If the polls are correct, we’re heading for either a good year for Democrats or a very good year for Democrats.
They’ll do well in House contests, probably getting the 23 seats they need for a majority and possibly a lot more. They’ll win a bunch of races for governor, and regain up to half the state legislature seats they’ve lost since 2008. They’ll likely fall short of a Senate majority, but only because the playing field is tilted against them: Democrats will win more than half the Senate elections this year, and the vast majority of votes cast, but that probably won’t be enough to make up for the Republican lead in seats that aren’t up for election.
If, that is, the polls are correct. Contrary to a lot of myths, polling usually works pretty well. But it isn’t foolproof, and this year the risk of getting it wrong is higher than normal.
So what could happen?
A late surge. That’s what happened in 2016: Hillary Clinton’s lead over Donald Trump fell sharply in the last two weeks of the campaign. Although the polls picked this up, or at least most of it, it’s a good reminder that even mid-October surveys can miss late change. Could it happen again? It probably won’t, but it could—in either direction. Watch what’s in the news in the final days before the election.
Luck. All polls are subject to sampling error and other biases. Even when there are enough surveys to produce a mostly reliable average, it’s possible that random chance won’t even out and the average will end up being off. In House elections, there’s also the question of distribution. Even if surveys get the overall vote for the House correct, the distribution of those votes could wind up favoring one party or the other. That could be because of targeting, demographics or some other substantive reason. But it could also simply be luck of the draw. Just as pollsters misread the distribution of Trump’s vote, partly because there were fewer quality state polls than national ones, it’s possible that they’ll misjudge House contests this year because there are few reliable polls in most districts. Good forecasters can extrapolate from national surveys and other districts to project what will happen in under-polled races. But there’s always the chance it won’t work.
Missing the electorate. Both late surges and bad luck are risks to polling in every cycle. What’s especially tricky about 2018 is that so many unusual things are happening at once. Money raised is off the scale. Early voting is off the scale. The president’s unpopularity is off the scale. Trump has actually rallied a bit lately, but at 43.1 percent approval (as calculated by FiveThirtyEight), he still ranks third lowest of any president in the polling era though 641 days, and his disapproval number is dead last. There’s never been a president who spent virtually all of his first two years having more than half the nation disapprove of him.
No one knows how all this will translate into votes. Normally, various indicators of voter interest can be helpful in making predictions, but when so many indicators are off the scale, it’s harder to extrapolate. Forecasts based on fundamentals and projections based on reporting from the field are similarly vulnerable when the inputs aren’t within the normal range of experience.
Remember, too, that these mistakes could go in either direction. Take early voting. Record totals might indicate that the same people who would normally vote in a midterm are simply more enthusiastic this year. Or they could predict record turnout. Both outcomes would look pretty much the same at this point. It’s also possible that reasons for early voting differ by party, state or region. So anyone attempting to predict the eventual size and composition of the electorate using early voting risks either over or underestimating the effect.
In most elections, such risks tend to be overblown or to cancel each other out. But you can’t know that in advance. So, yes, the polls usually get most things right—except when they don’t.
This article was provided by Bloomberg News.