One talking point I’ve heard lately circulating on Twitter is this: Ron DeSantis has lost half his support since January - therefore, he is not very electable. This claim is at best wishful thinking and at worst deceptive marketing, because it refers to polls that ask Republican voters about Trump and Desantis. This says nothing about how the voters as a whole think about Trump versus Biden.
Polls that ask Republican voters who their favorite Republican candidate is don’t capture the question of how voters in general will decide between that Republican candidate and a Republican candidate. Ron DeSantis being the first choice of fewer Republican voters in a few horse race polls doesn’t mean he isn’t electable. Today, I’m going to look at the type of poll that does try to get at this question: Head to head polling.
Even for trying to forecast a primary election, open “horse race” polls using a plurality vote and featuring a large number of candidates don’t mean much - results can radically change when some of those candidates drop out. In a crowded field, approval polls are a more robust measure of support.
What head-to-head polls tell us about the Republican contenders
FiveThirtyEight conveniently publishes data from most polling published in the United States. The central problem of using head-to-head polls to measure electability is that different polls tend to have very different underlying levels of Republican and Democratic support. Only polls that test multiple head-to-head matchups can be effectively used for comparing candidates of the same party.
Aggregating the 243 head-to-head polls in FiveThirtyEight’s data set using this code, what we have is this picture plotting relative support for candidates. The candidate’s expected level of support compared to the median Republican candidate is plotted on the x axis. Their opponent’s expected level of support (again, compared to the median) is plotted on the y axis.
Candidates nearer the lower-left corner, such as Liz Cheney, don’t have firm levels of support or opposition - they’re either not very well known or don’t provoke strong opinions. Candidates near the upper-right corner, such as Ted Cruz, are well-known and voters have clear opinions on them.
Weaker candidates will be closer to the upper-left corner; this includes Nikki Haley and Mike Pompeo. Stronger candidates will be near the lower-right corner. Chris Sununu and Glenn Youngkin show up there. However, there are some caveats.
Refining a polling measure
First, not all polling evidence is equal. There are hundreds of polls testing Trump’s performance in head-to-head contests. Chris Sununu and Glenn Youngkin’s stellar polling performance is based on a small number of polls conducted in their home states.
Polling support for Chris Sununu and Glenn is positive but weak, in other words. Weighting by a factor roughly proportionate1 to the square root of the number of polls and discounting state-level polling gives us a more plausible “polling performance” score that will have a stronger impact on candidates with better polling numbers.
One critique I have to acknowledge was a flaw in my similar analysis of the 2020 primary field: In 2020, most Republican voters didn’t care which Democrat was in the race; they were ready to support Donald Trump regardless. Similarly,most Democratic voters don’t care which Republican will be nominated in 2024; they are ready to support Joe Biden regardless. This means that in polls, measured changes in Biden’s vote share are smaller2 than measured changes in Republican contenders’ vote shares.
After adjusting for these issues, the resulting adjusted polling score is most positive for Donald Trump - and most negative for Ron DeSantis, who has been included in more polls than any other non-Trump contender.
Notably, this gives Liz Cheney and Ted Cruz sharply different polling scores in a way that makes sense: Ted Cruz seems like a near-certain loser against Biden (poll respondents undecided between Trump and Biden broke against him by a factor of almost 2:1), while Liz Cheney’s level of support is much less certain.
Wrapping up
In the first part of this series, I talked about how the different governors who are considered likely contenders for the Republican nomination stack up in three ways:
How they performed in past elections.
What their approval ratings look like.
How valuable their home state advantage is.
After that, I looked at candidates who were formerly part of the Trump administration, particularly including Trump himself, and talked about how candidates from different backgrounds tend to perform differently. Former presidents like Trump have done badly at getting back into office.
After considering polling evidence available thus far - as limited as it may be - Donald Trump looks closer to the middle of the pack than at rock bottom.
There are definitely more electable alternatives, and Ron DeSantis still looks more electable than Trump. After all, the head-to-head polling evidence isn’t conclusive, as Ron DeSantis still attracts fewer votes against him than Trump does. What the evidence thus far suggests, though, is that the most electable potential Republican contenders are ones that haven’t filed for candidacy.
Square root of sample size and a time decay factor that mostly doesn’t matter in this window, but which I will probably adjust a little bit later in this series.
And arguably also more important, but I picked a weight of 1.5 to make the variation in Biden support merely as important as variation in Republican support.