My latest magazine piece looks at the state of the presidential race after the conventions. One item from that column I’d like to expand on is the 2018 elections. I’ve typically tracked midterm elections by looking at the RealClearPolitics polling averages over time, and seeing how the undecided voter poll shrinks — i.e., where those voters break to one party or the other. I did that after the 2018 elections. The polls called their share of races well in 2018, but they did have a few embarrassing misses, getting the final results wrong in the Florida, Indiana, and Arizona Senate races and the Florida, Ohio, Iowa, and Nevada governors’ races, and massively understating the margins of victory of Marsha Blackburn and Josh Hawley, among others. If we focus on states that had more than one Senate/governor’s race (Mississippi had two Senate races, Minnesota had two and a governor’s race), and average how the outcome diverged from the final poll average, see if you notice a pattern:
The list of states with the biggest poll misses underestimating Republicans is conspicuously heavy on states that Donald Trump carried narrowly in 2016 and needs badly to win again (Florida, Ohio, Michigan, Wisconsin, Pennsylvania), supplemented by Tennessee (one of Trump’s best states) and Maine (where Trump won an electoral vote in one of the state’s congressional districts). What these states all have in common is a lot of white working-class voters, many of them Midwesterners. (Note that Minnesota is not on that list.) Then you look at the opposite end of the list — New Mexico, Nevada, Texas, Mississippi, Arizona . . . a lot of Sun Belt states where Democrats rely heavily on non-white working-class voters who also tend to be harder to poll.
There is always a temptation to fight the last war in poll analysis. The more you applied the lessons of 2012 in 2016, the more wrong you were apt to be, and the same was true of applying the lessons of 2004 in 2008. But two consecutive election cycles showing the same poll failures in the same places is a valid basis for expecting a similar trend to persist at least so long as national politics revolves around Trump. That does not mean we should disregard bad news in the polling, but it does suggest that Trump may be likely to pull out some key swing-state races in these states where the polls show Biden ahead by only a few points at the end — and that Biden is likely to be stronger than his poll showing in the Sun Belt states outside of Florida.