Hillary Clinton's lead over Donald Trump has dipped to 2 points according to a new survey released by Quinnipiac University — a difference that's sharply at odds with the 12-point lead Clinton enjoyed in the most recent Washington Post/ABC News poll.

Does this suggest that Trump's seen a sudden, dramatic surge in support since the Post/ABC poll was completed June 23? Well, no.

With the amount of attention paid to the 2016 presidential race, it's only natural that we obsess over each new national poll, laying out its entrails in the hopes of divining the November outcome. For those who've placed a bet on one of the horses in the contest, there's as much reason to trumpet or lament a good or bad poll as there is to celebrate a filly's sharp move to the rail or quick stumble. But polls are photos of a rapidly moving horse race in progress, meaning that you get more information from photos taken at multiple angles and multiple points in the proceedings. Which means — setting aside that exhausted metaphor — that it's more useful to consider the average than any individual outcome.

RealClearPolitics tracks an average of polls over time. Since the beginning of April (a randomly selected point), Clinton's almost always been in the lead, seeing that edge decline after Trump secured enough delegates for the nomination but widen again after she did the same.


We've marked individual polls on there, so you can see the work the average is doing. Individual surveys have results all over the place; the average gives a much better sense of how all of the polls say the race has changed. It's a brief bit of video cobbled together from all of those photos, if you will.

So why is the Quinnipiac poll so much closer than the Post/ABC one? One reason may lie in the way the question was asked. The Quinnipiac poll prompts respondents to answer this question:

If the election for President were being held today, and the candidates were Hillary Clinton the Democrat and Donald Trump the Republican, for whom would you vote?

Post/ABC asks this:

If the presidential election were being held today and the candidates were (Hillary Clinton, the Democrat) and (Donald Trump, the Republican), for whom would you vote? Would you lean toward (Clinton) or toward (Trump)?

That last bit in italics is likely one of the reasons for the difference. Eighteen percent of respondents in the Quinnipiac poll offered a non-Clinton/Trump response. In ours, only 10 percent did. (Our poll also included more Democrats, which is discussed here.)

Notice that the Quinnipiac results from the end of May were above the RealClearPolitics average while the new one is below. The change between the two isn't really that large. Most of the difference lies in Clinton's support dropping more than Trump's over the month. (Among those 65 and over, Clinton went from trailing by 1 point to trailing by 16.) If we look at the average for each candidate instead of the margin of support, you can see that dip in Clinton support thanks to the Quinnipiac poll clearly — as well as the big dip Trump saw earlier in the month.


The picture that the Quinnipiac poll paints is of two candidates that are reasonably popular with their own parties, but that's it. Asked if they would definitely, probably or never support Trump or Clinton, independents were more likely to say "never" to both.


(In a reassuring sign for Democrats worried about a Bernie Sanders schism, 58 percent of Democrats said they'd definitely vote for Clinton, slightly more than the percentage of Republicans who will definitely vote for Trump.)

People view the campaign at large as a negative. Sixty-one percent of respondents to the Quinnipiac poll also said that the race had increased the level of hatred and prejudice in the United States, with two-thirds of that group blaming Trump for the increase. (That includes a quarter of Republicans.)

So it's probably not a big surprise that, not pushed to pick between the two major party candidates, a lot of people simply didn't. We should, of course, remember that national polls are less informative over the long run than state ones (given how we select a president). But the story we should take away from the new Quinnipiac survey is, first, that neither Clinton nor Trump is terribly well-liked, earning 53 and 57 percent unfavorable ratings respectively. And, second, predicting the winner of a horse race from a photograph taken in the back stretch is a good way to lose a bet.