Political Science Speaker Series: Walter Mebane

Election forensics is the use of statistical methods to determine whether the results of an election accurately reflect the intentions of the electors.  I introduce and investigate a statistical model called eforensics, which is a measurement model that I argue produces valid aggregation unit estimates of the incidence and magnitude of election frauds, using as input aggregation unit (e.g., polling station) counts of eligible voters and of votes cast for the ballot alternatives.  While valid, the model's estimates of realized frauds are imperfect:  eforensics cannot detect procedural defects, measures only frauds that benefit one alternative and is sensitive to things---particularly elector strategic behavior and election administration weaknesses---not produced by the kinds of malevolent distortions of elector intentions that constitute genuine frauds.  The parameters of and other quantities computed from the finite mixture eforensics model specification, which are estimated using Bayesian methods, to some degree support distinguishing results of malevolent distortions from results of strategic behavior and from lost votes.  The method's success and its ambiguities are due to its being informed by dependencies between electors that malevolent distortions of elector intentions induce, but other normal features of electoral politics such as elector strategic behavior and lost votes also induce dependencies.