Statistics Students Learn Every Vote Counts
Faculty across country asked to participate in 2014 project
Mary W. Gray, American University
As a member and now chair of the ASA Scientific and Public Affairs Advisory Committee, I am always eager to interest statisticians in public policy issues. A project my students have carried out in the last two election cycles is a good example of how to involve students in the political process. In 2012, we conducted an exit poll in the District of Columbia, Montgomery County (suburban DC), and Northern Virginia. We did not ask how the interviewees voted, but rather whether they had any trouble voting, either because of voter ID laws (relevant mainly in Virginia) or other problems. This past fall, the only local jurisdiction where there was a statewide contest was Virginia, so we repeated the exit poll project there.
My class in survey sampling selected the precincts, and students from several of American University’s undergraduate basic statistics classes conducted the poll. This is a course taken by students majoring in a wide variety of subjects—arts and humanities, communication, sciences, and social sciences. American University has a strong social science emphasis and a very politically involved student body, so many students were eager to participate in a nonpartisan political process that would actually make use of what they were learning.
The precincts chosen came from Arlington, Fairfax, and Loudon counties, as well as the city of Alexandria. Initially, the students were somewhat disappointed that less than 2% of those interviewed as they were leaving the polls reported having problems; any concern about suppression of voter turnout seemed unfounded. However, at the end of the count the day after the election, it turned out that in the race for attorney general—one of three statewide offices—the margin of victory was 165 votes out of a total of more than 2.2 million votes cast. A recount weeks later increased the margin to around 900 votes, but no one was left in doubt about the importance of individual votes.
We did do a bit of stratification by income level in this year’s project and found relatively more difficulties with voting in lower-income precincts, but we didn’t draw any conclusion with our sample size. We also found younger (<25) and older (>65) voters encountered somewhat more problems than the middle group; one can think of various reasons why this might be the case.
Students who participated, and some of their friends who didn’t, have already asked whether we are planning to repeat the project in 2014. Yes, of course, as all local jurisdictions will have crucial races. However, we would like to see widespread participation across the country, particularly in those states with highly restrictive voter ID laws.
In 2012, I wrote to a number of statistician friends on faculties around the country to try to engage others in the exit poll project, but only statisticians and their students from George Mason University in Northern Virginia joined in. Locals know that Northern Virginia is not representative of the state as a whole, and the other jurisdictions reached easily by American University are also not very representative of the country as a whole with respect to voter ID laws—or other characteristics for that matter. We would like to have results from a wide variety of states and regions within states.
What I would like to do is to persuade statistics faculty in institutions around the country to engage their students in this project next year. There’s plenty of time to plan for this fall, so please let me know if you would like to know more about this project and perhaps join with us.