The Washington PostDemocracy Dies in Darkness

Biden comes into the Democratic convention in an unusually strong position

To hear Trump tell it, it’s only because most people lie to pollsters

Former vice president Joe Biden gives a thumbs up as he walks on a picket line in Las Vegas in February. (Patrick Semansky/AP)
Placeholder while article actions load

As polls began to close on Election Day four years ago, polling averages offered a prediction on what to expect: Hillary Clinton led Donald Trump nationally by 3.9 points, according to FiveThirtyEight, or by 3.2 points, according to RealClearPolitics. Both were wrong, which is to say neither were precisely right. Clinton won the national popular vote by 2.1 points, though thanks to narrow losses in several Rust Belt states, she lost the electoral vote.

Despite what you may have heard, polling four years ago was unusually accurate at a national level. That’s worth keeping in mind when considering now-President Trump’s responses to what the polls are saying currently.

As it stands, Trump trails former vice president Joe Biden by 8.2 points (FiveThirtyEight) or 7.7 points (RealClearPolitics). The differences stem largely from how the averages are calculated, but each tells the same story: Trump trails.

Comparing that margin to the past 50-plus years of presidential polls, as also compiled by FiveThirtyEight, we see that Biden’s lead is stronger coming into the Democratic convention than any Democratic candidate over that period. In 1976, Jimmy Carter came into the convention with a 6.9-point lead, according to FiveThirtyEight’s re-created average of polls at the time. In 1996, Bill Clinton’s lead entering the Democratic convention was 5.1 points. Both of those candidates went on to win.

What should be more worrisome for Trump is that, even if the conventions were proceeding as normal, the bump candidates have gotten from conventions in the past 20 years have been modest. Since 2000, the biggest shift in the polling average between the start of the convention and a week later was seen by then-Vice President Al Gore, who improved by 11 points. That lead eventually faded.

(To the point above: Notice that each of the blue/Democrat lines starts above the dotted blue line marking where Biden enters this year’s convention.)

It’s been common that convention bumps fade fairly quickly. You can see the effect of conventions in 1976, for example, when the Democratic convention drove Jimmy Carter’s lead upward, with the Republican convention later helping the incumbent, Gerald Ford. That bumps haven’t lasted long is good news for Trump, since his convention comes the week after Biden’s. That could damp the effect of the Democrat’s bump — which is what happened to Trump four years ago.

That’s if there’s a bump at all. The virtual nature of this year’s conventions adds enormous uncertainty to the effects we might expect.

There are other caveats specific to the tool itself. This is FiveThirtyEight’s effort to establish what a polling average would have looked like in each year using its current methodologies and assumptions. It’s also dependent upon polling in each year and, by extension, the quality of those polls. On average, though, the reconstructed polling averages differ from the actual results by only 2.7 points. Since 2000, the difference has averaged only 2 points.

One thing we can be confident about, though, is Trump’s presentations of what the polls say can be broadly discounted. Trump has repeatedly disparaged the 2016 polling, conflating national polling (which was mostly accurate) with inaccuracies in state polling that led to misleading forecasts.

In an interview on Fox News on Monday, Trump went further, alleging national polling was meant to suppress enthusiasm for his candidacy. There’s no way to describe this claim besides “ludicrous.” Polls conducted by responsible media outlets are accurate, statistically valid measures of public sentiment. Besides, if one’s goal were to suppress support for Trump, you’d do that when people are actually voting, and there’s a point to do it.

Over the weekend, Trump introduced a new argument to support his claim that his supporters are loath to be honest with pollsters. (This is somewhat at odds with his “suppression polls” claim, but consistency is not one of Trump’s hallmarks.) He said during a news briefing and again on Fox that 62 percent of people say they’re not honest with pollsters — apparently an effort to retrofit a July poll from the Cato Institute, which dealt broadly with people’s comfort levels in being honest in their political discourse to polling specifically.

In other words, Cato found more than half of Americans have withheld their opinions at some point out of concern they would be considered offensive. To Trump, this apparently means 62 percent of Americans tell pollsters they support Biden instead of him.

As always, Trump isn’t trying to accurately convey information about polling; he’s trying to blur that reality. In this case, the reality is that Biden has a healthy lead over Trump, one which is probably stronger coming into the convention than that seen by any Democrat since at least 1968.

Since 2000, Democrats have seen their position in the polls improve by 5 points on average in the week after their conventions begin. Again, luckily for Trump, the odds Biden will see that large a bump given the nature of the conventions and the muted effect of conventions in recent years seem low.

But who knows? Maybe that 62 percent of people who he says lie to pollsters will suddenly tell the truth. Weirder things have happened.