WASHINGTON – President Trump is getting his wish: It’s all about him.

The election, that is.

New evidence indicates that the midterm elections in seven weeks will be the clearest referendum on a president in at least 80 years.

But while it may delight the narcissistic president that the 2018 midterms are entirely about him, this is precisely what his fellow Republicans were hoping to avoid. With Trump’s support at historic lows – 60 percent overall disapprove of his performance, including 59 percent of independents – Republicans scrambling to hold the House and Senate have been struggling in vain to make the election about other issues: tax cuts, Democrats’ personal foibles – anything to avoid the election being about Trump.

This has failed, bigly.

Midterm elections have generally come to be seen as the electorate’s reaction to a presidency. But this one is on a whole different level. “In no previous election,” Gary Jacobson, a University of California political scientist who crunched the numbers, tells me, “has the linkage between opinions of the president and how people are likely to vote been as strong as this time.” Jacobson’s research goes back to the 1930s, before which there was no polling and therefore no ability to compare.

Jacobson, who presented his findings to the American Political Science Association recently and provided me with updated data, found in 93.1 percent of cases this year, voters’ approval or disapproval of the president is correlated with their planned votes for or against the president’s party in House races. That’s an all-time high. It averaged 86 percent in recent elections, 74 percent in the 1980s and 1990s.

And it isn’t just correlation – it’s causation. Using regression analysis, Jacobson determined that for every percentage point movement in Trump’s job approval rating, votes for Republican House candidates in the midterm elections move by 0.75 percentage points – the highest effect ever seen. For Barack Obama, it was 0.50 percentage points. For George W. Bush and Bill Clinton, closer to 0.25 percentage points. There isn’t as much data about Senate voting, but the relationship has been virtually identical.

This has been caused by a combination of historical trends and Trump’s uniquely polarizing status.

Jacobson, working with data from Gallup, the American National Election Studies and others, writes that as recently as 1990, there was only a 31-point difference between how Republicans and Democrats rated the president’s performance in midterm election years. That jumped to a roughly 70-point gap during the midterms of 2006, 2010 and 2014. This year? Seventy-eight percent.

At the same time, party loyalty – the tendency of Democrats and Republicans to vote for their own party’s congressional candidate – has grown from the mid-70s in the 1980s to 90 percent in the past decade. This year? Ninety-six percent.

This does not necessarily mean a Democratic wave, or even a victory. Democrats now lead by about 9 points in the generic House ballot – which asks respondents which party they would vote for in a congressional election. But because of gerrymandering and the distorting effects of urban districts with Democratic supermajorities, Democrats would need to win between 53 percent and 55 percent of the popular vote, Jacobson calculates, to pick up the necessary two dozen House seats. And this year’s Senate map is even more daunting.

But Trump’s unpopularity seems to offset the benefit Republicans should get from the strong economy. Using results from previous midterms and factoring in the president’s approval and the growth in real disposable income, Jacobson reports that in conditions close to the current situation – 2 percent income growth and Trump’s approval at 40 percent – Republicans would, by historic models, lose 33 House seats.

Republicans in competitive races are in a bind. Among independent voters they need to win, Trump is a pariah. But among the Republican voters they need to turn out in high numbers, Trump has 78 percent approval.

Their dilemma was evident on Thursday when Trump made the outrageous and false claim that the official death toll of 2,975 from last year’s storms in Puerto Rico was inflated by Democrats “to make me look as bad as possible.” (Even storm deaths are all about him.) Delicately, Republican candidates in Florida, who had been trying to win over Puerto Rican voters, tried to step away from Trump. Gov. Rick Scott, running for Senate, tweeted: “I disagree.” Rep. Ron DeSantis, running for governor, issued a similar statement.

Good luck with that. Trump, later in the day, repeated his insulting and bogus claims.

As Trump continues to repel, opinions of him drop and support for a Democratic Congress rises. It has the makings of a wave, but one that could recede before the election. We are destined for one of two outcomes: a massive repudiation of Trump, or an unthinkable affirmation of him.

The stakes could hardly be higher.

Dana Milbank is a member of the Washington Post Writers Group. Follow him on Twitter @Milbank.