Skip to main content
Menu
Most Primary Polls So Far Are Internal Polls. That’s A Problem.

Welcome to Pollapalooza, our weekly polling roundup.


Interested in who’s going to win the Republican primary for U.S. Senate in Ohio in a few months? Well, we’ve got good news and bad news for you. The good news is that there are already 22 polls of the race! But the bad news is that 21 of them are internal polls.

Internal polls are polls sponsored by either a candidate running in the race or an outside group that has endorsed one of the candidates. In other words, these are polls that have an agenda behind them — and that agenda is usually to make their preferred candidate look good. As a result, they are usually too good to be true for their sponsor. FiveThirtyEight editor-in-chief Nate Silver found in 2017 that internal and partisan polls conducted during the last three weeks of U.S. House general elections overstated their party’s candidate by an average of 4 or 5 percentage points.

This poses a serious problem for people (like us) just trying to get an honest read of a race — especially if internal polls are all (or most) of what we have to go off. And unfortunately, that’s not just the case in Ohio this year; it’s true in primaries nationwide.

We hunted down every poll we could find of 2022’s Senate, House and governor primaries (as of Feb. 10 at 3 p.m. Eastern), and a majority of them (at least 118 out of 229) are internal polls. And while we have polling data for a healthy 78 primaries so far (that’s good), that polling is 100 percent internal in 23 of those primaries (that’s bad).

So far, more than half of primary polls have been internal

Share of polls sponsored by a candidate or biased interest group from all polls in 2022 Senate, House and governor primaries, as of Feb. 10, 2022, at 3 p.m. Eastern

*Alaska, California and Washington have all-party primaries where all candidates, regardless of party, run on the same ballot.

Source: Polls

For an example of how internal polls can mislead, look no further than that Ohio Senate race. The Republican primary is a crowded affair with tons of undecided voters, creating an opportunity for several candidates to claim they’re doing well. For example, the campaign of former state Treasurer Josh Mandel recently released a WPA Intelligence poll that gave himself 28 percent and businessman Mike Gibbons 17 percent. But a Cygnal poll conducted for Gibbons around the same time put Gibbons at 16 percent and Mandel at only 13 percent.

Meanwhile, in early January, a Moore Information Group survey for former Ohio GOP Chair Jane Timken trumpeted that Timken was “statistically tied” with Mandel, 16 percent to 18 percent. And back in October, a Fabrizio Lee poll for a super PAC supporting author J.D. Vance similarly put Vance (16 percent) in a strong second place behind Mandel (19 percent).

Clearly, these polls are all telling their own stories. But that doesn’t mean they’re totally useless. For instance, the fact that Mandel leads in even a couple of his opponents’ polls suggests that he is a front-runner. (The only independent poll of the race, conducted in December by the Trafalgar Group, concurs: It showed Mandel at 21 percent, Vance at 15 percent, Gibbons at 12 percent and Timken at 10 percent.)

Likewise, internal polls’ trendlines can often be useful. For instance, in two Fabrizio Lee polls since October, Vance has slumped to 10 percent and 9 percent, suggesting he is losing support over time. (In fact, it’s likely that these polls were released not to convince the public that Vance is doing well, but to convince Vance that his campaign needs to step it up.) Same with Mandel, whose own campaign’s polling staked him to an even bigger lead back in September (37 percent to Vance’s 13 percent).

In Ohio, we don’t have a lot of independent polling to validate those internal polls. But when we do, it can be a useful reality check. Take this year’s Republican primary for Texas governor. According to a recent poll by Paradigm Partners on behalf of former Texas GOP Chair Allen West’s campaign, West leads incumbent Gov. Greg Abbott, 43 percent to 34 percent. But two independent polls taken around the same time (one from the University of Texas at Tyler/Dallas Morning News and one from YouGov/University of Houston) put Abbott way ahead with 58-59 percent, while West is way back at 6-11 percent. That’s such a big disagreement that we reached out to Paradigm Partners to ask more about their methodology, and it turns out there’s a reason their poll is such a big outlier: While most polls are weighted to ensure they are demographically representative of the electorate, their survey was completely unweighted. 

So you can see why it’s a problem that internal polls are such a large share of the primary polling data we have so far. And in fact, that problem may be understated. While we did our best to identify every internal poll so far this cycle, that “118” number is almost certainly an undercount. Just because we didn’t mark a poll as internal doesn’t mean it’s truly independent and unbiased. For instance, according to the New York Post, this In the Field Global poll of the Democratic primary for governor of New York was paid for by an unknown “moderate political group” — hardly an uninterested actor in a race that looks like it will be another proxy war between Democrats’ moderate and progressive wings.

One of the more unusual cases we identified is a quartet of Public Policy Polling surveys of Republican primaries sponsored by the Democratic Governors Association. These aren’t internal polls because the DGA certainly isn’t endorsing GOP candidates — but they’re clearly not independent polls either. 

If anything, their purpose is probably to sow mischief in GOP primaries, with the hope that it will help Democrats in the general election. For instance, back in October, the DGA released a poll of the Republican primary for governor of Massachusetts showing that conservative former state Rep. Geoff Diehl led moderate Gov. Charlie Baker, 50 percent to 29 percent. That poll may have contributed to doubt that Baker could win the primary, which in turn may have contributed to Baker’s December decision to retire. (Though to be clear, Baker almost certainly conducted his own polling and analysis to inform that decision.) Regardless, though, that decision rocketed Massachusetts to the top of the list of governorships most likely to flip in 2022.

And in November, the DGA also sponsored a survey of the GOP gubernatorial primary in Pennsylvania — the results of which promoted a narrative that there is no clear front-runner and that that could lead to the nomination of far-right state Sen. Doug Mastriano, who participated in the Jan. 6 insurrection at the U.S. Capitol. Mastriano took 18 percent in the survey, more than any other candidate. And in the absence of more recent polling of the race, we have no independent confirmation of how close that result is to reality.

Clearly, there are plenty of reasons to be wary of internal polls — and even some non-internal polls. But unfortunately, they’re most of the data that we currently have for 2022 primaries, so sometimes using them is unavoidable. That’s OK as long as you do so with your eyes open to the fact that they are trying to sell you an agenda. And once you factor in that agenda, sometimes internal polls can tell you something interesting after all — if not about the actual state of the race, then about what one of the actors in that race wants you to think.

Other polling bites

  • A majority of Americans (62 percent) backed a pathway to citizenship for undocumented immigrants and believed more broadly that immigrants strengthen American society (56 percent), according to a recently published survey from the Public Religion Research Institute, conducted last September. But even though a majority of Americans from most religious backgrounds that PRRI surveyed supported creating a pathway to citizenship — with Black Protestants most in favor, at 75 percent — there was one notable exception: White evangelical Protestants were the least likely (47 percent) to support a pathway to citizenship.
  • The share of Americans extremely satisfied with their romantic relationship dropped by 10 percentage points since a year ago, according to a Jan. 20-24 poll from Monmouth University. But divorce lawyers shouldn’t get too excited: Majorities were still extremely satisfied with their relationship (60 percent) and felt that their partner is extremely important to their happiness (54 percent). In fact, the decline in relationship satisfaction puts the current number within the range of pre-pandemic figures, after an 11-point rise in Monmouth’s May 2020 survey.
  • On that romantic note, Americans aren’t enthused by Valentine’s Day. Fifty-eight percent of adults said that Valentine’s Day is not a “real special occasion,” according to a Jan. 27-31 poll from YouGov. Of the nine other holidays YouGov asked about, Americans preferred every one of them more than Valentine’s Day. Christmas was the most popular, leading Valentine’s Day by 69 points, and Americans even preferred the second-least-popular holiday, Labor Day, over Valentine’s Day by 8 points.
  • Twelve percent of registered voters ranked “division in the country” as the most important issue facing America, just narrowly behind the rising cost of living (13 percent), according to a Jan. 22-27 poll conducted by Lake Research Partners and The Tarrance Group for Georgetown University’s Institute of Politics and Public Service. Two-thirds of voters said that the pandemic had led to politics getting worse, while 58 percent felt that the Jan. 6 Capitol attack had made violent political protests more likely to happen in the future. In total, 43 percent said that politics had become less civil during Biden’s presidency.
  • Fewer Americans believe that Trump bears “some” or “a lot” of responsibility for the Jan. 6 Capitol attack, according to a Jan. 10-17 poll from the Pew Research Center. The share of Americans assigning blame to Trump declined from 75 percent in January 2021 to 67 percent in January 2022. This change resulted from a smaller share of Americans (43 percent this January, compared with 52 percent last January) saying Trump bore “a lot” of responsibility, while the share assigning him “some” responsibility was nearly identical.

Biden approval

According to FiveThirtyEight’s presidential approval tracker,1 41.3 percent of Americans approve of the job Biden is doing as president, while 52.6 percent disapprove (a net approval rating of -11.3 points). At this time last week, 41.7 percent approved and 52.6 percent disapproved (a net approval rating of -10.9 points). One month ago, Biden had an approval rating of 43.1 percent and a disapproval rating of 51.5 percent, for a net approval rating of -8.4 points.

Generic ballot

In our average of polls of the generic congressional ballot,2 Republicans currently lead by 2.0 percentage points (44.5 percent to 42.6 percent). A week ago, Republicans led Democrats by 1.9 points (44.3 percent to 42.4 percent). At this time last month, voters preferred Republicans by 0.5 points (42.4 percent to 41.8 percent).

Mary Radcliffe contributed research.


Most Americans just aren’t that into politics | FiveThirtyEight Politics Podcast

Footnotes

  1. As of 5 p.m. Eastern on Thursday.

  2. As of 5 p.m. Eastern on Thursday.

Nathaniel Rakich is a senior elections analyst at FiveThirtyEight.

Jean Yi is a politics intern at FiveThirtyEight.

Comments