F or the last few weeks, the New York Times Upshot has been conducting a highly interesting experiment: They’ve been polling battleground districts and displaying the results of their calls, their estimates of where the different races are and more as the data comes in . It’s a very cool project—you can see more here . And now that The Upshot has finished a couple dozen House polls, I thought I’d be worth taking a swing through some of the data and coming up with a few takeaways. Trump Looms Large In House Races Historically, midterms are referendums on the incumbent president and his party (in this case, Trump and the GOP).
And this data suggests that that pattern is holding in 2018 This graphic shows each of the districts where a poll has been completed. The trend is simple —if Trump’s approval rating is high in a district, the margin tends to be better for the Republican. And if the voters really dislike Trump, then they tend to favor the Democratic candidate more. Obviously Trump doesn’t explain everything here. Incumbents tend to run more strongly than non-incumbents, and individual candidates, idiosyncratic districts, weird local issues and more can move (and do) move the needle. But the common thread between most midterm elections is the president, and right now Americans aren’t too fond of Trump .
The GOP Retirement Avalanche Matters Earlier this year, many analysts noted that Republican House members were retiring at a breakneck pace. Some of these congressmen left safe districts (places where there re-election was almost assured) because they no longer liked the job or wanted to run some other office, but others left swing-ier districts (maybe for the same reasons, or maybe to avoid a tough re-election fight). And right now the GOP is feeling the impact of those retirements.
In the open districts that The Upshot polled (ones where no incumbent is running), the Republican candidate’s share of the vote was on average just a bit lower than Trump’s overall approval rating in that district. But in the districts where an incumbent was running, the Republican’s vote percentage was higher than Trump’s approve rating by three points on average. There’s a lot of variation in these estimates. Some incumbents ran behind Trump’s approval rating and others ran well ahead of it.
One open-seat Republican (Carol Miller) ran far behind Trump’s approval rating, but that’s probably because she both has a strong opponent in Democrat Richard Ojeda and West Virginia gets more Democratic as you move down the ballot. But the point is that, in general, the polls are picking up on the downstream effects of this year’s retirements. Hispanic Districts — a GOP Bright Spot? Republicans do have a bright spot—as others have noted , Republicans in heavily Hispanic districts have been polling better than one might expect just based on past presidential results.
In Florida’s 26 th District (Miami), Republican Carlos Curbelo leads his Democratic opponent, Debbie Mucarsei-Powell, by three points in a district that voted for Hillary Clinton by 16 points . Texas’s 23 rd District (which contains parts of both San Antonio and El Paso) is also heavily Hispanic (though it’s less Democratic—Clinton won it by three points in 2016) but Republican Rep. Will Hurd leads Democrat Gina Ortiz Jones by eight points. My guess (though I’ll want to see election results before committing to this point) is that low Hispanic/Latino turnout plus strong candidates might help the GOP in some of these districts.
Obviously, polling well in heavily Hispanic districts is a plus for the GOP. But it’s worth noting that there are only so many heavily Hispanic districts on the House map . A lot of Hispanic/Latino voters are packed into a relatively small number of districts in the Southwest, major cities and the tip of Florida. Put simply, a good showing in heavily Hispanic districts would help the GOP. But it’s probably not enough by itself to save the Republican majority. How the Sausage Gets Made I think the Upshot’s live polling project is great. It does an excellent job of showing readers how the sausage is made, so to speak, and that is a crucial part of communicating honestly about our data.
In my experience, people tend to think about polls as either a) complete crap or b) an infallible, 100 percent precise read of public opinion. But neither of those pictures are right. Polling is our best tool for prediction elections and quickly figuring out what mass public opinion looks like. But, as the Upshot’s project shows, pollsters need to make complicated decisions about how to design a poll, how to reach the right people and how to communicate about the random error inherent to the enterprise.