Donald Trump has 39 percent of the vote in our Pennsylvania polling average, 37 percent in California, and 39 percent in Maryland. If this were February or early March, that would leave him without much to worry about. Even if Trump picked up zero undecided voters, he’d be pretty much guaranteed a win with the rest of the vote divided between a half-dozen opponents.
But those days are over. In Wisconsin on Tuesday, Trump had 35 percent of the vote — the same share that allowed him to win New Hampshire easily in February, and a larger percentage than he got in winning South Carolina. But not only did Trump fail to win Wisconsin — he got crushed by Ted Cruz.
In many respects, this is an old story. One of the main reasons for our initial skepticism of Trump’s candidacy last summer and fall is that his high unfavorable ratings implied he’d have trouble gaining ground as the field winnowed, potentially allowing other candidates to overtake him.
But I think people may under-appreciate the degree to which this is no longer just a theoretical problem for Trump. It’s become an actual problem, as is readily apparent in the data from the states that have voted so far. The threshold Trump needs to win states is increasing considerably faster than the share of the vote he’s getting, which isn’t increasing much at all. Technically, Trump is chasing delegates, not wins, but most of the remaining states award at least some delegates to the statewide winner (and there are still five winner-take-all contests left on the GOP calendar).
So that we can be more precise about this, I’m going to define a statistic called the Minimum Winning Vote Share. As its name implies, it shows the smallest percentage of the vote a candidate could receive and still win a state. Here’s how we’d calculate it for Trump in South Carolina, for instance. We start by listing the number of votes received by every candidate except Trump:
Marco Rubio was the top non-Trump candidate, receiving 166,565 votes. In order to win, Trump needed one more vote than Rubio, or 166,566 votes. If Trump had gotten that many votes, with all other candidates staying the same, he’d have won South Carolina with only 25 percent of the vote. That’s Trump’s Minimum Winning Vote Share.
In actuality, Trump got 240,882 votes in South Carolina, winning the state fairly easily. But those extra votes were superfluous; a quarter of the vote would have gotten it done. Trump’s Minimum Winning Vote Share is typically much higher than that now, however. Here it is for all states (plus the District of Columbia) to have voted so far:
|TRUMP VOTE SHARE|
|DATE||STATE||MINIMUM REQUIRED TO WIN||ACTUAL|
Trump’s lowest Minimum Winning Vote Share was in New Hampshire, where he could have gotten away with just 19.5 percent of the vote and still beat John Kasich.1 On Super Tuesday, Trump’s average Minimum Winning Vote Share was just 31.2 percent.
But as I said, it’s been increasing steadily. It was 37.4 percent on average in the five states to vote on March 15. And it’s averaged 40.3 percent in the three states to vote since then, including 42.6 percent in Wisconsin. Here’s the same data in graph form:
While there will continue to be some variance from state to state, Trump is now usually going to have to be in the 40s to win. That’s a problem, because as you can see from the bottom half of the chart, it’s not clear that his performance is improving much at all. (This is also apparent in national polls, where Trump’s share of the vote has grown only to 40 percent from 35 percent before Iowa.) Of the six states where Trump’s Minimum Winning Vote Share has been at least 40 percent – Maine, Mississippi, Missouri, Ohio, Utah and Wisconsin – Trump has won only two.
Part of the problem is that Republican voters seem to be behaving tactically, gravitating toward the most viable non-Trump alternative in their state. In Wisconsin on Tuesday, Cruz beat his polling average by about 10 percentage points. But Kasich underperformed his by 5 points, suggesting the presence of a #NeverTrump vote.
Furthermore, because Trump tends to do poorly with late-deciding voters and doesn’t have much of a turnout operation, Trump has tended to hit his polling averages right on the nose instead of gaining from undecided voters. If he’s at 37 percent in the polling average in a state, that’s a reasonably good estimate of his election-day vote; you don’t necessarily want to round up a few points to account for undecideds, as you would for most candidates.
Listen to the latest episode of the FiveThirtyEight politics podcast.
Trump will probably still be fine in the Northeastern states to vote later this month. In New York, Trump’s above 50 percent in polls, so he doesn’t care how Kasich and Cruz split up the rest of the vote. And in Pennsylvania and Maryland, Cruz and Kasich have competing claims for being the best anti-Trump, which could complicate tactical voting.
Once we leave the Northeast, however, Cruz is liable to be much more viable than Kasich, allowing him to gain from tactical voting. I’d be nervous about California if I were Trump, for example – Cruz is only 6 or 7 percentage points behind, there are two months to go, and 33 percent of voters are either undecided or say they’ll vote Kasich. If past states are any guide, a pretty decent fraction of those votes could wind up with Cruz.
It may still wind up being too little and too late for Trump’s opponents, who would have benefited from a smaller field early on. But Trump will have to work harder to win states the rest of the way out.