The geography of giving: Nonprofit philanthropy, public parks and land conservation in the United States

Faced with budget cutbacks, state and local governments struggle to adequately fund public parks and support land conservation. “Support nonprofits” are charitable organizations soliciting voluntary contributions of money, property, and service to protect parks and natural systems (Gazley, 2015). From 2000 to 2015, the number of support nonprofits increased by 188% and total nonprofit revenue increased by 263%, outpacing growth in public charities nationwide (NCCS). The rapid growth in support nonprofits prompts key questions about access and equity: Which communities benefit from support nonprofits? Do support nonprofits advocate for public land conservation? Using a novel national dataset pairing large nonprofits with the counties they serve, the article has three key findings: (1) counties in metropolitan areas with well‐educated, liberal residents are more likely to have a support nonprofit fundraising for public land, and (2) there is no correlation between public funding for parks and support nonprofit presence. In addition to philanthropy, (3) counties with support nonprofits are more likely to pass ballot initiatives funding public land conservation. It is imperative that policy‐makers and conservation advocates consider how nonprofit philanthropy maps on to—and augments—existing inequality in access to parks and natural resource conservation.


| INTRODUCTION
The Great Recession dramatically impacted state and local governments fiscal health: revenue shortfalls ushered in a "fiscal ice age" for spending on public services in the United States (Alm and Sjoquist 2014;Gordon 2012;Kiewiet and McCubbins 2014, 105). Funds for parks and natural resource conservation were disproportionately cut and have been slow to recover, suggesting a permanent shift in spending priorities (Barrett, Pitas, and Mowen 2017). Nonprofits are stepping in to aid state and local governments in provision of public services, and nonprofit philanthropy is increasingly important in protecting and maintaining public land (Gazley 2015). This article focuses on the distribution of large support nonprofits, charitable organizations organizing volunteers and donations to conserve public land in the United States.
From 2000 to 2015, the number of support nonprofits in parks and conservation increased by 188% and total nonprofit revenue increased by 263%, outpacing growth in public charities nationwide. 1 Support nonprofits primarily engage in fundraising and voluntary activities: maintaining trail infrastructure, creating educational programming, helping visitors, engaging the community and leading advocacy efforts (Gazley, Cheng, and Lafontant 2018;Murray 2010;Nisbet and Schaller 2020;Walls 2014). The relationship between support nonprofits and local governments varies. Municipal or county parks and conservation departments create nonprofit organizations to help protect parks and natural resources (Nisbet and Schaller 2020). In other instances, support nonprofits represent "citizens' taking initiative" for managing and maintaining natural resources through "friends" of parks and conservation efforts (Murray 2010, 189). Support nonprofits focus on local, city, or municipal public parks or specific parcels of protected land like estuaries or rivers. 2 This article focuses on "large" support nonprofits that are well-resourced, formalized, and long-standing (Grønbjerg 2002). Support nonprofits range in size. For example, the Partnership for the City Parks generated $13.4 million in revenue to improve the park system in New York City's five boroughs in 2015. Most organizations are smaller: the Clinton County Conservation Foundation in Iowa generated $71,003 in revenue to purchase equipment and supplies for an environmental education program organized by the county government. 3 Nonprofit mobilization for public parks and land conservation is "neither a temporary nor a short-term trend", and public parks increasingly rely on private funding (Gazley, Cheng, and Lafontant 2018, 9). For every $8.5 dollars a national park unit spent on operations, the support nonprofit raised and spent $1 (Gazley, Cheng, and Lafontant 2018). The rapid growth in support nonprofits prompts questions about access and equity: Which communities benefit from support nonprofits? Do support nonprofits advocate for public land conservation?
This manuscript uses a longitudinal dataset to study the distribution of support nonprofits in the United States. First, I review key theories on supply, demand, and the distribution of nonprofits, as well as theories on nonprofits and political mobilization. Next, I describe the data and methodology. Using logistic regression, I model the association between large nonprofits and county socioeconomic, demographic, and fiscal characteristics. To understand nonprofit's role in mobilizing voters for public land conservation efforts, I use county characteristics and nonprofit density to predict the success of county-level ballot initiatives for conservation funding. Support nonprofits are primarily located in large metropolitan areas with more liberal, educated residents. In addition to helping protect public land through philanthropy, nonprofit presence is correlated with successful ballot initiatives for land conservation.
The rapid growth-and permanency-of support nonprofit philanthropy has important implications for conservation policy. Scholars have raised concerns about the uneven capacity of the nonprofit sector (Paarlberg & Yoshioka, 2016). Support nonprofits play an important role in financing and shaping public services through philanthropy, yet they are not evenly distributed to improve equity in access to public services (Cheng 2019;Cheng, Yang, and Deng 2021;Gazley 2015;Gazley, LaFontant, and Cheng 2020). Support nonprofits likely impact land conservation initiatives through voter outreach and mobilization. Alternatively, voters may engage in nonprofit philanthropy as a replacement for public funding, undermining support for government funding to protect public land (Holmes 2012;Walls 2014). Equitable access to public land is the responsibility of state and local governments; support nonprofits do not have the same democratic mandate to ensure equal access. As state and local governments increasingly rely on support nonprofits to fund parks, natural resources, and land conservation efforts, it is imperative that policy-makers and conservation activists understand how nonprofit distribution impacts parks and conservation governance.

| OVERVIEW OF THE LITERATURE ON SUPPORT NONPROFIT PHILANTHROPY AND PUBLIC LAND
In 2015, at least 10.3% of counties have one or more large support nonprofit advocating for neighborhood, municipal, or county public parks, natural resources, and conservation efforts (see Figure 2) (National Center for Charitable Statistics 2015). Existing research analyzes the distribution of nonprofits funding state parks (Gazley, LaFontant, and Cheng 2020), federal parks (Gazley, Cheng, and Lafontant 2018), and large metro-areas (Cheng, Yang, and Deng 2021). This project focuses on county variation in support nonprofits.
Understanding county parks and natural resource conservation efforts are important for several reasons: first, state and local budgets were hit hardest by the Recession (Alm and Sjoquist 2014;Perlman 2010). As a result, county conservation efforts may depend on nonprofit philanthropy to fill in gaps in public funding. Second, counties increasingly provide public services due to declines in state and federal funding (Paarlberg and Zuhlke 2019). Third, local conservation efforts are impacted by uneven patterns in public funding: the F I G U R E 2 Distribution of support nonprofits by county, 2015. Sources include the National Center for Charitable Statistics Core Trend Data File. quality of park systems varies by community income and racial/ethnic composition (Rigolon, Browning, and Jennings 2018). Support nonprofit advocacy may map on to-or alleviate-existing inequities in access to public land.
2.1 | What are the community conditions that foster nonprofits?
Two key theories explain the conditions that foster nonprofits: government failure and interdependence theory. In government failure theory, nonprofits form in response to government failure to meet public demand for services (Paarlberg and Gen 2009). When the public has heterogeneous policy preferences, government responds to the demands of the perceived average voter (Paarlberg and Zuhlke 2019;Weisbrod 1977;Zuhlke 2022). Nonprofit organizations emerge as a response to address diverse policy preferences. Scholars use community racial diversity as a proxy for heterogeneous policy preferences (Cheng, Yang, and Deng 2021;Gazley, LaFontant, and Cheng 2020;Paarlberg and Zuhlke 2019).
Government failure may also occur when government cuts funding for public services (Hill, Kiewiet, and Arsenault 2014). During the Great Recession, state and local governments faced revenue shortfalls, and public spending for parks and land conservation has yet to rebound to pre-Recession levels (Barrett, Pitas, and Mowen 2017). Nonprofit organizations emerge to supplement government funding when politicians ignore, or fail to fund, environmental issues (Delfin Jr and Tang 2006;Yandle, Noonan, and Gazley 2016).
In interdependence theory, nonprofits partner with government (Lecy and Van Slyke 2013;Salamon 1995). Governments subsidize nonprofit activities by outsourcing delivery of public services to the nonprofit sector (Lecy and Van Slyke 2013;Salamon 1987). Existing research provides evidence for this theory: government spending through grants in social services impacts nonprofit sector density (Lecy and Van Slyke 2013).
Building on previous research, I use the Annual County Resident Population Estimates (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) to construct a measure of the percent of racial and ethnic minorities by county. This serves as a proxy to measure how heterogeneous policy preferences impact the distribution of support nonprofits. To test the theory that public investment impacts nonprofit density, I measure Parks and Recreation expenditures using the Annual Survey of State andLocal Government Finances (2000-2015). Per government failure theory, counties that experience declines in funding-measured by a decrease in annual expenditures-are more likely to have nonprofits fundraising for public land. Per interdependence theory, in counties that experience increases in government funding-measured by an increase in annual expenditures-nonprofits are more likely to fundraise for public land. 4 Support nonprofit philanthropy requires financial resources. Wealthy and well-educated individuals are more likely to volunteer and donate (Gilens 2012;Schlozman and Tierney 1986;Schlozman, Verba, and Brady 2012;Verba et al. 1993;Verba, Schlozman, and Brady 1995). Perhaps as a result, support nonprofits are often located in well-resourced communities (Gazley, LaFontant, and Cheng 2020;Nelson and Gazley 2014;Paarlberg and Gen 2009). There are notable exceptions. In Los Angeles, support nonprofits worked in partnership with public agencies to address disparities in park access (Rigolon 2019). To examine the relationship between community wealth, education and support nonprofit presence, I use the Small Area Income and Poverty Estimates to measure median household income and the U.S. Census Bureau's American Community Survey to measure educational attainment, or percent of the population with a bachelor's degree or more (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015).
Variation in county characteristics leads to inequality in access to safe, public green space. Wealthy communities have better access to protected public land (Wen et al. 2013;Williams et al. 2020). I posit that nonprofit philanthropy compounds this inequity: counties with more protected land are more likely to benefit from support nonprofits. I use data from the National Land Trust Census to measure the percent of land that is protected by conservation by county for 2015 (National Land Trust 2015).
Finally, I measure county partisan composition using the National Neighborhood Data Archive from 2004 to 2015 to control for the well-established connection between partisanship and support for environmental policy (Casola et al. 2022;McCright, Xiao, and Dunlap 2014;Switzer 2019). Counties with more Democratic voters are likely to have an active support nonprofit. The analysis includes state-fixed effects to control for variation in state laws regulating nonprofits and land conservation (Owley and Rissman 2016). 2.2 | How does nonprofit mobilization impact political advocacy for public land conservation?
As partners in public service, support nonprofits may influence parks and land conservation policy. Existing research on nonprofit mobilization outlines three ways in which nonprofit organizations participate in public policy (Fyall 2017). First, nonprofits help overcome collective action issues that keep voters from organizing, providing an avenue for individuals to express their policy preferences and mobilize around public issues (Berry and Wilcox 2015;Child and Grønbjerg 2007). Nonprofit advocacy can take many forms, including lobbying, voting, protests, demonstrations, and education campaigns (LeRoux 2007;Prentice and Brudney 2017). Second, nonprofits are embedded in policy communities or networks with shared interest (Fyall 2017;Zahariadis 2014). Support nonprofits may use their networks to mobilize for conservation ballot initiatives. Third, support nonprofit leadership or volunteers may act as policy entrepreneurs, advocating for public funding for land conservation. Policy entrepreneurs promote policy change through coalition building (Mintrom and Norman 2009). Following these theories, it is likely that support nonprofits are influential in getting conservation issues on the policy agenda-in this case, on the ballotand insuring their success through voter mobilization.
Alternatively, nonprofit philanthropy may undermine public mobilization for land conservation initiatives. Support nonprofits are fulfilling a critical public responsibility, subsidizing "the provision of public services through the use of their own resources", such as volunteers and donations (Schatteman and Bingle 2015, 77). Governments and public officials may prefer "donation over taxation", pushing for nonprofit philanthropy to replace government funding for public land (Schatteman and Bingle 2015, 75). Last, voters may perceive nonprofit philanthropy as more efficient and responsive than government funding to protect public land (Holmes 2012;Walls 2014). Relying on nonprofits to provide public services weakens government capacity (Cammett, MacLean, and Gough 2014). In sum, county reliance on support nonprofits to maintain public land may undermine institutional support for public land conservation.
These theories offer a mechanism by which nonprofits affect county conservation initiatives. The analysis focuses on outcomes: examining the correlation between support nonprofit philanthropy and vote share in favor of land conservation efforts and ballot success. To measure nonprofit impact on vote share and ballot success, I use The Trust for Public Land's LandVote Database, which provides a comprehensive history of conservation finance measures on county ballots (Landvote.org from TPL 2000. The Trust for Public Land categorizes the primary purpose for each conservation finance ballot initiative. From 2000 to 2015, the 313 conservation finance measures in the sample focused on preserving open space, farmland, watershed protection, wildlife habitat, and parks and recreation. The majority of conservation finance ballot initiatives occurred pre-Recession, with 261 initiatives from 2000 to 2008, and 52 initiatives from 2009 to 2015. County conservation finance initiatives are rare: only 0.62% of counties voted for conservation issues through ballot initiatives.

| METHODOLOGY
The National Center for Charitable Statistics provides a sample of nonprofits dedicated to philanthropic support of parks, natural resources, and land conservation from 2000 to 2015. The sample includes all active nonprofits that are required to file an IRS form 990 (Lampkin and Boris 2002), including organizations with $25,000 or more in gross receipts in 2000 and more than $50,000 in gross receipts in 2010. Because filing requirements changed during the time period, I restrict the sample to organizations with gross receipts over $50,000 (Bowman, Tuckman, and Young 2012;Calabrese 2011;Froelich, Knoepfle, and Pollak 2000). The median nonprofit revenue for 2015 was $162,635. Relying on data from the IRS excludes many nonprofit organizations from the analysis, including nonprofits that do not meet the filing threshold of $50,000 in gross receipts, nonprofits that are IRS-registered but are not reporting, and nonprofits that register with the state and not the IRS (NCCS; Grønbjerg, 2002). As a result, the dataset is biased towards large, formalized, and long-standing nonprofits, undercounting informal grassroots organizations that rely on volunteers. 5 Nonprofits were selected based on the National Taxonomy of Exempt Entities (NTEE), the system used by the IRS and NCCS to classify public charities by activity (Child and Grønbjerg 2007;Joassart-Marcelli and Wolch 2003;Twombly 2002). Recent research highlights issues using NTEE codes to define a sample for analysis: for example, organizations can only select one NTEE code to encompass all of their activities, and NTEE codes are static and do not change over time (Fyall, Moore, and Gugerty 2018). To manage possible discrepancies within the NTEE classifications, I use a dual classification method. First, I subset the data based on relevant NTEE codes, including single organization support groups focused on fundraising and fund distribution for parks and land conservation. 6 Next, I audit organizations using a text categorization formula, which categorizes text strings based on keywords with an array formula. For example, the formula "reads" nonprofit names and flag nonprofits with national, international, and global in the title. The nonprofits are matched to the counties they serve based on zip code. 7 For more information on this process, see Data S1.
To examine the characteristics of counties with large support nonprofits, I match nonprofits with county characteristics, like median household income, percent of residents with a bachelor's degree or more, racial and ethnic composition, and urban and rural codes from the Small Area Income and Poverty Estimates (SAIPE) (US Census Bureau, 2000b), the Annual County Resident Population Estimates, and the American Community Survey from 2000 to 2015 (US Census Bureau, 2015). In addition, I control for county partisan composition using the National Neighborhood Data Archive from 2004 to 2015 (Chenoweth et al. 2020). Finally, I include data on parks and conservation funding and access: I measure state-level changes in total parks expenditures over time from the Annual Survey of State and Local Government Finances (US Census Bureau, 2000a), and a measure of the percent of land protected by conservation trusts by county from the National Land Trust Census. All continuous independent variables are logged. The key dependent variable measure variation over time, with a binary variable measuring whether or not a county has a support nonprofit that files taxes with the IRS from 2000 to 2015 (0/1). The data are analyzed using logistic regression.
To illustrate the relationship between support nonprofits and political advocacy, I use county-level data on political initiatives to raise public funds for land conservation from The Trust for Public Land's LandVote Database (Landvote.Org from TPL, 2022). The database provides a comprehensive history of conservation finance measures on county ballots from 2000 to 2015. The key dependent variables are longitudinal: (1) whether or not a county passed a conservation ballot initiative (0/1) and (2) vote share in favor of public funding for county conservation ballot measures (0-100). I use logistic and ordinary least squares linear regression, and all models include statelevel fixed effects. State fixed effects help to address variation: states have diverse legal, social, political, and economic characteristics which impact subnational governments funding decisions related to conservation and parks, as well as nonprofit formation and advocacy (Kiesecker et al. 2007;Owley and Rissman 2016). In addition, access to public land is not evenly distributed across states, and state-fixed effects partially accounts for this variation.

| RESULTS
The analysis proceeds in two parts: first, I present descriptive statistics on counties with one or more support nonprofit focused on public parks and conservation efforts. I use logistic regression to test for associations between large nonprofits and county socioeconomic status and demographic characteristics. Second, I examine the relationship between nonprofit presence and vote share in favor of land conservation. Together, the T A B L E 1 Characteristics of counties with and without support nonprofits.

With support nonprofits
Without support nonprofits Two sample T-test analysis highlights how nonprofit philanthropy and mobilization shapes access to public land and conservation efforts.

| Characteristics of counties with support nonprofits focused on protecting public land
Counties with support nonprofits have higher median household income, fewer residents living in poverty, and higher proportion of the population with a bachelor's degree or more. Among counties with support nonprofits, the average median household income is $58,028-over $10,000 more than counties without support nonprofits. Disparities persist in education, as well: the average percent of the population with a bachelor's degree or more is 33.0% in counties with support nonprofits, compared to 21.5% in counties without a nonprofit. Counties with support nonprofits are much more likely to be in a metropolitan area (68.9%), compared to counties without a support nonprofit (31.1%). Counties with support nonprofits are more diverse, with 28% of the population identifying as a racial/ethnic minority (compared to 22.6% in counties without support nonprofits). Finally, clear differences appear in the parks and conservation landscape in counties with and without support nonprofits. Counties with support nonprofits have about 17% of land that is protected by land trusts, compared to 12% in counties without support nonprofits. These differences are statistically significant across the two samples. While state and local park expenditures are higher in counties with support nonprofits, this difference in means is not statistically significant. The average number of support nonprofits by county is 1.55 (standard deviation, 1.32) ( Table 1). I estimate the probability of having a large support nonprofit (0/1) that files annually with the IRS using county socioeconomic, demographic, and political data: median household income, percent of the population that identifies as an ethnic/racial minority, percent of the population with a bachelor's degree or more, and whether or not the county is in a metro area. I include variables measuring county partisan composition. To examine whether support nonprofits mobilize to fill in the gaps in public funding after the Great Recession, I include a variable that measures the annual change in state government expenditures for parks and recreation and a dummy variable to indicate the period before and after the Great Recession. Finally, I include a variable that measures the percent of land that is protected by land conservation efforts by county. The model is a logistic regression with state-fixed effects. Following best practices for nonprofit data (Tinkelman and Neely 2011), I use the "robust" command to correct for heteroskedasticity (Rogers 1994). The results are presented in Odds Ratios in Table 2.
What characteristics predict whether counties have a support nonprofit? Key theories of nonprofit distribution include government failure and interdependence theory. Nonprofits form due to government failure to provide services, as a result of heterogeneous policy preferences (measured by racial and ethnic diversity) or changes in financial support (measured by annual change in state expenditures for parks and recreation). Diverse policy preferences, proxied by racial/ethnic diversity, increase the odds of a county having a support nonprofit. An increase in the logged percent of the population in the county that identifies as a racial or ethnic minority (logged scale: À.9 to 4.6, original scale: .39-97.9 percent) makes it 1.5 times more likely to have a support nonprofit, all else equal.  There is little evidence that support nonprofits are mobilizing to fill in gaps in state and local expenditures for parks, and marginal evidence in favor of interdependence theory. The dummy variable measuring the post-recession period is negatively associated with odds of having a support nonprofit, although the coefficient is not statistically significant. An increase in state and local government annual expenditures increases the odds that a county has a support nonprofit, although it does not reach conventional levels of statistical significance.
Community characteristics associated with support nonprofit activity include income (measured by median household income), education (measured by percent of residents with a bachelor's degree or more), and existing conservation land (measured by the percent of land that is protected in conservation trusts). Median household income is positively correlated with the odds a county has a support nonprofit, although the coefficient does not reach conventional levels of statistical significance. Education has the greatest impact on the distribution of support nonprofits. A one-unit increase in the logged percent of population with a bachelor's degree or higher (logged scale: .39-4.37, original scale 0-79.14 percent) increases the odds that a county has a support nonprofit ninefold, all else equal.
T A B L E 3 County characteristics and percent voting in favor of ballot initiative for conservation finance.

Logistic regression-Probability ballot passed
Linear regression-Predicted vote share Counties with a greater percentage of land protected by conservation trusts are more likely to have a support nonprofit: a one-unit increase in the percent of protected land (logged scale: À10.7-4.58, original scale 0.00-97.18 percent) correlates with a 21% increase in the odds of a county having a support nonprofit, holding all else constant. Perhaps unsurprisingly, counties with more land in conservation trusts are more likely to benefit from support nonprofits.
Several control variables are related to the distribution of support nonprofits. Counties in large metropolitan areas are more likely to have one or more support nonprofits mobilizing for public land, compared to counties outside of metropolitan areas. Partisanship also emerges as a predictor of support nonprofit distribution: counties with a large percentage of Democratic voters are much more likely to have a support nonprofit, all else equal.

| Support nonprofits and political mobilization for public land conservation
Support nonprofits may impact political mobilization for public conservation initiatives. Voters may view nonprofit philanthropy as a replacement for government land conservation efforts, eroding support for public spending for conservation. Alternatively, support nonprofits may provide a channel for voters to engage in the political process and increase demand for public funding. Voting opportunities for land conservation are not evenly distributed across the United States. There are statistically significant differences among counties with and without ballot initiatives: counties with conservation finance measures on the ballot are wealthier, more diverse, better educated, more liberal, and more likely to be in a metropolitan area. Counties with conservation ballot initiatives have more land protected by conservation trusts and higher average state and local expenditures for parks and recreation. The descriptive statistics on the distribution of county ballot initiatives are presented in Data S1 (Table 3).
The analysis examines the probability of success among counties with conservation issues on the ballot. The key dependent variables include: (1) whether or not the land conservation ballot initiative passed (0/1), and (2) the percentage of votes in favor of public funding for county conservation efforts (0-100). Model 1 and Model 3 estimate the probability of passing a land conservation ballot initiative and predicted vote share, with a dummy variable measuring whether or not the county has a support nonprofit 8 as well as state fixed effects. Model 2 and Model 4 include important control variables, such as logged household income, percent of the population identifying as a racial/ethnic minority, percent with a bachelor's degree or more, percent Democratic voters, percent land protected by land conservation trusts, parks and recreation expenditure annual change, and the finance mechanism proposed to fund the ballot initiatives. Adding these important explanatory variables limits the sample size.
In the bivariate analysis with state-fixed effects, there is a clear correlation between support nonprofits and mobilization for public conservation initiatives. Counties with one or more support nonprofit are more likely to pass public initiatives in favor of land conservation, and have higher vote share in favor of conservation legislation. The odds of legislative success in counties with support nonprofits are 175% higher than counties without support nonprofits. There is a correlation between support nonprofits and vote share, as well: counties with support nonprofits have 4.43% more votes in favor of funding for public land.
Model 2 and Model 4 include important control variables which impact the statistical power of the model. Including county demographic, socioeconomic, and political characteristics, the relationship between support nonprofits and political mobilization for land conservation is no longer statistically significant. In Model 2, the coefficient remains substantively significant: the odds of passing legislation in favor of land conservation efforts in counties with support nonprofits are 205% greater than counties without support nonprofits, all else equal. Including control variables in Model 4, counties with one or more support nonprofits have 3.9% more votes in favor of conservation legislation. The coefficients are not statistically significant, and no other predictors meet conventional levels of statistical significance.
In sum, there is evidence that support nonprofits are engaging in political mobilization for conservation policy: support nonprofits improve the odds of successful ballot initiatives and increase votes in favor of land conservation funding. While this manuscript focuses on outcomes, future research should examine the mechanism by which support nonprofits shape conservation policy: for example, support nonprofits may engage volunteers in collective action or leverage organizational networks to mobilize voters around conservation issues. Alternatively, support nonprofits may act as policy entrepreneurs, advocating to put conservation finance ballot initiatives on the agenda and to insure their passage.

| DISCUSSION
The prevalence of support nonprofits brings up key questions of equity and access: Which communities' benefit from support nonprofits? Does nonprofit presence suppress or increase votes in favor of public funding for land conservation? The manuscript brings together quantitative data from multiple sources, including nonprofit data from the NCCS, county level demographic data from the Census Bureau, The National Land Trust Census and The Trust for Public Land's LandVote data on county level ballot initiatives for public funding for land conservation.
Before discussing key findings, there are important limitations of the study. First, the sample of support nonprofits is selected from the NCCS, which includes nonprofits that registered with the Internal Revenue Service (IRS) as 501c3 public charities with gross receipts over $50,000. Data availability restricts the findings to large, formalized support nonprofits. This has two distinct limitations. First, the analysis likely undercounts the number of support nonprofits over time. Second, inferences from the analysis may not apply to small nonprofits that are not required to file with the Internal Revenue Services. Existing research shows no statistically significant relationship between government funding, poverty, education and nonprofit formalization (Grønbjerg 2002). While it is possible that small support nonprofits focus on lowincome communities and redistribution, it is likely that small support nonprofits also form in counties that are more liberal and well-educated.
There are issues of representativeness and accuracy in the NCCS data. Representation issues include categorization of organizational activities and geographic location. To address categorization issues, I use a dual categorization method to ensure accuracy of the sample. Geographic issue include differences between nonprofits registered tax location and their service area (McDougle 2015). Support nonprofits are established to protect and maintain protected land, and focus on local, municipal or county public parks or specific parcels of land. The physical space that parks, beaches, watersheds, lakes, estuaries and forests occupy will not always match perfectly with administrative boundaries and jurisdictions. This manuscript follows best practices for NCCS nonprofit research and matches nonprofits to the county level, to mitigate errors in geographic location by zip code (Bowman, Tuckman, and Young 2012). Aggregating to the county level also makes intuitive sense for the study of parks and land conservation efforts, as benefits accrue mostly to close residents but also impact broader communities. For additional discussion of geographic issues, please see Data S1.
With these limitations in mind, this project makes important contributions to the research on nonprofit advocacy and public land. Building on research on state and federal support nonprofits by Gazley et al. (2018), this manuscript documents the rapid growth of local support nonprofits. First, the manuscript makes a novel contribution towards understanding the geography of support nonprofits. Demographically diverse and liberal counties in metropolitan areas are more likely to have one or more support nonprofit fundraising for public parks and land conservation. In light of these findings, it is tempting to argue that support nonprofits are remedying endemic issues of racial equity in access to public land: racial disparities exist in park acreage and quality (Rigolon, Browning, and Jennings 2018). These findings should be interpreted with caution, as the focus on counties may obscure withincounty variation in support nonprofit advocacy. For example, research that examines the distributional impact of support nonprofit philanthropy and park access in US cities finds that "nonprofit sector promotes park access of all racial/ethnic groups, [but] more benefits accrue to whites than to other racial/ethnic groups" (Cheng, Yang, and Deng 2021, 3). There are important differences across racial and ethnic groups in use of, and experience in, parks and natural resources. People of color experience discrimination and racial conflict in public spaces (Gobster 2002;Holson 2018;Rigolon 2019). Equitable access to public land is the responsibility of state and local governments, not support nonprofits.
There is no evidence that support nonprofits are helping communities hit hardest by the Great Recession: although the results are not statistically significant, counties that saw increases in state and local government funding for parks and recreation were more likely to have support nonprofits fundraising for parks and land conservation. This provides partial evidence for interdependence theory: support nonprofits work in partnership with state and local government, and are located in counties that have more government funding for public land. Support nonprofits may supplement public services when government fails to meet demand or complement government through mutual partnerships, depending on context (Young 1999(Young , 2000.
Second, counties in metropolitan areas with welleducated residents have increased odds of support nonprofits fundraising for parks and conservation efforts. Education has a substantively and statistically significant impact on the distribution of support nonprofits: all else equal, an increase in the percent of population with a bachelor's degree or more improves the odds that a county has a support nonprofit ninefold. Counties with more land in protected trusts are more likely to have support nonprofit, all else equal. In sum, counties with well-educated residents and existing land protected by conservation are much more likely to benefit from support nonprofit advocacy and philanthropy. There is no clear evidence that support nonprofits are serving a redistributive purpose.
Third, the analysis suggests that counties with support nonprofits engage in voter mobilization for county conservation efforts. In counties with support nonprofits, the odds of successful conservation ballot initiatives are almost three times greater than counties without support nonprofits. Counties with one or more support nonprofits have 4.43% more votes in favor of conservation legislation, all else equal. The multivariate analysis did not yield statistically significant results, so the results should be interpreted with caution. In addition, the analysis does not include other factors that likely influence the success of conservation finance measures, such as local institutional context, economic development, or prevalence of manufacturing industries (Le on-Moreta 2021).
The uneven distribution of support nonprofits has implications for equity and access to public land. Existing research has demonstrated the link between community wealth and nonprofit philanthropy in cities (Brecher and Wise 2008;Shi and Cheng 2021), and this analysis builds on and challenges these findings with a national, longitudinal perspective. While the descriptive statistics showed clear differences in median household income in counties with and without support nonprofits, education, partisanship, and diversity are the most important predictors of support nonprofit presence in the multivariate analysis.
Local nonprofits act as civic intermediaries: community-based nonprofits encourage individuals to vote, especially in diverse communities (LeRoux 2007). I argue that support nonprofits embody the role of provider and advocate simultaneously in public land conservation policy (Fyall 2017, 121). By acting as partners in public service, support nonprofits work with parks and recreation departments and land conservation trusts to maintain public land through volunteer efforts and donations. In addition, support nonprofits serve as advocates, mobilizing for public initiatives to improve land conservation efforts. Nonprofits shape public land through private philanthropy and public advocacy, benefitting highly-educated, Democratic counties and possibly augmenting existing inequalities in access to public land. Support nonprofit philanthropy shapes conservation policy in myriad ways: support nonprofits impact the fiscal health of parks and land conservation efforts (Gazley, LaFontant, and Cheng 2020), support nonprofits coordinate volunteers to provide services (Murray 2010), and support nonprofits likely mobilize voters to increase public funds for land conservation resources. In their roles as philanthropist, advocate, and service provider, support nonprofits are influential actors in conservation policy. Conservation researchers and policymakers should consider the consequences of relying on support nonprofit advocacy to maintain and protect public land and natural resources.  Figure 1. In contrast, the growth rate for all public charities filing Tax Form 990 s with the Internal Revenue Service from 2000 to 2015 also increased but at a slower rate: the number of organizations increased by 78% and total nonprofit revenue increased by 159%. 2 This manuscript examines the distribution and impact of two types of organizations: (1) Support nonprofits for conservation focus on conservation biodiversity and have two main activities: fundraising for conservation, "habitat management and restoration. They also take the lead in mobilizing public concern about biodiversity through education and outreach." (Armsworth et al. 2012, 271). (2) Support nonprofits for parks focus on improving access to, and advocating for, public parks through fundraising, community outreach, and voluntarism (Baker et al. 2010). Both types of nonprofits address shortfalls in spending for conservation and park access by state governments (Duda et al. 2022;Gazley, Cheng, and Lafontant 2018;Lerner, Mackey, and Casey 2007;Shaffer, Scott, and Casey 2002). 3 For more information on Clinton County Conservation Foundation, please see the organization's most recent Guidestar profile online: https://www.guidestar.org/profile/39-1878166 4 Existing research on interdependence theory by Lecy and Van Slyke (2013) uses variation in private funding (grants and donations) and public funding (government grants) to predict nonprofit density. This data is derived from the NCCS digitized files measuring direct public and private support. Unfortunately, these data are not available for all counties-years in the analysis. Instead, the article uses a variable measuring annual change in state and local expenditures for parks and recreation to proxy variation in public funding for parks and land conservation. 5 How might limitations in available data impact the descriptive analysis? Grønbjerg and Paarlberg (2002) use comprehensive data from state and federal registries to describe the nonprofit landscape and analyze county-level variation in nonprofit formalization. County characteristics are not significantly related to nonprofit formalization: there are few statistically significant differences in county characteristics for small, informal nonprofits and large, formalized nonprofits. However, relying on IRS registries of nonprofits results in a conservative sample of nonprofits. Growth in the support nonprofit sector is likely more dramatic than this manuscript reports, because the analysis is limited to the distribution of "large" support nonprofits, with substantial revenues and formalized operations. 6 NTEE codes include: C11, C12, C30, C32, C112, C1130, C1132, C1134, C1230, C1232, C1236, N1232, N1230, N1132, N1130. Additional information can be found in the online appendix. 7 Froelich, Knoepfle, Pollak (2001) examine the reliability and consistency of IRS 990 form filingsand find that most nonprofits accurately report total revenue and total expenses, but the data are inconsistent. As a result, I stick to simple counts of support nonprofits as an outcome variable, rather than total revenue. The supplemental materials, summarizes sample selection, data cleaning and issues related to ecological fallacy. 8 To measure the impact of support nonprofits on the probability of success for land conservation ballot initiatives, I use a binary independent variable noting presence of one or more support nonprofits rather than a count measure of nonprofits. Unfortunately, there is not enough variation in the number of large support nonprofits: 35% of counties that passed ballot initiatives have a support nonprofit, and the majority of those (20% of counties) have just one large support nonprofit, and 9.5% have two support nonprofits. In sum, there are very few counties that (a) passed ballot initiatives and (b) have more than 2 large support nonprofits.