They Just Do Not Vote Like They Used To: A Methodology to Empirically Assess Election Fraud


Political Science and International Affairs

Document Type


Publication Date



Objectives. In contemporary U.S. elections there is no shortage of allegations concerning election fraud. These claims are, however, based in large part on anecdotal evidence, unsubstantiated assertions, or the study of reported complaints. The absence of a general methodology to actively search for evidence of election fraud has resulted in policy arguments devoid of empirical data and systematic analyses.

Methods. In this article, we present a general methodology to study contemporary election fraud based on the Knowledge Discovery in Databases (KDD) process. We then apply this approach to a case study of a particular type of fraud.

Results. After examining approximately 2.1 million votes cast during the 2006 general election in Georgia, we find no evidence that election fraud was committed under the auspices of deceased registrants.

Conclusion. In this article, we have introduced a general methodology for the scientific study of election fraud. We urge social scientists to make use of such a framework to investigate the prevalence of different types of fraud across varying election cycles and jurisdictions.