Just as it is rare for rivers to reverse their flow, in the last six months before a midterm election the clear direction of that campaign almost never reverses. Midterm elections are almost always referenda on the sitting president and governing party, with the cumulative impressions of how that president and the party in charge are performing gradually hardening. Conditions can improve or deteriorate, but the direction, generally speaking, doesn’t reverse. With the results from the Gallup Organization’s May survey now in hand, President Biden’s Gallup approval ratings have slipped below those of Donald Trump’s at this point of their respective presidencies, and are the lowest of any elected president in the post-World War II period in May of their second year.

Obviously, things can change, but recent history certainly looks ominous for Democrats. Biden’s 41 percent Gallup approval is 1 point below that of Trump in the May before the 2018 midterms, and 2 points below Trump’s final, pre-midterm-election Gallup survey. As one might expect under those conditions, Republicans lost 42 House seats in that election, although the GOP actually had a net gain in the Senate of two seats, thanks in part to a very favorable map.

To put Biden’s approval numbers in a historical context, his current Gallup approval ratings are 7 points below Bill Clinton’s 48 percent in May of his second year, and worse yet, 5 points below where Clinton’s were going into the 1994 midterms, when Democrats lost 54 seats in the House. Barack Obama was 7 points higher at this point, and going into the 2010 midterm Obama’s approval rating was 4 points higher than the current president. His party still lost 64 seats in that 2010 election.

With inflation running at nearly a 40-year high and fears of a recession growing, this is certainly a combustible situation for any party in power. A Quinnipiac University national poll earlier this month indicated that 85 percent of Americans expect a recession in the next year. This column has previously discussed “issue contamination”—that is, when voters are mad at a president on one issue, it colors their view of that president’s performance on other issues as well. The public’s memory of Biden and his administration insisting that there would not be an inflation problem is not hazy in the least bit. It’s akin to Trump’s dismissiveness about the coronavirus. It is hard for any other issue to provide any buoyancy when one big issue is pulling a president down.

It is difficult to see even how a reversal of Roe v. Wade or the gun issue can trump economic concerns in general and inflation in particular this November. Abortion, guns, or both may very well come to haunt Republicans in the future, perhaps 2024 or more likely 2026, but in the near term, nothing is likely to overshadow the economy.

Political scientists use various models to project midterm-election outcomes, though many are not released until the American Political Science Association’s fall meeting around Labor Day, roughly nine weeks before Election Day. One well-regarded political scientist who has offered a model is Seth Masket of the University of Denver, a frequent contributor to FiveThirtyEight and the Mischiefs of Faction blog. In late December, Masket produced his first cut modeling the 2022 midterms, his computations incorporating economic growth in the form of real per-capita disposable income, controlling for a president’s Gallup approval ratings and the number of House seats that president’s party holds. Using available data at year end, Masket’s model suggested a net loss for Democrats of about 25 seats. Earlier this month, Masket updated the model again with new data, which produced a new projection of a Democratic loss of about 64 seats, coincidentally the same level of losses that Democrats suffered in Obama’s first midterm election.

While Democrats are likely to lose both their House and Senate majorities, it is doubtful that House losses will reach 64 seats. Indeed, Masket even offers some reasons it might not. One is that after three consecutive cycles of very aggressive, even audacious gerrymandering by both parties, the number of competitive districts is much smaller, arguably reducing the volatility.

Another reason is that Democrats lost a dozen House seats in 2020. Just as the ‘A’ seat on an airliner is always a window seat, a party cannot lose a seat they don’t have.

In modern times, big wave elections have tended to come from a party well behind in seats. Republicans’ House gain of 54 seats in 1994 was from a starting point of just 174 seats; their 64-seat pickup in 2010 was from 178 seats. When Democrats gained 42 seats from Republicans in 2018, they started with just 194 seats. Allocating the currently vacant seats into the column they had come from (and will likely return), Democrats hold 222 seats and Republicans 213, well above the GOP levels going into 1994 and 2010 and Democrats in 2006.

Most current estimates of likely GOP House gains range from as low as a dozen seats (seven more than necessary for the barest majority) to about three dozen. The current outlook from David Wasserman, The Cook Political Report with Amy Walter’s expert on the House, is a Republican gain of between 20 and 35 seats.

One of the foremost political scientists specializing in congressional elections, James Campbell of SUNY-Buffalo, points out that a gain of 34 seats would equal Republican’s high-water mark since the 1928 election, implying that if a party has not exceeded 247 in 90 years, the odds are against it doing so now.

Then again, caution may be in order in applying historic patterns from a period before partisan polarization became as extreme as it is today, with defections among partisans quite rare and pure independents exceedingly fickle and prone to buyer’s remorse.

We are left wondering whether this will be a bad election for Democrats, very bad, or very, very bad. As they say in the markets, the downside risk for Democrats is grave.

The article was originally published for the National Journal on May 31, 2022.

More from the Cook Political Report