Acres for the Affluent: An Interactive Model of Nonprofit Resources and Demand Heterogeneity

Abstract

According to the theory of government failure, nonprofit organizations emerge when governments fail to provide goods or services to a public with heterogeneous demands. This study approaches this fundamental theory of the nonprofit sector from a pluralist political viewpoint, marrying the theory of government failure to resource-driven nonprofit arguments via updated modeling and measurement strategies. This article proposes a new, conditional demand heterogeneity hypothesis, wherein the effect of demand heterogeneity on the nonprofit sector increases in the presence of increasing resources: nonprofit service delivery is most likely when those experiencing government failure have access to resources. This article is the first to test supply and demand nonprofit arguments in tandem using an interactive modeling specification and the first to operationalize demand heterogeneity using policy-based measures. I employ a finite distributed lag model with an interactive term in a time series, cross-sectional analysis of public and nonprofit land conservation in the United States. I find that nonprofit land conservation increases when citizens express greater environmental concern but only in the presence of increasing disposable income. Examining nonprofit theory within the context of land conservation provides a comparable measure of government and nonprofit service provision, controls for the nature of the good provided by these institutions, and allows for a policy-driven measure of demand heterogeneity, improving upon previous studies’ employment of diversity-based proxy measures. The results advance our understanding of how to test and measure demand heterogeneity nonprofit arguments and suggest that access to resources conditions which interests find expression in nonprofit organizations.

Introduction

A key theory of the nonprofit sector is the theory of government failure: nonprofits provide services when governments fail to meet citizen demand for their preferred quality or quantity of service delivery. According to Weisbrod (1977, 1988), governments cater their services to the median voter (Downs 1957). Citizens whose preferences differ from those of the median voter establish nonprofits to satisfy their policy preferences. Ultimately, the theory of government failure theory can be summarized by the demand heterogeneity hypothesis: as the demand heterogeneity of an area increases, the size of the nonprofit sector in the area will increase.

Despite Weisbrod’s (1977) theory being over 40 years old, nonprofit scholars hold no consensus on its ability to explain the nonprofit sector. Tests of the demand heterogeneity hypothesis produce mixed results (Matsunaga and Yamauchi 2004). Some studies find demand heterogeneity increases the size of the nonprofit sector (e.g., Ben-Ner and Van Hoomissen 1992; Corbin 1999), while others find the opposite (e.g., Grønbjerg and Paarlberg 2001; Paarlberg and Gen 2009) or no relationship at all (e.g., Matsunaga, Yamauchi, and Okuyama 2010; Pryor 2012). This mixed support could be due to the various ways demand heterogeneity is measured (Matsunaga and Yamauchi 2004), the nature of the good being provided, or the ultimate misspecification of the theory (Paarlberg and Zuhlke 2019).

In this article, I argue that demand heterogeneity is a necessary but insufficient predictor of nonprofit activity. Demand for nonprofit organizations may exist, but demand may fail to translate into nonprofit organizations in the absence of resources. Conversely, resources may be available to start nonprofits, but nonprofits may fail to emerge if demand for their services is not present. A parallel line of nonprofit research argues that resources and other supply-side factors are the major driver of the nonprofit sector. However, few studies examine the effect of resources on the nonprofit sector in conjunction with demand-side factors like demand heterogeneity (e.g., Ben-Ner and Van Hoomissen 1991; Gazley, LaFontant, and Cheng 2020; Lecy and Van Slyke 2012; Paarl-berg and Gen 2009). Furthermore, no studies examine how supply-side factors may condition the effect of demand heterogeneity on the nonprofit sector using an interactive modeling strategy, which most accurately tests conditional relationships (Brambor, Clark, and Golder 2006).

I approach Weisbrod’s classic theory of government failure through the perspective of pluralist political theory—that politics is the unequal distribution of resources across groups—and the rich literature on supply-side nonprofit arguments to justify a new, conditional demand heterogeneity hypothesis: the effect of demand heterogeneity on the nonprofit sector is conditional on the resources available to those experiencing government failure. The nonprofit sector will be largest when those experiencing government failure have access to greater resources.

I test this conditional demand heterogeneity hypothesis using new measures of demand heterogeneity and nonprofit service outcomes. Previous mixed empirical support for the traditional demand heterogeneity hypothesis could stem from poor operationalization of the independent variable. Demand heterogeneity, as theorized by Weisbrod (1977, 1988), is diversity of policy preferences. However, no nonprofit studies to date operationalize demand heterogeneity as such. Rather, studies proxy for policy preferences using diversity variables (e.g., race, education, religion, age, language). While diversity-based proxies may correlate with policy preferences, they are not direct measures of policy preferences. More concerning, if proxy measures correlate with theoretically important omitted variables like resources, then proxies may produce biased estimates thereby driving the literature’s mixed results. Often, different proxies produce conflicting inferences within the same demand heterogeneity test (e.g., Bae and Sohn 2018; Chang and Tuckman 1996; Kim 2015; Liu 2017; Marcuello 1998).

Similarly, very few existing studies examining government–nonprofit relations measure the nonprofit sector in terms of nonprofit outcomes (e.g., Grant and Langpap 2019). Governments respond to market failure by producing goods and services or financing their production via payments or incentives (Steinberg 2006; Weisbrod 1988, 20). Most studies analyze nonprofit outputs like nonprofit expenditures or measures of the size of the nonprofit sector like the total number of nonprofits. Fewer studies test the demand heterogeneity hypothesis against nonprofit outcomes. This trend is unsurprising, given the difficulty of measuring nonprofit service provision. Weisbrod (1989, 244) suggests this measurement problem is inherent to the sector, discussing the difficulty in measuring nonprofit performance like “‘tender, loving care’ in nursing homes, and ‘good judgement’ in the care of prisoners.” However, failing to operationalize nonprofit outcomes overlooks testing nonprofit theory on a fundamental component of the nonprofit sector. Furthermore, examining nonprofit outcomes has important consequences for understanding equity within the sector (Gazley, LaFontant, and Cheng 2020).

The contributions of this article are threefold: the theoretical contribution of the conditional demand heterogeneity hypothesis, the methodological recommendation to replace problematic diversity-based proxies with policy-based measures, and the identification of US land conservation as a case for analyzing nonprofit outcomes. Government and nonprofits play a critical role in US land conservation, in part through conservation easements (Doremus 2003). Protected land provides a valid and reliable measure of nonprofit outcomes, comparable to government service provision.

This study argues both demand heterogeneity and resource availability are necessary for interests to find expression within the nonprofit sector, and advances our understanding of the demand and supply mechanisms behind nonprofit outcomes. In doing so, the paper contributes to the long-standing nonprofit debate between supply- and demand-side arguments—with updated modeling, the dichotomy between these arguments fades. Theoretically, this research updates Weisbrod’s (1988) traditional demand heterogeneity hypothesis by proposing a new, conditional demand heterogeneity hypothesis that tests supply-side and demand-side factors in tandem using an interactive modeling specification. Empirically, this study updates our tools for testing government–nonprofit theory. This article is the first to measure policy demand using a policy-based measure of demand heterogeneity rather than diversity-based proxies. This article is also one of the first to measure nonprofit activity using measures of nonprofit outcomes. I use a novel measure of nonprofit service delivery: acres of land conserved by nonprofits within US states. These new measurement strategies address previous issues identified within the literature, allowing for an updated test of the demand heterogeneity hypothesis.

A Conditional Theoretical Specification

Weisbrod (1997, 542, italics added) explains the demand heterogeneity hypothesis as:

When government provides these services in forms and amounts that voters want, there will be little role for nonprofits. However, when populations are very diverse, services that satisfy the majority may leave many people severely undersatisfied; nonprofits are thus understandable as an alternative mechanism for providing collective services.

Weisbrod (1977, 1988) argues that undersatisfied individuals fulfill their preferences via the nonprofit sector. However, tests of the demand heterogeneity hypothesis produce mixed results (Matsunaga and Yamauchi 2004). I argue the literature’s mixed findings regarding the demand heterogeneity hypothesis partly stem from theoretical misspecification. Weisbrod (1977, 1988) assumes that all dissatisfied voters have an equal ability to form nonprofits. However, just as there is demand heterogeneity, there is heterogeneity among groups whose preferences differ from those of the median voter. Pluralist political theory argues that groups have unequal access to resources (Dahl 1961; Hayes 1992), and articulates politics as the unequal distribution of resources (Dahl 1961). According to Hayes (1992), these resources can be tangible (e.g., money, public opinion) or intangible (e.g., expertise, legitimacy, strategic position, internal cohesion, political skill of the group’s leadership). Applied to the demand heterogeneity hypothesis, citizens dissatisfied with government services may have the impetus to seek out nonprofit services but not necessarily the means, due to the unequal distribution of resources within society. Just as citizens vary in their preference for levels of government service provision, dissatisfied citizens vary in their access to resources. Dissatisfied citizens with lower levels of resources will have a harder time raising capital, navigating the political system, or organizing themselves in order to produce the desired service through a nonprofit institution. On the other hand, dissatisfied citizens with higher levels of resources will be able to capitalize on their assets and expertise in order to achieve their desired outcomes via nonprofits. Thus, unequal access to resources results in an unequal likelihood of dissatisfied citizens being able to support or seek out nonprofit institutions.

Nonprofit scholars have demonstrated many times over that supply-side factors like resources are determinants of the nonprofit sector. In general, organizations depend on resources to survive (Pfeffer and Salancik 1978). As organizations, nonprofits rely on tangible financial resources like private contributions, government grants, and revenue-raising commercial activities (Froelich 1999). Ben-Ner and Van Hoomissen (1991, 520) argue the “most important supply factor is the ability of some demand-side stakeholders (consumers, sponsors, or donors) to ensure that the nonprofit organizations of interest to them perform according to their wishes within economic feasibility constraints.” Weisbrod (1997, 543) wrote that “Nonprofits, however, face an enormous obstacle: They lack government’s power to tax, and so when they are confronted by increased demands, they do not have the commensurate resources to meet those demands.” Nonprofits also rely on intangible resources like capacity to design institutional features (Ben-Ner and Van Hoomissen 1991), social cohesion (Ben-Ner and Van Hoomissen 1991; Ben-Ner and Hoomissen 1992), or social capital (Saxton and Benson 2005). These resources determine a group’s ability to coalesce and take action within the nonprofit sector. In the case of US parks, Cheng (2019a) examines when nonprofit organizations become involved in local parks and recreation services provision. Cheng (2019a) finds that organizational-level characteristics (size) and community-level characteristics (human and financial resources, community stability, weak government capacity) increase the likelihood of large nonprofits participating in the provision of park services. These findings suggest that nonprofits require resources to provide services.

However, demand heterogeneity studies often omit supply-side variables like resources. Omitting resources from modeling specifications may be to blame for the previous mixed empirical support of Weisbrod (1977), given this omitted variable likely correlates with existing proxy measures of demand heterogeneity (i.e., race, education). As such, omitted variable bias could account for the alternating positive (e.g., Ben-Ner and Hoomissen 1992) and negative (e.g., Grønbjerg and Paarlberg 2001) effects of demand heterogeneity on the nonprofit sector. When supply-side factors are included within tests of the demand heterogeneity hypothesis, the inclusion is often the result of a measurement decision rather than a theoretical specification. For example, supply-side resource measures are often used as proxies to operationalize demand heterogeneity concepts. These resource-based proxies are typically found to increase the size of the nonprofit sector. For example, income heterogeneity often increases the density of the nonprofit sector (Grønbjerg and Paarlberg 2001; Kim 2015; Liu 2017). Operationalizing demand heterogeneity using proxy measures of resources muddies the theoretical clarity between supply- and demand-side factors.

Other studies articulate the importance of both supply and demand factors on the nonprofit sector, but leave unexamined how unequal access to resources conditions the effect of demand on nonprofit organizations. For example, Ben-Ner and Van Hoomissen (1991, 520) articulate that the “confluence of demand and supply factors determines the incidence of nonprofit organizations.” Their argument reflects Ostrom (1990), wherein the benefits of forming an association must outweigh the costs of forming a nonprofit institution. While this perspective articulates the interdependencies between supply and demand factors, Ben-Ner and Van Hoomissen (1991) leave unexamined how access to resources may disrupt this cost–benefit equation. Ben-Ner and Van Hoomissen (1991, 542) briefly address that higher levels of resources like income and education increase demand for collective goods and decrease the cost of forming a nonprofit but do not tease apart these two mechanisms. Given systemic differences in resource levels across groups, examining how access to resources affects the interaction between supply and demand is a logical next step. DiMaggio and Anheier (1990, 145) observe: “Heterogeneity alone is insufficient to generate effective demand if different groups are markedly unequal in political power, however. The extent to which religion, ethnicity, and language are correlated with one another and with social status (Blau 1977) should condition the effect of social heterogeneity on [nonprofit sectors].” However, neither Ben-Ner and Van Hoomissen (1991) or DiMaggio and Powell (1990) test their proposed conditional models. To the author’s knowledge, no studies to date test how resources condition demand heterogeneity’s effect on the nonprofit sector. Testing a conditional relationship implies the use of an interactive model because the interactive term allows for testing conditional hypotheses (Brambor, Clark, and Golder 2006).

Rather, studies examining supply and demand factors in tandem (e.g., Gazley, LaFontant, and Cheng 2020; Lecy and Van Slyke 2012; Paarlberg and Gen 2009) examine supply and demand using separate models. When exploring the effect of demand and supply variables on public education, Paarlberg and Gen (2009) conclude that supply-side factors (i.e., financial resources) are more important than demand-side (i.e., social homogeneity measured as racial diversity) factors in predicting the nonprofit sector. Paarlberg and Gen (2009) test their argument using two models: the second model is run on a subset of the organizations (i.e., those providing ≥$25,000 to schools) within the sample. This strategy mimics an interactive specification. In this instance, the “conditioning” variable is the level of financial support provided by nonprofits. Similarly, Gazley, LaFontant, and Cheng (2020) discuss the importance of including both supply and demand explanations of the nonprofit sector, but test these competing factors in two different models against two different dependent variables operationalizing supply and demand.

Other studies include supply and demand measures within the same models as control variables. While control variables account for important alternative explanations, these additive models fail to assess conditional relationships by only examining the direct effects of variables (Dhima 2020). Lecy and Van Slyke (2012) compare interdependence theory (Salamon 1995) to the theory of government failure (Weisbrod 1977) in one of the first panel tests. The authors include various measures of nonprofit resources (e.g., private donations, government grants, program revenue) to adjudicate between these theories. While an important study, Lecy and Van Slyke (2012) does not account for the conditioning nature of resources on demand for nonprofits. Rather, demand heterogeneity is controlled for via unit fixed effects. While this approach tests both demand and supply in tandem, it does not assess how access to resources conditions the effect of demand heterogeneity on the nonprofit sector. Similarly, Cheng (2019a) controls for demand heterogeneity (i.e., social diversity) when testing supply-side theory against nonprofit parks and recreation provision. Ultimately, models that account for alternative explanations via the inclusion of control variables do not allow for tests of context or interactive effects. These modeling specifications do not assess how the effect of demand heterogeneity on nonprofits might change in the presence of increasing resources.

Pluralist political theory and supply-side nonprofit research suggest that dissatisfied citizens do not have an equal ability to start, seek out, or provide nonprofit services due to the unequal distribution of group resources. If dissatisfied citizens do not have the means to start or support a nonprofit, then they are less able to seek out or generate nonprofit services to satisfy their unmet demand. Conversely, dissatisfied citizens with access to resources are better positioned to start, seek out, or provide services via nonprofits. Thus, nonprofit service provision is simultaneously determined by both supply-side (i.e., resources) and demand-side (i.e., demand heterogeneity) factors. This proposed relationship suggests these variables should be tested as conditional upon one another. Despite the prevalence of existing tests examining supply and demand inputs, very few examine how access to resources conditions the effect of demand heterogeneity on the nonprofit sector. To test this relationship, I propose the conditional demand heterogeneity hypothesis: The effect of demand heterogeneity on nonprofit service outcomes increases in the presence of increasing resource availability. This hypothesis refines the traditional demand heterogeneity hypothesis to account for the changing effect of demand heterogeneity in the presence of changing resources. Demand heterogeneity alone does not guarantee the formation or activity of a nonprofit; resources are also necessary.

Problematic Proxies

Existing studies measure demand heterogeneity using variables that proxy for heterogeneous preferences through diversity-based measures like race (e.g., Bae and Sohn 2018; Chang and Tuckman 1996; Paarlberg and Gen 2009), religion (e.g., Grønbjerg and Paarlberg 2001; Matsunaga, Yamauchi and Okuyama 2010), income (e.g., Kim 2015; Liu 2017; Paarlberg and Gen 2009), education (e.g., Ben-Ner and Hoomissen 1992; Marcuello 1998), age (e.g., Feigenbaum 1980; Matsunaga and Yamauchi 2004), and language (e.g., James 1993; Salamon and Anheier 1998). Many of these studies include multiple proxy measurements within the same model (e.g., Bae and Sohn 2018; Kim and Kim 2016; Matsunaga and Yamauchi 2004), often resulting in conflicting inferences. For example, Chang and Tuckman (1996) observe a positive relationship between racial diversity and nonprofit externalities, but a negative relationship for religious diversity. To date, no studies employ policy-based measures of preference heterogeneity despite Weisbrod’s (1977) argument being grounded in policy preference. The lack of direct policy preference measures is surprising, given Weisbrod’s (1977) recommendation that:

…diversity of population characteristics is not, ipso facto, equivalent to heterogeneity of quantities demanded. As a first approximation, though, it seems reasonable to assume that differences across political units in the degree of population heterogeneity are useful proxies for differences in the degree of undersatisfied demand. (56)

Measuring demand heterogeneity using diversity proxies introduces measurement error into the predictor variable, which may induce bias or inefficiency into estimates (Carroll et al. 2006). More specifically, measurement error can induce bias into estimations if the error correlates with omitted variables, thereby mimicking omitted variables bias. When demand heterogeneity studies employ proxies to measure demand heterogeneity and omit resource variables in their model specification, the measurement error in the proxies that correlates with omitted resource variables may bias estimates of demand heterogeneity. Given many demand heterogeneity studies omit supply-side variables, this scenario is likely. Diversity-based measures like race, education, and income correlate with resources in a US context, so estimation bias caused by proxies could produce the literature’s mixed results. Thus, the use of proxies is problematic. I suggest policy-based measures, like those used by political scientists, offer a superior measure of undersatisfied demand. These measures more accurately reflect Weisbrod’s (1977) argument and avoid the problems associated with diversity-based proxies.

Another possible explanation for the literature’s mixed support is the operationalization of the dependent variable, nonprofit service delivery. Though Weisbrod’s (1977) theory bases nonprofits’ emergence on their ability to provide service more in line with policy preferences, very few studies employ measures of nonprofit service delivery (i.e., outcomes). Examining outcomes is important as Weisbrod’s (1977) theory addresses both outcomes and outputs. Furthermore, understanding nonprofit outcomes is important for understanding equity implications within the sector (Gazley, LaFontant, and Cheng 2020). Nonprofit research examining schools sometimes defines the nonprofit outcomes as pupil-teacher ratios (Downes and Greenstein 1996) or student enrollment (James 1993), but this practice is rare. Rather than measuring service outcomes, which is a difficult and costly endeavor, most studies measure nonprofit sector outputs such as the presence (e.g., Paarlberg and Gen 2009) or number of nonprofits (e.g., Corbin 1999; Lecy and Van Slyke 2012; Marcuello 1998), nonprofit density (e.g., Grønbjerg and Paarlberg 2001; Kim 2015; Matsunaga and Yamauchi 2004) nonprofit expenditures (e.g., Cheng 2019b; Kim and Kim 2016), and employment (e.g., Bae and Sohn 2018; Ben-Ner and Hoomissen 1992; Matsunaga, Yamauchi, and Okuyama 2010). Given the policy preference mechanism behind Weisbrod’s (1977) theory, wherein individuals are dissatisfied with the quality or quantity of government services, tests employing nonprofit outcomes as the dependent variable can provide exceptional insight into the theory’s proposed mechanisms. However, the service outcomes of nonprofits are difficult to measure and even harder to compare across individual nonprofits, sectors, and government.

Similarly, Weisbrod’s (1977) mixed support may be due to the type of good provided by the nonprofit, making case selection very important: nonprofits may be less likely to form around pure public goods, given the diluted benefits and concentrated costs dissatisfied citizens incur, and more likely to form around quasi-public goods, given dissatisfied citizens are the direct beneficiaries of the goods (Paarlberg and Zuhlke 2019). Many studies do not directly distinguish between tests examining collective and quasi-public goods, despite parsing tests across different nonprofit sectors.¹ Similarly, studies examining nonprofit goods across sectors often demonstrate inconsistent effects across said sectors. Notably, nonprofit sectors are often defined by the type of good provided by the nonprofit. For example, Marcuello (1998) observes that as income heterogeneity decreases, the number of education nonprofits decreases, whereas there is no effect on cultural or welfare-oriented nonprofits. Grønbjerg and Paarlberg (2001) find that increasing religious diversity decreases the density of advocacy nonprofits and mutual benefit nonprofits, but has no effect on the number of charitable nonprofits. Tests of the demand heterogeneity hypothesis need to be intentional in their case selection because case selection informs theoretical expectations.

In short, I argue the current mixed support of the demand heterogeneity hypothesis is the consequence of the use of proxy measurements, which often correlate with theoretically important, but oft omitted, supply-side variables. Similarly, the traditional operationalization of dependent variable as nonprofit outputs rather than outcomes overlooks testing the demand heterogeneity hypothesis on an important component of Weisbrod’s (1988) argument: service provision. Empirically, grounding demand heterogeneity measures in policy preferences and nonprofit sector measures in nonprofit outcomes is necessary in order to assess the viability of the demand heterogeneity hypothesis. Additionally, controlling for the type of good provided by the nonprofit is necessary in order to isolate the effect of demand heterogeneity on the nonprofit sector.

Hypotheses and Expectations

First, I test the traditional demand heterogeneity hypothesis: As demand heterogeneity increases, nonprofit service outcomes will increase (Hypothesis 1). Next, I test the conditional demand heterogeneity hypothesis: The effect of demand heterogeneity on nonprofit service outcomes increases in the presence of increasing resource availability (Hypothesis 2).

Prior to testing the hypotheses, I demonstrate the pitfalls of employing traditional proxy measurements to illustrate the problems with proxies and increase the relevancy of this research to prior work. I expect that the proxy measures will produce conflicting inferences when applied to the traditional demand heterogeneity hypothesis. I also demonstrate how proxies may mimic omitted variables bias when variables measuring resources are omitted from models. I expect that the proxies’ statistical significance will disappear when supply-side measures are no longer omitted.

After demonstrating the problems with proxies, I test Hypotheses 1 and 2 using a new measure of heterogeneous policy preference. With regard to the traditional demand heterogeneity hypothesis (Hypothesis 1), I expect the new, policy-based measure of demand heterogeneity to have a positive but statistically insignificant effect on nonprofit outcomes. A positive but statistically insignificant direct effect suggests that demand heterogeneity is not a sufficient predictor of nonprofit outcomes. However, if the effect of demand heterogeneity on nonprofit service outcomes becomes statistically significant once interacted with resources, then this finding suggests that both demand and resources are necessary for nonprofit outcomes. Importantly, I expect the effect of demand heterogeneity on nonprofit outcomes to increase in the presence of increasing resources. The effect of demand heterogeneity on nonprofit service delivery will be strongest when resource availability is highest.

Research Design

Case selection is critical given the need to improve variable measurement, control for the type of good produced by nonprofits, and test supply and demand factors in tandem. US public land represents a public good, similar to public education, to test nonprofit theory against (Gazley, LaFontant, and Cheng 2020, 350). A growing number of studies employ parks and recreation as a case to test government–nonprofit theory. Cheng (2019b, 239) notes that local parks and recreation services are a “particularly rich setting for studying government-nonprofit interactions.” Some studies specifically use parks to test government-nonprofit theory (Brecher and Wise 2008; Cheng 2019a, b; Gazley, LaFontant, and Cheng 2020), while other studies discuss the role of park-supporting charities in the United States and their influence on public provision of services (Harnik and Martin 2015; Yandle, Noonan, and Gazley 2016).

In the United States, access to public land varies depending on land management practices—public land may be excludable. For example, only 78% of land trusts allow public access to their properties (Lieberknecht 2009).² I analyze three cases of US protected land to test the hypotheses against a range of good types: (1) land held in fee or easements by US public agencies and nonprofits, (2) fee and easement lands that are not open to extraction, and (3) fee and easement lands specifically managed for biodiversity protection. Land open to extractive purposes, like mining or logging, reflects attributes of private or club goods. Land specifically managed for biodiversity purposes more closely reflects a case of public goods. Biodiversity is a public good, in part because its benefits are not limited geographically (Stone 1995) and are nonrivalrous and nonexcludable in their scientific, ecological, and existence value (Doremus 2003, 220).

In line with prior work (Ben-Ner and Van Hoomissen 1991; Paarlberg and Zuhlke 2019), I expect the effects of demand heterogeneity and resources on the nonprofit sector to vary by the type of good in question. I expect the effect of demand heterogeneity on the nonprofit sector to be strongest on goods reflecting private-like qualities and to lessen as the nature of the good becomes more public, due to the diluted dispersion of benefit.

US Protected Land: Fee and Easement

Nonprofits and government play an important role in land conservation (Doremus 2003). To date, no existing studies test theories of government–nonprofit relationships using measures of physical land set aside for conservation. Parks and recreation studies typically examine nonprofit outputs like expenditures (e.g., Cheng 2019a). However, other studies employ physical environmental measures as nonprofit outcomes. For example, Grant and Langpap (2019) use water quality indicators to measure the outcomes of nonprofit lake associations.

Two major avenues of land acquisition in the United States are fee-simple purchasing and easement agreements. In the United States, land is associated with a “bundle” of rights. Easement agreements “unbundle” these rights to set aside land for specific purposes, while still allowing certain activities to take place on that land as long as those activities are not in conflict with the mission of the easement. Conservation easements are a conservation mechanism wherein land owners transfer a range of property rights to an easement “holder” on a limited or permanent basis to achieve conservation aims (Owley and Rissman 2016; Rissman et al. 2017). Easement holders, or the entities that enforce and manage the easement agreement, are typically governments or nonprofit organizations (e.g., land trusts); however, state law defines holders (Owley 2015).³ Under conservation easements, certain property rights are leased depending on “those rights necessary to protect the target species or ecosystem while leaving the owner free to use the land in ways compatible with conservation” (Doremus 2003, 219). Land trusts are generally viewed as effective land conservationists (Armsworth and Sanchirico 2008).

In return for entering an easement, land owners may receive money, tax deductions, or development permits (Owley and Rissman 2016), though development rights are typically restricted upon the creation of a conservation easement. Most land that enters into conservation easements is donated by the landowner (Byers and Ponte 2005; Farmer et al. 2011). Just like individuals make donations to nonprofits for various reasons (Bekkers and Wiepking 2011), individuals place their land into conservation easements for various reasons. In a survey of landowners who entered into conservation easements, Farmer et al. (2011) find that “a desire to contribute to the public good” underlies private landowners decisions to enter conservation easements. Attachment to the land is the biggest motivator, while financial is the smallest motivator. To enter a conservation easement, a landowner contacts a potential easement holder to initiate the process, which includes biological reviews to determine whether the land is fit for conservation purposes. The timeline of the process ranges depending on the organization and state. For instance, the Northcentral Pennsylvania Conservancy states their typical timeline to set up an easement is 6 months,⁴ whereas the Western Landowners Alliance states the process can take between 9 and 18 months.⁵

Conservation easements differ from fee-simple land parcels. Under fee-simple purchases of land, all property rights are taken over from the landowner (Doremus 2003, 219). When government purchases land, it can do so via the voluntary sale of land owners or condemnation via eminent domain. For example, the National Park Service (NPS) may acquire lands following an act of Congress. Fee-simple ownership provides the NPS with the greatest leeway for management, but the NPS uses condemnation as a “last resort” to acquire land (National Park Service 2001). Eminent domain is rarely used by the four major federal land management agencies—the NPS, the Forest Service, the Fish and Wildlife Service, and the Bureau of Land Management—because of its controversial nature (Vincent et al. 2019). Nonprofit organizations have no such power and therefore purchase fee-simple land from “from willing sellers” (Doremus 2003, 219).

Acres of land held in fee or easements provide the opportunity to directly measure nonprofit outcomes and policy preferences as dependent and independent variables. As a measure of policy preference, environmental conservation provides a specific policy area over which to assess citizens’ policy preferences. As an outcome, acres as a unit of land represent a valid and reliable measure of nonprofit and government outcomes across geographic space and time. That acres are comparable between nonprofits and government is important given the need to control for government activity when testing the demand heterogeneity hypothesis (Corbin 1999). Furthermore, using physical land extends previous work testing nonprofit theory against parks and recreation expenditures, which are limited to nonprofits with available financial data (e.g., Cheng 2019a). Employing physical land measures may capture the behavior of small and large organizations. Similarly, other studies examining parks and recreation expenditures are sometimes limited to urban areas for data purposes (e.g., Cheng 2019b; Harnik and Martin 2015). Physical land measures present an opportunity to extend this work to rural areas.

To create the three aforementioned cases of US protected land for the analysis, I draw upon the USGS PAD-US Data, version 2.1 (US Geological Survey [USGS] Gap Analysis Project [GAP] 2020). PAD-US is the nation’s “official inventory of U.S. terrestrial and marine protected areas... that are dedicated to the preservation of biological diversity and to other natural, recreation and cultural uses, managed for these purposes through legal or other effective means.” (USGS GAP 2020). PAD-US is a multi-year data collection effort to create a geospatial inventory of protected land parcels in America. At present, PAD-US is the most comprehensive inventory of public lands in the United States ever assembled, over time and space.⁶ The full PAD-US v2.1 fee and easement data are presented in figure 1. Though these data are extensive, they ultimately still constitute a sample. PAD-US is a work in progress; not all US protected parcels are included in the data. Furthermore, parcels within the PAD-US data may be missing important attribute data, like the parcel’s year of establishment.

Table 1 describes the land qualities associated with each of the three cases. To create Case 1 reflecting private/common good qualities, I subset the PAD-US data to all fee-simple and easement land parcels. This strategy excludes “Designated” areas, or managed protected lands not tied to title documents. In the United States, Congress and the executive branch can designate areas, based on land characteristics, value, use, and a variety of management and regulatory features. Typically, Designated land areas apply to federally owned land (Comay et al. 2018). Consequently, the Designated areas in the PAD-US data typically overlap fee parcels (USGS GAP 2020). I exclude these areas from the analysis to avoid double-counting.⁷ Similarly, the PAD-US data contain “Proclaimed” areas, which “have been set aside and reserved for” certain purposes (Vincent et al. 2019, 5). Proclaimed areas include approved acquisition boundaries, which typically subsume existing land parcels. I also exclude these areas from the analysis to prevent double-counting. Finally, I also exclude marine protected areas. This selection strategy includes land that is open to extractive purposes (i.e., logging or mining), and therefore represents private or quasi-public goods. To create Case 2 reflecting club good qualities, I remove land that is subject to extractive purposes from the Case 1. The PAD-US data contain information on the protection of each parcel, including whether a land parcel is open to extractive purposes. I use the same information to isolate land that is specifically set aside for biodiversity purposes to create Case 3, reflecting public good qualities.

To account for endogeneity concerns when testing government-nonprofit relations, I use a time series cross-sectional data set (N = 1,776) on land protection in US states from 1975 to 2011. Accounting for order helps shore against alternative theories of government-nonprofit relations, like voluntary failure (Salamon 1987). Furthermore, the use of conservation easements in the United States has evolved overtime. First introduced in the 1970s, conservation easements were intended as a “low-cost alternative to fee simple acquisition” (Owley and Rissman 2016, 83). Conservation easements increased in earnest following the addition of the conservation easement tax deduction to the US federal tax code in 1980 and the passage of the Uniform Conservation Easement Act (UCEA) in 1981 (Owley and Rissman 2016).

The unit of analysis is the US state.⁸ As Owley (2016, 171) writes: “Conservation easements are a creature of state law.” States have unique histories and policies regarding the implementation of conservation easements, adopting UCEA at different points in time (Owley and Rissman 2016, 78). There are other reasons that land conservation may systematically vary at the state level. In a study of parks and recreation services, Gazley, LaFontant and Cheng (2020) find strong state effects they attribute to differences in state budgeting processes. Some states offer conservation education to landowners while others do not (Doremus 2003, 218). Additionally, available data on conservation easements may vary by state according to data-sharing laws (Rissman et al. 2017, 6). Analyzing data at the state level also minimizes potential spillover effects; most land parcels are conserved within state boundaries. Despite these justifications, a state-level analysis lacks the granularity inherent to a higher-resolution analysis (e.g., county or city-level). Analyzing relationships at the state-level risks committing ecological fallacy. Specifically, analyzing the conditioning effect of resources on demand at the state level masks whether “demanders” and resource-holders are the same. Thus, this analysis should be interpreted as a good first cut at testing the conditional demand heterogeneity hypothesis. Future work and testing is needed across various levels of analysis, resource types, and cases.

Variables

This study employs a measure of nonprofit outcomes as the dependent variable: the total new acreage in a state conserved by nonprofits at time t. A measure of service delivery, using new total nonprofit acres is consistent with Weisbrod’s (1988) argument that government can produce services or finance private provision. As detailed above, this measure was calculated using the PAD-US data under three different case selection criteria. In each case, the total amount of new land protected by nonprofits in a state in time t was calculated. PAD-US contains a measure of time, noting the date of establishment for each land parcel. I sum the total acres of land newly protected by nonprofits within a given year. Not all land parcels within the PAD-US data have a year of establishment associated with them. These parcels were dropped from the analysis.

I employ various independent variables to measure demand heterogeneity. To illustrate the problems with proxies, I employ two common demand heterogeneity proxies: racial diversity and income inequality. Though concerns exist over whether inequality measures are equivalent to heterogeneity measures (Harrison and Klein 2007; Reardon 2009), nonprofit research employs income inequality as a proxy for demand heterogeneity (e.g., Kim 2015; Liu 2017). I estimate the effects of these two proxies to illustrate how demand heterogeneity proxies may produce conflicting inferences and demonstrate that this problem is present within the sample. Following Kim (2015), I measure racial diversity as the percent of nonwhite individuals within the state population. Following Kim (2015) and Liu (2017), I measure income inequality using a gini coefficient. Both of these measures were downloaded from the Correlates of State Policy Project v2.2 (Jordan and Grossmann 2020). Notably, some studies find that as income in an area increases, the size of the nonprofit sector decreases (Bae and Sohn 2018; Matsunaga and Yamauchi 2004), while others find the opposite effect (Kim 2015).

To improve upon traditional proxy measures, I employ a new measure of demand heterogeneity based on heterogeneous policy preferences. I adopt Kim and Urpelainen’s (2018) measure of US environmental public opinion. This measure captures the percent of citizens in a state, over time, who believe we are spending “too little” on the environment. The specific question employed by Kim and Urpelainen (2018) to create the measure reads:

We are faced with many problems in this country, none of which can be solved easily or inexpensively. I’m going to name some of these problems, and for each one I’d like you to name some of these problems, and for each one I’d like you to tell me whether you think we’re spending too much money on it, too little money, or about the right amount. Are we spending too much, too little, or about the right amount on [improving and protecting the environment]? (Kim and Urpelainen 2018, 94)

Measuring demand heterogeneity as the proportion of the citizenry that thinks government is not doing enough on a particular issue area is appropriate given the theory of government failure’s emphasis on the behavior of individuals who are undersatisfied with government service provision. As Weisbrod (1988, 26, italics added) notes: “The undersatisfied demand for collective-type goods is a governmental ‘failure’ analogous to private market failure.” Weisbrod (1977) argues that oversatisfied individuals—individuals who think government is doing too much in a particular area and who desire a lower level of service from the government—may do nothing, move elsewhere (Tiebout 1956), or “exert political pressure” to change the tax rate or lower the quality or quantity of services provided by the government (Weisbrod 1977, 57). On the other hand, undersatisfied individuals who desire a higher quantity or quality of service from the government can move elsewhere (Tiebout 1956), form a lower-level government to increase their service provision (Burns 1994), or satisfy their demand by turning to the private market or nonprofit sector (Weisbrod 1977, 57). In short, undersatisfied—not oversatisfied—individuals support nonprofits. Kim and Urpelainen’s (2018)’s measure is particularly salient to demand heterogeneity applied to the case of protected land, because it measures individuals’ undersatisfaction with government performance regarding environmental protection.

To measure the conditional effect of resource availability on demand heterogeneity, I use a measure of disposable income per capita. To create this measure, I inflation-adjust an annual measure of total disposable income within each state (Jordan and Grossmann 2020; Klarner 2013) to 2011 USD ($), then divide this inflation-adjusted measure by the annual population in a state. This process produces a per capita measure of disposable, or after-tax, wealth. Access to resources could be operationalized many different ways. I operationalize resources as financial resources, given my case selection. Land conservation, particularly fee-simple acquisition, requires financial resources to purchase land.

To control for government service provision, I include the total new area in a state protected by special districts, local, state, federal government in time t as individual variables. Similarly, I also include the total new area in a state protected by other groups, like Native American tribes, in time t. Similar to land conserved by nonprofits, I calculated these data from the PAD-US dataset. According to Weisbrod (1977) and related supplemental models (Young 2000), government land conservation should exhibit an inverse relationship with nonprofit land conservation.

To control for the heterogeneous value of land, I include a measure of the average farm real-estate value in each state for all the years included in the study, also inflation adjusted to 2011 USD ($). Farm real-estate value is “the value at which all land and buildings used for agricultural production, including dwellings, could be sold under current market conditions, if allowed to remain on the market for a reasonable amount of time.” (National Agricultural Statistics Service 2020, 20). These values are calculated by the USDA National Agricultural Statistics Service in an annual survey. I also control for years after 2000. In an examination of 269 conservation easements from six states, Owley and Rissman (2016) find that conservation easements established in the 1980s and 1990s are less complex than conservation easements established after 2000. After 2000, conservation easements rose sharply (Rissman et al. 2017, 2) due to knowledge transfer and the increased availability of funds via federal programs and local bond initiatives (Owley and Rissman 2016). Given this study’s time extent (1975–2011), I use a dummy variable to control for these shifts. I also include controls for a state’s annual population density (measured as persons per acre of land), annual total population (logged), and state fixed effects to account for unobservable, unique state factors.⁹ I discuss the application of state fixed effects below. A descriptive summary of the variables described for all fee and easement land representing private/common good qualities (Case 1) is available in table 2. Descriptive tables of the variables for land without extraction rights, thus representing club good qualities (Case 2), and biodiversity specific land representing public good qualities (Case 3) are available in supplementary appendix.

Model Specification

I estimate a finite distributed lag (FDL) model with state fixed effects to test both hypotheses, given my theoretical expectations (Beck and Katz 2011). I estimate a FDL model, lagging control variables like new government land (i.e., federal, state, local, special district), in order to account for the time it takes to protect land via nonprofit organizations in response to government land protection. I include an additional time lag (t−2) of federal acres, based on its exhibited autocorrelation (De Boef and Keele 2008). Accounting for the order of government and nonprofit activity within the model specification also helps addresses endogeneity concerns. Statistical tests confirm a lack of serial autocorrelation but the presence of cross-sectional error dependence within the panel. To account for this dependency, I use a form of panel corrected standard errors (Beck and Katz 1995, 2011); specifically, Driscoll–Kraay standard errors (Hoechle 2007).

To test the traditional demand heterogeneity hypothesis (Hypothesis 1), the model I employ is:

wherein subscript i references the state and t references the year, NA is the total acres of new land protected by nonprofits within state i in time t, DH is demand heterogeneity present within state i in time t, and R is access to resources, measured as disposable income per capita in state i at time t. PL is a vector of total acres of new land protected by federal, state, and local government and other institutions (respectively) in state i at t − 1 by, FL is total acres of new land protected by the federal government in state i at t − 2, X is a vector of additional control variables, and δ are state fixed effects. To test the conditional demand hypothesis (Hypothesis 2), I include an interactive term between DH and R within the same the modeling specification (Equation 2). Specifying the conditional demand heterogeneity hypothesis as an interactive model, rather than an additive model like Equation 1, allows for testing the conditional relationship between supply-side and demand-side factors. Interaction terms are an appropriate strategy for testing conditional hypotheses (Brambor, Clark, and Golder 2006).

The inclusion of state fixed effects (δ) accounts for time invariant factors like overall state size. Additionally, state fixed effects account for land protected in a state prior to 1975 and state location (e.g., western states). Similarly, there may be other, unobservable differences between states that the inclusion of fixed effects addresses. Given the inclusion of state fixed effects, interpretations are limited to effects within states over time.

Results and Discussion

Table 3 displays the estimates illustrating the problems with demand heterogeneity proxies using the traditional demand heterogeneity hypothesis (Hypothesis 1). The estimates displayed utilize Case 1 data, representing private/common good qualities. The full set of illustrations across all three types of land cases are available in supplementary appendix; given the consistency of results, I only report Case 1 estimates.

First, consider Model 1 (table 3) wherein proxy measurements for demand heterogeneity are employed while measures of resources (i.e., disposable income per capita) are intentionally omitted. Similar to prior work, the proxy measures of demand heterogeneity produce conflicting results. As the racial diversity of a state increases, new acres of land protected by nonprofits decrease (though this effect just barely fails to achieve statistical significance). On the other hand, as the income inequality within a state increases, new acres of land protected by nonprofits increase. These findings illustrate how different proxies of demand heterogeneity produce different inferences within the same model. Furthermore, Model 1 intentionally omits a measure of resources: disposable income per capita, measured in $1,000s and inflation-adjusted to 2011 USD ($). When resources are no longer omitted from the modeling specification (Model 2), the statistically significant effect of income inequality dissipates. This comparison between Model 1 and Model 2 demonstrates how proxy measurements may replicate omitted variables bias when measures of resources are omitted from modeling specifications. Thus, table 3 demonstrates that the problems with proxies described earlier are present within the study data.

Given these problematic proxies, I test Hypothesis 1 and Hypothesis 2 with a policy-based measure of demand heterogeneity. Hypothesis 1 posits that as demand heterogeneity increases, nonprofit service outcomes will increase (traditional demand heterogeneity hypothesis). Hypothesis 2 posits that the effect of demand heterogeneity on nonprofit service outcomes increases in the presence of increasing resource availability (conditional demand heterogeneity hypothesis). The percent of nonwhite residents in a state and income inequality, previously used as proxies, are included as control variables because they relate to both nonprofit service provision and environmental public opinion. The estimates across all three cases—private/common goods (Case 1), club goods (Case 2), and public goods (Case 3)—are provided in table 4. If demand heterogeneity is not a sufficient predictor of nonprofit service output, then its coefficient should be positive but not statistically significant when testing Hypothesis 1. If the effect of demand heterogeneity is conditional upon resource availability, then demand heterogeneity should become statistically significant once interacted with measures of resource availability in Hypothesis 2. Table 4 confirms this expectation across Case 1 (private/common goods) and Case 2 (club goods), but not Case 3 (public goods).

When land represents private/common good qualities, as in Models 1 and 2, the effect of demand heterogeneity on the nonprofit sector fails to achieve statistical significance (Model 1) until it is interacted with resource availability (Model 2). This evidence suggests the demand heterogeneity is a necessary, but insufficient predictor of the nonprofit sector. Figure 2 plots the marginal effect of demand heterogeneity across increasing levels of resource availability: as disposable income per capita increases by $5,000, the marginal effect of a 1% increase in citizens who do not think the government is doing enough to protect the environment increases nonprofit land conservation by roughly 98 acres.¹⁰ As expected, the effect of undersatisfied demand is highest in the presence of the greatest resource levels. That the effect of undersatisfied demanders on nonprofit acres is negative at lower resources is likely a consequence of the imposed linear form of the model against data that are truncated at zero (the minimum amount of new land conserved by any institutional form in time t is zero [table 2]). This suspicion is supported in an examination of predicted outcomes (figure 3). In applied terms, when 80% of a state’s population thinks the government is not doing enough to protect the environment and individuals have $25,000 in disposable income, 3,397 new acres of land will be protected by nonprofits (p < 0.01). When individuals within a state have $45,000 in disposable income, 21,313 acres of new land will be protected by nonprofits (p < 0.01).¹¹ Nearly six times the amount of land is conserved by nonprofits when individuals with the same public opinion have increased access to resources.

The conditional effect of access to resources on demand heterogeneity holds for the models wherein land represents club good qualities (Case 2), but not public good qualities (Case 3). However, the basic substantive relationship—that the effect of demand heterogeneity increases in the presence of increasing resources—is maintained within the public goods case (Case 3). As expected, the substantive effects of the interactive term decrease as the nature of the good being provided becomes more public in type. These results are consistent with the logic that benefits cluster on the individual as goods provided by nonprofits become more private in nature (Ben-Ner and Van Hoomissen 1991; Paarlberg and Zuhlke 2019). Notably, even when land represents a public good, resources have a substantive conditioning effect on demand heterogeneity’s effect on nonprofit outcomes, though this estimate fails to achieve statistical significance. Furthermore, the measure of fit improves when comparing the modeling specifications within modeling pairs across good types.

Conclusion

This study advances an interactive, conditional model of the nonprofit sector and contributes to a long-standing debate between supply- and demand-side arguments about nonprofit organizations. Merging Weisbrod’s classic theory of government failure with supply-side, resource-driven arguments and pluralist political theory, I argue the effect of demand heterogeneity on the nonprofit sector is conditional on the resources available to those experiencing government failure. Specifically, nonprofits are most active when those experiencing government failure have access to resources. Traditionally, supply-side and demand-side factors explaining the nonprofit sector are tested in separate models and studies. When supply-side and demand-side factors are tested within the same model, the relationship is not specified as conditional. Testing these factors in an interactive model allows for a model specification that better reflects the theoretical interdependence of these factors.

This study also makes important methodological contributions. First, it articulates how diversity-based proxy measurements of demand heterogeneity employed by previous studies may be the culprits behind the literature’s mixed support of the theory of government failure. Diversity-based proxies may produce biased estimates if they correlate with theoretically important but omitted variables, like resources or other supply-side measures. Despite this problem, proxies for demand heterogeneity are widely employed within demand heterogeneity tests. Existing literature testing the demand heterogeneity hypothesis fails to operationalize the main independent variable as theoretically intended—undersatisfied policy preferences. Given the policy-based theoretical mechanisms in Weisbrod’s (1977) theory of government failure, future tests of the theory of government failure should employ policy-based measurements of demand heterogeneity. These measures are widely used within political science.

Second, future studies should test nonprofit theory against measures of nonprofit outcomes, like service delivery, in order to fully understand the implications of the government failure thesis. This study identifies a way to measure nonprofit service outcomes using acres of land conserved by nonprofits in the United States. It is difficult to identify reliable and valid measures of nonprofit service provision; as evidenced by recent work (e.g., Grant and Langpap 2019), environmental data may offer some relief to this issue. The case of US public land presents an opportunity for such a test. In particular, the universe of US land conservation data provided by the PAD-US data provide a wealth of opportunity to test theories of the nonprofit sector across different good types, as well as political contexts, space, and time.

However, this study is not without its limitations. First, a state-level analysis risks committing ecological fallacy. Future work should examine the interactions between undersatisfied demanders and resources at a higher-resolution (e.g., county or city-level) in order to more robustly connect those that demand with access to resources. The results of this study should be interpreted as evidence that the conditional demand heterogeneity hypothesis merits further investigation. Second, the case of land conservation itself may be unique. Land is a financially intensive endeavor; in particular, fee-simple land requires resources to purchase. Less resource-intensive goods may have a different data generating process. Future tests should explore the conditioning effects of other important types of resources on demand heterogeneity, as well as any nonlinear relationships between resources, demand heterogeneity, and nonprofit outcomes. Similarly, whereas this study examines nonprofit outcomes, future work should test the conditional demand heterogeneity hypothesis on measures of nonprofit outputs like expenditures. Furthermore, environmentalism within the United States is an important political force; as a movement, it may create unique government-nonprofit dynamics. This study’s emphasis on fee-simple and easement land may overlook important political dynamics created when the US government designates or proclaims an area important for an environmental or cultural reason, thus beginning the process of land acquisition. To counter some of these case-specific dynamics, this study takes advantage of land’s multidimensional properties in order to understand how government–nonprofit relations are affected by the nature of the good (i.e., private/common, club, public) being provided. However, future tests should examine the conditional demand heterogeneity hypothesis across different issue areas to provide a better understanding of its generalizability.

Despite these limitations, these results suggest that a conditional demand heterogeneity hypothesis is a promising evolution of the canonical demand heterogeneity hypothesis. Though nonprofit arguments combining supply and demand factors are not new (e.g., Ben-Ner and Van Hoomissen 1991; Paarlberg and Gen 2009), this study provides a long-overdue discussion on and update to the measurements and modeling strategies employed to test these important arguments. As a first test, this study provides evidence from a single case study that the dichotomy between supply-side and demand-side arguments is not concrete. Rather, demand and access to resources are both necessary components of nonprofit outcomes.

The necessity of resources for the production of nonprofit outcomes raises equity concerns within the nonprofit sector. Nonprofits are important societal actors because of their distribution of services and redistributive properties. However, the findings presented here suggest that nonprofits only conserve land when an undersatisfied demander has roughly $35,000 in disposable income. This threshold drops slightly to $25,000 when a majority of residents (80%) feel that the government is not doing enough to protect the environment. Future research should investigate trade-offs between tangible and intangible resources, like finances and social capital. However, despite these potential trade-offs, these findings suggest that tangible resources play a dominant role in the nonprofit sector. Previous research demonstrates that philanthropy exacerbates equity issues in the case of state parks (Gazley, LaFontant, and Cheng 2020). This research similarly suggests that the distribution of nonprofit outcomes is strongly affected by demanders’ access to resources. Future work should continue the examination of nonprofit outcomes, particularly the distribution of their costs and benefits, in order to better investigate the role nonprofits play in alleviating or exacerbating equity problems.

Footnotes

Acknowledgments

Previous versions of this article were presented at Environmental Politics and Governance Online (2020), the 6th Annual Duck Family Workshop in Environmental Politics and Governance (2020), the annual meeting of the American Political Science Association (2020), and the Texas Methods Annual Meeting (2020). I thank Manuel Teodoro, Laurie Paarlberg, Guy Whitten, Scott Cook, David Switzer, and discussants for their helpful advice. I also thank the three anonymous reviewers for excellent comments that greatly improved the quality of the manuscript. This research was made partly possible thanks to the Texas A&M Dissertation Fellowship.

Data Availability

The data underlying this article are available in the supplemental materials. The dataset was derived from sources in the public domain: USGS PAD-US v2.1 (https://www.usgs.gov/core-science-systems/science-analytics-and-synthesis/gap/science/pad-us-data-download?qt-science_center_objects=0#qt-science_center_objects) and Institute for Public Policy and Social Research (IPPSR)’s Correlates of State Policy Project v2.2 (http://ippsr.msu.edu/public-policy/correlates-state-policy).

References