To subscribe to the monthly C&E email newsletter and event announcements click here.
I feel bad for any pollster who was trying to track the Alabama special Senate election.
One of the keys to an accurate survey is talking to the right people. Pollsters typically start with past turnout behavior and then adjust the turnout model depending on the circumstances of the current election.
For the 2018 cycle, for instance, most pollsters will look at who voted in 2014 as well as 2016 and 2012. These last two where both presidential election years, which must be factored in. This process is part science and part art, and every pollster tends to take a slightly different approach to determining what their universe will be.
The special election in Alabama was going to be tricky for pollsters to begin with. But the allegations against Roy Moore made this race a near impossibility to “get right.” Alabama is one of the reddest states in America. Indeed, President Trump won with 62.7 percent of the vote in 2016, which was his fifth highest win from a percentage basis.
If Sen. Luther Strange had won the GOP primary, I am sure the number of polls would have been few and far between since the outcome would have been inevitable. But as we all know, this wasn’t the result in the primary and Moore was on the ballot.
Before Moore won the primary, expectations on turnout typically were around 10 percent of the electorate based on previous special elections in the state. Last night ended up being closer to 40 percent, with Democrat Doug Jones winning with 49.9 percent of the vote to Moore’s 48.4.
Given all these factors, it’s not surprising to see that most polls were off the mark. Indeed, Monmouth University was the only public poll in December that had the race tied (fairly close to the end result), and Fox News was the only outlet to have Jones ahead (although they had him winning by 10 points).
It should be mentioned that Monmouth also ran the numbers based on higher or lower turnout, which swung the numbers from a 48-to-45 Jones win using a high turnout model to a 48-44 Moore win using a lower turnout model. SurveyMonkey also did a very interesting study showing that their data on the race ranged from Jones winning by 8 points to Moore winning by 10 points depending on how they modeled the turnout.
While both SurveyMonkey and Monmouth had similar views on the difficulty of getting this race right, when we look at the internals of the two surveys we see some interesting discrepancies. Monmouth had Moore winning white voters by a 64-to-28 percent margin. SurveyMonkey had it much closer among white voters (56-to-41 percent for Moore) and the exit polls had it at 68-to-30 percent for Moore. Both surveys had overwhelming support for Jones among African-American voters as did the exit polls.
Fox News doesn’t release their internal numbers, but from the press release they had Moore winning white voters 55 to 35. If Fox News had the proportions similar to the 70-30 split between white and African-American voters respectively, then Fox News had Jones winning African-American voters by a 92-to-8 margin, which was basically what the exit polls had. This is just a guess, but interesting to see this trend of pollsters having a tighter race among white voters.
We will need to wait to see who ended up turning out for this election before we can truly understand if the proportions of turnout are correct in the exit polls. Moreover, if the proportions were correct and Moore “grew” his support among white voters, one hypothesis would be a bit of incorrect self-reporting. That would mean white voters said they were not going to support Moore, but ended up doing that once they were in the privacy of the voting booth.
Looking back, this was going to be an incredibly tough race for the pollsters to get right. While the average across the polls showed a close race, there were some very different views of what was likely to happen on Election Day. The view on turnout was no doubt the biggest driver of these differences, and I am sure any pollster who was involved in surveying voters in Alabama are all happy this election is over. But by the same token, I think we all are happy we no longer need to talk about Roy Moore.
Stefan Hankin is founder and president of Lincoln Park Strategies, a Washington D.C.-based public opinion firm. Follow him on Twitter at @LPStrategies.