What Elections are Worth Recounting?

Posted on October 13th, 2011 by

When: Wednesday, October 19, 11:30am – 12:20pm
Where: Olin Hall 320
Presenter: Max Hailperin

Minnesota law provides for publicly-funded recounts of close elections. This raises the question of what exactly constitutes a “close election.” Is the current standard appropriate? Two recent high-profile statewide recounts have drawn considerable attention to this question, though in fact multiple smaller-scale recounts occur each year. Attempts to change the law have been made in each of the two previous legislative sessions and another attempt is expected in the next session.

The goal isn’t just a correct election outcome, it is public confidence in a correct election outcome. That makes this a question as much about rhetoric as about statistics. Nevertheless, there isn’t much sense in debating possible changes to the recount laws without some contextual understanding of the statistical questions. For example, in past recounts, what fraction of ballots were found to be non-machine-readable, yet have human-discernible voter intent? If one assumes (questionably) that this fraction represents a uniform probability that applies independently to each ballot, then what is the largest number of problem ballots one might plausibly encounter, as a function of the election size? And what is the largest change in vote margin that might plausibly result from counting those extra ballots, given that some cancel each other out? Finally, what does experience tell us about other sources of error that affect entire batches of ballots?

Lunch will be served.

This presentation is part of the MCS Seminar series; please see the calendar of upcoming events. Also, please email suggestions for future seminars as well as any questions.

 

Comments are closed.