Skip to main content
Election Update: Is Gary Johnson Taking More Support From Clinton Or Trump?

We’ve hit a little bit of a lull in polling for the 2016 presidential election.

The recently released tracking and weekly national polls, conducted by firms such as Ipsos, Morning Consult, SurveyMonkey and YouGov, continue to show Democrat Hillary Clinton ahead of Republican Donald Trump (with the exception of the Republican-leaning Rasmussen Reports). A new Pew Research Center poll also has Clinton ahead of Trump 45 percent to 36 percent, with Libertarian Gary Johnson at 11 percent.

We’ve also had a few new state polls, which basically confirm what we thought we knew about the race:

  • A Field Poll from California gave Clinton an unsurprising 50 percent to 26 percent lead over Trump, with Johnson at 10 percent. The FiveThirtyEight polls-only forecast gives her a 98 percent chance of winning the state.
  • An Icitizen survey from Oregon found Clinton up 46 percent to 32 percent over Trump. The FiveThirtyEight polls-only forecast has Clinton with an 86 percent chance of winning Oregon.
  • A Fairbank, Maslin, Maullin, Metz & Associates poll from Vermont has Clinton ahead of Trump 39 percent to 24 percent, with Johnson at 10 percent. That is a smaller-than-expected margin — President Obama won the Vermont vote by 36 percentage points in 2012. Perhaps there is resistance to Clinton among Bernie Sanders supporters in his home state. Either way, Clinton is still ahead by a wide margin, and our polls-only forecast gives her a 94 percent chance of winning Vermont.

Overall, Clinton has a 78 percent chance of winning the presidency, according to the latest FiveThirtyEight polls-only forecast. Polls-plus gives Clinton a 72 percent chance. Those are both a tick downward since we launched our forecast models last week.

Because so little new polling has come in, let me discuss something more fundamental: The effect of pollsters including third-party candidates in their surveys.

Pollsters have different views on when and whether to include third-party candidates. On the one hand, pollsters should want to be as inclusive as possible. On the other, third-party candidates are often unknown, and voters may be selecting them in polls right now as a protest — so that support may not materialize on Election Day. Of the 23 pollsters who have done nationwide surveys since June 1, 87 percent have polled the presidential race with Johnson at least once, and 57 percent have polled the race with Green Party candidate Jill Stein at least once.

Right now, pollsters that include Johnson and, less frequently, Stein are showing Clinton with a slightly smaller lead than pollsters that test only Trump and Clinton. You can see this by looking at the national polls taken since June 1. According to the FiveThirtyEight polling database, 18 pollsters have taken a national poll that asked about the presidential race with only Clinton and Trump offered as an option and in a separate question asked about the race with at least Johnson included. Here is the average margin by which Clinton is ahead of Trump in those polls, with and without third-party options.

CBS News/New York Times +6.0 +7.0 +1.0
Ipsos +10.5 +11.1 +0.5
YouGov +4.5 +5.0 +0.5
Pew Research Center +9.0 +9.0 0.0
Public Policy Polling +4.0 +4.0 0.0
Quinnipiac University +2.0 +2.0 0.0
Saint Leo University +7.9 +7.6 -0.3
Fox News +4.5 +4.0 -0.5
Morning Consult +3.3 +2.5 -0.8
CNN/Opinion Research Corp. +5.0 +4.0 -1.0
Monmouth University +7.5 +6.5 -1.0
Suffolk University +5.2 +4.0 -1.2
SurveyMonkey +4.6 +3.2 -1.4
ABC News/Washington Post +12.0 +10.0 -2.0
Zogby Interactive/JZ Analytics +8.0 +6.0 -2.0
IBD/TIPP +4.0 +1.0 -3.0
Rasmussen/Pulse Opinion Research +4.0 +1.0 -3.0
NBC News/Wall Street Journal +5.0 +1.0 -4.0
Average +5.9 +4.9 -1.0
How including a third-party candidate affects Clinton’s lead in national polls

Results averaged when a pollster sampled more than one population (e.g., registered voters and likely voters).

The majority of pollsters (12) have Clinton’s margin over Trump shrinking when at least one third-party candidate is included. The difference in margins, however, varies among pollsters, and a few, such as Ipsos, have Clinton’s lead rising by the tiniest of bits when at least Johnson is included. Overall, including third-party candidates takes about 1 percentage point away from Clinton’s margin, on average.

We can argue about the significance of a single percentage point. It’s not a very big deal when Clinton is leading by 5.5 percentage points in the FiveThirtyEight national polling average and is projected to win the national vote by 6.3 percentage points in the FiveThirtyEight polls-only model. (Note that our model prefers the versions of polls that include Johnson. Otherwise, Clinton’s advantage would be slightly larger.1) The discrepancy could, however, become an issue if the race becomes tighter. Although Clinton has been hurt by the inclusion of third-party candidates over the past month, it hasn’t been consistent:2


The fact that there is a difference puts some pressure on pollsters to decide whether a full ballot or partial ballot test is giving them a more accurate representation of the race. As I noted above, third-party candidates often fade later in the campaign. If that happens this year, the inclusion of third-party candidates in the polls could be artificially inflating Trump’s chances of winning at this point.

But it’s also possible (and, I would argue, probable) that because Clinton and Trump are two of the most disliked presidential candidates of all time, third-party candidates are going to do better than usual. Johnson looks especially likely to peel votes from Clinton and Trump because he will probably achieve ballot access in all 50 states, which is unusual for a non-major-party candidate.

That’s part of the reason why FiveThirtyEight is including Johnson’s chances in these projections. We aren’t explicitly projecting Stein’s vote, in part because polls include her less often than they include Johnson, and in part because she probably won’t be on the ballot in some states. You’ll notice, however, that the projected vote shares for Clinton, Trump and Johnson usually don’t add up to 100 percent. (In Missouri, for instance, they sum to 98.7 percent.) That’s because the model reserves a small share for “other” candidates, including Stein, in states where we expect at least one of them to appear on the ballot.

As I’ve said before, it’s far from certain that Clinton will continue to do worse when voters can select third-party candidates. For instance, Clinton and Sanders are in talks about the possibility of Sanders endorsing the presumptive nominee. An average of YouGov surveys over the past two weeks indicates that Sanders primary voters who say they will vote for a third-party candidate are, by about a 3-to-1 ratio, more likely to choose Clinton than Trump when the third-party candidates aren’t an option. A full-throated Sanders endorsement could push many of these voters into the Clinton column, even with third-party candidates on the ballot. Their shift would make Clinton’s margin over Trump in the national polls about the same with third-party candidates as without.

Of course, events in the weeks to come could shift the results in unexpected ways. As my colleague Nate Silver wrote on Tuesday, we have the vice-presidential selections and the party conventions coming up. We don’t yet know how those will affect the race.

Dhrumil Mehta contributed research.


  1. The model also adjusts polls that don’t include Johnson, projecting what his vote “should” be in the state and taking votes away from Clinton, Trump and undecided and giving them to him. Where this occurs, the model takes an equal number of votes away from Clinton and Trump.

    Shouldn’t the model be taking more votes away from Clinton, based on what I’ve written here? Maybe, but doing this would introduce a number of complications. For one thing, as I’ve mentioned, the extent to which third-party candidates are drawing votes from Clinton and Trump has shifted over time, and it will probably shift again. For another, it might vary from state to state; it’s plausible that third-party candidates hurt Clinton more in blue states but hurt Trump more in red states, for example. Finally, the polls-plus model (but not the polls-only model) assumes that the third-party vote will fade down the stretch run, and it gives some of the vote back to Clinton and Trump. If we program the model to take more votes away from Clinton when polls don’t include Johnson, should we also have it give more third-party votes back to Clinton when polls do include him, on the assumption that the third-party vote will decline?

    We prefer to sidestep all of these complications by assuming the third-party vote is drawn evenly from each candidate instead. From our vantage point, the most rigorous approach is for pollsters to include Johnson and (in states where she’ll appear on the ballot) Stein explicitly — so we don’t have to make any assumptions at all.

  2. This chart does not include surveys from Ipsos, which polls so frequently that including them swamps out the other data.

Harry Enten is a senior political writer and analyst for FiveThirtyEight.