Published online by Cambridge University Press: 30 November 2022
While quantitative methods are routinely used to examine historical materials, critics take issue with the use of global regression models that attach a single parameter to each predictor, thereby ignoring the effects of time and space, which together define the context in which historical events unfold. This problem can be addressed by allowing for parameter heterogeneity, as highlighted by the proliferation of work on the use of time-varying parameter models. In this article, I show how this approach can be extended to the case of spatial data using spatially varying coefficient models, with an eye toward the study of electoral politics, where the use of spatial data is especially common in historical settings. Toward this end, I revisit a critical case in the field of quantitative history: the rise of electoral Populism in the American West in the period between 1890 and 1896. Upending popular narratives about the correlates of third-party support in the late nineteenth century, I show that the association between third-party vote share and traditional predictors such as economic hardship and ethnic composition varied considerably from one place to the next, giving rise to distinct varieties of electoral Populism—a finding that is missed by global models, which mistake the mathematically particular for the historically general. These findings have important theoretical and empirical implications for the study of political action in a world where parameter heterogeneity is increasingly recognized as a standard feature of modern social science.