That's the projection from a new model built for Vox by Josh Putnam — an expert on the primary process, a lecturer at the University of Georgia, and proprietor of the Frontloading HQ blog.
Our projection shows Trump on track to finish with 1,279 delegates in the state races we've chosen to include. However, our variables also indicate he's likely to win at least 100 more from contests we've chosen to exclude from the model due to the complexity of their rules. So he has a bit of room to spare in this scenario.
Now, consider our model a ballpark projection of how the race would look if what's happened so far keeps happening in the states still to vote. It's a starting point for analysis, not an endpoint. We'll update and tweak these projections in the coming weeks, and incorporate the results of newer contests.
And let's be clear — this is not a model that means to predict anything going forward, especially individual states. For instance, the model doesn't know or care that Ohio is Kasich's home state, something that could give him a much better shot to win there Tuesday (we have Trump winning). And it doesn't try to account for how the race could change going forward. For instance, if Rubio drops out of the race, we don't know whether Trump, Cruz, or Kasich would get the biggest boost, so we've chosen not to speculate about that.
To come up with our projections, we looked backward to see what factors best predicted the GOP candidates' success in states so far. Putnam gathered the data and ran the numbers for several different demographic and political variables. But in his analysis, a relatively simple combination of just three seemed to work best:
- Nonwhite population: As others have observed, Trump performs better among Republican primary electorates in less white states. Since those Republican electorates are still overwhelmingly white, this appears to be a racially tinged reaction or backlash effect in regions like the Northeast and the South that doesn't exist in many of the heavily white Midwestern states that have voted so far. Putnam found that Ted Cruz, too, does unusually well in these districts (and Rubio unusually poorly).
- Obama-voting states and districts: We looked at Charlie Cook's Partisan Voting Index (PVI), which compares how a state or district's vote in the past two presidential elections differed from the national average. Putnam noticed that John Kasich in particular did better in areas where Obama did better.
- The Interior West: So far this year, Ted Cruz has performed best (and Trump has appeared to stumble) in states like Iowa, Kansas, Oklahoma, and Idaho. So Putnam came up with a regional "dummy variable" that separates out states that are west of the Mississippi but not on the West Coast or along the border with Mexico. This is, admittedly, fairly arbitrary and we'll see how it holds up later on (for instance, in Missouri today), but this variable does a good job accounting for results so far.
So Putnam ran a regression of those variables and the vote share for each candidate in states and congressional districts so far, and used them — and his intricate knowledge of the delegate allocation rules — to project the outcomes in upcoming states and congressional districts, and therefore the overall delegate count. (For a lengthier explanation of Josh's methodology, check out this post at FrontloadingHQ.)
Overall, our projection has good news for Donald Trump:
- It shows him winning 18 out of the 23 remaining states in our model — from Northeastern states like New York, Connecticut, and Pennsylvania to Midwestern states like Wisconsin and Indiana to West Coast states like California, Oregon, and Washington.
- It shows him winning the vast majority of congressional district delegates in the states he wins — for instance, that he'll win all 172 delegates from California, even though the vast majority of them are split up by congressional district, with three delegates going to the winner in each of the state's 53 districts.
Finally, it shows him surpassing his target of 1,237 delegates with 1,279 overall. Now, that may seem at first glance to be cutting it extremely close even in a scenario where Trump wins a great many victories. However:
- There are another 141 delegates from Illinois, Pennsylvania, and West Virginia who are directly elected (meaning people have to vote for the delegates themselves, not the candidates). We've chosen not to include them in the model because of this complication, but given our variables for those states and districts, Trump would probably pick up most of these delegates.
- There will also be some uncommitted delegates floating around from states or territories without traditional primary and caucus contests, like North Dakota, whom Trump (or his rivals) could potentially win over.
So Trump has some room to spare. But not a ton of room. If his support declines somewhat, he could certainly fall short of that target. And if he starts falling short in several states we expect him to win — like Ohio, which votes Tuesday — he could be in serious danger of missing that target.
As for the other candidates, our projection shows Ted Cruz winning all five of the states that aren't won by Trump — Missouri, Utah, Nebraska, Montana, and South Dakota. Unfortunately for Cruz, those states are only worth 161 delegates, and though he picks up a few more here and there, our projection shows him stalling out at 601 delegates — less than half Trump's total. The takeaway from this is that Cruz cannot win if he only wins states similar to those he's won so far, because there aren't many of them left. If he can't manage to broaden his appeal (perhaps in a winnowed field), he'll lose by a wide margin.
Marco Rubio and John Kasich, meanwhile, pick up only a paltry number of delegates, in keeping with their weak performance so far. That's unsurprising, since Rubio has only won Minnesota, Puerto Rico, and the District of Columbia so far, and Kasich has won nothing. If Rubio and Kasich do stay in for the entire rest of the race, our projection would have Rubio finishing with 235 delegates and Kasich with 78.