IE 11 is not supported. For an optimal experience visit our site on another browser.
Image: Tudor Dixon, Republican gubernatorial candidate for Michigan.
Tudor Dixon, Republican gubernatorial candidate for Michigan.Tristan Wheelock / Bloomberg via Getty Images file

Primary polls can often vary widely, so proceed with caution

With small sample sizes, more undecideds and concerns about response rates, polls in primary races are tricky.


Last week, a poll on the Michigan Republican gubernatorial primary field showed that conservative commentator Tudor Dixon had vaulted into an 11-point lead in the race. The advantage, which was outside the poll’s margin of error, suggested she was the clear front-runner.

This week, another poll of the race showed Dixon up by four points, well inside the poll’s margin of error in a race the pollster called “a complete toss-up.”

The difference in those polls, which is not easily attributable to any changes in the race, is another example of why polling for primary races, especially for Republicans in 2022, can be exceedingly difficult and why people should be very cautious in reading them. 

To be clear, the margins of error in the two polls, +/- 3.75 points in the July 11 Mitchell Research poll and +/- 4.4 points in the July 18 Glengariff Group poll, mean that the surveys could line up. For instance, once you figure in the MOE, Dixon’s support could be somewhere around 22% or 23% in both polls.

But there are other reasons to be skeptical of the numbers.

First off, polling at the state level is often tricky because drawing a representative sample can be difficult. The samples are usually smaller (these were 683 and 500 “likely voters” respectively) and getting the correct geographic mix can be challenging. Communities that sit next to each other can be quite different.

Second, both of these Michigan polls have a massive number of undecided respondents. In fact, in both polls “undecided” is actually leading. That means there is a lot of potential for swings before the actual vote on August 2.

And third, the challenges around polling of late merit a special warning on the Republican side of the equation. One of the biggest issues in the 2016 and 2020 presidential races was concern over whether Republicans — particularly those who supported Donald Trump — were even participating in polls. When the sample a pollster is trying to gather is all Republican, that refusal to answer pollster's questions can sharply heighten the degree of difficultly.

We have already seen how such complications have yielded primary polls that have missed the mark in this cycle.

In Pennsylvania’s Republican Senate Primary, for instance, some late polls showed eventual winner Mehmet Oz with a narrow lead over hard-charging Trump supporter Kathy Barnette. But in the results Oz narrowly held off former hedge fund manager Dave McCormick by just .1 percentage points.

And just this past week in Maryland, a conservative Trump backer, Delegate Dan Cox, won that state’s Republican Gubernatorial Primary comfortably, by 16 points. A poll in late June showed the race neck-and-neck.

Those results should serve as a reminder that all polling in general is not easy, but polling for Republican primaries is an especially fraught task.