Expert Commentary

Election Beat 2022: What we can learn from predictive models and polls

Will 2022 be a 'normal' year for voting on House seats? Academic statistical models predict a Republican win in the 2022 U.S. midterm elections, but opinion polls add a cautionary note.

(PublicDomainPictures from Pixabay)

The congressional campaign for the U.S. midterm elections has been going on for months, with election spending already topping $5 billion.  With all that effort and money and the accompanying news coverage, it might be thought that the campaign determines whether Democrats or Republicans will prevail. That can be true in the close contests, but it’s the larger and less heavily reported factors outside the campaign that are normally decisive, as can be seen from the predictive accuracy of academic statistical voting models that omit campaign effort and spending in their estimations.

Predictive models are used to reduce complex phenomenon to a limited number of explanatory factors, such as the condition of the economy. The usefulness of a model is judged by the accuracy of its predictions – in this case, whether a model’s prediction of the outcome of an election aligns with the actual outcome.

The “surge-and-decline model” is one example. It starts with a recognition that the previous presidential election brought out a “surge” of voters who don’t show up two years later for the midterm. These presidential-year voters were especially responsive to the political mood of the presidential race and disproportionately voted, not only for the winning presidential candidate, but also for the congressional nominees of that candidate’s party.

Two years later, these voters stay home (the decline), erasing the advantage that the in-party’s candidates had in the presidential election. The result is that the outcome in most of the House districts that were narrowly won by the in-party in the presidential election flip to the out-party, enabling it to gain House seats in the midterm.

The surge-and-decline model fits with the historical pattern of midterm elections. In 22 of the 25 midterm elections in the past century, the out-party has gained House seats. The average pickup during this period has been 30 seats.

A second model of midterm election is “the referendum model.” It is premised on the idea that the midterm is a referendum on the performance of the president and, by implication, of the president’s political party. This model also predicts that the in-party will typically lose seats in the midterms. The basis for this prediction is that, once in office, presidents pursue policies and inherit policy problems that are likely to make more enemies than friends.

Presidents do indeed tend to lose popularity with time. Since 1960, only two presidents (Bill Clinton in 1998 and George W. Bush in 2002) have had a higher approval rating in the second year after their election than in the first year of their term. It is especially bad news for the in-party when the president’s approval rating slips below 50% in the second year. In the eight midterms since the end of World War II where the president’s approval rating has been that low, the out-party has picked up 26 or more House seats each time, except 2014, when it gained 13 seats.

Joe Biden’s approval rating is in the low 40s. Using previous presidents’ approval ratings and midterm outcomes as the basis for a regression analysis, I estimated Republicans’ gain this time. It came in at roughly 30 House seats, which is in the same range predicted by the surge-and-decline model.

The question then is whether the 2022 midterm will be typical of midterms. Is there anything that would offer the Democrats a glimmer of hope of holding onto the House?

If the midterms had been held in the Spring, nothing would have saved the Democrats. Generic polls at the time – “If the election were held today, would you vote for the Democratic or Republican candidate for Congress?” – averaged roughly 53-47% in favor of Republicans. If the national House vote were to divide in that way, Republicans would pick up more than 30 House seats in the midterm election, based on historical averages.

The situation has changed since Spring. Recent generic polls indicate that voters are divided roughly 50-50 in their congressional preference. Democrats have also closed the “enthusiasm gap,” which is the polling surrogate for estimating whether Republicans or Democrats are more likely to turn out on Election Day. In the Spring, Republicans were more than 10 percentage points more likely than Democrats to say they were “very” or “extremely” interested in voting in the midterm. Today, Democrats and Republicans are in a statistical tie on that indicator.

What’s changed that appears to give Democrats a chance of bucking the historical pattern? I would single out the “referendum” model. Typically, the midterm is a referendum on the in-party. The 2022 midterm has become a referendum on both parties.

Two developments have made the out-party’s performance an issue in the midterms. One is Donald Trump’s deep involvement – almost unprecedented for a former president – in Republican primaries, his continued false claim that the 2020 presidential election was stolen, and his unveiling by the January 6 House Committee. If Biden is hovering over the 2022 midterm ballot, so is Trump. And Democrats gain an edge because Trump’s approval rating is worse than Biden’s.

The second development is the Supreme Court’s Dobbs abortion ruling, accompanied by the perception that its Republican appointees are acting less like arbiters of the law than as agents of the Republican Party. And that also is working in Democrats’ favor. The Court’s favorability ruling is at lowest recorded point in history and below that of Biden’s.

Does all this mean that control of the House is fully up in the air, that Democrats have as much of a chance of holding onto the House as Republicans do of retaking it? There’s yet a third prediction model that tips the balance in Republicans’ favor. It factors in the state of the economy at the time of the election. This model favors the in-party when the economy is on the upswing and favors the out-party when the economy is trending downward. America’s economy is struggling and is creating a headwind that’s likely too strong for Democrats to overcome whatever the amount of their effort and spending between now and that fateful Tuesday in early November.

Thomas E. Patterson, Bradlee Professor of Government & the Press at Harvard’s Kennedy School, is the founder of The Journalist’s Resource and author of several books, most recently “Is the Republican Party Destroying Itself?” JR has posted a new installment of his Election Beat 2022 series every week leading up to the midterm elections. Patterson can be contacted at

Further reading:

Angus Campbell, “Surge and Decline: A Study of Electoral Change,” Public Opinion Quarterly 24 (1960):  397-418.
James Campbell, “The 2022 House Midterm by the Numbers,” Real Clear Politics, July 5, 2022.

John R. Petrocik and Frederick T. Steeper, “The Midterm Referendum,” Political Behavior 8 (1986): 206-229.