Twelve years ago, when all the leading pundits were telling us how the Democrats were going to lose badly in the 2010 midterms, I was initially skeptical. In fact, I posted on a blog I was writing at the time that I thought a leading DC prognosticator was off-base and that Democrats were actually going to make gains that year.

I was much younger then and not as well-versed in the ebb and flow of U.S. politics as I should have been, and fortunately, I realized that my statement was foolish even before the elections took place. (I still underestimated the GOP House pickups by 10 seats that year, having predicted they would net 53 House seats; they netted 63. Until 2020, that was the only election in which I missed the final House margin by double digits.)

At the time, I was saying many of the things that naive Democrats on Twitter are saying on a daily basis today–that these midterms were different, that the pundits were all wrong, that voters would see through the Republican rhetoric, etc. It was wishful thinking then, and it’s wishful thinking now.

In fact, predicting that the sitting president’s party will lose seats in a midterm election is the safest bet in U.S. politics. There have been 39 midterm elections since the Civil War ended, and in 36 of those midterms, the sitting president’s party has suffered a net loss of seats in Congress. In short, in 92.3% of midterm elections held since 1865, the president’s party has lost seats.

The three exceptions are notable because they didn’t just happen randomly. Since 1865, the president’s party has only gained seats in midterm elections when the president has been extraordinarily popular. In 1934, Franklin Roosevelt was flying high after unemployment fell from 25% at the start of his term to 14% by the midterms. In 1998, Bill Clinton was polling around 65% approval ratings as Republicans prepared to impeach him for what most of the public considered an unimpeachable offense. And in 2002, George W. Bush was over 65% approval in the wake of the September 11, 2001 terrorist attacks.

That’s it. Those are the only three midterms in the last 157 years in which the president’s party gained seats. And those only happened when the president was extremely popular.

Now let’s fast forward to today. Joe Biden is hovering around 40% approval ratings despite a level of job creation in his first 14 months that has put several generations of prior presidents to shame. High inflation and gas prices have turned the public sour on the new president and his party.

Even if Biden were over 50% approval right now, his party would still be expected to lose seats. But with a president hovering around 40%, the likelihood of very big losses is high. Since World War II, when midterms took place with the president’s approval ratings slumping, the president’s party has averaged U.S. House losses of nearly 44 seats. In the Senate, when the president has been underwater, his party has lost an average of just under 6 seats.

Democrats currently have a margin of five seats in the House and are dead even with Republicans in the Senate. If Democrats this year only lose their post-Civil War average of seats, Republicans could be expected to gain 34-35 House seats and 3-4 Senate seats. (Senate losses are only calculated since 1914, because U.S. Senators were appointed by state legislatures prior to that year.) This would mean Republicans would easily take control of both houses of Congress in 2023–even if they don’t win at the levels that could be expected with a president well under 50% approval.

But wait, says the Twitterverse–this year is going to be different!

How, exactly? As we have seen, the only circumstances under which the president’s party has netted seats in Congress since the Civil War have been when the president’s popularity is stratospheric. Those conditions do not exist this year.

And none of those three presidents became stratospherically popular in a vacuum–there were special circumstances in all three cases that led to their high popularity: the start of a recovery from the worst economic depression in the country’s history; the overreach of Republicans against a popular president because of his personal peccadilloes; and the worst terrorist attack in the country’s history. None of those conditions exist this year, either.

Now, there are those who say that the Republican Party’s complicity in the January 6, 2021 insurrection created special circumstances, but at this point, who is even talking about the insurrection except Democrats? Right now, the public is obsessed with inflation and gas prices, not an episode that most people decried and then promptly forgot.

Another point that is getting too much credence is the fact that Democrats appear to have done better in redistricting than expected. But redistricting successes do not protect an incumbent president’s party against the normal midterm losses they suffer 12 times out of every 13 midterms, on average.

In short, there is no reason to expect that Democrats will gain seats in the November midterms, or even suffer small losses. All signs point to very large Republican gains in November, and anybody who says otherwise should not be taken seriously. They’re substituting wishful thinking for a pattern that has prevailed for more than a century and a half and is all but certain to hold true this November as well.