Donald Trump looks set to lose the presidential election to Hillary Clinton by a fair margin. But what about Congress?

The GOP’s Senate majority is looking shaky. The Rothenberg & Gonzales Political Report currently lists eight GOP seats as vulnerable, compared to only one Democratic seat. For the GOP to hold the Senate (assuming Clinton wins the presidency), they'd have to win five of these nine—not impossible, but a heavy burden when Trump is dragging down the rest of the ticket.

What about the House? The conventional thinking at this point is that the House is in reasonably good shape for the GOP. At the moment, Rothenberg & Gonzales lists 35 seats as currently in play—29 Republican seats and 6 Democratic seats. The Democrats would have to run the table to take the House, based on these ratings. That is virtually impossible.

Rankings such as these are based upon a granular analysis of the conditions on the ground. These conditions can shift, sometimes late in the cycle. The worry for the GOP is that the current analysis, while accurate for this moment in time, gives a false sense of security—that by September or October, a lot more GOP seats will show signs of slipping away.

As a way to get in front of such a scenario, it makes sense to look at the big picture. Historically speaking, does a congressional party run ahead of its losing presidential candidate, and if so, by how much? To answer this question, I examined the congressional vote for parties that lost the presidential election in the postwar era (1948-2012). Blue dots trace how far ahead congressional Democrats ran ahead of their losing nominee; red dots trace the same for Republicans.

The postwar average is about 3 percent—meaning that the congressional party runs about 3 points ahead of its losing presidential nominee. But that long-run average elides the recent trends.

Between 1948 and 1988, there were some pretty wild differences between the presidential vote and the congressional vote. Democrats in Congress tended to run very far ahead of their presidential nominee, while Republicans usually ran behind. This was a consequence of widespread split-ticket voting, especially in the South. Voters would disproportionately favor Republican presidential nominees over Republican congressional candidates.

That began to change in 1992. Since then, the House vote and the presidential vote have been more or less in sync, with very little divergence between them—the congressional party has run a little less than one percentage point ahead of the presidential nominee.

We can get at the recent trend from another direction, via the exit polls.

The exit poll for the 2000 presidential election is unreliable, so this chart tracks four of the last five elections. As you can see, split-ticket voting has been pretty minimal—just 10 percent of the total vote, usually divided equally between the two sides. This is why the House vote and presidential vote have been in sync for the last generation.

That is bad news for the GOP in 2016, but the 1996 results offer a ray hope. Bob Dole lost by a fairly large margin, but there was a substantial rate of defection from Clinton voters to the congressional GOP. This was enough for the GOP to win 50 percent of the two-party House vote while Dole won 45 percent of the two-party presidential vote.

The final result was a fairly comfortable, 227-208 majority for Republicans heading into the next Congress in 1997. Typically, incumbent parties win the House majority on a split vote. That was also the case in 2012—the House GOP won 49.4 percent of the two-party vote for the House, but won 234 seats.

Bear in mind that this analysis overlooks the kind of granularity offered by analysts like Rothenberg & Gonzales. So, it does not take into account the current partisan tilt of congressional districts, the number of retirements, and the quality of Democratic challengers. The goal is instead to lay down some broad parameters for assessing the state of play.

With that in mind, the GOP can probably retain its majority if it loses the two-party vote by 2 points or less—so it needs to reach about 49 percent of the two-party vote. Recent history suggests that in the best-case scenario it can run about five points ahead of the presidential nominee. By this reckoning, Trump needs to win 44 percent of the two-party vote.

Right now, the Real Clear Politics average has him at 41.2 percent of the total vote, with a fairly large undecided bloc of voters. If they swing overwhelmingly to Clinton, the House Republicans could very well lose their majority. If they split evenly between Trump and Clinton (or vote for a third party), the House GOP should be able to hold its majority.

Unfortunately, the operative question—can the GOP hold the House?—has no answer just yet. It looks as though Trump is set to lose in November, and the main question for House Republicans is how badly he will lose. If it is a 1996-style defeat, the House GOP should be able to retain control (although that cannot be admitted as a certainty). If the magnitude of the defeat is on the order of 1984, 1972, or 1964, then Nancy Pelosi stands a good chance of reclaiming the gavel. In this modern age there are simply not enough split-ticket voters for the Republicans to hang on.

There are two practical implications from this analysis. First, Republican candidates need to be ready to bail on Trump—perhaps not explicitly, for fear of alienating Trump's hard core of support. But they need to begin communicating to voters that splitting their tickets is a good way to check the liberalism of Nancy Pelosi and Hillary Clinton.

Second, they should not presume that they can liberate themselves entirely from their nominee. Yes, congressional parties were able to hold their majorities despite wipeouts at the top of the ticket in '56, '72, and '84—but these elections were a long time ago. Over the last quarter-century, no party has run better than five points ahead of its presidential nominee. Maybe, because Trump is sui generis, they can outperform the recent trend—but they should not presume this.