Over the course of fifty years, governments have sought to devolve the production of human services to private providers.  Nonprofit organizations have become the primary means of delivering vocational, youth-oriented, mental counseling, and daycare services.  More than 50% of human service nonprofit revenues are from public sources: government grants, contracts, vouchers, and fees-for-services.  With that said, there is great variation across jurisdictions as to the support for these organizations.  This paper answers two questions: first, is there evidence of fiscal spillover (e.g. Case, Rosen, and Hines, 1993; Brueckner, 2000) in the support for human service nonprofits across jurisdictions?  Second, do jurisdictions with people more in need of human services provide more support for those services? 

One of the distinguishing features of human services is that they are largely place-based.  While the online revolution of the past twenty years has made it possible to learn, transact, and play from a distance, in large part human services requires face-to-face interactions that cannot easily be moved to a virtual form.  As scholars have repeatedly noted, place matters (e.g. Allard, 2009; Hillier, 2008, Peck, 2009).  Individuals in poverty are particularly limited in their ability to move, as well as face socially-constructed heuristics about what types of organizations are for “their people” (Hillier, 2008).  It matters that human service nonprofits locate close to their potential clients.

While this question is both practically and theoretically important, this paper also seeks to introduce a relatively under-used tool in policy analysis: spatial regression.  Ordinary Least Squares (OLS) regression has become one of the dominant tools in the social sciences, and particularly in nonprofit studies.  Place is important in our field, yet it can potentially produce fundamental problems in the use of OLS.  Clustering over space leads to the autocorrelation of error terms, which in turn leads to biased OLS estimators.  Conceptually similar to time series analysis, where one attempts to control for serial autocorrelation by lagging variables over time, spatial regression uses a series of tools to account for clustering over space.  By using a spatial-lag or spatial-error model, one is able to account for the autocorrelation which in turn results in unbiased estimators (e.g. Anselin 1988).

The data used in the analysis is the National Center for Charitable Statistics Core Files for 2007-2009. Our dependent variable is human service expenses per capita in Northeastern counties, with independent variables representing SES values measured through the American Community Survey.  Using GeoDa and R, we construct a series of six spatial regression models.  The results indicate that counties with low SES have lower levels of expenses on human service nonprofits, while also exhibiting a high level of spatial clustering.  The result confirms the fiscal spillover model.