We’re almost done. After two debates full of memorable moments such as Donald Trump saying he’d put Hillary Clinton in jail and mentioning his 10-year-old son as a potential expert on computers and cybersecurity, the final presidential debate of 2016 is upon us. And Trump needs to win it. Clinton now leads Trump by 7 percentage points, according to the FiveThirtyEight polls-only forecast. (That translates to about an 88 percent chance of winning the election Nov. 8.) A clear Trump win in the third debate would presumably help him close that gap, but he now trails Clinton by a large enough margin that winning the third debate would be only step one in what would have to be a multistep comeback: The few third presidential debates we’ve had in past elections haven’t moved the polls much.
Historically, third debates have had less of an impact on the polls than first or second debates. Before we begin exploring those debates, though, a caveat: There have been only six third presidential debates before this year’s. You can’t draw ironclad conclusions from a sample size of six. This year, moreover, has been so odd that we can’t rule out something unexpected occurring. Still, given Trump’s unwillingness to prepare for debates and the shrinking number of undecided voters, I’d be surprised if the third debate moved the polls more than normal.
Here’s how the polling averages shifted between the week just before1 and the week just after the six previous third presidential debates:
| INCUMBENT PARTY AVERAGE POLLING LEAD | |||
|---|---|---|---|
| YEAR | PRE-DEBATE | POST-DEBATE | ABSOLUTE CHANGE |
| 1976 | -3.0 | -2.4 | 0.6 |
| 1992 | -16.0 | -9.9 | 6.1 |
| 2000 | -2.1 | -2.4 | 0.3 |
| 2004 | +0.7 | +0.8 | 0.1 |
| 2008 | -8.4 | -7.1 | 1.3 |
| 2012 | +0.5 | +1.0 | 0.6 |
| Average | 1.5 | ||
All but one year saw a shift of about 1 percentage point or less. The average shift has been only 1.5 percentage points, compared with 2.1 points for second debates and 2.6 points for first debates.
So although we have a small sample, all this makes sense. Third debates are closer to Election Day, so you’d expect more voters to be locked into their choices. Also, there have already been two debates; voters have seen the candidates in this setting. If one candidate is an amazing debater, then you’d expect that advantage, at least partly, to already be baked into the polls. Finally, viewership rarely peaks for the third debate. In 2012, for example, 6 million fewer people watched the third debate than watched the second.
What about the 6.1 percentage point shift that occurred after the third debate in 1992? As I’ve noted before, 1992 was an odd election because of independent candidate Ross Perot. Perot rose in the polls in October and participated in the presidential debates. Polls tend to be more volatile in a three-candidate race. Moreover, if you look at the polls before and after the third debate in 1992 by candidate (instead of the margin between Democrat Bill Clinton and Republican George H.W. Bush), you’ll see that Bill Clinton lost most of his ground to Perot, not Bush, after the third debate. It’s less of a jump for voters to go from their candidate to a moderate independent or third-party candidate like Perot than to a candidate of the opposite party. This year, Trump doesn’t have that luxury. Only the two major-party candidates are in the debates, and the third-party candidates are losing, not gaining, ground in October.
Trump’s more realistic goal tonight would be to make up a couple of percentage points, not overtake Hillary Clinton. If Trump can, for instance, cut his deficit in half, then he’d be within the range of a normal polling error in the final few weeks of the campaign. Here’s the difference between the polling the week after previous third debates and the election results:
| INCUMBENT PARTY ADVANTAGE | |||
|---|---|---|---|
| YEAR | POST-DEBATE POLLING | ELECTION RESULT | ABSOLUTE DIFFERENCE |
| 1976 | -2.4 | -2.1 | 0.3 |
| 1992 | -9.9 | -5.6 | 4.3 |
| 2000 | -2.4 | +0.5 | 2.9 |
| 2004 | +0.8 | +2.5 | 1.7 |
| 2008 | -7.1 | -7.3 | 0.2 |
| 2012 | +1.0 | +3.9 | 2.8 |
| Average | 2.0 | ||
The average change between post-third-debate polls and the result (2.0 percentage points) is bigger than the average change in the polls from before to after the third debate. Changes of about 3 points or more occurred three of six times. Still, there’s never been a shift of more than about 4 points. It’s very rare for polls to miss by wide margins in the final few weeks.
Simply put, 7 percentage points is a big deficit. Forget third debates. Counting this year, there have been 21 presidential debates for which we have polling immediately before and after. Not one of them resulted in a shift in the polling of more than 6.1 percentage points. Tonight’s debate is very unlikely to vault Trump into the lead, but it’s probably Trump’s last shot to start giving Clinton a run for her money.