The Influence of Spatial Density of Nonprofits on Crime

Abstract

Nonprofits have been posited to have benefits for communities and neighborhoods, including reduction of crime. Empirical research has provided mixed results. Drawing on both criminological and theories of public organization and organizational ecology, this article examines the effects of nonprofit density on neighborhood crime in Denver, Colorado, between 2010 and 2015. Controlling for demographic and environmental influences, as well as the effects of spatiotemporal autoregression, results suggest that a higher density of both place-based and generic nonprofits ameliorate rates of crime. In contrast, nonprofits focused on crime reduction have a significant positive curvilinear effect on several types of crime. Findings also suggest that while nonprofits at the micro-ecological level increase crime to a point, the impact is negligible compared with other factors. Moreover, a relatively high density of place-based or generic nonprofits may have some benefits at higher densities, indicating a synergistic effect.

Keywords

Bayesian spatiotemporal nonprofit crime social capital social disorganization

Introduction

There has long been an interest in the effect of nonprofits and voluntary organizations on neighborhood crime, with social disorganization and social capital theories suggesting these organizations should reduce crime by improving collective efficacy and informal social controls (Bursik 1988; Morenoff, Sampson, and Raudenbush 2001; Sampson and Groves 1989; Shaw and McKay 1942). However, support for this proposition has been weak and inconsistent (Jacoby 2018; Morenoff, Sampson, and Raudenbush 2001; Peterson, Krivo, and Harris 2000; Wo 2016). While most studies find some benefit, if not for crime reduction then for other social processes (Peck 2008; Small and McDermott 2006; Small et al. 2008), others find that nonprofits may attract crime (Groff and Lockwood 2014; Slocum et al. 2013). In part, these findings may be attributable to the ways in which past studies have both conceptualized and measured nonprofits.

Research on nonprofits and neighborhood crime could be improved in three ways. First, improved theoretical development is needed to link organizational perspectives with criminological theories and inform how nonprofit density may affect neighborhood crime. Second, more refined methods should be utilized to capture whether organizations are engaging in substantive efforts likely to have an impact on crime. Third, improvement is needed in measurement of micro-ecological contexts which go beyond demographics to better account for environmental factors likely to influence the spatial distribution of both nonprofits and crime. Using longitudinal spatiotemporal data from Denver, Colorado, this article examines how a higher density of nonprofits with a place-based or crime-reduction focus effect crime at the micro-ecological level, independent of the effects of demographics, environmental factors, and spatiotemporal autoregression. The article seeks to advance both theory and research on the effects of nonprofits on crime by advancing a methodological focus that seeks to overcome limitations in existing measurement and theory.

Past Research on Nonprofit Influences on Neighborhood Crime

At least since the resurgence of interest in social disorganization and social capital theories in the late 1980s, there has been an interest in the relationship between voluntary and charitable organizations and neighborhood crime rates. From the social disorganization perspective, nonprofits are theorized to increase collective efficacy along with local social controls, thereby mitigating social disorganization and reducing crime (Bursik 1988; Bursik and Grasmick 1993; Sampson and Groves 1989; Skogan 1988; Tomeh 1973). Related theories of social capital have also posited an indirect effect on crime through the increased bonding social capital resulting from organizational connections (Coleman 1988; Putnam 2000).

Despite expected theoretical benefits, empirical evidence for these assertions is weak, with critics attributing the absence of anticipated effects to either inadequate methodology or competing theoretical explanations (Hall 1992; Sharkey, Torrats-Espinosa, and Takyar 2017; Slocum et al. 2013; Wo, Hipp, and Boessen 2016). Contrary to social capital perspectives, those advocating environmental theories posit that nonprofits may increase crime by attracting crime-prone individuals or those who are at higher risk for victimization. This suggests that nonprofit density, especially for those providing direct services locally, may increase crime (Brantingham and Brantingham 1995; Cohen and Felson 1979).

Others have noted that early work primarily focused on the presence of nonprofits without distinguishing between functions and activities (Morenoff, Sampson, and Raudenbush 2001; Peterson, Krivo, and Harris 2000). This is problematic because nonprofits and volunteer organizations are so varied that they barely form a coherent sector (Hall 1992), so the presence of a generic nonprofit does not necessarily have a consistent theoretical connection to neighborhood crime (Sharkey, Torrats-Espinosa, and Takyar 2017; Wo, Hipp, and Boessen 2016). Consequently, such simple operationalizations likely mask more complex phenomena, at least partially accounting for the disparate findings in the literature.

Recognizing that the features of nonprofits are important, several recent studies have sought to better measure nonprofit types. Slocum et al. (2013) looked at the impact of nine types of organizations, such as places of worship, government agencies, schools, charitable organizations, vocational training centers, and organizations that act as bridges to the larger community (Roman and Moore 2004). Consistent with past research, the authors find no relationship between a simple measure of the number of organizations in an area and crime. However, in support of social capital and social disorganization theories, they find that the presence of bridging organizations correlates with a decrease in violent crime, whereas the presence of family/child welfare organizations correlates with decreased property crime. Consistent with environmental theories of crime, the authors also find a positive association between schools and property crime (see also Kautt and Roncek 2007; Roncek and Faggiani 1985; Willits, Broidy, and Denman 2013). In addition, they find a positive correlation between violent crime and organizations serving high-risk clients along with charitable religious organizations in disadvantaged areas, but find a negative correlation in more affluent areas suggesting the importance of context along with organizational purpose.

Wo, Hipp, and Boessen (2016; Wo 2016) offer one of the first analyses to utilize a multicity sample to examine how characteristics of nonprofits influence crime at the neighborhood level. They examine how age and type of organization influence crime within a half-mile radius of census blocks. Their findings support a modest reduction of crime near nonprofits, although these effects differ based on the type of organization and type of crime, with some nonprofit types correlated to increases in some crimes. Likewise, Wo, Hipp, and Boessen (2016) describe a “delayed impact scenario” in which nonprofits over 10 years old reduce crime while younger nonprofits do not. From this, it can be inferred that those neighborhoods with nonprofits that have successfully weathered the initial stages of nonprofit formation to achieve a stable presence, and especially those that engage in efforts to build social capital, are likely to be most beneficial for communities (Chambré and Fatt 2002).

In another recent study, Jacoby (2018) assesses the degree to which organizations that engage in bridging activities influence crime. She argues that, unlike nonprofits in general, bridging organizations actively engage in community decision-making and are therefore more likely to reduce crime. In support of this, she finds that both the presence and capacity of these organizations reduce crime rates in New York City neighborhoods, especially property crimes. Interestingly, areas with the highest and lowest baseline crime rates show these effects, but nonprofits in moderate crime areas have no effect on crime. This study suggests that beyond type, organizational practices and relationships are likely to matter.

Looking at the macro-level effects of nonprofits, Sharkey, Torrats-Espinosa, and Takyar (2017) find that the decline in crime rates between 1990 and 2010 was greater in cities with an increase in nonprofit organizations targeting violent crime and community development. Their study accounts for endogeneity between crime and these types of organizations by employing a longitudinal design and instrumental variable approach to show that organizational growth led to a reduction in both property and violent crime in 264 U.S. cities. Their work suggests that at aggregate levels, a greater concentration of nonprofits has beneficial effects. What remains unclear is why there is not more robust support for such an effect at lower levels of aggregation.

Although the improved methodologies of recent studies have been more supportive of the social capital/social disorganization perspectives, strong support has yet to emerge. Even where support has been found, effects on crime rates are often small, being fractions of a percentage to several percentage points annually. Likewise, some types of nonprofits continue to produce a net increase in crime, more in line with ecological perspectives. For instance, Groff and Lockwood (2014) find a criminogenic effect of halfway houses and drug treatment facilities in Philadelphia neighborhoods. And, Slocum et al. (2013) find that, in some contexts, certain types of organizations are associated with increased crime in the South Bronx. Thus, despite advances, theory and measurement remain deficient. Seeking to address some of these issues, we employ an approach that integrates social capital and public management theories to better measure the effect of nonprofits on crime.

Linking Criminological and Public Organizational Theories

There is wide variation in nonprofit types, capacities, and levels of effectiveness (Hall 1992; Jacoby 2018; McQuarrie and Marwell 2009; Slocum et al. 2013). Even for organizations with an explicit crime-reduction focus, not all are engaged in the immediate area, but instead may focus on administration operations which address issues on the other side of the city, state, or globe (da Costa 2016). This variety in nonprofits has only recently been acknowledged in the criminological literature, and as a result, theory and measurement have lagged. As such, we must seek to better understand the “black box” of organizational functioning and organizational ecology (Ingraham and Donahue 2000; McQuarrie and Marwell 2009) to theorize how and why nonprofits may affect crime rates at the neighborhood level. Likewise, we must measure and link these organizational processes more broadly with criminological theories to adequately understand how variations across nonprofits and their spatial distribution are likely to influence neighborhood crime.

Past findings indicate that any benefits from nonprofits are conditioned by the focus and activities of nonprofits. However, when variation in nonprofit functions are taken into account, researchers typically use National Taxonomy of Exempt Entity (NTEE) codes (Wo 2016; Wo, Hipp, and Boessen 2016; see Jacoby 2018, for an exception). While an improvement over past research, which treated all nonprofits as identical, this strategy also poses a number of potential problems. First, it assumes that most nonprofits within selected categories engage in beneficial activities and few in omitted categories do so. We find this assumption largely untenable, as nonprofits are multidimensional entities not easily defined by what a single “primary function” declared upon initially filing for tax exempt status. Therefore, we argue that such distillation of complex functions fails to distinguish between the types of organizations and their influence on crime in neighborhoods.

As noted above, even if the dominant categorization of nonprofits is substantively aligned with factors likely to reduce crime, this does not necessarily mean that these benefits will translate to the immediate surroundings (e.g., if a nonprofit location serves as an administrative offices). In other words, a unidimensional classification does not tell us enough about whether these activities have any relation to the surrounding geography. Likewise, exclusion of categories risks leaving out nonprofits with potential local benefits for crime reduction. Even if an organization is not primarily focused on crime prevention or social services, it could be argued that the presence of organizations with a significant investment in a neighborhood should have greater benefits for crime reduction, if social capital perspectives hold merit. Thus, further efforts must be undertaken to operationalize the relevant missions and activities of nonprofits, to assess how/whether those features are likely to influence crime at a neighborhood level.

Drawing on past research, likely features of nonprofits for closer examination would be those which have either an explicit focus on improving the neighborhood around them and/or an explicit interest in crime reduction at a neighborhood level. This aligns with both social capital and environmental perspectives, both of which agree that local influence is important, although these perspectives diverge in their predictions. Social capital and disorganization theorists argue that crime-reduction-focused organizations likely tighten social controls or boost collective efficacy resulting in reduced crime, while environmental perspectives argue that such organizations likely serve as crime generators or attractors. Unfortunately, few studies go beyond surface-level NTEE categorizations to assess these factors at a level deep enough to pair actual missions and activities with these theoretical connections (see Jacoby 2018 for an exception, and see below for our strategy for addressing this issue).

Another limitation in the existing literature is failure to measure the diffusion of effects of nonprofits across the urban landscape. Most studies discretely measure the number of nonprofits within aggregate units and correspondingly examine the effects of nonprofits within the same unit on levels of crime. In doing so, these studies fail to capture how the broader network and landscape of nonprofits surrounding a neighborhood are likely to have an aggregate impact, not only within units where nonprofits are located but also in surrounding units. Evidence suggests that nonprofits tend to cluster according to both needs and resources, and that organizational effectiveness can benefit from these arrangements (Bielefeld and Murdoch 2004; da Costa 2016). Bielefeld and Murdoch (2004) find evidence of agglomeration economies, or the co-location of nonprofits with shared or complementary missions. There is further evidence these organizations collaborate with one another, which may have additional benefits for both the organizations themselves and the individuals and communities they serve (Baum and Haveman 1997; McDougle and Lam 2014). Such benefits are also evidenced in the trend toward collaborative one-stop assistance centers (Austin 2000).

Likewise, to benefit disadvantaged and crime-prone areas, nonprofit entities need to maintain close ties to the local community, but these very ties are conditioned by the lack of resources and social capital in disadvantaged neighborhoods (Lam and McDougle 2016; Small and McDermott 2006; Walker and McCarthy 2010). Therefore, organizations must develop other means of attaining resources to fulfill their missions. One critical factor is the degree to which organizations are integrated with both their local communities and other organizations (McDougle and Lam 2014; Moldavanova and Goerdel 2018; Schneider 2009). Such integration is likely to result in improved social capital as organizations, their members, and community residents collaborate in ways that create connections between nonprofits and outside resources (Austin 2000; Baum and Oliver 1996). Thus, higher spatial density or clustering of organizations with similar missions is both a product of and benefit for organizational stability and the creation of public services like crime reduction. Likewise, a higher density, especially sustained over time, is indicative of the ability of a number of organizations to persist despite potentially adverse conditions. Therefore, we expect the overall impact on crime to be stronger in areas with higher concentrations of nonprofits, and that this effect will be amplified as density increases.

In sum, there may be a cumulative increase in social capital resulting from multiple nonprofits operating in the same areas over time; however, this is not easily detected with current methodologies which fail to account for the distribution of nonprofits with shared missions across the urban landscape. Considering deficits in existing literature, and drawing on public organizational and criminological theory and available empirical research on the relationship between nonprofits and crime, we can identify a number of organizational factors likely to influence crime at the neighborhood level. First, the substantive focus and activities of organizations is critical for any impact. As suggested by theory and evidenced by past literature, those organizations which are focused on the development of social capital in their proximal surroundings or directly involved with reducing crime broadly within the community are most likely to have an impact on local crime (Austin 2000; Baum and Oliver 1996; Sampson, Raudenbush, and Earls 1997; Sharkey, Torrats-Espinosa, and Takyar 2017; Skogan 1988). In addition, it is likely that organizations with similar missions and concerns cluster within the same areas, and that these entities are working together to improve the social problems in the area and build ties with the community. Higher densities are also indicative of the development of sufficient underlying social capital to sustain multiple organizations. Therefore, a higher concentration of nonprofits should have benefits that are greater than the sum of their parts.

Hypotheses

Building on the theoretical framework presented above, we propose two main hypotheses about the relationship between organizational factors and crime at the neighborhood level. These hypotheses reflect the anticipated benefits of having a concentration of nonprofits with functions that are deemed most likely to have an impact, that is, those that are focused locally (place-based) and those focused on crime reduction within an intra-city geography. Although these hypotheses are developed from a social disorganization/social capital perspective, environmental perspectives, such as routine activities, theory may affect these benefits, reflecting the complex reality of neighborhood dynamics.

Hypothesis 1: Neighborhoods with higher concentrations of place-based nonprofit organizations, likely to produce increased neighborhood social capital, will have lower rates of crime than other neighborhoods.

Hypothesis 2: Neighborhoods with higher concentrations of nonprofits whose mission includes a focus on the local reduction of crime, and bolstering of social controls and collective efficacy, will have lower rates of crime than other neighborhoods.

In addition, we integrate perspectives on organizational ecology by assessing the potential cumulative benefits of nonprofit density and the broader benefits of agglomerations of nonprofits. Drawing on research which suggests that clustering of nonprofits is likely to lead to collaboration (Austin 2000; Baum and Oliver 1996), there are likely to be multiplicative benefits in areas with a high density of nonprofits.

Hypothesis 3: The effect of nonprofit density will be nonlinear such that neighborhoods with increasing levels of nonprofit density will have a decelerating rate of crime.

Finally, our methodology accounts for organizational stability and persistence by measuring both crime and nonprofit density longitudinally. To the degree that organizations persist or fail, our methodology accounts for the changing landscape of nonprofit organizations at the local level over time, and links this to changes in the local crime rate.

Method

Issues with Commonly Used Data, Measures, and Methods

One significant problem with past studies is reliance on unverified data from the National Center for Charitable Statistics (NCCS 2019). In a recent critique of research on nonprofit placement, McDougle (2015) notes a number of deficiencies in the NCCS data, criticizing nonprofit research for continued utilization of these data without adequate measures to correct for them. Foremost among these problems is the fact that address data reported to the Internal Revenue Service (IRS) and utilized by the NCCS are often inaccurate, with McDougle (2015) finding that up to 27% of addresses are incorrect. A high proportion of cases also report postal boxes, which are commonly omitted from analyses under the assumption that there is no identifiable operating address. Another fundamental problem with the NCCS data is that it only reports one address per entity; therefore, the activities of nonprofits with multiple service locations are not adequately reflected.

Seeking to determine the effects of inaccuracies in the NCCS data, McDougle (2015) found that address corrections also made a significant difference on the measurement of nonprofit concentrations at the zip-code level. Specifically, nonprofits are more likely to be undercounted in areas with low income and a high percentage of African-Americans. In addition, not only did the number of entities operating in these areas increase with correction but so did their fiscal capacity (Lam and McDougle 2016). Uncorrected, such inaccuracies have significant implications, especially when these effects are correlated with demographic factors widely known to be related to crime. As a result, any analysis utilizing uncorrected address data from the NCCS is likely to underestimate the effects of nonprofits, especially in disadvantaged areas.

As noted previously, the NTEE codes utilized by the IRS and NCCS also have known conceptual issues (Grønbjerg 1994). Because the codes are seldom changed after initial paperwork is filed, NTEE codes do not reflect either the multidimensional or evolutionary nature of nonprofit work (Hall 1992; Keating and Frumkin 2003; Lampkin and Boris 2002). Grønbjerg (1994, p. 18) finds that while organizations with an identified purpose of arts, culture, or recreation are the most likely to isolate their services to a single purpose, almost 43% delved into other areas including building social capital and working with at-risk populations. The percentage of organizations serving multiple purposes ranged from 63% to 80% for other types of organizations. Other studies show that the use of typologies from NTEE codes is not reliable at the subnational level, let alone within municipalities (Grønbjerg 1994; Lampkin and Boris 2002). Yet, nonprofit and criminological research has widely persisted in utilizing these unidimensional categorizations as the sole basis for measuring effects of nonprofits by function.

In our data, there was clear evidence of discrepancies between the NTEE categories and actual activities. For example, when compared with publicly available mission statements, only 33% of social service organizations were coded in the NTEE as providing services in their immediate neighborhoods. Only 75% of organizations identified in the NTEE with a crime or legal function focused on crime prevention or reduction, whereas a surprising 15% of organizations coded as “food, agriculture and nutrition” (NTEE Category K) espoused a crime-prevention element.

Taken together, the extant research suffers from a number of deficits that may account for inconsistency and weak effects in past work. Use of NCCS codes represents advances over early research which merely measured presence, but these data suffer from their own weaknesses. Both address data and functional categories utilized by NCCS are likely to contain inaccuracies which may attenuate effects. More troubling, these inaccuracies are likely to be especially pronounced in disadvantaged and high crime areas resulting in underestimation of nonprofit activity and capacity in the very areas they are theorized to affect the most.

This study corrects for these deficiencies in a number of ways. First, we rely on mission statements rather than NCCS/NTEE typologies to differentiate nonprofits by function and focus on factors theoretically likely to improve neighborhoods and reduce crime. Mission statements have been shown to better reflect organizational aspirations for nonprofit organizations (Kirk and Nolan 2010) and have been used to predict organizational sustainability, improved organizational outcomes, and other organizational functions (Brown and Yoshioka 2003; Frumkin and Andre-Clark 2000; Levine and Kim 2016; McDonald 2007). These studies indicate that mission statements are appropriate proxies for the types of services that nonprofits provide and are more accurate than NTEE codes alone.

Second, we make extensive efforts to manually verify and validate address data to ensure that addresses accurately reflect the distribution of nonprofit activities and services throughout the city. These efforts are documented in more detail in the methods section below. Finally, much of the research in this area utilizes cross-sectional data. Instead, we employ longitudinal analyses to allow for appropriate temporal ordering of changes in nonprofit density across the urban landscape, and any changes in crime rates that occur subsequently (Jacoby 2018; Wo 2016; Wo, Hipp, and Boessen 2016). The longitudinal and geographic nature of the data, however, introduce the possibility of dependency among units across space and time; therefore, both spatial and temporal autocorrelation must be addressed (Lee, Rushworth, and Napier 2018; LeRoux 2000; Rushworth, Lee, and Mitchell 2014). As described in the next section, we advance on past studies by utilizing Bayesian spatiotemporal multilevel time series models to address these issues (Hughes, Schaible, and Jimmerson 2019). These methods allow us to better measure the effects of the changing landscape of nonprofits on crime over time.

Data and Measures

Data were collected on the spatial distribution and levels of nonprofits, crime, and socioeconomic control variables for the City and County of Denver, Colorado, between 2010 and 2015 (with longitudinal data points for each year). Focusing on a single city allowed corrections to the NCCS data to overcome the problems noted above and address our hypotheses (McDougle 2015). We started by downloading all 4,900 nonprofits that filed as 501(c)3 in Denver. Of those, we eliminated nonprofits that were not theoretically relevant such as trade or business organizations like chambers of commerce, and national headquarters with no direct services provided to Denver. This resulted in the retention of 532 nonprofits with 168 NTEE codes reflecting services likely to be connected to human services in the Denver area. For these nonprofits, we visited websites if available to investigate their mission statements and their service addresses. For the remainder of cases, we used archived newspapers and press releases to identify official statements indicating addresses and mission statements. For a few nonprofits, addresses and missions were retrieved by calling registered agents (fewer than 10 nonprofits) to verify their information. Of the remaining nonprofits, mission statements were identifiable for 501 unique nonprofit locations within Denver.

During the process of identifying nonprofits to be included, we also verified service locations. Of the 501 nonprofits, 84 addresses were corrected from postal boxes to geographic locations. In addition, we identified 87 addresses that were incorrect because they were outdated, included typos, or simply listed no address, along with 23 that listed intersections where we identified an actual address. In one instance, a park that served as the location of a standing mobile soup kitchen was used in our data as the correct service address. All told, we corrected 34.77% of the addresses in the dataset to better reflect the empirical reality of the nonprofit landscape in Denver.

Focusing on Denver ensured that the nonprofits were operating within the same municipal and state sociopolitical environment (Beyerlein and Hipp 2005; Slocum et al. 2013; Small and Stark 2005). Although Denver is a mid-sized city, it is the largest in Colorado and the Rocky Mountain West and contains an extremely high level of nonprofits per capita, with sufficient variation in nonprofits and levels of crime across neighborhoods to address our hypotheses (NCCS 2019). Denver also makes crime data publicly available through their Open Data portal.

While most past literature has focused on census units, measurement at this level may attenuate the effects of nonprofits on neighborhood crime because of what has been referred to as the “modifiable areal unit problem” (MAUP). Closely related to the concept of ecological fallacies, MAUP suggests that results are likely to vary depending upon the unit of analysis examined (Hipp 2007; Ratcliffe 2004; Stucky and Ottensmann 2009). As such, theory should be utilized to inform what effects might be expected at what levels. In the present case, we argue the types of nonprofits under consideration should have the greatest effect on the area immediately surrounding them, with decaying effects as one gets farther away. This is reflected in our measurement strategy for nonprofits, which seeks to capture organizational activities with a neighborhood focus. To capture effects in the micro-ecological areas immediately surrounding nonprofits, we utilized 1,000 × 1,000 foot grid cells covering the study area (n = 3,981), a strategy that has been used to examine a range of other ecological effects on crime (see esp. Bernasco and Block 2011; Hughes, Schaible, and Jimmerson 2019; Stucky and Ottensmann 2009). Unlike traditional census units, because of their smaller size, grid cells are much less internally heterogeneous; therefore, it is easier to isolate the unique environmental features that characterize each cell (Smith et al. 2000; Stucky and Ottensmann 2009). For example, it allowed us to control for the spatial density of a number of nonprofit and business types operating in proximity to the grid cells. It also allowed us to control for other spatial features such as the presence of major roads or transit stops which have been shown to have an influence on crime (Brantingham and Brantingham 1993; Felson and Boba 2010). Thus, our measurement strategy provides a more granular examination of the urban landscape.

Using MS Access, ArcGIS, and R software, we calculated the counts, levels, or densities of variables for each year observed (2010 through 2015), yielding a longitudinal dataset with key spatial information on demographics, commercial and nonprofit activity, and crime between 2010 and 2015. The methods for creating each variable utilized in the analysis are provided below.

Dependent Variables

Crime

Using geocoded point data on crimes reported within Denver, and provided through Denver Open Data, counts of each of eight National Incident-Based Reporting System (NIBRS) defined crime types (public disorder, burglary, aggravated assault, robbery, larceny, theft from motor vehicles, drug and alcohol offenses, and auto theft) were calculated for each year and grid cell to create a longitudinal panel (grid cells = 3,981, n = 23,886, years = 2010–2015).¹ As with all official data, counts do not reflect crimes that were not reported to police, but the counts do provide the best available indicator of known levels of crime over time. Crime counts within units exhibit a Poisson distribution with a high number of observations equaling 0 or 1, and with diminishing probability of larger counts in any cell during a given year.

Independent Variables

As described above, this study captures the gamut of relevant nonprofits in Denver using NCCS data and then uses manual research and corrections to correct for known problems identified previously. The locations of nonprofits were geocoded to produce the XY coordinates. Nonprofits were considered active upon their first filing with the IRS and for two years after their last filing as some small nonprofits are not required to file in every tax cycle. Using these data, the spatial density² of all active nonprofits operating within half mile of each grid cell was calculated for each year between 2010 and 2015. Point data were paired with additional information on theoretically derived types (see below) to calculate densities for nonprofits.

Organizational mission and type

We coded organizational missions of the 501 nonprofits in the dataset as follows. Place-based organizations (n = 78) have a mission which focuses on the area directly surrounding the nonprofit (e.g., neighborhood associations, youth centers). Crime-focused nonprofits (n = 29) have a mission which includes the reduction or prevention of crime, whether place-based or citywide. Since types are not mutually exclusive, we analyzed all nonprofits, place-based, and crime-focused nonprofits in separate analyses as noted below.³

Densities were calculated for each 1,000 × 1,000 cell for each organizational type to calculate a smoothed probability density function (subsequently referred to simply as “density”) of nonprofits for each grid cell in the city. This measurement strategy accounts for nonprofits of similar types operating within half mile and improves on using coarsely defined census boundaries which are not amenable to the production of smoothed raster data. A similar strategy was employed to calculate the density of commercial activity across the city.

Control Variables

Commercial activity

Based on past research and environmental theories of crime, we sought to control for the influences of a wide range of business types reflective of both gentrification and commercial development occurring in Denver (e.g., coffee shops, yoga studios, fast food restaurants) and businesses associated with criminal behavior (e.g., pawn shops, easy money check cashing, and payday loans) (McCord et al. 2007; Steenbeek et al. 2012; Yu and Maxfield 2013). We utilized ListGrabber software to extract addresses from search results for each business type using the online Yellow Pages. Address data were then geocoded as described above for nonprofit type. Unfortunately, archival longitudinal data were not readily available on whether businesses were actively operating during each year; therefore, unlike the nonprofit data, data on commercial activities only provide a time-invariant snapshot of the business landscape. As such, the data should be viewed as a proxy for the persistence of business types over time.

Factor analysis of densities indicated that two factors explained a significant portion of the variance across items. The first factor (eigen > 2.773) included businesses typical of established commercial development and/or gentrification. A standardized summative scale was created (COMMERCIAL α = .90) using densities for businesses including fast food restaurants, liquor stores, coffee shops, and yoga studios.⁴ The second factor (eigen > 1.48) included business types that are most often associated with increased crime including pawn shops, easy money services, tattoo shops, and used car dealerships. The sum of standardized scores for each of these densities was incorporated into a scale measuring RISKY BUSINESSES (α = .68).

Major roads and transit

As an additional control for high volumes of nonresident traffic and commercial activity, we developed a dichotomous indicator of the presence of at least one major road, transit stop, or freeway off-ramp in the grid cell. The presence of major arterials and transit stops is positively associated with crime (Brantingham and Brantingham 1993; Yu and Maxfield 2013).

Disadvantage

Disadvantage is one of the most consistently supported predictors of crime (Peterson, Krivo, and Harris 2000; Sampson, Raudenbush, and Earls 1997). Consistent with past efforts to measure disadvantage, we utilize data from the 2013 American Community Survey (U.S. Census Bureau 2013), along with areal interpolation models to estimate the levels of key variables at the grid cell level (He, Páez, and Liu 2017; Qiu, Zhang, and Zhou 2012). Using the aerial interpolation tools within the spatial analyst extension in ArcMap, interpolation models were calibrated through iterative adaptation based on semivariograms. When compared, the interpolated estimates closely resemble the spatial patterns observed in the base block group data.

Using these interpolated estimates, we conducted a factor analysis on the following variables: household income, percentage in poverty, percentage of population that is male and between 15 and 25 years, percentage on public assistance, percentage renters, and percentage with a different house in the past year. Factor analysis results utilizing varimax rotation indicated two factors. One factor measured disadvantage (eigen = 1.9, α = .82) and included household income (reverse), percentage in poverty, and percentage on public assistance. Another factor measured residential mobility items (eigen = 1.4, α = .80) and included percentage living in a different house in the last year and percentage renters. Neither percentage female heads of household nor percentage male between the ages of 15 and 25 years loaded onto a factor, so these items were included separately as controls for the presence of crime-prone populations. Likewise, while percentage Latinx loaded somewhat with the disadvantage indicator, percentage Black did not load with any factor (likely due to recent heavy gentrification of historically Black areas in Denver). To capture the unique effects of high proportions of these two racial/ethnic groups, the largest in Denver, both percentage Latinx and percentage Black were entered into equations separately.

Logic of Analysis

Longitudinal and spatial data pose the potential for both spatial and temporal autocorrelation. Therefore, we utilized the CARBayesST package within R 3.4 because it is one of the few packages capable of simultaneously handling multilevel relationships and conditional autoregressive spatial and temporal structures along with count data. Using priors developed by LeRoux (2000), ST.CARar models include random terms for the joint and independent effects of temporal and spatial autocorrelation (Lee, Rushworth, and Napier 2018; LeRoux 2000; Rushworth, Lee, and Mitchell 2014). CARBayesST models utilize Markov chain Monte Carlo (MCMC) chains to estimate the posterior distribution of coefficients. Using this utility with 440,000 iterations, a burn-in of 40,000, and thinning by 50 to reduce autocorrelation between MCMC results, we produced an effective sample of 8,000 MCMC simulated estimates of the parameters for each model. Our formula for the Bayesian model is as follows:

Y_{i t} | μ_{i t} ~ Poisson (E_{i t} \exp (μ_{i t}))

where Y_it is the observed count of a given type of crime in neighborhood grid cell i during year t, E_it represents the expected number of crimes if these events are distributed in proportion to neighborhood grid cell population, and μ_it is the log-relative risk modeled as a linear combination of fixed covariate and random effects. The formula for the log-relative risk is as follows:

μ_{i t} = α + λ \cdot t + {X^{'}}_{i, t} β + θ_{i, t} + φ_{i, t} + ω_{i, t}

where α is the intercept, λ·t is a citywide linear time trend over the five-year study period, ${X^{'}}_{i, t}$ is a matrix of time- and space-specific predictor variables, β is a vector of fixed-effects coefficient estimates, θ_i,t denotes the random effects representing spatially unstructured heterogeneity in crime counts, φ_i,t denotes the random effects representing CAR spatial dependence, and ω_i,t denotes the temporal random effects to allow for unexplained variance in crime counts across years (see Hughes, Schaible, and Jimmerson 2019).

These models allow for the inclusion of a weighting matrix to account for the effects of neighboring units. Following Tobler’s first law of geography, which states that “near things are more related than distant things,” we utilized queen inverse-distance weighting to account for the effects of neighboring units within a half mile (Anselin 2010). As such, the Bayesian approach and the use of conditional autoregressive models provide a useful gauge of the likely range of effects of various nonprofit types on crime while controlling for spatial and temporal autocorrelation.

The resulting models provide 95% credible intervals for coefficients measuring relationships between nonprofit density and the identified crime types. Interpretation of these results is analogous to traditional interpretation of mixed-effects coefficients; however, contrary to frequentist interpretations, credible intervals suggest that the true coefficient lies somewhere within the 95% credible interval, rather than 95% of samples being likely to contain the estimate.

To assess our hypothesis that the effect of nonprofit agglomeration is greater than the sum of its parts, we also included a quadratic term to reflect possible nonlinear effects of nonprofit density for each type of nonprofit. This reflects the possibility that nonprofit effectiveness may be enhanced where a number of organizations are able to leverage resources when in proximity to other organizations, as discussed above.

Results

As a preliminary analysis, we examined descriptive statistics and correlations for each of the variables (see Table 1). Results indicated that there was a very strong correlation between nonprofit densities of various types (r ≥ .83) as noted previously. This is consistent with other research that shows that nonprofits of all types tend to be concentrated in similar areas over time (Bielefeld, Murdoch, and Waddell 1997; Peck 2008). Therefore, we ran our models for each type of nonprofit density separately to avoid multicollinearity and to address substantial overlap across types. All other control variables were approximately normally distributed did not appear to be collinear, with all correlations below .7. The most closely related variables were commercial density and the density of risky businesses (r = .696); however, the factor analysis supported dividing these indicators into two dimensions and since they are only controls, we are most interested in parsing out their variance rather than looking at their substantive effects.⁵ To further assess collinearity, we also ran variance inflation factors (VIFs) for each set of variables,⁶ and there were no values higher than 3. The population variable was logged to correct for positive skew and included in all equations as an offset so that rates could be calculated. The crime counts resembled a Poisson distribution with some evidence of overdispersion. Although there are no negative binomial variants of the models presently available for dealing with overdispersion, because the CARBayesST models include an observation-level random effect, overdispersion is modeled into the equations (Harrison 2014; Huang 2016).

Table 1.

Descriptives and Bivariate Correlations (N = 23,886).

	M	SD	Minimum	Maximum	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.	14.	15.	16.	17.	18.	19.	20.	21.	22.
1. Public disorder rate	1.93	3.83	0	81	1.00
2. Burglary rate	1.18	1.85	0	19	.47	1.00
3. Aggravated assault rate	0.38	1.27	0	35	.61	.34	1.00
4. Robbery rate	0.28	1.05	0	32	.54	.31	.60	1.00
5. Larceny rate	1.87	8.47	0	418	.44	.20	.40	.45	1.00
6. Auto theft rate	1.08	2.40	0	68	.47	.40	.43	.36	.32	1.00
7. Drug and alcohol rate	0.92	6.44	0	236	.50	.18	.47	.58	.35	.19	1.00
8. Theft from motor vehicle rate	1.63	3.05	0	70	.53	.44	.45	.39	.37	.57	.19	1.00
9. All active NPO density	0.17	0.27	0	2.03	.50	.35	.41	.40	.33	.40	.34	.52	1.00
10. Place-based NPO density	0.17	0.33	0	2.45	.46	.35	.39	.36	.26	.37	.30	.45	.90	1.00
11. Crime-focused NPO density	0.17	0.44	0	3.26	.40	.29	.34	.32	.23	.31	.27	.39	.83	.85	1.00
12. Population	181.67	112.94	51	866	−.08	−.09	−.03	−.03	.00	−.08	−.02	−.09	−.16	−.09	−.09	1.00
13. Population logged	5.09	0.43	3.93	6.76	−.08	−.09	−.02	−.02	−.01	−.08	−.02	−.08	−.16	−.11	−.09	.94	1.00
14. % Black^a	0.00	1.00	−1.60	3.57	−.12	−.15	−.05	−.05	−.07	−.15	−.03	−.17	−.15	−.01	−.01	.15	.11	1.00
15. % Latino^a	0.00	1.00	−2.14	3.02	.10	.11	.08	.03	−.05	.09	−.01	−.03	−.17	−.03	−.14	−.06	−.03	−.07	1.00
16. Disadvantage index	0.00	0.86	−2.49	4.17	.20	.16	.17	.12	.03	.16	.08	.07	.08	.25	.09	−.19	−.18	.10	.68	1.00
17. Commercial density	0.00	1.00	−1.64	3.70	.44	.41	.31	.32	.28	.39	.23	.49	.77	.62	.57	−.20	−.20	−.37	−.11	.12	1.00
18. Risky business density	0.00	1.00	−1.33	2.61	.39	.38	.30	.28	.18	.35	.18	.36	.47	.44	.35	−.23	−.23	−.38	.23	.45	.70	1.00
19. Residential mobility index	0.00	0.91	−2.07	4.39	.26	.15	.22	.20	.19	.24	.17	.26	.42	.39	.29	−.10	−.14	−.03	.07	.38	.40	.38	1.00
20. % Female headed households^a	0.00	1.00	−3.92	2.88	.03	.01	.02	.02	.02	.03	.02	.00	.06	.13	.07	−.29	−.40	.49	.01	.29	.00	.02	.35	1.00
21. % Male, 15–25 years^a	0.00	1.00	−2.40	6.70	.14	.10	.11	.08	.07	.14	.06	.09	.05	.08	.01	−.12	−.08	−.10	.53	.49	.11	.20	.33	.05	1.00
22. Road or transit stop	0.62	0.48	0	1	.22	.23	.16	.16	.10	.19	.09	.21	.27	.24	.20	−.07	−.04	−.19	.06	.16	.38	.34	.14	−.07	.14	1.00

Note. NPO = nonprofit organization.

Indicates a variable that has been standardized.

To simplify presentation of multivariate results, estimates for controls are not presented here, although they were included as a block in all equations. With few exceptions, the effects of controls were significant and in the expected direction across equations. The racial/ethnic composition, degree of disadvantage, commercial and risky business density, and the presence of a major road or transit exhibited significant effects in the expected direction. In contrast, the credible intervals for percentage female heads of household and mobility were largely nonsignificant (i.e., credible intervals included 0). Contrary to theory, percentage of the population male 15 to 25 years was significantly negatively associated with crime—likely an artifact of Denver’s rapidly growing professional population of millennials who have moved to the city for employment opportunities.

To assess the effect of nonprofits overall, we first examined the relationship between density of all active nonprofits and each type of crime. These results are presented in Table 2. Consistent with observed bivariate effects, there is a moderate positive relationship between density of nonprofits and all types of crime; however, the negative valence of quadratic terms indicates that the adverse effects of nonprofit density level off and eventually decrease. The only type of crime for which increased density of nonprofits as a whole appeared to exacerbate crime is with drug and alcohol offenses. This effect is relatively flat when predicted within the range observed for nonprofit densities, suggesting that nonprofit density would need to be very high to result in adverse effects (see Figure 1).

Table 2.

Comparison of Nonlinear Effects of Density of All Active, Place-Based, and Crime-Focused Nonprofits on Crime Types.

	Public Disorder	Burglary	Aggravated Assault	Robbery	Larceny	Auto Theft	Drug and Alcohol Crime	Theft from Motor Vehicle
All active NPO density	2.22	1.85	−0.07	−0.70	0.01	1.28	0.04	1.71
All active NPO density	[1.40, 3.02]	[1.00, 2.83]	[–0.94, 1.04]	[–1.95, 0.67]	[–0.37, 0.34]	[0.20, 2.08]	[–0.20, 0.30]	[1.03, 2.43]
All active NPO density squared	–1.10	–0.90	−0.27	−0.45	–0.30	–0.68	0.16	–0.94
All active NPO density squared	[–1.46, –0.68]	[–1.34, –0.49]	[–0.88, 0.15]	[–1.10, 0.19]	[–0.42, –0.18]	[–1.04, –0.26]	[0.05, 0.26]	[–1.32, –0.58]
Goodness-of-fit statistics
DIC	60,138.08	52,685.87	25,101.41	19,437.44	51,163.66	46,815.38	27,252.29	58,563.35
LMPL	−24,237.61	−21,424.22	−9,827.18	−7,434.15	−20,368.81	−18,702.46	−10,222.27	−23,448.78
Place-based NPO density	1.11	1.24	0.07	−0.31	0.32	0.70	0.79	0.58
Place-based NPO density	[0.54, 1.52]	[0.57, 1.71]	[–0.64, 0.90]	[–1.00, 0.41]	[0.03, 0.60]	[0.21, 1.19]	[0.53, 1.06]	[–0.08, 1.06]
Place-based NPO density squared	–0.41	–0.45	0.00	0.08	–0.23	–0.28	–0.27	–0.23
Place-based NPO density squared	[–0.62, –0.12]	[–0.69, –0.10]	[–0.37, 0.38]	[–0.32, 0.42]	[–0.36, –0.10]	[–0.51, –0.04]	[–0.37, –0.17]	[–0.46, –.005]
Goodness-of-fit statistics
DIC	60,185.58	52,699.22	25,129.65	19,459.32	51,214.80	46,859.70	27,288.67	58,547.68
LMPL	−24,249.00	−21,426.78	−9,839.55	−7,448.18	−20,389.02	−18,706.64	−10,250.70	−23,452.06
Crime-focused NPO density	−0.20	−0.09	0.13	–0.81	–0.50	−0.34	−0.08	−0.11
Crime-focused NPO density	[–0.54, 0.14]	[–0.51, 0.32]	[–0.36, 0.65]	[–1.30, –0.25]	[–0.70, –0.32]	[–0.79, 0.03]	[–0.32, 0.12]	[–0.50, 0.33]
Crime-focused NPO density squared	0.15	0.07	0.12	0.30	0.23	0.17	0.27	0.09
Crime-focused NPO density squared	[0.03, 0.26]	[–0.07, 0.21]	[–0.06, 0.31]	[0.12, 0.48]	[0.16, 0.30]	[0.04, 0.32]	[0.21, 0.34]	[–0.07, 0.25]
Goodness-of-fit statistics
DIC	60,192.38	52,696.68	25,131.48	19,446.00	51,148.08	46,770.72	27,309.69	58,561.16
LMPL	−24,236.60	−21,420.40	−9,837.56	−7,448.57	−20,361.83	−18,708.29	−10,242.56	−23,449.82

Note. Median estimates are reported with confidence interval for estimates in brackets. All models include controls for variables in Table 1. Significant estimates are in bold (p < .05). NPO = nonprofit organization; DIC = Deviance Information Criterion; LMPL = Log-Marginal Pseudo Likelihood.

Figure 1.

Nonlinear effect of nonprofit density on selected crime types.

More typical is the effect shown for burglary in Figure 1, where increases in all nonprofit density as well as place-based organizations result in an increase in reported crime at low-to-moderate nonprofit densities, but then a decline at higher densities. In addition, the effect of crime-focused nonprofit density is minimal or non-significant (as is also the case for public disorder, aggravated assault, theft from motor vehicle, and auto theft; see Table 2). Robbery, aggravated assault, and larceny show a slightly different pattern—nonprofit density of any type shows little to no relationship with these offenses.

Similar patterns are observed for the effects of place-based nonprofits; however, these effects are most pronounced for disorder, burglary, larceny, auto theft, and theft from motor vehicles. There were no significant effects of the density of place-based nonprofits on any violent crimes, nor were there any effects on drug and alcohol offenses. In addition, unlike generic nonprofits, no effects of place-based nonprofits can be interpreted as significant and positive. When utilizing the observed range for place-based nonprofit densities to predict rates, the effects are slightly less pronounced relative to those observed for generic nonprofits. Increased density of place-based nonprofits does not increase crime as much as increased density of generic nonprofits, but neither does it decrease crime as much after the point of diminishing marginal effects. Therefore, overall, an increased density of place-based nonprofits is likely to produce either very similar or somewhat lower levels of crime relative to generic nonprofits.

Contrary to our hypotheses about the potential effects of crime-focused nonprofits, there was little evidence of substantively important negative or positive effects on crime. The positive effects of the quadratic term were significant for disorder, robbery, auto theft, larceny, and drugs and alcohol; however, these effects would seem to necessitate an incredibly high level of density, outside the observed range of nonprofit densities, to be detected. Exponential effects of nonprofit density for all types of crime only begin to diverge from the patterns observed for other types of nonprofits around seven standard deviations above the mean density for nonprofits; thus, such instances are likely to be extraordinarily rare, even amid very high concentrations of crime-reduction-focused nonprofits. Stated in terms of our actual data, such cells would have to have 20 times the average density of crime-oriented nonprofits before one saw a rate even close to the effects predicted for place-based and generic nonprofit densities. Thus, while effects are significant and positive, we should be cautious in interpreting these with regard to their substantive importance.

Limitations

As with any study, the present research has a number of limitations. Among the most important of these is the fact that this is but one study of one city during a limited period of time. While this has the benefit of ensuring that differences in local polity cannot account for the findings (Slocum et al. 2013), it cannot fully account for changes occurring within the study area during the period examined. During the period examined, for example, recreational marijuana was legalized in Colorado (see Hughes, Schaible, and Jimmerson 2019). Simultaneously, Denver has had one of the strongest recoveries and fastest rates of population growth since the “Great Recession” (Murray 2017; Svaldi 2017). Thus, Denver has rapidly changed in ways that may be largely undetectable in presently available census data. While our study utilized mixed effects, which reduces concerns about omitted variable bias by focusing on effects of the broader landscape of nonprofits and business within units, it does not completely eliminate these concerns. Given this, future studies must seek to replicate this study utilizing alternative controls and data from other jurisdictions which offer variation in local sociopolitical factors.

Although utilization of Bayesian conditional autoregressive models can be considered a strength and contribution to the literature, it is not without its limitations. Bayesian results are only valid inasmuch as the priors utilized are valid representations of the actual processes, and like many assumptions, this is not easily assessed. That being said, the priors selected go farther than other traditional fixed- and mixed-effects models in controlling for the unique challenges of autocorrelation found in longitudinal spatial data. Likewise, the simulations have the benefit of identifying the parameters within which effects are likely to occur rather than testing against a theoretical distribution. Although not presented here, the findings were comparable to traditional mixed-effects models incorporating random effects for both spatial and temporal factors; however, only through further replication and refinement of priors can the true effects be assessed.

In any longitudinal study, the ability to distinguish between causation and correlation is also of concern. While it is impossible to establish causation in the absence of an experimental design, the methods employed here offer a number of advantages. First, we control for the effects of factors identified by both social capital (e.g., demographics) and environmental perspectives (e.g., business landscape) that are likely to condition the effects of nonprofits and their placement. This is not something ordinarily included in nonprofit research, with the tendency to focus on controls for either environmental or social disorganization variables, not both. Second, the models employed are explicitly designed to account for both spatial and temporal autocorrelation. In doing so, the models control for not only the influences of neighboring units but also the effects of units for the preceding time period, and their joint spatial and temporal influences. Finally, the models intentionally measure how the shifting landscape of nonprofits change over time, and how these changes affect crime in each time period. Therefore, the measures utilized here assess how changes in the landscape and distribution of nonprofits effect crime, rather than examining a stagnant view of nonprofits. While these procedures cannot completely rule out initial selection effects, it is a significant advance beyond other published studies. Thus, while there is always room for further advancement, the present results better control for omitted variable bias and other factors likely to interfere with causal inference.

Conclusion

Taken together, the findings suggest that there is likely to be an impact of nonprofit placement on crime. All other factors being equal, as density of nonprofits increases, places are likely to have slightly increased rates of crime; however, this effect is curvilinear such that once a place reaches a critical density of nonprofits (a probability density function of around 1 per 5 square miles), rates of crime begin to decline with increasing density of either generic or place-based non-profits. In support of environmental perspectives, and contrary to our hypotheses that crime-oriented nonprofits should have greater beneficial effects, increased densities of these types of nonprofits appear to significantly increase rates of crime at extremely high densities. This may be an unintended consequence of having a high concentration of nonprofits which either serve crime-prone populations or serve as crime generators or attractors themselves. That being said, the magnitude of these increases is substantively suspect, with little or no observable effect of the density of crime-oriented nonprofits, until the density is almost seven standard deviations (20×) above the mean density of nonprofits. In addition, seldom does the effect predicted for crime-oriented nonprofits exceed the predicted effects of other types of nonprofits. Therefore, we can conclude that while nonprofits tend to have a significant positive effect on crime, this effect declines with higher density and varies such that increased density of nonprofits focused on crime or place is likely to be less pronounced than nonprofits in general. In other words, we do not generally see the kinds of exponential growth that might be predicted to result from increased density of nonprofits by environmental theories of crime.

This study also finds interesting patterns with regard to the relationships between nonprofit types and crime types (disorder, property, and violent) that deserve further exploration. As discussed above, past research has found some mixed evidence that nonprofits may reduce violent crime; however, we found few significant effects of any type of nonprofit on violent crimes. Furthermore, only crime-focused nonprofits exhibited a significant negative effect on robbery, though this curvilinear effect is not especially pronounced or distinguishable from the effects of other nonprofits unless observed at high densities. In contrast, we found that a higher concentration of nonprofits in general, and especially those which are place-based, seems to have a consistent effect on both disorder and property crimes which results in declining prevalence of these crimes as the density of nonprofits increases. The same cannot be said of crime-reducing nonprofits. Taken together, and consistent with Sharkey, Torrats-Espinosa, and Takyar (2017) and Slocum et al. (2013), these findings support a social capital perspective, which argues that increased bonding and bridging efforts among community organizations likely have benefits. Likewise, it suggests some support for environmental perspectives with regard to reducing violent crimes, perhaps as a result of focusing on populations of young, highly motivated offenders and providing them with prosocial activities. Subsequent research would be beneficial for more thoroughly exploring the robustness of these relationships, and through what mechanisms nonprofits are most likely to affect specific types of crime.

In general, those organizations with a specific mission to reduce crime seem to have the least adverse impact on surrounding neighborhoods at lower densities. The results for place-based nonprofits are slightly more like those of generic nonprofits, but there appears to be something of substantive importance about the manner in which these types of organizations influence their surrounding neighborhoods. It is likely that the mechanisms behind these effects are tied to factors theorized by both environmental (e.g., increased guardianship) and social capital theories (e.g., social control, collective efficacy). Much more research and further refinement of measurement for nonprofit activities need to be conducted to directly evaluate the influence of these types of nonprofits on theoretical factors.

Given the current findings, it might be tempting for some to utilize this to aid in their case that nonprofits increase crime. However, we strongly caution against such a simplistic interpretation. While this study does suggest some impact, as noted previously, it further suggests that impacts are likely to vary substantially. And, in particular, it suggests that crime-focused and place-based organizations, which are most likely to be subjected to local opposition, are perhaps the most desirable types of nonprofit organizations to have in one’s neighborhoods, as their impacts are actually less than other run of the mill nonprofits (assuming typical densities). The present findings further suggest that a higher density of commercial activity has a much more substantial impact on crime relative to nonprofit density, and a higher density of risky business types in particular has an effect that is much higher and more troubling than nonprofits. Thus, those concerned about the placement of nonprofits should first consider what is likely to fill space that might otherwise be occupied by a place-based or crime-focused nonprofit. If the space is likely to be occupied by general commercial interests or crime generating or attracting businesses, then this is likely to come with higher crime than nonprofits are generally likely to produce. It is likely that nonprofits will bring with them other benefits, especially inasmuch as they address local problems or increase opportunities. Considering the findings and limitations presented here, it is clear that much more research needs to be conducted that takes a finer grained perspective on the relationship between nonprofits and crime and corrects for the mismatch between measures and theory. While this study strongly suggests that factors, such as the concentration of nonprofits, focused on place or crime influence local crime rates, it only begins to hint at the degree to which organizational processes and features operate to influence crime. Future studies should seek to identify, develop, and collect better measures of organizational functioning and practices over time to evaluate how these interact with variables of importance to both social capital theories and environmental theories of criminology. Drawing on public organizational theories and theories of organizational ecology, efforts should be made to more directly identify how specific organizational practices and concerns are likely to influence things like neighborhood cohesion, collective efficacy, and capable guardianship. In doing so, there is a great deal to potentially be gained by nonprofit organizations, communities, and researchers alike.

Supplemental Material

Appendix_A – Supplemental material for The Influence of Spatial Density of Nonprofits on Crime

Supplemental material, Appendix_A for The Influence of Spatial Density of Nonprofits on Crime by Lonnie Schaible, Lucy Dwight and Nuri Heckler in Urban Affairs Review

Footnotes

Authors’ Note

Full models for this article are available on request from the lead author.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Lonnie Schaible

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Lonnie Schaible is an associate professor in the School of Public Affairs at the University of Colorado Denver. His research interests include policing, communities, and criminological theory.

Lucy Dwight is a clinical assistant professor in the School of Public Affairs at the University of Colorado Denver. Her research interests include race, crime, and gun ownership.

Nuri Heckler is an assistant professor in the School of Public Administration at the University of Nebraska at Omaha. His research examines public and nonprofit management, race, and gender.

References

Anselin

Luc

. 2010. Spatial Econometrics: Methods and Models. Dordrecht: Springer.

Austin

James E.

2000. The Collaboration Challenge: How Nonprofits and Businesses Succeed Through Strategic Alliances. San Francisco: Jossey-Bass.

Baum

Joel A.C.

Haveman

Heather A.

1997. “Love Thy Neighbor Differentiation and Agglomeration in the Manhattan Hotel Industry, 1989-1990.” Administrative Science Quarterly 42 (2): 304–334. doi: 10.2307/2393922.

Baum Oliver

1996. “Toward an Institutional Ecology of Organizational Founding.” Academy of Management Journal 39: 1378–427. doi: 10.5465/257003.

Bernasco

Wim

Block

Richard

. 2011. “Robberies in Chicago: A Block-Level Analysis of the Influence of Crime Generators, Crime Attractors, and Offender Anchor Points.” Journal of Research in Crime and Delinquency 48 (1): 33–57. doi:10.1177/0022427810384135.

Beyerlein

Kraig

Hipp

John R.

2005. “Social Capital, Too Much of a Good Thing? American Religious Traditions and Community Crime.” Social Forces 84 (2): 995–1013. Accessed February 15, 2020. https://www-jstor-org.web.bisu.edu.cn/stable/3598488.

Bielefeld

Wolfgang

Murdoch

James C.

2004. “The Locations of Nonprofit Organizations and Their for-Profit Counterparts: An Exploratory Analysis.” Nonprofit and Voluntary Sector Quarterly 33 (2): 221–46. doi:10.1177/0899764003260589.

Bielefeld

Wolfgang

Murdoch

James C.

Waddell

Paul

. 1997. “The Influence of Demographics and Distance on Nonprofit Location.” Nonprofit and Voluntary Sector Quarterly 26 (2): 207–25. doi:10.1177/0899764097262007.

Brantingham

Patricia L.

Brantingham

Paul J.

1993. “Environment, Routine and Situation: Toward a Pattern Theory of Crime: Advances in Criminological Theory.” In Routine Activity and Rational Choice, edited by Clarke

Ronald V.

Felson

Marcus

, vol. 5, 259–94. New Brunswick: Transaction Publishers.

10.

Brantingham

Patricia L.

Brantingham

Paul J.

1995. “Criminality of Place.” European Journal on Criminal Policy and Research 33:5–26. doi:10.1007/BF02242925.

11.

Brown

William A.

Yoshioka

Carlton F.

2003. “Mission Attachment and Satisfaction as Factors in Employee Retention.” Nonprofit Management & Leadership 14 (1): 5–18.

12.

Bursik

Robert J.

1988. “Social Disorganization and Theories of Crime and Delinquency: Problems and Prospects.” Criminology 264:519–52. doi:10.1111/j.1745-9125.1988.tb00854.x.

13.

Bursik

Robert J.

Grasmick

Harold G.

1993. “Economic Deprivation and Neighborhood Crime Rates, 1960-1980.” Law & Society Review 27 (2): 263–83. doi:10.2307/3053937.

14.

Chambré

Susan M.

Fatt

Naomi

. 2002. “Beyond the Liability of Newness: Nonprofit Organizations in an Emerging Policy Domain.” Nonprofit and Voluntary Sector Quarterly 31 (4): 502–24. doi:10.1177/0899764002238098.

15.

Cohen

Lawrence E.

Felson

Marcus

. 1979. “Social Change and Crime Rate Trends: A Routine Activity Approach.” American Sociological Review 44 (4): 588–608. Accessed February 15, 2020. https://www-jstor-org.web.bisu.edu.cn/stable/2094589.

16.

Coleman

James S.

1988. “Social Capital in the Creation of Human Capital.” American Journal of Sociology 94:S95–120. Accessed February 15, 2020. https://www-jstor-org.web.bisu.edu.cn/stable/2780243.

17.

da Costa

Marcelo M.

2016. “What Influences the Location of Nonprofit Organizations? A Spatial Analysis in Brazil.” Voluntas: International Journal of Voluntary and Nonprofit Organizations 273:1064–90. doi:10.1007/s11266-016-9682-7.

18.

Felson

Marcus

Boba

Rachel

. 2010. Crime and Everyday Life. 4th ed. Thousand Oaks, CA: Sage.

19.

Frumkin

Peter

Andre-Clark

Alice

. 2000. “When Missions, Markets, and Politics Collide: Values and Strategy in the Nonprofit Human Services.” Nonprofit and Voluntary Sector Quarterly 29 (1): 141–63.

20.

Groff

Elizabeth R.

Lockwood

Brian

. 2014. “Criminogenic Facilities and Crime across Street Segments in Philadelphia: Uncovering Evidence About the Spatial Extent of Facility Influence.” Journal of Research in Crime and Delinquency 51 (3): 277–314. doi:10.1177/0022427813512494.

21.

Grønbjerg

Kirsten A.

1994. “Using NTEE to Classify Non-profit Organisations: An Assessment of Human Service and Regional Applications.” Voluntas: International Journal of Voluntary and Nonprofit Organizations 53:301–28. doi:10.1007/BF02354038.

22.

Hall

Peter D.

1992. Inventing the Nonprofit Sector and Other Essays on Philanthropy, Voluntarism, and Nonprofit Organizations. Baltimore: The Johns Hopkins Univ. Press.

23.

Harrison

Xavier A.

2014. “Using Observation-Level Random Effects to Model Overdispersion in Count Data in Ecology and Evolution.” PeerJ 2:e616. doi:10.7717/peerj.616.

24.

Páez

Antonio

Liu

Desheng

. 2017. “Persistence of Crime Hot Spots: An Ordered Probit Analysis.” Geographical Analysis 49 (1): 3–22. doi:10.1111/gean.12107.

25.

Hipp

John R.

2007. “Block, Tract, and Levels of Aggregation: Neighborhood Structure and Crime and Disorder as a Case in Point.” American Sociological Review 72: 659–80. doi: 10.1177/000312240707200501.

26.

Huang

Guowen.

2016. “Quantification of Air Quality in Space and Time and Its Effects on Health.” PhD thesis, Univ. of Glasgow, Glasgow.

27.

Hughes

Lori A.

Schaible

Lonnie M.

Jimmerson

Katherine

. 2019. “Marijuana Dispensaries and Neighborhood Crime and Disorder in Denver, Colorado.” Justice Quarterly. Published electronically February. doi:10.1080/07418825.2019.1567807.

28.

Ingraham

Patricia W.

Donahue

Amy K.

2000. “Dissecting the Black Box Revisited: Characterizing Government Management Capacity.” In Government and Performance: New Perspectives, edited by Heinrich

Carolyn J.

Lynn

Laurence E.

Jr., 292–318. Washington, DC: Georgetown Univ. Press.

29.

Jacoby

Annette

. 2018. “Social Service Organizations, Discretionary Funding, and Neighborhood Crime Rates.” Crime & Delinquency 64 (9): 1193–214. doi:10.1177/0011128716688884.

30.

Kautt

Paula M.

Roncek

Dennis W.

2007. “Schools as Criminal ‘Hot Spots’: Primary, Secondary, and Beyond.” Criminal Justice Review 32(4): 339–57. doi:10.1177/0734016807311559.

31.

Keating

Elizabeth K.

Frumkin

Peter

. 2003. “Reengineering Nonprofit Financial Accountability: Toward a More Reliable Foundation for Regulation.” Public Administration Review 63 (1): 3–15. Accessed February 15, 2020. https://www-jstor-org.web.bisu.edu.cn/stable/977517.

32.

Kirk

Gary

Nolan

Shabnam Beth

. 2010. “Nonprofit Mission Statement Focus and Financial Performance.” Nonprofit Management and Leadership 20 (4): 473–90.

33.

Lam

Marcus

McDougle

Lindsey

. 2016. “Community Variation in the Financial Health of Nonprofit Human Service Organizations: An Examination of Organizational and Contextual Effects.” Nonprofit and Voluntary Sector Quarterly 45 (3): 500–25. doi:10.1177/0899764015591365.

34.

Lampkin

Linda M.

Boris

Elizabeth T.

2002. “Nonprofit Organization Data: What We Have and What We Need.” American Behavioral Scientist 45 (11): 1675–1715. doi:10.1177/0002764202045011005.

35.

Lee

Duncan

Rushworth

Alastair

Napier

Gary

. 2018. “Spatio-Temporal Areal Unit Modeling in R with Conditional Autoregressive Priors Using the CARBayesST Package.” Journal of Statistical Software 84 (9): 1–39. doi:10.18637/jss.v084.i09.

36.

LeRoux

Brian G.

2000. “Modelling Spatial Disease Rates Using Maximum Likelihood.” Statistics in Medicine 19(17–18):2321–32. doi:10.1002/1097-0258(20000915/30)19:17/18<2321::AID-SIM572>3.0.CO;2-#.

37.

Levine

Daniel J.

Kim

Mirae

. 2016. “When Missions, Markets, and Politics Collide: Values and Strategy in the Nonprofit Human Services.” Administration & Society 50 (7): 947–72.

38.

McCord

Eric S.

Ratcliffe

Jerry H.

Garcia

R. Marie

Taylor

Ralph B.

2007. “Nonresidential Crime Attractors and Generators Elevate Perceived Neighborhood Crime and Incivilities.” Journal of Research in Crime and Delinquency 44 (3): 295–320. doi:10.1177/0022427807301676.

39.

McDonald

Robert E.

2007. “An Investigation of Innovation in Nonprofit Organizations: The Role of Organizational Mission.” Nonprofit and Voluntary Sector Quarterly 36 (2): 256–81.

40.

McDougle

Lindsey M.

2015. “The Accuracy of the Core Files for Studying Nonprofit Location: How Many Nonprofits Are There?” Nonprofit and Voluntary Sector Quarterly 44 (3): 609–24. doi:10.1177/0899764015573873.

41.

McDougle

Lindsey M.

Lam

Marcus

. 2014. “Individual- and Community-Level Determinants of Public Attitudes Toward Nonprofit Organizations.” Nonprofit and Voluntary Sector Quarterly 43 (4): 672–92. doi:10.1177/0899764013479830.

42.

McQuarrie

Michael

Marwell

Nicole P.

2009. “The Missing Organizational Dimension in Urban Sociology.” City & Community 8 (3): 247–68. doi:10.1111/j.1540-6040.2009.01288.x.

43.

Moldavanova

Alisa

Goerdel

Holly T.

2018. “Understanding the Puzzle of Organizational Sustainability: Toward a Conceptual Framework of Organizational Social Connectedness and Sustainability.” Public Management Review 20 (1): 1–27. doi:10.1080/14719037.2017.1293141.

44.

Morenoff

Jeffrey D.

Sampson

Robert J.

Raudenbush

Stephen W.

2001. “Neighborhood Inequality, Collective Efficacy, and the Spatial Dynamics of Urban Violence.” Criminology 39:517–58. doi:10.1111/j.1745-9125.2001.tb00932.x.

45.

Murray

Jon

. 2017. “Denver’s Population Has Swelled in the Last 7 Years.” The Denver Post, September 28. Accessed February 15, 2020. https://theknow.denverpost.com/2017/09/28/denver-neighborhoods-growth-2017/160032/.

46.

National Center for Charitable Statistics. 2019. “National Taxonomy of Exempt Entities (NTEE) Codes.” The Urban Institute, National Center for Charitable Statistics, Washington, DC. Accessed February 15, 2020. http://nccs.urban.org/classification/national-taxonomy-exempt-entities.

47.

Peck

Laura R.

2008. “Do Antipoverty Nonprofits Locate Where People Need Them? Evidence from a Spatial Analysis of Phoenix.” Nonprofit and Voluntary Sector Quarterly 37 (1): 138–51. doi:10.1177/0899764006298963.

48.

Peterson

Ruth D.

Krivo

Lauren J.

Harris

Mark A.

2000. “Disadvantage and Neighborhood Violent Crime: Do Local Institutions Matter?” Journal of Research in Crime and Delinquency 37 (1): 31–63. doi:10.1177/0022427800037001002.

49.

Putnam

Robert D.

2000. “Bowling Alone: America’s Declining Social Capital.” In Culture and Politics, edited by Crothers

Lane

Lockhart

Charles

, 223–34. New York: St. Martin’s Press.

50.

Qiu

Fang

Zhang

Caiyun

Zhou

Yuhong

. 2012. “The Development of an Areal Interpolation ArcGIS Extension and a Comparative Study.” GIScience & Remote Sensing 49 (5): 644–63. doi:10.2747/1548-1603.49.5.644.

51.

Ratcliffe

Jerry H.

2004. “Geocoding Crime and a First Estimate of a Minimum Acceptable Hit Rate.” International Journal of Geographical Information Science 18: 61– 72. doi: 10.1080/13658810310001596076.

52.

Roman

Caterina Gouvis

Moore

Gretchen E.

2004. “Measuring Local Institutions and Organizations: The Role of Community Institutional Capacity in Social Capital.” Washington, DC: Urban Institute Justice Policy Center.

53.

Roncek

Dennis W.

Faggiani

Donald

. 1985. “High Schools and Crime: A Replication.” The Sociological Quarterly 26:491–505. doi:10.1111/j.1533-8525.1985.tb00240.x.

54.

Rushworth

Alastair

Lee

Duncan

Mitchell

Richard

. 2014. “A Spatio-Temporal Model for Estimating the Long-Term Effects of Air Pollution on Respiratory Hospital Admissions in Greater London.” Spatial and Spatio-Temporal Epidemiology 10:29–38. doi:10.1016/j.sste.2014.05.001.

55.

Sampson

Robert J.

Groves

W. Byron

. 1989. “Community Structure and Crime: Testing Social-Disorganization Theory.” American Journal of Sociology 94 (4): 774–802. Accessed February 15, 2020. https://www-jstor-org.web.bisu.edu.cn/stable/2780858.

56.

Sampson

Robert J.

Raudenbush

Stephen W.

Earls

Felton

. 1997. “Neighborhoods and Violent Crime: A Multilevel Study of Collective Efficacy.” Science 277 (5328): 918–24. doi:10.1126/science.277.5328.918.

57.

Schneider

Jo Anne

. 2009. “Organizational Social Capital and Nonprofits.” Nonprofit and Voluntary Sector Quarterly 38 (4): 643–62. doi:10.1177/0899764009333956.

58.

Sharkey

Patrick

Torrats-Espinosa

Gerard

Takyar

Delaram

. 2017. “Community and the Crime Decline: The Causal Effect of Local Nonprofits on Violent Crime.” American Sociological Review 82 (6): 1214–40. doi:10.1177/0003122417736289.

59.

Shaw

Clifford R.

McKay

Henry D.

1942. Juvenile Delinquency and Urban Areas. Chicago: Univ. of Chicago Press.

60.

Skogan

Wesley G.

1988. “Community Organizations and Crime.” Crime and Justice 10:39–78. doi:10.1086/449143.

61.

Slocum

Lee Ann

Rengifo

Andres F.

Choi

Tiffany

Herrmann

Christopher R.

2013. “The Elusive Relationship Between Community Organizations and Crime: An Assessment Across Disadvantaged Areas of the South Bronx.” Criminology 51 (1): 167–216. doi:10.1111/1745-9125.12001.

62.

Small

Mario Luis

McDermott

Monica

. 2006. “The Presence of Organizational Resources in Poor Urban Neighborhoods: An Analysis of Average and Contextual Effects.” Social Forces 84 (3): 1697–724. doi:10.1353/sof.2006.0067.

63.

Small

Mario Luis

Stark

Laura

. 2005. “Are Poor Neighborhoods Resource Deprived? A Case Study of Child Care Centers in New York.” Social Science Quarterly 86 (s1): 1013–36. doi:10.1111/j.0038-4941.2005.00334.x.

64.

Small

Mario Luis

Jacobs

Erin M.

Massengill

Rebekah Peeples

. 2008. “Why Organizational Ties Matter for Neighborhood Effects: Resource Access through Childcare Centers.” Social Forces 1: 387–414. doi: 10.1353/sof.0.0079.

65.

Smith

William R.

Frazee

Sharon Glave

Davison

Elizabeth L.

2000. “Furthering the Integration of Routine Activity and Social Disorganization Theories: Small Units of Analysis and the Study of Street Robbery as a Diffusion Process.” Criminology 38: 489– 524. doi: 10.1111/j.1745-9125.2000.tb00897.x.

66.

Steenbeek

Wouter

Völker

Beate

Flap

Henk

Oort

Frank van

. 2012. “Local Businesses as Attractors or Preventers of Neighborhood Disorder.” Journal of Research in Crime and Delinquency 49 (2): 213–48. doi:10.1177/0022427810397948.

67.

Stucky

Thomas D.

Ottensmann

John R.

2009. “Land Use and Violent Crime.” Criminology 47 (4): 1223–64. doi:10.1111/j.1745-9125.2009.00174.x.

68.

Svaldi

Aldo

. 2017. “Denver Created Jobs in Recovery, but Productivity Was Missing.” The Denver Post, March 2. Accessed February 15, 2020. https://www.denverpost.com/2017/03/02/denver-economic-recovery-stronger/.

69.

Tomeh

Aida K.

1973. “Formal Voluntary Organizations: Participation, Correlates, and Interrelationships.” Sociological Inquiry 43 (3–4): 89–122. doi:10.1111/j.1475-682X.1973.tb00004.x.

70.

U.S. Census Bureau. 2013. “American Community Survey.” 1-Year-Estimate Summary File Data. Accessed February 23, 2017. https://www.census.gov/programs-surveys/acs/data/summary-file.2013.html.

71.

Walker

Edward T.

McCarthy

John D.

2010. “Legitimacy, Strategy, and Resources in the Survival of Community-Based Organizations.” Social Problems 57 (3): 315–40. Accessed February 15, 2020. https://www-jstor-org-443.web.bisu.edu.cn/stable/10.1525/sp.2010.57.3.315.

72.

Willits

Dale

Broidy

Lisa

Denman

Kristine

. 2013. “Schools, Neighborhood Risk Factors, and Crime.” Crime & Delinquency 59 (2): 292–315. doi:10.1177/0011128712470991.

73.

James C.

2016. “Community Context of Crime: A Longitudinal Examination of the Effects of Local Institutions on Neighborhood Crime.” Crime & Delinquency 62 (10): 1286–312. doi:10.1177/0011128714542501.

74.

James C.

Hipp

John R.

Boessen

Adam

. 2016. “Voluntary Organizations and Neighborhood Crime: A Dynamic Perspective.” Criminology 54 (2): 212–41. doi:10.1111/1745-9125.12101.

75.

Sung-Suk Violet

Maxfield

Michael G.

2013. “Ordinary Business: Impacts on Commercial and Residential Burglary.” British Journal of Criminology 54 (2): 298–320. doi:10.1093/bjc/azt064.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.34 MB