Election Update: Early Polls Suggest A Post-Debat

Quote:

---

Election Update: Early Polls Suggest A Post-Debate Bounce For Clinton

http://fivethirtyeight.com/features/election-.../undefined

By Nate Silver

Every scientific poll we’ve encountered so far suggests that voters thought Hillary Clinton beat Donald Trump in Monday night’s debate. In fact, some of them showed her winning by a wide margin — wide enough to make it a good bet (though not a guarantee) that she’ll gain in horse-race polls against Trump over the next week or so.

But so far, we’ve seen just two polls released that tested Clinton’s standing against Trump after the debate. They have pretty good news for Clinton, but I’d recommend some caution until we get more data.

The first poll is from Morning Consult, which shows Clinton leading by 3 percentage points in a matchup that includes third-party candidates — that’s a 4-point swing toward Clinton from the 1-point Trump lead that Morning Consult showed before the debate. In a head-to-head matchup against Trump, Clinton leads by 4 points, up from a 2-point lead before the debate.

The other survey is from Echelon Insights, and it shows Clinton leading Trump by 5 percentage points. In theory, that would be consistent with a bounce for Clinton, since she led Trump by just 1 to 2 points overall before the debate, based on FiveThirtyEight’s projection.

But it’s hard to know for sure because Echelon has surveyed the race only once before — just after the Republican convention, when they showed Clinton leading Trump by 1 point.

Apart from these polls, the only other data we have is from the USC Dornsife/Los Angeles Times and UPI/CVOTER national tracking polls, but I’d discourage people from paying very much attention to them.

It’s nothing against the polls themselves — in fact, I’ve defended the USC/LA Times poll’s methodology in the past — it’s just a matter of timing.

Each poll uses a 7-day field period, which means that only about one-seventh of their interviews were conducted after the debate. I’d wait a couple of days before making too much of these surveys — until they consist mostly of post-debate interviews.

There are other reasons to be cautious, too. Polls conducted over a one- or two-day period, like the Morning Consult and Echelon Insights polls, can suffer from low response rates, since the pollsters won’t have time to recontact voters who they missed the first time around.

That could plausibly bias the poll toward whichever candidate has the most enthusiastic supporters at the time of the poll, making it less representative. Many traditional pollsters prefer their polls to be in the field for three or four days, and we won’t see any results from polls like those until Friday at the soonest.

Another complication is that it can be hard to separate voters’ reaction to the debate itself from their reaction to the media’s reaction to the debate. By that I mean: Clinton has had some tough news cycles lately, so getting some better headlines could help her, and that could plausibly also affect the polls. Or maybe not, since Trump has a knack for turning the news cycle on its head.

One last admonition: When evaluating a post-debate bounce, consider whether the poll was an outlier before.

For instance, the most recent NBC/Wall Street Journal poll showed Clinton ahead by 6 percentage points nationally, on the high end of her range heading into the debate. By contrast, the most recent Rasmussen Reports poll had Clinton trailing Trump by 5 points.

Clinton is more likely to improve her numbers in the next Rasmussen poll than in the next NBC poll, but that could reflect reversion to the mean as much as a debate bounce.

Overall, there are some tentatively positive signs for Clinton — but not more than that, yet. At the moment, our polls-only model shows Clinton with a 58 percent chance of winning; polls-plus shows her with a 56 percent chance.

But our forecast models don’t make any special assumptions about the debate, and they’ll take several days to catch up to whatever impact it has or hasn’t had. You can track the latest polls as they’re added to the model here.

---