Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC


February 15th, 2020, 12:47am by Sam Wang

A reader asked the following question in comments:

Let’s get right to it (aka the $64K question): Who has the BEST chance to beat Trump? Your personal opinion and/or statistically.

The answer is that as far as I can tell, basically it doesn’t matter.

More than anything else, a presidential election is a referendum on the incumbent. The one predictive variable is presidential approval. National popular vote share has ended up within 0-4 points of national job approval, as noted by Kyle Kondik:
The breakeven point for Trump, in terms of the Electoral College, is about 48% (assuming normal levels of support for minor-party candidacies; 2016 had abnormally high levels of minor-party support for Stein and Johnson and had a lower threshold, 46%).

Therefore the number to watch is Trump’s approval. If it gets to 46% then he has even odds of re-election. His range in 2018-2019, as per FiveThirtyEight, has been 39-44%.

If one assigned the entire 4-point difference to the quality of the opponent, then that’s a ceiling on how much that person’s identity matters. From the above table, then, that logic would suggest that Ronald Reagan was a weak opponent in 1980, as was Bill Clinton in 1992. I don’t believe that.

Also, in case you haven’t noticed, Donald Trump has taken partisan polarization to extremes and represents the apex of trends that began in the mid-1990s. His Administration is radically shaping what the executive branch does, sometimes in authoritarian directions. It’s going to be hard to focus this year’s campaign on anything other than his performance as an incumbent.

Bottom line, it might make a small difference who gets the nomination, but really, Donald Trump’s approval/disapproval rating is moving in a very narrow range. Consequently that number is likely to be the best information we have about November. You should not be asking about “electability,” but re-electability.

One final and essential point: Trump’s approval is remarkably stable. The standard deviation is about 0.7 percentage point, the lowest since Eisenhower. To recover from where he is now, he has to bust out of that pattern. That will require something to expand his appeal beyond his hardcore base. For example, with the impeachment trial over, his rating is at the top of its range so far. If that is sustained, it could get him above the necessary threshold.

Tags: 2020 Election · President

15 Comments so far ↓

  • Steven

    Wait, so incumbents have achieved popular vote shares that are 0-4 percentage points greater than their approval ratings (even when they lost), except for Clinton who did 6 points worse. Right?

    So… with Trump’s approval just shy of 44% we could expect him to get 44-48% of the popular vote in 2020? Since he won w 46% in 2016 (albeit, as you said, w a substantial number of votes siphoning to third parties), then would it be fair to say we are in for a squeaker? Assuming Russian interference has a similarly strong effect in this cycle as last…

    Sounds like a collective Democratic strategy of jointly chipping at Trump’s approval would be helpful.

  • Chaz

    I think the 1980 results can be explained by the likelihood that a sitting president in a virtual two-candidate race is going to receive at least 40% of the vote no matter where his approval rating is. 1992 and 1996 both a relatively competitive third party candidate (Perot), which could explain, in part, the lower share of the national vote in those years. GHB would probably have cracked 40% if Perot didn’t run in ’92.

  • Marco Ciocca

    Guess what? It is going to be Wisconsin, Michigan and Pennsylvania. If Trump stays at 43-44, Democrats have a chance. If the Orange one goes near 45, we’re…going to be upset.

    • Mark F.

      Polling has been very good for Trump in Wisconsin. If he holds all other 2016 states, loses PA and MI but gets one electoral vote from Maine like last time, he wins with 270 Electoral votes. Sanders will need all three states, unless he gets one 2016 Trump state like Arizona.

  • Zack Kilpatrick

    Sam — Couldn’t these data hide something in that an incumbent’s approval rating at the time of the election could be _correlated_ with the quality/electability/approval of the challenger? Therefore, incumbent approval could still be a good proxy for election outcome, but that approval rating is shaped by which challenger won the primary.

    • Sam Wang

      This year, I guess if Trump’s approval changes much after Memorial Day, it might reflect a variety of factors, including his opponent’s qualities.

      But…a big problem with your suggestion is that it’s plausible, but untestable, i.e. there’s no way to tell if it’s true or false. That I can think of anyway.

    • Zack Kilpatrick

      Sam — Perhaps one way to test would be to look at historic incumbent approval time series and see how they respond to the timing and strength of the final challenger nomination decision, as you say we could look this year at Trump approval post Memorial Day. On the whole though, I tend to agree with your analysis above, and think it’s a point that’s really being overlooked by most.

    • Pechmerle

      There is also, I think, a way in which “this year is different from all other years.” I.e., the states that vote on Super Tuesday (first Tues. in March) are significantly different this year from any prior year I remember. California now has its primary then, instead of the former June date. This can have a major impact on who the Dem nominee is. Though, interestingly, Bernie has had a recent surge in the polling in Calif.

      I also wonder if there is a quantitatively discernible difference in years in which the out-party candidate is clear by, say, June 1, and years in which the out-party choice remains contested to a much later point (e.g., the now-rare brokered convention).

  • nceladean

    I suspect that using Trump’s approval rating versus that of other presidents to predict the outcome in the upcoming election may not provide an accurate result because of the unusually high level of enthusiasm of Trump voters. Consider:

    1. First, according to FiveThirtyEight, Trump’s aggregate approval rating just among registered and likely voters is currently at 44.7% (the 43.9% figure is the aggregate of all types of polls) [1]

    2. Second, among those registered voters who approve of Trump, 79% are “extremely” and “very” enthusiastic about Trump [2]. To put things in perspective, 30% of Republicans think Trump is the greatest president of all time. Not just a good president, but the greatest president. [3] Empirical studies have shown that enthusiastic voters are more likely to turn out, and when one party has a clear advantage in enthusiasm, it can impact election outcomes. [4]

    3. Voter turnout in US presidential elections from 1972 to 2016 ranged from 51.7 to 61% of eligible voters. [5]

    If the Democrats can’t match the enthusiasm of Trump voters (i.e. the Democratic nominee doesn’t generate much enthusiasm and/or anti-Trump sentiment is not enough to carry the day), then all else equal and assuming the above approval and enthusiasm numbers maintain and turnout is within the “typical” range for a presidential election, then Trump already has the votes to be re-elected, despite the low overall approval raising.






    • Pieter Van Tol

      That post has to set the world’s record for number of assumptions in favor of Trump.

  • JW

    Hello, is the approval data used in the study that you mention “all polls” or “polls of likely or registered voters”? Right now on 538, they show Trump’s approval for both types (landing page defaults to “all polls” but you can change it) and there is a significant difference between the two. As of this writing, all polls is showing as 44.3 approval vs 45.8 for likely/registered voters.

  • Amitabh Lath

    The pollsters are making a lot of assumptions about demographics. Look at the difference between Emerson and ABC/Washington Post. A straight average or even weighted average (like the 538 poll Sam shows) assumes that in aggregate the pollsters are covering the entire phase space of voters, even if individual pollsters are getting it (very) wrong. If that doesn’t bother you, it should.

    Rachel Bitecofer has raised some interesting points about turnout. I don’t know if I buy her line either, but it’s true that this time around we do not know who is going to show up. This echoes the 2016 problem, and no, the fact that pollsters got 2018 right does not mollify me.

  • Scott

    As the nation has become more polarized over the past 2 decades, and as polling methodologies have evolved as well, does it make sense to use NET Approval (Approval – Disapproval), for this correlation?
    If so, what does that do to the predictive quality of the available data?

  • Lee

    I am curious as to what effect the 15% threshold in states is having on the delegate counts. In many (all?) states, a candidate who doesn’t get 15% gets zero delegates. Suppose instead that each candidate gets delegates proportional to the vote count in primary states, and proportional to the first round tallies in caucus states. Would that change the delegate counts significantly?

  • Sacto Joe

    I’m getting the sense that, just in time for Super Tuesday, the Democratic nomination process is starting to gel. Bloomberg may hang in for a while, and may pull more moderate votes from Biden, but I feel strongly that the huge choice availability Democrats have had to winnow down is about to narrow to a field of two: Biden and Sanders.

    I’d also note that, in terms of the raw vote count, Biden has now passed Sanders. That may be less important than the final raw vote count of “moderate versus liberal”.

    I do think that the Sanders’ run makes more possible a Biden/Warren ticket or a Sanders/Buttigieg ticket I had thought previously….

Leave a Comment