The Wisconsin Recount: Learning about the Accuracy of Humans and Machines

• In Event: Electoral Legitimacy and Representation

Abstract

We examine the presidential election recount in Wisconsin in 2016, primarily to assess the relative accuracy of vote-counting methods (hand-count vs. machine-count). For the past two decades, there has been a running controversy over the relative accuracy of different methods of counting ballots. This question can rarely be adjudicated empirically, because there is rarely a basis to compare the quality of hand-counting vs. machine counting. Wisconsin 2016 provides such an opportunity. A significant fraction of Wisconsin employs hand-counting of ballot. Using the results from the 2016 recount, we explore the question of which method was more likely to get the count right the first time, hand- or paper-counting. This is only the beginning of the analysis, however. One question that has gone unexplored is the degree to which voting technology inaccuracies are due to problems with the machines themselves (e.g., hanging chad) or due to problem with administrative procedures associated with the use of machines (e.g., transcription errors on reporting forms). Our analysis of administrative records for the Wisconsin recount will help provide a first-time systematic glimpse at the frequency of counting errors that can be attributed to tabulation errors, per se, and reporting errors that are caused by administrative problems.

Authors

• Stephen D. Ansolabehere, Harvard University

• Kenneth R. Mayer, University of Wisconsin, Madison

• Charles Stewart, Massachusetts Institute of Technology