Last week, Kentucky made national headlines as Senator Mitch McConnell easily swept by his Democratic challenger Secretary of State Alison Lundergan Grimes. While polls predicted the race to be close in the week leading up to the election, showing a 5-9 point lead for McConnell, he ended up with a 15-point win on election night. In this blog post, I will explore why there was such the discrepancy in predicted polls and the margin of victory as well as see what we learned from the 2014 Kentucky Senate election.
Why Did Polls Report a Close Race when There Was a 15-Point Margin of Victory?
According to Andrew Gelman and Gary King, when opinion-based polls circulate, people tend to respond with “enlightened answers.” Enlightened answers, to them, are not rational. Especially with polls in the beginning stages, people are responding with imperfect information—they do not exactly know what either candidate stands for, but instead make assumptions and do not take their uncertainty into account when responding to opinion-based polls. Thus, especially in the early stages, polls are inaccurate predictors of rational voting behavior (see here and here for information about poll accuracy). This, for Gelman and King, would explain the horserace of opinion polls throughout campaigns—individuals are continually gaining more and more information and thus their answers are becoming more rational as time progresses.
Was that the case with McConnell and Grimes? Looking at the aggregate polling data, as time progressed, Grimes was less and less likely to be leading in the polls. In the month leading to the election, Grimes polled ahead of McConnell a single time, and even then, it was well within the margin of error. Looking at a time period between May 2013 and June 2014, Grimes polled ahead of McConnell 9 times, and was tied with McConnell one time. Thus, most of the data used in the horserace media culture presenting the race as a tight one came from the beginning months of the election. In contrast, the one time Grimes polled ahead of McConnell during the last month of the election, it was within the margin of error, which should have been a clear predictor of McConnell’s victory; the media, however, didn’t focus on that—they often gave the impression that the race was continuing to be close.
So now that we know it wasn’t as much as a horserace as the media made it out to be, why was the margin of victory higher than the predicted poll? KET asked college professors in Kentucky for their answers to this question. Responses ranged from the idea that perhaps undecided voters broke out at the last minute, to the idea that conservative Republicans who had strayed away from McConnell ended up voting for him while Grimes lost traction with liberal Democrats, to the idea that conducting polls is harder nowadays due to caller ID and cell phones even though polling methodology has improved. In a separate interview, Centre College’s Professor of Politics Benjamin Knoll says that whatever the cause for McConnell’s over-performance relative to the polling averages, it was not limited exclusively to Kentucky. “Polling averages tended to underestimate Republican advantages in several Senate elections this year,” he said, “and thus Kentucky’s outcome fits into a wider national pattern.” Thus, there are apparently many reasons why the polls may not have matched the margin of victory.
Takeaways from the Kentucky Senate Race:
Grimes fell severely short in places where she should have won: on election night Grimes lost half of the ten most Democratic counties by voter registration—all of which are in Eastern Kentucky (see here for statistics on turnout by county)
Counties with the highest turnout were not turning out for Grimes: of the 39 counties that saw turnout to be above 50%, Grimes only won two (see here for statistics on turnout by county)
Obama helped McConnell win votes, but voters turned out for reasons other than to vote against Obama: One of every four voters said they were angered with the Obama administration; further, of those voters, ninety percent voted for McConnell; however, 46% of Kentuckians said Obama had no part in their reason for showing up to vote
Kentucky thinks the country is on the wrong track: 69% of Kentuckians think the country is on the wrong track; interestingly, around the same percentage said the same thing 6 years ago in McConnell’s previous election
Most Kentuckians already knew for whom they were voting: when asked when voters had made up their minds, 70% of voters had made up their minds more than a month before the election; interestingly, 10% of voters said they made up their minds in the three days leading up to the race, with an even distribution between those deciding to vote for any of the three candidates (all less than 1% of the candidates’ total vote percentage)
Last, but not least—41% of Kentuckians claimed they were not contacted by either Grimes’ campaign or McConnell’s: let’s say this together… “You lucky dogs!”