PS: Political Science & Politics
Harbers, Imke, and Matthew C. Ingram*
Most outcomes that social scientists care about—including democracy, development, institutions, participation, and violence—are not distributed randomly across geographic space. Similar units are often located near one another so that phenomena of interest tend to cluster or exhibit similar patterns in space, making these phenomena spatially dependent. This clustering of outcomes (as well as any clustering in explanatory variables, including omitted variables that might lurk in the error term of statistical models) is no accident and should matter in how scholars understand and explain each outcome. That is, clustering is often a symptom of some underlying spatial process (e.g., diffusion), so scholars should take these underlying processes seriously in developing explanations of the related outcome of interest. Further, because the study of interdependent social phenomena is at the heart of social science, arguably all social science data are inherently spatial or—at the very least—spatial data are central to the social sciences (Darmofal 2015, 11–13). Nevertheless, the spatial dimensions of political data rarely receive explicit attention in how multimethod scholars design and conduct research.
* Denotes CSDA Associates, Affiliates, and Staff