSA_MATCH, TASSET, and GRANT_INT. By reducing contracting costs, these factors are expected to increase the probability of observing grant restrictions. Zij represents a vector of proxies that represent increases in the potential for the manager to pursue sub-goals inconsistent with the desires of the firm. These variables include PROG_HHI, LEGAL, and UNRES. V is the set of firm controls. Ij is a vector of dummy variables defining each of the 26 NTEE sub-sectors, described in Table 2, to control for industry specific effects. To account for the binary-dependent variable, the model is estimated using Logit. Table 5 reports the change in the odds ratio with a unit increase in the independent variables. The model is estimated both with and without NTEE sub-sector fixed effects.

Table 5 Regression output. Logistic regression Objective match Tangible asset MSA match Grant intensity Legal Program HHI % unrestricted fund balance Firm age Total revenue

4.2. Descriptive statistics Table 4 offers summary statistics for the relevant variables. The table demonstrates that three-fourths of grants within the sample are offered with material restrictions, implying that only a minority of grants are unrestricted. Forty-one percent of grants go to nonprofits 19 The Yetman & Yetman (2004) legal governance index includes a sequence of indicator values equal to unity if the following occur: 1) Does the state statute distinguish nonprofits from for-profits? 2) Are liquidating distributions restricted to other nonprofits only? 3) Must the Attorney General be notified of asset sales? 4) Are there limitations on re-incorporating as a for-profit corporation? 5) Must the Attorney General be notified of any suits involving charities? 6) Do parties other than Attorney General have standing to bring legal actions? 7) Do the courts have Cy pres authority? 8) Is the Attorney General the enforcing authority?

Sector fixed effects† N Pseudo R2

Odds ratio

Odds ratio

1.35⁎⁎ (2.57) 1.20⁎ (1.89) 1.01 (0.06) 2.15⁎⁎ (3.16) 1.01 (0.66) .81 (− 1.43) 1.35⁎ (1.93) 0.98⁎ (− 1.93) 1.00⁎⁎ (2.23) N 2510 .03

1.44⁎⁎ (2.60) 1.19⁎ (1.84) 1.02 (0.15) 2.05⁎⁎ (2.85) 1.01 (0.01) .87 (− 0.93) 1.39⁎⁎ (2.07) 1.01 (0.04) 1.00⁎⁎ (2.00) Y 2510 .04

Table displays logistic regression coefficients when an indicator variable for material restrictions on a grant (RESTRICT) is regressed on a set of proxies for contractibility and opportunism. The table reports odds-ratios for the Logit model. The right hand column includes NTEE sector dummies in the regression. Notes: Odds-ratios are reported. t-stats are in parentheses. ⁎⁎ and ⁎ denote statistical significance at the 5 and 10% level, respectively. † sector dummies based on NTEE classifications have been suppressed, but are presented in Table 6. For logistical regressions, pseudo R2 are reported, as calculated by STATA. Robust standard errors were clustered by grant-making foundation.

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

5.1. Cross-section results

Table 7 Summary of psuedo panel dataset.

Taking the coefficients in the order described previously, the odds of observing a restricted grant rise by 35% when the stated objectives of the foundation match the objectives of the recipient nonprofit. This finding is statistically significant and robust across specifications. Increases in tangible assets also have a positive and statistically significant influence. A 10 percentage point increase in tangible assets is associated with a 2 percentage point increase in the odds of receiving a restricted grant. As assets become more tangible (hence observable), foundations are better able to construct meaningful restrictions on their use. Another significant influence comes from the fraction of the nonprofit's budget accounted for by the grant (Grant Intensity). A one percent increase in grant intensity would result in a two percentage point increase in the odds of receiving a restricted grant. Geographic proximity to the recipient nonprofit does not appear to have a statistically significant impact on gift restrictions. Results for those variables which are consistent with an increase in the potential for moral hazard offer additional support for the TCE framework. Increasing the concentration of program service expenditures (PROG_HHI) by ten percent, results in the odds of observing grant restrictions falling by 2%. The result is not statistically significant at normal levels of alpha, but the sign and magnitude are consistent with our previous hypothesis. The presence of unrestricted funds appears to result in meaningful increases in subsequent grant restrictions. A 10% increase in unrestricted fund balances is associated with a roughly 4% increase in the probability of a grant restriction. The metric for state legal oversight over nonprofits is not statistically associated with a change in grant restrictions. The previous analysis lumped all nonprofit organizations together into a single regression. However, it is likely that both nonprofit firms and foundation behavior vary substantially across nonprofit industries. Table 6 presents coefficients of the sub-sector fixed effects based on the broadest NTEE classification system. It should be noted that several subsectors contain an insufficient number of observations to estimate the coefficient. Others should be examined with caution for the same reason. The majority of nonprofit sub-sectors demonstrated increased odds of restricted grants relative to Arts organizations (the reference Table 6 Sub-sector fixed effects.

B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Education Environment Animal related Healthcare Mental health/crisis Diseases & disorders Medical research Crime & legal related Employment Food & agriculture Housing & shelter Public safety Recreation & sports Youth development Human services International Civil rights & advocacy Community imp. Philanthropy Science & technology Social science Public benefit Religion-related Mutual membership Unknown

51

Odds ratio

t statistic

N

% of firms

1.69⁎⁎ 2.68⁎⁎ 1.79 1.53⁎ 2.09⁎⁎ 1.90 2.02 3.35⁎⁎ 2.96⁎⁎ 2.45⁎ 1.70 0.73 4.64⁎⁎ 1.18 1.47⁎ 1.28 0.99 1.58⁎ 0.77 2.71 2.84⁎⁎ 2.44⁎ 1.24 N/A N/A

2.77 3.16 1.06 1.74 2.06 1.5 1.33 2.81 2.12 1.67 1.55 − 0.24 2.76 0.55 1.95 0.76 − 0.02 1.64 − 1.1 1.23 2.47 1.79 0.43 N/A N/A

634 106 26 186 54 34 22 55 31 25 66 3 34 85 298 56 50 115 116 12 42 35 23 2 3

25.26 4.22 1.04 7.41 2.15 1.35 0.88 2.19 1.24 1 2.63 0.12 1.35 3.39 11.87 2.23 1.99 4.58 4.62 0.48 1.67 1.39 0.92 0.08 0.12

This table presents industry sub-sector effects of the regression model presented in Table 5. Arts organizations & museums (sector A) is the base case, containing 397 firms. ⁎⁎ and ⁎ denote statistical significance at the 5 and 10% level, respectively.

Restricted grant Objective match Tangible asset MSA match Grant intensity Program HHI Legal % unrestricted fund balance Firm age Total revenue

1998

1999

Additional draw

Additional draw

2000 Original sample

Mean

N†

Mean

N

Mean

N

0.79 0.36 0.51 0.31 0.30 0.60 6.10 0.35

1273 1273 1273 1273 1273 1273 1273 1273

0.80 0.34 0.49 0.34 0.30 0..69 6.10 0.31

1320 1320 1320 1320 1320 1320 1320 1500

0.76 0.41 0.47 0.31 0.30 0.71 6.09 0.33

2590 2590 2590 2590 2590 2590 2590 2590

34.33 3.08E + 07

1327 1273

33.49 3.66E + 07

1320 1320

34.00 3.51E + 07

2590 2590

The original cross-section was augmented with two additional random draws of 1500 rant contracts each from years 1998 and 1999. The resulting unbalanced panel contains 832 unique foundations offering grants to 3768 unique nonprofit firms. Basic summary statistics are offered below. † The additional random draws included 1500 observations each. However, 227 grant contract were deleted from 1998 and 180 from 1999 due to either missing financial data or the EIN number could not be paired with the NCCS–Guidestar database.

case). Recreation & Sports (N) along with Crime & Legal organizations (I) represented the most dramatic, and statistically significant, increase in the odds of a restricted grant relative to the base case. In fact, no sub-sector demonstrated a statistically significant decrease in the odds of a restricted grant, relative to arts organizations. In summary, empirical results for a random cross-section of grant contracts lend support to a TCE framework for understanding funding relationships between foundations and nonprofits. Generally, increases in the empirical proxies for contractibility correlate with a higher probability of grant restrictions. The most significant factors impacting the probability of grant restrictions include: the alignment of objectives between the foundation and nonprofit (OBJ_MATCH), the fraction of assets which are tangible (TASSET), the financial impact of the grant on revenues (GRANT_INT), and the mix of restricted/ unrestricted funds available to the manager (UNRES). These findings are consistent with the predictions of a transaction costs approach to understanding restricted grants. Subsequent tests for influential observations and problems with functional form did not reveal significant problems with the original specification. However, there remains the important issue of the potential for endogenous matching between contract participants. It is possible that unobserved characteristics such as management competency or founder preference will influence the propensity for a foundation to offer, or a nonprofit to accept, a restricted grant. Ackerberg & Botticini (2002) suggest the construction of panel data to identify unobserved heterogeneity that is consistent for agents across multiple contracts. Because of how the data is recorded, it is not possible to observe individual grant contracts over time. Instead, the original data is expanded to include additional random draws over multiple years. This process creates a pseudo panel of repeat random draws which can be used to identify unobserved foundation characteristics in the sample. 6. Pseudo panel (repeated cross-section) This section addresses concerns regarding unobserved heterogeneity among foundations that may influence endogenous matching among foundations and nonprofits. Particular attention is paid to unobserved characteristics of foundations, such as a preference for restricted grants. Recall that the unit of observation, the grant contract, is observed only at the time the grant is given. Yet, within any given year, as well as across years, unique foundations are observed issuing multiple grants to different nonprofits. Similarly, unique nonprofits are observed receiving grants from multiple

52

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

foundations. Distinct foundation/nonprofit pairings occur repeatedly, both within and across years. To the extent that foundations are consistent in their preferences over multiple grants, it is possible to identify foundation-specific characteristics influencing the grant process. The original dataset is expanded to include two additional random draws of 1500 grant contracts (one each from 1998 and 1999) from the Foundation Center database. As with the original sample, these grants were matched with their associated nonprofit and foundation financial data from the Guidestar–NCCS database. Adding these draws to the original sample creates a multi-year, repeated cross-section of 6000 randomly selected grant contracts. Again it was necessary to remove observations that either could not be matched with nonprofit financial data or had missing values in the merged data. From the two additional draws, 227 observations for 1998 and 180 observations for 1999 were case-wise deleted. The resulting expanded sample of 5183 grant contracts includes 832 distinct foundations offering grants to 3678 unique nonprofits observed over three years. The average foundation in the sample distributes 7.2 grants to nonprofits within the sample. The average nonprofit receives 2.4 grants from foundations within the sample. Table 7 offers a summary of the expanded dataset. 6.2. Pseudo panel results Both random effects and fixed effects estimators were used to test the influence of foundation effects. The results are largely similar between the two. The FE coefficients tend to be smaller in magnitude than the RE estimates. Only Grant Intensity and Program HHI differ in statistical significance when NTEE sub-sector effects are included.

Table 8 Pseudo panel results with foundation random and fixed effects. Logistic regression RE Objective match Tangible asset MSA match Grant intensity Legal Program HHI % unrestricted fund balance Firm age Total revenue NTEE sub-sector fixed effects† N Log likelihood

Odds ratio 1.30⁎⁎ (2.61) 1.46⁎⁎ (2.67) .88 (− 1.02) 1.55⁎⁎ (2.33) 1.01 (.13) 0.71⁎⁎ (− 2.09) 1.39⁎⁎ (2.28) 1.00 (1.55) 1.00⁎⁎ (4.07) N 5183 − 2041

FE

RE

FE

1.26⁎⁎ (2.12) 1.37⁎⁎ (2.25) .87 (−.96) 1.41⁎ (1.76) 1.00 (.73) 0.71⁎⁎ (− 1.97) 1.44⁎⁎ (2.31) 1.00 (1.55) 1.00⁎⁎ (3.57) N 2862 − 1015

Odds ratio 1.19 (1.34) 1.49⁎⁎ (2.79) .94 (− 0.45) 1.41⁎ (1.75) 1.01 (0.28) 0.77⁎ (1.65) 1.37⁎⁎ (2.00) 1.00⁎ (1.67) 1.00⁎⁎ (3.41) Y 5183 − 2009

1.12 (.82) 1.38⁎⁎ (2.17) .91 (− 0.60) 1.29 (1.26) 1.03 (0.74) 0.80 (1.26) 1.44⁎⁎ (2.19) 1.00⁎ (1.93) 1.00⁎⁎ (3.21) Y 2862 − 954

Using the expanded dataset described in Table 8, the model was re-estimated controlling for foundation heterogeneity by including foundation effects. In the expanded sample we observe 4652 grant contracts issues by 832 unique foundations. Each of these foundation offers (on average) 7.2 grants to different nonprofits within the sample. We estimate the model both with and without NTEE sub-sector dummies. Both Random Effects (RE) and Fixed Effects (FE) models are reported. A Hausman test indicates that the Fixed Effects estimator may be more consistent, but deletes a significant number of observations due to insufficient variation in the dependent variable. Notes: Odds-ratios values are reported. t-stats are in parentheses. ⁎⁎ and ⁎ denote statistical significance at the 5 and 10% level, respectively. † Sector dummies based on NTEE classifications have been suppressed for presentation.

While a subsequent Hausman test indicates that the fixed effects estimation may be more efficient, it also drops a significant number of observations due to a lack of within foundation variation of the dependent variable. Consequently, both sets of estimates are reported in Table 8. The models are first estimated without NTEE sub-sector effects. Overall the pseudo panel findings support the cross-sectional results. Even when foundation effects are included, the sign, magnitude, and statistical significance of the coefficients are consistent with the cross-section. Increases in matching objectives, tangible assets, and grant intensity all result in positive, statistically significant, and economically meaningful increases in the odds of a grant restriction. This is consistent with the hypothesis that increases in contractibility (lower transaction costs) should result in more complete contracts. Similarly, increases in the proxies for nonprofit opportunism appear to result in increased grant restrictions. A 10% point increase in the dispersion of program expenditures (lower Program HHI) reduces the odds of observing a grant restriction by roughly three percentage points. A 10% point increase in unrestricted fund balances increases the odds of a grant restriction by four percentage points. As with the cross-sectional model, the legal environment (LEGAL) and geographic proximity (MSA Match) are not significant predictors of contract form. These results indicate that unobserved foundation heterogeneity does not significantly influence the use of grant restrictions. 7. Conclusions and limitations This paper examines the structure of grant contracts between charitable foundations and recipient nonprofit firms. While theoretical advances in understanding the role of incomplete contracts in nonprofit formation have been made, demonstrating empirical evidence has been more difficult. The empirical literature has bypassed the nonprofit sector, whose very existence is theoretically tied to incomplete contracts. The nonprofit context is unusual because it examines contractual information among nonprofit firms without direct profit maximization objectives. Furthermore, nonprofit institutions face idiosyncratic institutional constraints distinct from their forprofit counterparts. This paper lends empirical support to a transaction costs framework for interpreting variation in grant contracts. Foundations place material restrictions on the use of grants to mitigate opportunistic behaviors by nonprofit managers. These behaviors may include perquisite consumption or cross-subsidization of programming. However, grant restrictions are not costless to construct or enforce. Restrictions may force nonprofit managers to allocate gifts suboptimally as operating conditions change. I propose that foundations will economize on these transaction costs in a predicable way. While far from complete, this approach opens the door to a more systematic understanding of a wide variety of issues within the nonprofit sector including: nonprofit finance, firm growth, and incentive alignment within and among nonprofit agencies. Using proxies for contractibility, the paper finds that foundations indeed follow the predictions offered by a TCE framework. More restrictive contracts are offered as output becomes increasingly contractible or as the potential for ex-post opportunism rises. More importantly, the findings are robust across a broad spectrum of nonprofit types and empirical specifications. While empirical results match up reasonably well with the predictions of TCE, several limitations in this analysis should be noted. Given the current structure of the data offered by the Foundation Center, a true panel structure (where an individual grant is observed longitudinally) for the data is not viable. A true panel would allow for the observation of relationship-specific capital. Similarly, true panel data would allow for the possibility that grant contracts may be renegotiated. Currently, the contract (grant) is

J. Thornton / Int. J. Ind. Organ. 28 (2010) 44–53

observed only at the time of the gift while subsequent renegotiations of the gift are not. The analysis conceives of the case where foundations offer a take-it-or-leave-it contract. In fact, nonprofits or foundations could maintain bargaining power to negotiate or renegotiate the grant restrictions. The Foundation Center does not track the status or duration of the grant after the initial bequest. Should this data ever become available, a complete panel structure would be a natural extension to this work. Second, the dependent variable itself is an imperfect measure of contractibility. Not all restricted grants are completely constrained, nor are all grants for general operating support fully flexible. The fungible nature of grants categorized as “general operating support” makes them, for the purposes of this analysis, ex-post unrestricted. Discriminating among finer gradations of program restrictions on foundation grants would also be a fruitful avenue of future research. This study likely masks differences among the wide variety of charities operating within this market. It appears that many of the empirical predictions originally proposed by TCE in a for-profit context can be extended to the nonprofit sector. Given the limitations described above, the analysis should be considered an initial test, rather than conclusive evidence, of the concept's applicability. That TCE can be applied as a robust framework for understanding grantsmanship is good news for both foundations and nonprofits. There clearly exists both asset specificity and the potential for opportunism in the grant-making process. As such, the full elimination of grant restrictions is not a reasonable objective by either party. Indeed, the implication of applying a TCE framework is not the reduction of grant restrictions, but to increase their efficiency. The policy consequences of the approach are that improvements in transparency and more robust legal environments should enhance the donor's ability to offer more efficient restrictions. References Ackerberg, D., Botticini, M., 2002. Endogenous matching and the empirical determinants of contract form. The Journal of Political Economy 110 (3), 564–591. Aggarwal, R.M., 2007. Role of risk sharing and transaction costs in contract choice: theory and evidence from groundwater contracts. Journal of Economic Behavior and Organization 63, 475–4796. Alchian, Armen, Demsetz, H., 1972. Production, information costs, and economic organization. American Economic Review 62, 777–795. Arrow, K., 1963. Uncertainty and the welfare economics of medical care. American Economic Review 53, 941–973. Baker, G., Hubbard, T., 2003. Make or buy in trucking: asset ownership, job design and information. American Economic Review 93 (3), 551–572. Barzel, Y., 1982. Measurement cost and the organization of markets. Journal of Law and Economics 25 (1), 27–48.

53

Bilodeau, M., 1992. Voluntary contributions to united charities. Journal of Public Economics 48, 119–133. Blum, D.E., 2002. The ties that bind. Chronicle of philanthropy. Castaneda, M., Garen, J., Thornton, J., 2008. Competition, contractibility and the market for donors to nonprofits. Journal of Law, Economics, and Organization 24 (1), 215–246. Dreazen, Y., 2005. More Katrina woes: incidents of fraud at Red Cross centers. The Wall Street Journal B1 October 19. Fisman, R., Hubbard, G., 2005. Precautionary savings and the governance of nonprofit organizations. Journal of Public Economics 89 (11–12), 2231–2243. Freemont-Smith, M., 2004. Governing nonprofit organizations: federal and state law and regulation. Belknap Press, Caimbridge. Freemont-Smith, M., Kosaras, A., 2003. Wrongdoing by officers and directors of charities: a survey of press reports 1995–2002. Hauser Center for Nonprofit Organizations at Harvard University, Caimbridge. Glaeser, E., Shleifer, A., 2001. Not-for-profit entrepreneurs. Journal of Public Economics 81, 99–115. Hansmann, H., 1980. The role of the nonprofit enterprise. The Yale Law Journal 89, 1633–1638. Hopkins, B.R., 2003. The Law of Tax Exempt Organizations. John Wiley & Sons, Hoboken. Jensen, M., Meckling, W., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305–360. Kaplan, S., Stromberg, P., 2004. Characteristics, contracts, and actions: evidence from venture capitalist and analyses. Journal of Finance 59, 2173–2206. Khumawala, S.B., Kraut, M.A., Meade, J.A., Gordon, T., 2007. The quality and reliability of Form 990 data: are users being misled. Academy of Accounting and Financial Studies Journal 11, 27–50 Special Issue. Krashinsky, M., 1986. Transaction costs and a theory of the nonprofit organization. In: Rose-Ackerman, S. (Ed.), The economics of nonprofit institutions: studies in structure and policy, pp. 114–132. Oxford. Krehely, J., House, M., 2005. Not All Grants Are Created Equal: Why Nonprofits Need General Operating Support from Foundations. National Committee for Responsive Philanthropy, Washington DC. Krishnan, R., Yetman, M., Yetman, R., 2006. Expense misreporting in nonprofit organizations. The Accounting Review 81 (2), 399–420. Steinberg, R., 2006. Economic theories of nonprofit organizations. In: Powell, W., Steinberg, R. (Eds.), The Nonprofit Sector: A Research Handbook. Yale University Press, New Haven. Stigler, J., 1961. The economics of information. The Journal of Political Economy 69, 213. Weisbrod, B., Schlesinger, M., 1986. Public, private, nonprofit ownership and the response to asymmetric information: the case of nursing homes. In: RoseAckerman, S. (Ed.), The Economics of Nonprofit Institutions: Studies in Structure and Policy. Oxford University Press, New York, pp. 133–151. Whoriskey, P. (2002, Nov. 21). Report Says United Way Withheld $1.3 Million; Group's New Leaders Acting on U.S. Audit. The Washington Post , p. 1. Williamson, O., 1975. Markets and Hierarchies, Analysis and Antitrust Implications: A Study in the Economics of Internal Organization. The Free Press, New York. Williamson, O., 1985. The Economic Institutions of Capitalism. The Free Press, New York. Williamson, O., 2002. The theory of the firm as governance structure: from choice to contract. The Jounral of Economic Perspectives 16 (3), 171–195. Williamson, O., 1979. Transaction-cost economics: the governance of contractual relations. Journal of Law and Economics 22, 233–261. Yetman, M., Yetman, R., 2004. The effects of governance on the financial reporting quality of nonprofit organizations. Retrieved from SSRN Working Paper Series. http://ssrn.com/abstract=590961.