The Washington PostDemocracy Dies in Darkness

The bad news for Donald Trump: Third debates don’t move the polls

The Trump Taj Mahal casino resort is seen in Atlantic City, New Jersey, on May 8, 2016. (JEWEL SAMAD/AFP/Getty Images)

With the clock ticking down to Election Day, Donald Trump is in a position that patrons of his casinos know well: Needing one last, big win to break even. He's been at the table for over a year, and has been throwing good money after bad for a few weeks now. Since the first debate on Sep. 26, really, which was the point at which his fortunes really started to tank.

Trump's problem is that making up that difference on a single hand is a long-shot, no matter what. But the other problem is that first debates are high stakes, high reward affairs. Third debates, like the one tonight? Not so much. To continue the belabored analogy above (hey, the debate's in Vegas), Trump is down $10,000 and is about to sit down at the penny slots.

Gallup polls debate viewers to get a sense of how they think the candidates fared. There have been six third debates in presidential history, with Gallup polling after five of them. Four have occurred since 2000, a period for which we have decent polling averages available. We took Gallup's data, Nielsen viewership data and national poll numbers to try and figure out how much of an effect the debates had on the election. Since polling shifts aren't instant, we looked at the national poll numbers two weeks after the debate, comparing the margins for the eventual victor (or Hillary Clinton) with the margin on the day of the debate itself.

We also made a rough metric combining viewership and debate victory margins in an attempt to overlap how well or poorly a candidate did with the number of people who saw the performance.

There's not really any correlation between viewership and poll shifts, which makes sense. There is a stronger correlation between how people felt the candidates did and how the polls moved. Combine viewership with those Gallup numbers, and the correlation is even stronger.

Which debate is under consideration matters. Over the last four cycles, the polls have moved much more after the first debate than the second, on average, and more after the second than the third. (On average, the movement after the first debate was 3.7 points. After the second, 1.3 and after the third, 1.0.)

Because of that, the correlation between debate performance and poll shifts is stronger for the first debate than the second and third. Remove the first debate from the calculation and there's almost no correlation between performance and what happens in the polls -- even less correlation than between poll movement and viewership. In other words: A good performance in a third debate means far, far less for polls than a good performance in the first debate.

The sample size here is small: Three debates. Maybe Trump can hit it big tonight and reverse his fortunes. Maybe he'll drop in his penny and hit the jackpot, making up his losses and then some. It's the dream Vegas is built on.

But you know the saying: The house, the owner of the casino, always wins. The odds are always against the big payoff, particularly in lower-stakes games.

The good news for Trump is that sometimes casino owners go broke.