| Abstract: | A growing number of papers have appeared which report that the partisanship of the local context influences the party attachments of individual residents. The present paper criticizes this finding and the contextual approach to the study of partisan choice. First, the general mathematical model of contextual partisanship is derived from existing research. Next, a multiple data file, consisting of survey and aggregate evidence, is used to examine the generality of certain specific models. Finally, the same data file is employed to consider the effects of two assumptions common to contextual models, "random mixing" and nonselection. Only one of four specific models is initially found to fit observations. Further analysis reveals that individuals' party allegiances are spuriously attributed to contextual influences as a result of assuming random mixing and nonselection. |
