When Boundaries Collide

Abstract

Scholars have long merged election and census geography to correlate census demographics and election results to infer political behavior (Ogburn and Goltra 1919; Gosnell and Gill 1935; Key 1949). Increasing accessibility of geospatially defined election data provides a valuable tool to understanding voting behavior in the United States at geographic levels unavailable to previous scholars. Here, we describe these data and examine four methods to merge spatial data when precinct and census boundaries are non-conforming: Areal weighting, dasymetric mapping, point kriging, and kriging-based areal interpolation. Through a case study of sixteen states and the District of Columbia, we find that dasymetric mapping-a method that uses external data to construct more accurate and realistic weights for areal weighting, in this case the National Land Cover Database-is the best method to estimate demographic characteristics of precincts when census block boundaries do not conform to precinct boundaries. We apply dasymetric mapping to generate a publicly available national database of merged election results and census data for precincts.