Between Models and Cases: Turning Regression Modeling Inside Out for Comparative Historical Analysis

Eric Schoon, The Ohio State University
Ronald Breiger, University of Arizona

Linear regression analysis is a powerful tool for understanding the relationships among variables in a dataset and it has long been used in comparative-historical research on states, politics and society to identify patterns across both space and time. Yet, regression models impose homogenizing assumptions, such that any patterns identified by the regression model are presumed to apply equally to all cases in the dataset. More broadly, regression models render the cases themselves invisible, thereby masking the mechanisms that explain statistical associations. For researchers interested in drawing conclusions about cases—especially cases that are not statistically and theoretically exchangeable, such as states, geographical territories, organizations, or social movements—these features of the linear regression models present meaningful limitations. In this paper, we provide an overview of a novel approach to regression analysis, termed regression inside out (RIO), which brings cases to the fore and allows researchers to explore the relationships between cases and variables through the lens of a regression model. We provide an overview of RIO, explaining how it can be used to facilitate dialogue between the conventional results of a regression model and knowledge of specific cases. We then demonstrate how adopting this case-oriented approach to regression analysis can yield powerful new insights. We do this through a re-analysis of Wimmer, Cederman and Min’s (2009) analysis of the impact of ethnic diversity and armed conflict. We show how RIO can bridge the gap between aggregate findings and individual cases, offering exciting new directions for deeper engagement with both the baseline regression model and the cases themselves.

No extended abstract or paper available

 Presented in Session 217. Beyond Models and Archives: Reconstructing Historical Complexity through Methodological Innovations