A constant refrain in political analysis is, essentially, that nothing matters. Sometimes, it can feel like not even a massive gaffe or scandal can affect how voters feel about a candidate, and not even major news events can move public opinion on issues.
But sometimes, things do matter. Presidential debates are one of those things. Last week's debate between Vice President Kamala Harris and former President Donald Trump produced fewer policy specifics than voters wanted, but along the way Harris succeeded in painting Trump in a bad light. In the week since the debate, polls have shown a slight increase in support for the Democratic nominee. As a result, on Monday, Sept. 16, Harris's chance of winning the election according to 538's forecasting model reached 61-in-100 — surpassing her previous high from Aug. 26.
If we look at all the polls that have been conducted entirely after the debate and compare them with each firm's pre-debate survey, Harris has made small gains in most of them. Harris gained 1 percentage point on margin according to an Ipsos/Reuters national poll released late last week; 2 points in RMG Research's polling; 2 points according to Morning Consult; 2 points according to Big Village; and 1 point according to SoCal Strategies. The only exceptions are Leger, Redfield & Wilton Strategies and Ipsos/ABC News, whose post-debate polls showed Harris losing 1 point, standing pat and standing pat, respectively, compared with their last surveys.
538's averages of national and state polls agree that Harris has made marginal gains since the debate. As of 9 a.m. Eastern on Tuesday, Sept. 17, Harris leads Trump by an average of 2.9 points in national polls, up ever so slightly from her 2.5-point lead on debate day. That 0.4-point gain is considerably smaller than the average post-debate change in election polls going back to 1952 — but then again, with polarization, most changes in opinion are smaller now. Notably, Harris is still down about 1 point from her late-August high in national polls.
In national polls, the real boon for Harris is not that she has gained four-tenths of a percent of support; it's that she has put a stop to Trump's gains. Before last week, Trump was gaining about half a point of popular-vote margin nationally on Harris since late August. If he had continued on that path, he would have been favored to win the race in 538's forecast by later this month — perhaps as soon as this week. Instead, Harris is the one who has modestly improved her chances of winning the race.
But because presidential elections are decided by the Electoral College, it's the state polls, not national polls, that really matter. And in addition to the national polls we've already mentioned, state polls also show real gains for Harris. In New Hampshire, a poll from Saint Anselm College Survey Center found Harris up by 8 points over Trump, 2 points better than Harris's pre-debate average in the state. Polls in Alaska, Arkansas and New Mexico also came in above average. A poll from the right-leaning Trafalgar Group released Monday showed Harris up by 2 points in Nevada. Polls from InsiderAdvantage and the highly rated Marquette University Law School found Harris up by 2 and 5 points, respectively, among likely voters in Wisconsin. And on Sunday, the always-closely-watched Selzer & Co./Des Moines Register poll showed Trump leading Harris by just 4 points in Iowa, a state our model had been expecting Trump to win by 8.
And yes, shifts in non-swing states matter, even if Harris is unlikely to win them in the end. Polling movement is often correlated from state to state, meaning that if Harris is gaining in Iowa and Arkansas — not to mention nationally — she's probably also gaining in Pennsylvania and North Carolina.
All of this nets out to a pretty sizable shift toward Harris in our election forecast (which can be a bit more aggressive than our polling averages in reacting to correlated movement in polls). Our model's predictions for the final margin between the two candidates in the big seven swing states are about 1 point better for Harris today than they were last week. Our forecast in Wisconsin has gone from Harris+2 to Harris+3; in Michigan, from Harris+1.5 to Harris+2.2; in Pennsylvania and Nevada, from Harris+0.1 to Harris+1.5; and in Georgia, North Carolina and Arizona, from Trump+1 to Harris +0.2.
In fact, the polls are so (comparatively) good for Harris that the model is actually forecasting a national popular vote that is more Democratic than the current national polling average. Our model calculates an implied national popular vote based on estimated turnout in each state, and today that number is closer to Harris+4.2 than Harris's 2.9-point lead in national polls. This echoes 2012, when state polls looked better for President Barack Obama than national polls did (Obama ended up beating his national polling average over Republican nominee Mitt Romney by more than 3 points).
This difference between the national electorate in the national polls and the one implied by state polls may sound small, but in fact 1.5 points could end up deciding the whole election. The forecasted margin between Harris and Trump is currently equal to or smaller than 1.5 points in Arizona, Georgia, Nevada, North Carolina and Pennsylvania — a bloc that is together worth 68 electoral votes.
In the 2020 election, now-President Joe Biden ended up needing to win the national popular vote by at least 3.8 points (assuming a uniform national swing) in order to carry enough states to win 270 electoral votes. This year, thanks to her stronger standing in state polls than in national ones, our forecast thinks Harris needs to win the national popular vote by only about 2.1 points in order to be favored to win. However, that still means the Electoral College will probably help Trump, meaning that it's essential for Harris to maintain her current lead in the northern battleground states especially. That's why our forecast gives Harris a 61-in-100 chance of winning instead of 72-in-100 (her current chances of winning the national popular vote).
Ultimately, though, it is still early days: We have seven weeks left until Election Day, and many major firms have yet to release any post-debate data. We will have to keep watching the data to see how Harris's and Trump's chances change. As history has shown, campaign events can unearth new information, reinforce perceptions of candidates, or deprive underdogs of a last chance to shake up the campaign. That's truer than ever in tightly contested national elections, regardless of higher levels of polarization. 538's polling averages and election forecasting model will continue to update with the latest polls as they are released.