President Trump speaks at a Make America Great Again rally in Cleveland on Nov. 5. (Jim Watson/AFP/Getty Images)

If you ask Republicans, which the Pew Research Center did, they’ll say that President Trump’s support and his steady stream of rallies before Election Day were mostly helpful to Republican candidates. Interestingly, younger Republicans were less likely to pronounce Trump’s advocacy as beneficial, with more than half saying that either Trump’s support was a wash or that he mostly hurt Republicans on the stump. Overall, though, 60 percent of Republicans say that, yes, Trump helped candidates of his party get across the finish line.

This, of course, is Trump’s argument. As the midterms approached, he talked about his strong track record of supporting candidates over the course of the year, intentionally conflating Republican primary contests — where his popularity was an asset — with Democrat-vs.-Republican contests, where results were much more mixed. Our first look at Trump’s tweeted endorsements showed that his candidate won about half the time, meaning that an endorsement from Trump was about the same as an endorsement from George Washington (in coin form, once flipped).

This discrepancy between Trump’s effect on Republicans (good) and on everyone else (less good, as a survey in Colorado showed) is probably why Republicans thought he was a benefit to Republican candidates. To bolster the case that Trump wasn’t a drag on his party, counselor Kellyanne Conway — whose job seems heavily to entail giving cable-news interviews — appeared on “Fox and Friends” the day after the election to argue that Trump was an asset.

“Even though [the Democrats] have won a few seats here and there,” she said, “this president has made history yet again last night. In the last 80 years, the party in power has only picked up eight Senate seats, and last night, the president — through his leadership and his engagement on the ground for these candidates again and again — has produced three new Senate seats.” She later amended, “Three, possibly five as the count still goes on.”

Those “few seats” the Democrats won now total 35, with three more wins possible. The three new Senate seats is one seat, with the race in Florida too close to call and a runoff in Mississippi scheduled in a few weeks' time.

“That last day alone,” she continued, “the three people he went and campaigned for” — in Ohio, Indiana and Missouri — “huge pickups, particularly in those two Senate seats [in Indiana and Missouri], those flipped from Democrat to Republican, on the strength of the president showing up to packed houses, historic rallies and overflow crowds.”

This is the question: Did Trump make the difference? Did he boost turnout among his base, powering those wins?

At first glance, we have to note that Republicans should win Senate races in those three states. In the 2016 presidential election, Ohio, Indiana and Missouri backed Trump for president by 8, 19 and 19 points, respectively. In Ohio, Conway noted that Trump campaigned for Mike DeWine (R), now governor-elect. But he also campaigned for James B. Renacci (R), who was challenging Sen. Sherrod Brown (D). Renacci lost, despite that advantage Ohio showed Republicans two years ago.

That Trump closed out the election in red states probably wasn’t a coincidence. The House was obviously out of reach, so Trump hit the trail mostly to stump for statewide candidates in states that he himself had won.

We decided to evaluate turnout and vote margin in states and House districts where Trump rallied in October and November to try to answer the question about how much of an effect he had.

In the House districts where he held rallies — usually not on behalf of the House candidates — the vote relative to the 2016 election was more heavily Democratic in 20 of 25 districts. If we consider a national shift of about three points to the Democrats (the 2016 national margin, compared with the national House vote as of Nov. 15, which favors the Democrats by about five points), 17 of 25 House districts where Trump rallied moved more to the Democrats than the country did overall.

(Philip Bump/The Washington Post) (Aaron Blake/(Philip Bump/The Washington Post))

Many of the candidates in those districts won, anyway — a mark of how heavily Republican most of the places Trump rallied are. On average, the Cook Political Report Partisan Voter Index score for these seats was R+6, leaning Republican. Take out the House district in Cleveland where Trump rallied on the eve of the election (and which then voted Democratic), and it climbs over seven points.

Comparing House vote totals in 2016 with totals this year, most of those races saw drops in voter turnout, though the national drop (clumsily comparing national House vote totals to the 2016 vote total) was larger.

(Philip Bump/The Washington Post)

The big exception to that rule was Arizona’s 9th District, where Democrats held the seat. Turnout likely surged in part because it was the home district of Rep. Kyrsten Sinema (D), who narrowly won election to the Senate in a state that backed Trump by four points two years ago.

But, remember: Trump was clearly picking rally locations based on statewide races. So how’d that go?

The results in Florida and Nevada moved further to the left from 2016, but not as much as the country on the whole (again looking at the national House vote in 2018). In Illinois, the incumbent Republican candidate for governor did worse than Trump had, but neither did very well.

(Two notes on the chart below. First, the results are for the biggest statewide contest, favoring Senate over gubernatorial races. Second, there were no statewide races in North Carolina or Kentucky, so we used statewide House vote totals.)

(Philip Bump/The Washington Post)

But notice how much more blue is on this chart! Trump visited only two states he didn’t win in 2016: Nevada and Illinois. Democrats picked up a Senate seat and a governor’s mansion in those states, respectively. They also got that Arizona Senate seat won by Sinema. Conway focused on the two Senate victories, but Democrats have taken as many seats in states Trump visited as Republicans did. (If and when Florida is called for Gov. Rick Scott (R), that will change.)

There’s an interesting point when we consider how vote totals in 2018, compared with 2016. The state where the total number of votes in the Senate race was closest to the votes cast in 2016 was Montana.

(Philip Bump/The Washington Post)

But you’ll notice that Montana also moved heavily to the Democrats, relative to 2016. Incumbent Sen. Jon Tester (D) narrowly won reelection. If we plot all of the statewide results as a function of change in turnout and change in partisan margin, Montana is at the top — and well to the left of the no-change line on partisanship.

(Philip Bump/The Washington Post)

Turnout was high — and the Republicans did worse than in 2016. What’s more, Trump visited the state twice, including on the weekend before the election.

The Wall Street Journal has a theory: There, Trump spurred an increase in non-Republican turnout, perhaps making the difference for Tester.

Not the story the White House is telling.