Untangling the DeSantis vs. Trump polling

In the last week, it looks like Donald Trump made a big recovery in the Republican primary polling. Emerson College showed Trump leading Florida Gov. Ron DeSantis 55 percent to 25 percent. But that’s nothing new for Emerson, which polled Trump at 55 percent in January, November and even in June. And, while consistency is a good thing in many facets of life, getting the same number time after time is usually a sign of bad polling.

But Emerson’s numbers are part of a larger problem when analyzing the polling for the GOP nomination: The polling is all over the map. While practically all polls show Trump leading a multi-way field, they disagree on the margin — and they diverge when Trump and DeSantis are matched up one-on-one. In addition, the state-level polls all have much better numbers for DeSantis than would be indicated by the national polls.

Trumpy polls

Of all the national polling, Emerson is by far the Trumpiest and the easiest to debunk and/or dismiss. First off, polling is not exact. It is an attempt to get an accurate picture by surveying a (hopefully) representative sample of the larger population. Good polling wobbles around the true number (that’s margin of error). A poll that delivers the exact same result time after time is a signal you have sampling problems. Emerson’s seemingly automatic 55 percent for Trump is not a testament to their genius; rather, it’s likely a bad indicator.

In addition, all other polls showed a significant drop in support for Trump after the Republican mid-term debacle — except Emerson. Lastly, Emerson has Trump beating President Biden by 4 points and DeSantis losing to Biden by 4 points — a swing against DeSantis of 8 points. However, no other poll shows the same dynamic. In fact, Morning Consult, Quinnipiac, Ipsos, Echelon and Yahoo all show DeSantis performing better against Biden than Trump. Only Harris has Trump improving on DeSantis — and by just 3 points.

Emerson may be the most egregiously Trumpy of the polls, but Morning Consult and Harris are also more favorable to Trump. Morning Consult runs a weekly online tracking poll that now includes a Trump v. DeSantis question with a sample of over 2,400. Throughout February, Trump tracked between 51 percent and 47 percent — which is how tracking polls should behave. But it’s higher than other polls, such as Fox News and Ipsos, which both had Trump at 43 percent.

The Harris poll has been running a few points in Trump’s favor. But the more significant problem with Harris is its forced ballot question. A forced ballot is when undecided voters are pressed to make a choice. That question is useful near election day when a decision has to be made, but it is useless a year out and just adds support based on name ID. Sure enough, the Harris forced ballot question pushes Trump to 56 percent — higher than any other poll, except for Trump’s own pollster.

Big disparities in other polling

In February multi-candidate surveys, Trump did best in the Fox News poll, leading DeSantis 43 percent to 28 percent — and did worst in the Monmouth and OnMessage polls, where the two were tied. Not all surveys tested head-to-head, but of those that did, Trump had the strongest showing in the Yahoo News survey, leading 47 percent to 39 percent and worst in OnMessage, trailing 53 percent to 38 percent.

Trump partisans celebrated the Yahoo poll, but it’s worth noting the small sample (n=444) and that moving from a multi-way field to heads-up, DeSantis gained 8 points while Trump only gained 4 — and Trump was still under 50 percent.      

The state of the states

What should concern Trump and the national pollsters are the state polls where several polls with good sample sizes show results well under the national numbers.

Since January, no state poll has Trump at 50 percent, either in a multi-way field or heads-up. Trump goes from winning to losing heads-up in California, Kansas, Maryland, Michigan and New York. In Virginia he leads in a multi-way poll just 39 percent to 28 percent and by 42 percent to 26 percent in Arizona.

California is a big problem. The 1,755-sample LA Times poll has DeSantis crushing Trump 50 percent to 33 percent heads-up. While California is a lost cause in the general election, the state sends 172 delegates to the National Convention, more than any other state. Winner-take-all by Congressional District — if this margin holds. Trump might not win a single California delegate.

With the state polls so much lower than the national polling, particularly the oddball Emerson poll, it presents the question: Where are all the Trump votes? Is Trump killing it in all the states not polled? Or are all the state polls wrong? That seems unlikely.

The verdict

With such variation, it is difficult to zero in on where the race really stands.

The answer is likely that the contest is a chaotic mess, with a dedicated “always Trump” faction, a “please, no more Trump” faction, and a big muddled middle of Republican voters not really engaged and ready for an alternative. There is not a single poll that shows a majority of GOP voters moving to Trump once the stragglers are dropped from the ballot test.

Some, if not most, of the disparity may also come down to sampling problems for the frequent polls. Getting a decent representative sample is very difficult. At an incidence of 1 percent for phone calls, a national poll would have to make 100,000 attempts to get a reasonable sample of 1,000 voters. And the response rate is actually below 1 percent these days.

What seems likely is that the frequent polls have developed an online pool, and are talking to the same people over and over rather than getting a truly “fresh” sample of voters. If so, a particularly “Trumpy” sample will keep giving you “Trumpy” results, whereas getting a new sample each poll would solve this problem. Any given poll has the potential for a non-representative sample, looking at several polls over time solves that problem. In the end, I tend to trust the infrequent polls — as their samples are more likely to be “fresh” — and polls with larger sample sizes. Samples that fall under 600 are dicey, and ones under 400, in my opinion, are worthless. In sum, the Morning Consult and Harris polls are probably a bit high on Trump, while Monmouth and OnMessage are probably a bit low. As for Emerson, they probably need to start over.

Keith Naughton, Ph.D., is co-founder of Silent Majority Strategies, a public and regulatory affairs consulting firm. Naughton is a former Pennsylvania political campaign consultant. Follow him on Twitter @KNaughton711.