Second and more importantly, presidential approval ratings in recent years have been a decent indicator of what will happen in the midterms. In the last four (2006, 2010, 2014, 2018), the incumbent president’s disapproval rating was higher than his approval, and in all four cases, the president’s party lost a sizable bloc of House seats. (The Senate results aren’t quite as tied to presidential approval.) The last time the president’s party gained House seats in a midterm election was in 2002, when George W. Bush had sky-high ratings in the wake of the Sept. 11, 2001, terrorist attacks. So, when we talk about the pattern that the president’s party nearly always loses congressional seats in the midterms, part of what seems to be happening is that the American electorate becomes somewhat disillusioned with a president after electing or reelecting him (or wants to check his power) and then backs the opposite party’s congressional candidates.

And presidential approval ratings are becoming even more predictive as American politics are increasingly partisan and president-centered. The Obama and Trump presidencies suggest that the overwhelming majority of voters lean toward either the Democrats or the Republicans and approve of presidents from their own party and disapprove of presidents from the opposite party.

And those mostly partisan approval numbers translate to mostly partisan voting: More and more, voters cast ballots for candidates from the same party in both presidential and congressional elections.