Racial context and voting behavior in one-party urban political systems

Abstract

Two Democratic primary elections for city-wide office in St Louis and one non-partisan election for city-wide school board positions are analyzed for the vote of seven candidates. Racial structuring of voting patterns is the focus of analysis and the precinct is the unit of analysis. Census information is aggregated or allocated to the precincts. Racial effects order the voting behavior in models based solely on race. When multivariate prediction models are formulated which include non-racial predictors the racial effects remain strong. Evidence for social class effects running contrary to the direction of race are found in the school board elections but not in contests for the mayoralty Democratic nomination. Spatial autocorrelation between observation units is assessed for voting and shown to be strong. Reassessment of spatial autocorrelation in the multivariate model residuals is used to evaluate model specification. Although very large reduction in spatial autocorrelation is achieved with multivariate event count models, the residuals retain statistically significant evidence of spatial structuring.