Surprising Causes: Propensity-Adjusted Treatment Scores for Multimethod Case Selection

Scholarship on multimethod case selection in the social sciences has developed rapidly in recent years, but many possibilities remain unexplored. This essay introduces an attractive and advantageous new alternative, involving the selection of extreme cases on the treatment variable, net of the statistical influence of the set of known control variables. Cases that are extreme in this way are those in which the value of the main causal variable is as surprising as possible, and thus, this approach can be referred to as seeking "surprising causes." There are practical advantages to selecting on surprising causes, and there are also advantages in terms of statistical efficiency in facilitating case-study discovery. We first argue for these advantages in general terms and then demonstrate them in an application regarding the dynamics of U.S. labor legislation.