Donald Trump has significantly improved his position in our general election forecasts as a result of state and national polls that show declining numbers for Hillary Clinton. Trump now has a 36 percent chance of winning the election, according to our polls-only forecast, and a 37 percent chance according to polls-plus, which also considers economic conditions.
With the conventions upon us — the Republicans’ starts Monday in Cleveland — let’s step back and ask some big-picture questions about where the race stands.
Who’s ahead in the polls right now?
Despite a relatively poor run of polls, Clinton is very probably still ahead of Trump right now. That doesn’t mean she’d be assured of winning an election held today, let alone one in November — there’s a lot of uncertainty (see the next question for more about this). But polls-only has her ahead of Trump by 3.4 percentage points nationally, similar to the margin by which Barack Obama beat Mitt Romney in 2012.
If Clinton has a 3.4-percentage-point lead, as our model surmises, that means we’ll sometimes see national or swing state polls that show her ahead by margins in the high single digits, such as the set of swing state polls that Marist College released this morning. We’ll also see some polls showing Trump with narrow leads, like the polls Quinnipiac University released earlier this week. All of this is pretty normal.
It’s not quite correct to characterize the race as a tossup, however. A relatively emphatic majority of recent swing state and national polls still have Clinton ahead, but often by narrower margins than before.
What’s the degree of uncertainty?
High, or perhaps very high, for a variety of reasons. As I noted above, the race isn’t that close — Clinton is matching Obama’s 2012 margin.But it’s early, so the outcome is highly uncertain. That uncertainty cuts both ways, meaning that Clinton could easily lose to Trump, and she could also easily end up winning in a landslide.
One reason for the high level of uncertainty is that polls are showing both a large number of undecided voters and a large number of third-party voters. An average of recent national polls shows Clinton at 40 percent or 41 percent and Trump at 37 percent or 38 percent, meaning that both candidates are struggling to get more than 40 percent of the electorate to commit to them.
Another challenge is that we’ve had a lot of political news lately — in some ways more than the polls can keep up with, even though we’ve had a lot of polls this week. The recent trend against Clinton could be a short-term blip caused by FBI Director James Comey’s testimony about her, the shootings of police officers in Dallas, or some other factor. Several other events that could move the polls — such as Bernie Sanders’s endorsement of Clinton, the terrorist attack in Nice, France, on Thursday, and Trump’s announcement of Mike Pence as his running mate — haven’t had time to show up in the polls yet.
At times like these, there are dangers both in overreacting and in underreacting to polling fluctuations. Our models are doing the best they can, but there’s inherently more uncertainty when the news cycle is fluid, versus when it’s relatively stable. It may be mid-August before the race settles into more of a steady state.
What’s the short-term trend in the polls?
As I’ve said, it’s toward Trump. Or if we’re being more precise, it’s away from Clinton; national polls show Clinton declining faster than Trump is gaining, with the number of undecideds increasing in some polls. Libertarian Party candidate Gary Johnson has also seen a slight improvement in his numbers.
What’s the medium-term trend in the polls?
It’s easy to become obsessed with how the polls have changed over the past day or the past week, but sometimes you’ll gain perspective by taking a longer-term view. Clinton’s position in the polls is about the same as it was on June 8, the day after she won the California primary. It’s better than it was for parts of May, when polls showed a very close national race. But it’s worse than it was for much of March and April, when polls had her ahead by high single digits. These types of fluctuations are fairly normal, especially early in the race; we were spoiled in 2012 by unusually stable polls.
Which states shape up as most important?
We recently relaunched our tipping-point state calculation, which describes a state’s likelihood of providing the decisive votes in the Electoral College in a close election. (There’s more detail on this in our methodology primer.) The top states are — these names will sound familiar — Florida, Pennsylvania and Ohio.
Florida is a must-win for Trump; in our simulations this afternoon, Trump won the Electoral College only 6 percent of the time that he lost the Sunshine State. Pennsylvania isn’t quite as crucial for Clinton, but it’s close; in our simulations, she won the election only 22 percent of the time that she lost Pennsylvania.
Virginia and North Carolina are fourth and fifth on the tipping-point list. Clinton’s polls have held up relatively well in those states, even as the national race has tightened. Winning them could help her compensate for losing states in the Rust Belt or the Upper Midwest.
Does one candidate appear to have an overall edge in the Electoral College, relative to his or her position in the popular vote?
Although either outcome is improbable, Trump is more likely to win the Electoral College while losing the popular vote than the other way around, according to the model. That doesn’t mean Trump is winning in swing states right now — but he’s behind by slightly less than he is nationally. While polls-only has Trump trailing by 3.4 percentage points nationwide, for example, it has him down 3.0 points in Pennsylvania, 1.2 points in Florida and by 0.6 points in Ohio. The rest of the swing states are more of a mixed bag — Clinton’s polling well in Colorado and Wisconsin, for instance. But the Florida-Ohio-Pennsylvania triad has been problematic for her, and that makes it hard for her to have a firewall of states that she can feel comfortable about.
How do the “fundamentals” look?
By fundamentals, I mean factors apart from head-to-head polls that potentially have predictive power, such as economic conditions, incumbency and presidential approval ratings. We’ve long had a love-hate relationship with the “fundamentals” at FiveThirtyEight. In theory, they help us to understand why voters act as they do (a better economy encourages them to reward the incumbent party, for example). But in practice, models that have attempted to predict the election without looking at polls have a middling-to-poor track record. So we’ve decided to compromise by issuing both a polls-only model and a polls-plus model, which blends polls with an economic index. Right now, the economic index gets about 35 percent of the weight in polls-plus; it will decline to zero by Election Day.
The economic index implies that the race should be close. But stronger economic data lately — in particular, a good jobs report this month — shows an economy that is very slightly above average. That translates into a very narrow advantage, about 1 or 2 percentage points, for Clinton. Obama’s approval ratings — not a factor our models use, but useful for context — are also very slightly above average, which could help Clinton at the margin.
In that sense, the recent polling movement has brought the race more in line with the fundamentals, and a narrow Clinton win — or a narrow Trump win — wouldn’t be that surprising.
How do FiveThirtyEight’s forecasts compare against prediction markets?
This is another useful gut-check. Right now, prediction markets imply that Trump has a 25 percent to 30 percent chance of winning the election. So our models are a bit bullish on Trump relative to the betting consensus, although it isn’t a huge difference. That may be because the bettors are discounting the recent polling movement toward Trump because they think it reflects a temporary shift — a blip caused by Clinton’s email scandal having been in the news, for instance. Perhaps that’s reasonable. I’d also read Sean Trende at RealClearPolitics, however, who reminds us that most elites — including many conservative elites — would prefer not to see Trump elected, so the conventional wisdom may be somewhat in denial about his chances.
What would keep me up late at night if I were Clinton?
We’ve gone over most of the reasons for concern. I’d be worried about those trend lines. I’d be worried about my numbers in Florida, Ohio and Pennsylvania, especially after having spent a lot of money on ads. In a big-picture sense, I’d be worried that Americans come to view the race as one between two equally terrible choices, instead of Trump being uniquely unacceptable.
What would keep me up late at night if I were Trump?
I’d be worried that after a bunch of events that would seem to play to my strengths — Brexit, Orlando, Dallas and the resurfacing of Clinton’s email problems — the best I could do was to narrow Clinton’s lead. I’d be worried that some of those undecided and Johnson voters will gravitate toward Clinton now that Sanders has endorsed her (although I’d be hopeful that I have some room to grow among my own base). And I’d be worried about my ground game. Maybe FiveThirtyEight’s model says that I’m more likely to win the Electoral College than the popular vote — but a disorganized turnout operation could eventually yield problems in the swing states.