The Dynamic Impact of Nonprofit Organizations: Are Health-Related Nonprofit Organizations Associated with Improvements in Obesity at the Community Level?

2.1 Market Failures and Nonprofit Entry/Exit

The most common explanations for the existence of nonprofit organizations are the failures of the market and government to efficiently or optimally provide certain goods or services (Douglas 1983; Hansmann 1980; Weisbrod 1975). For example, market failure theory predicts that services with significant positive externalities, such as education, should be provided by government because the market would provide education only to those willing to pay a market rate for it. Thus, we consider market and government failures to be instances when the market or government are not the ideal providers of a good or service, either due to inefficiency or sup-optimal provision. The health care market is a prime example of one specific type of market failure—contract failure (see Arrow 1963). Contract failure applies to situations of information asymmetry—when the provider of the service knows more about the quality of the service being provided than the recipient does (Hansmann 1987). This makes consumers fearful of being cheated or receiving a low-quality service. In these situations consumers turn to organizations that they consider to be trustworthy, presumably nonprofit organizations because of the non-distribution constraint, screening and selection of leaders, the presence of governing boards, and other structural aspects of the nonprofit form (Young 1989). Indeed, currently the United States has over 77,000 nonprofit hospitals and health-related organizations (McKeever 2018), indicating a substantial amount of nonprofit activity in the health care arena. Thus, while crude, nonprofit exit from and entry into the market can serve as indicator of nonprofit mobilization or activity.

Growing from and alongside failure theories, numerous nonprofit scholars have further examined how community/market factors are related to the size and structure of a community’s nonprofit sector, including how these factors are connected to the uneven distribution of nonprofit organizations across geographic communities (see, for example, Grønbjerg and Paarlberg 2001). This body of research suggests that certain contextual variables—such as the size of the community and the lack of financial resources—are related to both the demand for and supply of nonprofit organizations in a community (Bielefeld 2000; Corbin 1999; Grønbjerg and Paarlberg 2001; Marcuello 1998; Twombly 2003; Wolpert 1988). Reflecting these lines of inquiry, others have established that institutional (Baum and Oliver 1991; Twombly 2003) and social contextual factors (Saxton and Benson 2005) also relate to a community’s organizational makeup (Marquis et al. 2013).

Thus, our simple model of nonprofit entry and exit into the market is based on the failures of established institutional actors (governments and businesses) to efficiently provide services and corresponding community attributes that are known to shape the nonprofit sector in a community. The confluence of market and government failures and community attributes has given rise to a varied and growing landscape of nonprofit activity.

2.2 Mechanisms for Achieving Community Impact

The one thing that scholars seem to agree upon related to nonprofit effectiveness is the difficulty of measuring effectiveness, because of the wide variety of measures utilized and the difficulty of assessing effectiveness across different types of organizations and mission areas (Flynn and Hodgkinson 2013; Forbes 1998; Herman and Renz 2004, 2008; Sowa, Selden, and Sandfort 2004; Willems, Boenigk, and Jegers 2014). This process is further complicated by the fact that assessments of nonprofit effectiveness depend heavily on which stakeholders are viewing the nonprofits’ results (Balser and McClusky 2005; Herman and Renz 1997). Even more ambitious still is the task of assessing the collective, community-level impact of multiple nonprofits through their direct and indirect collaborations (Christens and Inzeo 2015; DiMaggio 2002; Epstein and Yuthas 2017; Flynn and Hodgkinson 2013; Kania and Kramer 2011). The focus in this article is more at the collective, or sector, level; we want to know whether the entry of new health nonprofits into a community affect a specific health outcome at the community level.

What mechanisms might explain the relationship between the nonprofit sector (and entry into the sector) and measures of community impact? We consider two possible mechanisms—nonprofit sector adaptability/innovation and collaborations. We briefly discuss how each mechanism could potentially operate and provide examples of nonprofits working on the obesity epidemic. Nonprofits’ adaptability has long been touted as an important feature of the sector. Indeed, it is because of nonprofits’ flexibility and adaptability that Salamon (2015) dubs it “the resilient sector.” Empirical research tends to support this assertion. Nonprofit organizations typically have more flexible rules and do not operate under as many constraints as government agencies (Eggleston and Zeckhauser 2002; Hall 1992). Nonprofit employees perceive much less red tape and greater personnel flexibility than employees in the public sector (Feeney and Rainey 2009). Nonprofits are also more likely to engage in participatory decision-making than for-profits (Kalleberg et al. 2006). Thus, nonprofit flexibility can appear in multiple ways in the organization, which often results from increased competition, greater expectations from donors, increasing costs, changing technology, and greater diversity of stakeholders (Wolf 1999).

Innovation is another sign of flexibility and adaptability. In developing his arguments about government failure theory, Douglas (1983) asserted “the absence of strong measures of accountability, far from being a weakness of the Third Sector, becomes a strength enabling it to undertake experiments, the benefits of which are too uncertain and too long term to be undertaken by either the commercial or the government sector” (p. 137). Other scholars have noted that nonprofits might have an advantage related to innovation because they are free from some of the constraints facing businesses and government agencies (Frumkin 2002; Light 1998; Payton and Moody 2008; Walker 2008). It is precisely because of their greater flexibility that nonprofits are able to innovate and experiment. Indeed, flexibility and adaptability are characteristics of innovative organizations (Hurley and Hult 1998; Kitchell 1995). Nonprofits organizations tend to value innovation and perceive themselves as innovative entities (Chinnock and Salamon 2002; Salamon, Hems, and Chinnock 2000). The capacity of the nonprofit sector to be flexible and adapt allows organizations to innovate and find new solutions to public problems (Smith 2002). These innovations pave the way for nonprofits to create lasting change at the community level.

One illustration of nonprofit adaptability is the Girl Scouts of America. The organization started a Girl Scout Research Institute to not only evaluate their own programs but to collect data on girls and their development that can be shared with other organizations and policy-makers (see https://www.girlscouts.org/en/about-girl-scouts/research.html). The Girl Scout Research Institute published a report in 2006 about girls’ perceptions of weight and healthy living. The preface to the report indicates the organization’s commitment to adaptability: “As the lives of girls continue to change, so does Girl Scouts of the USA—always adapting to ensure that girls in the twenty-first century are provided with a program that is engaging, interesting, and relevant to their needs” (Girl Scout Research Institute 2006, 4). As a result of this and other studies, the Girl Scouts created new programs focused on healthy lifestyles, such as GirlSports, a K-12 program designed to encourage girls to participate in sports. Girls Scouts sought information about an important health topic and subsequently adapted by developing new programs and offerings for the girls they serve in response to what they learned.

The entry of a new nonprofit organization can also be a signal of adaptations to meet needs in the environment. For example, the Obesity Action Coalition (OAC) is a nonprofit whose mission is “to elevate and empower those affected by obesity through education, advocacy and support” (OAC web page 2019). The organization was formed in 2005 when the founders were made aware of the need for such an organization: “During a meeting of legislators, a congressperson asked the question – “Who represents patients who are affected by obesity?” It was then that a legislator pointed out a serious need – a group whose only focus is on those affected by obesity. With this, the OAC was formed in 2005 with the goal of building a national coalition of those who are living with and/or affected by obesity” (OAC web page 2019). The entry of this nonprofit organization was in direct response to a perceived unfilled health need and illustrates a nimble response from the nonprofit sector when an unmet need became apparent.

The second potential mechanism is collaborations and partnerships. Public problems, like obesity, are complex and require multi-faceted solutions. Many time the best results arise when multiple organizations work together in a coordinated and systematic way. Formalized collaborations involve ongoing relationships among organizations through sharing resources and programs in some way (Kohm, La Piana, and Gowdy 2000). Among the primary reasons justifying a collaboration are cost savings and enhancements to service delivery, including both better outcomes for individuals and wider reach to more individuals (Arsenault 1998; Hill and Lynn 2003; Sowa 2009). Indeed organizations might enter collaborations because it is a vehicle for achieving outcomes they cannot achieve on their own (Snavely and Tracy 2002). Collaborations are also a method for diffusing important innovations through the nonprofit sector (Faems, Van Looy, and Debackere 2005). Collaborations can help reduce duplication in services, improve coordination of service-delivery, and enhance provider responsiveness (Martin et al. 1983). Through collaboration, organizations can more effectively deal with a more complex issue by providing a more wholistic solution and improving overall client access to necessary programs and services (Beatrice 1991; Farel and Rounds 1998: Poole and Van Hook 1997). Empirical research demonstrates that collaborations can yield many positive outcomes, including improved client outcomes, although success is often contingent on characteristics of the collaboration and participating organizations (Arya and Lin 2007; Gazley and Guo 2015; Jennings and Ewalt 1998; Linden 2003; O’Regan and Oster 2000; Selden, Sowa, and Sandfort 2006).

An example of the collaboration mechanism at work is Shape Up Somerville (https://www.somervillema.gov/departments/health-and-human-services/shape-somerville). Shape Up Somerville is a collaboration among 50 community-based nonprofits, government agencies, educational institutions, foundations, and other partners and supporters. The goal of the collaboration is to encourage healthy eating and active living among Somerville residents. Some of the collaboration activities included changing foods served in schools, improving physical education programming, and creating more access to healthy foods through an urban agriculture ordinance that changes zoning to allow for farming in city areas. The collaboration has been widely recognized for getting results (Economos et al. 2007, 2009; Burke et al. 2009; Flood et al. 2015; Goldberg et al. 2009). In fact, one recent study found a significant decrease in body mass index (BMI) among the parents of children who were getting a Shape Up Somerville program in their school (Coffield et al. 2015). The success of the endeavor is largely credited to the partnerships involved, including funders and political leaders (Shape Up Somerville 2013). Having multiple partners not only increased partner buy-in and support, but it allowed Shape Up Somerville to reach more residents because they are all involved in different organizations.

Sometimes the entry of a new nonprofit organization signals a successful collaboration. As a collaboration grows and formalizes over time, it might require a new organization to manage and help to coordinate the activities of the partner organizations. This was the case with Healthy Suffolk (http://healthysuffolkva.org/about/). In 1998, community leaders in Suffolk, Virginia convened a Suffolk Partnership for a Healthy Community to learn and understand the community’s health needs and to bring interested parties together. Over time the collaboration continued to grow and Suffolk Partnership for a Healthy Community (SPHC) received 501(c)3 status in 2006. SPHC’s partners include other community organizations, businesses, and government agencies. The organization rebranded itself as Healthy Suffolk in 2016. This example illustrates how a new nonprofit entering the sector can be an indicator of a successful ongoing collaboration—a collaboration with the potential to affect community health outcomes.

These two potential mechanisms can help to explain how the nonprofit sector—and entry and exit into the sector—might affect community-level outcomes. Considering the scarcity of models and large sample-size empirical analyses engaging the question of nonprofit community impact, our present purpose is to offer some preliminary tests of this question–in its simplicity–and add commensurate details in future work. We preface our description of data and analysis with a short section intended to better situate our research question in the present substantive policy area of interest: obesity.

2.3 Health Nonprofits and Community Impact: A Focus on Obesity

3.1 Data

The first source of data was the Core Financial Files from the National Center for Charitable Statistics (NCCS), which comprise nonprofit organizational and financial data from the United States’ Internal Revenue Service Form 990 that nonprofit organizations in the United States are required to submit annually. The second source of data was county-level obesity rates collected as part of the Behavioral Risk Factor Surveillance System (BRFSS) for years 2004–2012 (Centers for Disease Control and Prevention n.d.). The BRFSS is an ongoing, yearly, state-based telephone survey, supported by the Centers for Disease Control and Prevention, of non-institutionalized adult population.

We used the 2004 through 2012 NCCS nonprofit data, which includes 3,138,941 nonprofit organizations across nine years of data. Because our unit of analysis was the county-year (i. e. data for each county, by year), we collapsed this organization-level dataset into a dataset comprised of aggregated nonprofit data at the county level for each of the nine years. After the nonprofit data were collapsed, we used the county-level Federal Information Processing Standards (FIPS) code and year to append the health indicators, and demographic, data to the aggregated nonprofit dataset. We used the 2000 decennial Census data for years 2004–2009 (US Census Bureau 2000) and the 2010 decennial Census data for 2010–2012 (US Census Bureau 2010).

We excluded observations in Alaska because it has boroughs instead of counties and could not be comparably matched up with other data. To correct for missing data for the percent white, we extrapolated these data using the ipolate command in Stata (with the epolate option). This approach uses a linear function of the known responses for the variable over the nine years to fill in missing values. In other words, the values from the non-missing years were used to impute values for values from years with data, using information for the respective county. The final usable number of observations was 17,397 county-years. All data were publicly available so no ethics approval or informed consent was required for these analyses.

3.2 Variables

The dependent variable for our analysis was the percent of county residents 18 years and older who were classified as obese (BMI greater than 30 kg/m²). BMI was calculated using individuals’ self-reported heights and weights in the BRFSS. This measure was age-adjusted to the 2000 US standard population based on age groups 18–44, 45–64, and 65 years or older.

The first independent variable was the number of health-related nonprofit organizations per 10,000 individuals in the county. We used the National Taxonomy of Exempt Entities (NTEE) codes under the major group of “Health” (see http://nccs.urban.org/classification/ntee.cfm) from the NCCS data to obtain the number of health nonprofits in the county. When the NCCS data were collapsed at the county level, we summed all the nonprofit organizations with a relevant health related codes (see NTEE letters^[1]) assigned involved in health care (NTEE code of E), mental health (code of F), diseases and disorders (code of G), and medical research (code of H). This total was then divided by the population estimate for the county and multiplied by 10,000 to obtain the number of health nonprofits per 10,000 individuals in the county. In our analysis, we used the one-year and two-year difference in the number of health nonprofits per capita to represent the entry or exit of health nonprofits in the county.

The second independent variable was the mean fiscal health of all health nonprofit organizations in the county, calculated by subtracting a nonprofit’s end of year liabilities from its end of year total assets, and then dividing by the organization’s total expenditures. This measure reflects the length of time an organization can continue to operate in the absence of new income. The mean fiscal health variable is an aggregated measure representing the average financial health of all health nonprofits in the county.

The third independent variable was a dummy variable capturing whether there was a nonprofit hospital in the county, which reflects the health self-sufficiency of a county. It is also an important control for the fiscal health variable because nonprofit hospitals have such large budgets relative to other health nonprofits that they skew the fiscal health variable.

Other county-level covariates included in the model were: education (percent of people with a college degree); median household income; and percent of the population that was white.

3.3 Statistical Methodology

We used a cross-sectional time-series design using a random effects model, clustering the standard errors at the county level, and using the lead of the dependent variable. Specifically, we regressed the independent and control variables on the values of the dependent variable for the subsequent year(s). We used the lead variable because we expected it would take at least a year for any effects of the independent variable on the dependent variable to manifest. We estimated two models—the first had a one-year lead, and the second model had a two-year lead on the dependent variable. We also tested for interaction between urban/rural status and number of nonprofits. Stata was used for all statistical analysis and a p-value of 0.05 was used for statistical significance.