Organizational Density and Neighborhood Employment

Abstract

Previous research finds that organizations are important sites where knowledge about employment is diffused and social ties are embedded. This article builds upon this research by asking whether organizational density has implications for employment at the neighborhood level. More specifically, this study focuses on the relationship between ethnic organizational density and employment by the extent to which ethnic organizations influence neighborhood employment levels. Using post–World War II San Francisco as a case study, this article finds that the density of ethnic organizations positively affects neighborhood employment over time, though this effect is spatially tempered by their location in particular regions of the city.

KEYWORDS:

• quantitative data analysis
• poverty alleviation
• ethnicity
• organizations

Notes

1. Although these directories provide the most comprehensive listing of organizations in print, it is likely that very small organizations are not represented. Also it is possible that older organizations will be over-represented and younger organizations under-represented. For example, established Catholic or Protestant churches are more likely to appear in the directory than storefront churches with shorter histories in a neighborhood. Although this is certainly a concern, this does not bias the data because this research is interested in the overall effect of particular organizational types.

2. Other methods of estimations were attempted to represent the causal order. Structural equation models could not be estimated because there were too few cases (N = 111) in each cross-section of data to estimate effects for the number of relevant organizational and demographic variables. Fixed effects and generalized-estimated equations were also attempted; these methods are unable to correct for spatial autocorrelation and are not currently supported by spatial software packages.

3. Diagnostic tests available in this version of SpaceStat that interacts with ArcView revealed that adding a dummy variable that distinguishes tracts in the core from those in the periphery reduces heteroskedasticity. For this reason, spatial lag models are run using the GHET approach. The region variable was constructed based on exploratory spatial analysis tools (ESDA). Maps revealed that census tracts around the CBD exhibited high spatial autocorrelation—tracts with low employment were surrounded by tracts with low employment. The pattern was distinct however in the western part of the city. In the western region, tracts with high employment were surrounded by tracts with high employment. Given these patterns, a region variable was constructed where tracts eastern side of the city were given a value of 1.

4. I experimented with other time lags and found effects to be correlated with tract-level variables.

5. SpaceStat provides several statistics of fit including diagnostics for multicollinearity. The regression diagnostic indicates that a model is properly specified when the condition value is less than 30.

6. It would be interesting to examine separately employment in Black, White, and other ethnic minority neighborhoods. However, the small number of tracts in San Francisco makes this difficult.