Commentary

Prediction Science Does Not Support Biden Dropping Out

Political pundits can and often do express opinions without repercussions. They’re rarely held to account for what they say and there is little causal political research to prove them wrong.

It’s fair to say that Biden didn’t have his best performance during 2024's first presidential debate, but that doesn’t mean he should quit. Predictive analytics don’t support that outcome. 

We’re currently tracking several presidential election prediction methods – polling averages, a forecast based on the polls, prediction market results and a data-science methodology called the 13 Keys. Let’s look at the last first.

The good news for Biden is that according to the 13 Keys presidential election prediction system, debates have a low correlation with presidential election results.

advertisement

advertisement

Research done by a history professor at American University instead point to the incumbent white-house-party-candidate’s performance while in office as having the highest correlation to presidential election results.

The 13 Keys still predict a Biden popular vote victory. As far as I can tell, none of the keys have turned since the author, Allan Lichtman, made his 2024 election prediction this past April.

On the polling side, we’re following FiveThirtyEight's presidential election polling average. Biden did move ahead in the polls after Trump’s felony conviction; however, they flipped back to favor Trump after the debate.

You might think that the debate would have made a big difference in the polls, but that’s not the case. As of today, Sunday June 30th, the average only shows Trump up by 1.3 percentage points. When accounting for statistical error, the candidates are basically tied.

Polls provide answers to the hypothetical question – if the election were held today, who would win? FiveThirtyEight also produces presidential election forecast simulations that incorporate more data including poll results, economic and demographic information.

On June 25th, prior to the debate, FiveThirtyEight’s forecast simulations had Biden winning 50 out of 100 simulations and Trump winning 49 out of 100. We’ll continue to track FiveThirtyEight for changes.

As it relates to prediction markets, the University of Iowa’s electronic market (IEM) still predicts the democratic party candidate to win the presidency. However, the winner-take-all market moved closer to a tie after the debate.

IEM’s latest prices are 0.7 for a democratic win versus 0.355 for a republican victory. Last month, IEM prices were 0.748 and 0.270, respectively.

The PredictIt prediction market had the largest amount of movement out of all the predictive methods we’re tracking. It appears to be the most sensitive and least decisive.

In my experience, prediction markets bounce around a lot in the beginning when they’re first made public before they settle on a prediction one way or the other. Once settled, they tend not to change that much. IEM is case in point -- it has been predicting a democratic victory since the market was launched.

For its candidate-specific market, PredictIt is currently pricing a Trump win at 56 cents versus a Biden win at 30 cents. The numbers were a lot closer last month and Biden was recently ahead after Trump’s felony convictions before the debate occurred.

The GOP also moved ahead in PredictIt’s generic-party presidential election market. It’s pricing a GOP win at 57 cents and a DEM win at 47 cents.

PredictIt numbers for potential Biden replacements aren’t doing nearly as well as Biden even after the debate – Kamala Harris and Gavin Newsom are both trading at 10 cents and Pete Buttigieg is trading at 2 cents.

To put it all together:

  • The 13 Keys still predict Biden to win

  • The polls favor Trump, but are statistically tied

  • 538’s forecast hasn’t been updated since the debate so it’s not relevant

  • IEM continues to predict a democratic win

  • PredictIt favors Trump and GOP victories, but doesn’t support Biden replacement

The 13 Keys and IEM have the best predictive track records and they both still favor Biden and the democrats. 

The debates are only one of many events that will occur during the campaign. Just as Trump’s felony conviction didn’t end his chances, Biden’s poor debate performance is unlikely to end his.

The political experts calling for Biden to step down were premature as well as uninformed. According to the data, this race is far from over.

4 comments about "Prediction Science Does Not Support Biden Dropping Out".
Check to receive email when comments are posted.
  1. Tony Jarvis from Olympic Media Consultancy, July 8, 2024 at 3:37 p.m.

    Ed: A truly insightful piece.  In view of the ever present range of bounce in poll results and the fundmental flaws in using and reporting "standard error", I believe that if we use double the stated and misguided "standard error' to establish the reange of potential poll repsults we are on much safer ground.  If this approach is indeed more meaningful, we will likely have a statistical tie on any of the reputable polls from now until November.   

  2. Ronald Kurtz from American Affluence Research Center, July 8, 2024 at 3:38 p.m.

    The "over/under" bet for whether Biden will gain or lose voter confidence in his mental ability surely has greater risk for the lose result. I regret that, but in the absence of having and releasing the results of a proper medical exam, the odds are that his standing will only get worse. More press conferences and speeches will not get the job done. Time is of the essence. I hope the President will have a proper medical exam or withdraw from the nomination. If the latter, he should try to build a coalition around Harris and try to avoid party in fighting. She would have a good chance of winning if she redirected the Democratic party's agenda to appeal to the group positioned in the center right to the center left of the political spectrum. Polls show this is where the majority of voters reside and that they want evolutionary (not revolutionary) change and bi-partisan compromise to resolve the major issues facing us.

  3. Ed Papazian from Media Dynamics Inc, July 8, 2024 at 7:40 p.m.

    Good piece, Ed.

    Unfortunately, I think that this is a case where no matter what the varagies of the polls are that enough people who saw the shocking deterioration that has befallen Biden---especially likely GOP switch voters---from Trump to Biden---as well as moderate Democrat party faithful---will be unable to cast a POTUS vote for Biden. Many will still favor him in the polls and will vote for Democrat congressional and senatorial candidates --but pass on the POTUS line.

    As for Harris, were she to get the nomination replacing Biden and be properly coached so she gets her facts straight, stops giggling and acts like a serious candidate, I suspect that there is enough anti-Trump sentiment to, perhaps, see her through  to a narrow win. In this regard the selection of a serious and forceful middle -of-the-road VP, with governing experience and no baggage would be a major boost. The two would act as a team---not the VP quietly cheerleading for Harris in the background as is the usual practice---and that could reassure many who might question Harris's capabilities based on her past performance. 

  4. Ed DeNicola from MediaLytics, July 9, 2024 at 10:06 a.m.

    You make a very good point, Tony. Most lay people -- journalists, pundits, consultants and politicians -- treat polling numbers as if they're not subject to error. It's why I like to look at different data from a variety of research methodologies. I like your idea to use double the stated standard error.

Next story loading loading..