Donald Trump’s campaign is teetering on the brink. He had a horrible first debate, then started sliding in the polls, and then came the release of a 2005 tape in which Trump bragged about being able to sexually assault women. Scores of Republican officials are abandoning him, and Trump probably needs a strong second debate to get back into the race. Given Trump’s recent troubles and lack of debate prep, logic suggests that that will be a challenge. History too suggests that while a second-debate bounce is possible for Trump, it isn’t likely.
Second debates usually don’t move the needle much. Let’s look at how the polls shifted from the week before second debates to the week after:1
INCUMBENT PARTY AVERAGE POLLING LEAD | |||
---|---|---|---|
YEAR | PRE-DEBATE | POST-DEBATE | CHANGE |
1976 | -2.0 | -3.8 | -1.8 |
1984 | +16.8 | +17.0 | +0.2 |
1988 | +5.9 | +11.5 | +5.6 |
1992 | -10.3 | -16.0 | -5.7 |
1996 | +14.5 | +14.7 | +0.2 |
2000 | -0.8 | -2.1 | -1.3 |
2004 | +2.4 | +0.7 | -1.7 |
2008 | -6.2 | -8.4 | -2.2 |
2012 | 0.0 | +0.4 | +0.4 |
Seven out of nine times, the polls moved by less than about 2 percentage points. Trump is currently behind by 5.6 percentage points in the FiveThirtyEight polls-only popular vote forecast. So Hillary Clinton would still be up by a wider margin than she was heading into the first debate, if Trump got an average second-debate bounce. Still, the polls moved by 5 to 6 points twice, in 1988 and 1992, so such a shift isn’t out of the question.
Unlike first debates, there’s no pattern in which party tends to benefit, the party that holds the White House or the challenging party. Five times the incumbent party gained ground; four times the challenger did. The larger bounces fit this pattern as well. One (1988 for Republican George H.W. Bush) was for the incumbent party, and one (1992 for Democrat Bill Clinton) was for the challenging party.
There’s also no sign that candidates who lose the first debate are therefore more likely to rebound in the second debate. Here’s a chart that plots the change in the polls after the first debate compared to the change in the polls after the second debate:

Sometimes the candidate who does better in the first debate does better in the second too. Sometimes he does worse. Either way, there isn’t a tendency for the race to revert to some sort of pre-first debate polling equilibrium.
That’s not to say that Trump has no shot tonight at making up significant ground. Both of the largest largest bounces (Bush in 1988 and Clinton in 1992) were larger than any bounce coming out of first debates. Moreover, those debates both showed that a big moment can move the polls. The second debate in 1988 featured Michael Dukakis’s famous answer on the question of whether or not he would favor the death penalty if his wife was raped and murdered.
And Bill Clinton saw his poll numbers rise in 1992 after George H.W. Bush couldn’t really say how the national debt affected him personally.
That answer came in a townhall style debate (as tonight’s debate will be).
Even if tonight’s debate does move the polls, that’s no guarantee they’ll stay moved. Historically, the polls taken immediately before the second debate have been more predictive of the final election result than the polls taken immediately following it.
INCUMBENT PARTY EDGE | ABSOLUTE POLLING ERROR | ||||
---|---|---|---|---|---|
YEAR | PRE-DEBATE AVERAGE POLLING | POST-DEBATE AVERAGE POLLING | ELECTION RESULT | PRE-DEBATE | POST-DEBATE |
1976 | -2.0 | -3.8 | -2.1 | 0.1 | 1.7 |
1984 | +16.8 | +17.0 | +18.2 | 1.4 | 1.2 |
1988 | +5.9 | +11.5 | +7.7 | 1.8 | 3.8 |
1992 | -10.3 | -16.0 | -5.6 | 4.7 | 10.4 |
1996 | +14.5 | +14.7 | +8.5 | 6.0 | 6.2 |
2000 | -0.8 | -2.1 | +0.5 | 1.3 | 2.6 |
2004 | +2.4 | +0.7 | +2.5 | 0.1 | 1.8 |
2008 | -6.2 | -8.4 | -7.3 | 1.1 | 1.1 |
2012 | 0.0 | +0.4 | +3.9 | 3.9 | 3.5 |
Average | 2.3 | 3.6 |
Both Bush’s large bounce in 1988 and Clinton’s 1992 proved fleeting, for example. That doesn’t mean a Trump bump after tonight necessarily would be, but there hasn’t been a single time when the polls conducted after the second debate were significantly more predictive than those taken before it. Meanwhile, there are examples, such as 1988 and 1992, of the opposite being true.
Second debates simply haven’t been great venues in which to change the trajectory of the race. In this case, that’s bad for Trump, since he’s losing. But as with all studies of modern presidential campaigns, we have only a small sample of examples to work with. So this analysis doesn’t mean Trump won’t right the ship tonight. It just means you shouldn’t bet on.