Princeton Election Consortium

A first draft of electoral history. Since 2004

Is it in Donald Trump’s interest to win Ohio?

March 14th, 2016, 8:04am by Sam Wang

I have a new analysis at The American Prospect. Basically, it is in Trump’s interest to avoid the situation shown below.
Details of the model and additional comments to come. In the meantime, morning reading: (1) an excellent piece by Jamelle Bouie on Trump’s appeal to white voters, and (2) some data analytics at the Upshot on county-by-county correlates of support for Trump. Notable, at #2: ethnic self-identification as “American.”

Update: Twitter follower W. Dow Rieder points out the optimal anti-Trump strategy for Kasich: win Ohio, then drop out for later primaries where he is behind. Basically, this is Romney’s advice taken to an extreme. It has the highest likelihood of holding both Trump and Cruz below 50% of delegates each.

Tags: 2016 Election · President

37 Comments so far ↓

  • Amitabh Lath

    Rubio says he will not quit even if he loses Florida. Fig. 1 in the article (Trump v Cruz v Kasich v Rubio) may be relevant for some time.

  • Amitabh Lath

    Northern Mariana Islands gives 9 delegates to Trump. More importantly I believe this puts Trump over the limit specified by Rule 40(b) that specifies a candidate must win a majority of the delegates in at least eight states or territories.

    • JayBoy2k

      If neither Rubio or Kasich drop after losing their home states. does that not make Trump’s path that much easier.
      I have vaguely heard of rule 40b. That a candidate must win at least 8 states/territories to be nominated. How can this be? Only Trump and Cruz will pass that hurdle.
      Is this one of those rules that the Convention committee (made of GOP establishment types) can throw out/revoke as the Convention starts?

  • 538 Refugee

    Mondays are my night out. I learned first hand that Democrats crossing over to support Trump is real. Trump was the topic of conversation since he was here in Youngstown yesterday. One bartender didn’t understand that she couldn’t vote for him AND still vote Democrat in the primary. The other was still waffling. Wants to support Trump but he isn’t 100% sold ‘yet’. Sounded like he was fishing for affirmative support. He got none from me.

  • Sean Patrick Santos

    How do we feel about Kasich losing OH but staying in the race? I don’t think that it’s the most likely outcome, but if Ohio is close and Rubio drops out, Kasich may feel that it’s reasonable to go forward as the new establishment-anointed candidate, even having lost his home state.

  • Jim Hafford

    Can it be that delegate math is now as warped as campaign rhetoric?

  • Amit

    Sam, your analysis, while excellent as always, does not account for possible ‘momentum’ effects that an (unlikely) win in Ohio will cause. A look at Trump’s polling seems to indicate that his best narrative is victory itself – he seems to get a bump following victories. Sometimes a significant bump, like in FL (though admittedly other factors were in play as well).
    Unlike victories by Cruz / Rubio (or Sanders on the D side), which are regarded as ‘shocks’, a victory by Trump feeds into the leading the field / inevitability narrative.
    Even in your worst case mano-a-mano scenario, Trump is only about 260 delegates short, which is a gap that be may overcome on momentum alone. Any way to model that using a generic auto regressive parameter from other polls immediately following a Trump victory?

    • Amitabh Lath

      I am a little skeptical of this momentum thing, whether it actually exists, or is just a convenient narrative device for columnists looking to fill inches.

      Has (political) momentum ever been quantified? Maybe in exit polls in which voters admit they made their choice because of impressive wins in previous states?

      Anyway, always nice to meet another Amit.

    • Sean Patrick Santos

      I have a very different view on “momentum”. Winning the state has the most positive effect when a candidate has lower name recognition, or is considered by many people to be unable to win. Sanders and Cruz in this election (and Obama in Iowa, 2008) have benefited from wins at various points. Trump also did in New Hampshire, in that if he had underperformed both that state and Iowa, both “elites” and many voters would have seen him as a candidate with no chance. Kasich wasn’t even treated as one of the top candidates before he start overperforming, making him appear as a viable alternative to Rubio for establishment/traditional Republicans.

      But now that Trump is the indisputable frontrunner, I don’t think Ohio gives him a boost at all. Everyone expects him to do well, so I don’t see a win in Ohio as likely to boost his numbers much. Not unless voters switch from Kasich to Trump at a far higher percentage than anyone expects.

    • Amit

      Momentum will always be hard to quantify in a causal sense. On the other hand, it will easily show up in a ‘significant coefficient’ sense and provides an easy / lazy explanation for many phenomena social science.
      My only thought was that Trump’s entire campaign is built upon being a Winner (in business, negotiations and now in elections). Anything that feeds that narrative may sway more voters in his direction, and he does not need a whole lot. As Sam has constantly shown, a third of voters will do it.

    • Amitabh Lath

      “Narrative” is another one of those words that I wonder about, like momentum. Does it exist outside the tiny world of political obsessives? The usual trajectory is “candidate A did better than expected in state X which led to wins in state Y, Z”. Is that true? Or were X,Y and Z correlated, and the problem was simply underestimated expectations?

      I agree with S. P. Santos, Trump voters have made up their minds. They appear even more disconnected than usual voters from the usual media blather.

  • Josh

    Hey Sam,

    Can you comment on Ben Ginsberg’s assertion that if Trump doesn’t win Ohio tomorrow he likely won’t be able to win enough delegates to clinch the nomination?

    • Sam Wang

      He can’t be sure. Nine upcoming states have no polls. See my first histogram in the Prospect piece, plenty of possibilities. Also, those juicy uncommitted delegates…

    • Petey

      “He can’t be sure.”

      He’s not. He’s spinning.

      Ben Ginsberg is an incredibly important person in this process. He’s under-covered by the political press.

      He’s the institutional Party’s inside man in the anti-Trump effort.

      Right now, his role is passive. And that statement of his you mention is part of a pattern of smallscale dissembling on a variety of topics. His aims right now are just to cast aspersions on a 1st ballot Trump victory.

      But when the convention approaches and arrives, Ben Ginsberg will be ACTIVATED.

      If Trump walks in with 1,232 ‘bound delegates’, Ben Ginsberg will be the point man on not nominating him. And if Trump walks in with 1,300 ‘bound delegates’, well, if the Party is unified and determined, that’s where Ben Ginsberg will be the judge and jury for the nuclear scenario.

      (I’ve always thought Ginsberg would be the point man for the nuclear scenario, but to lie so baldly in that Politico piece made me more certain it’s real in their minds.)

      In short, read everything Ben Ginsberg says these days as propaganda, not analysis.

    • Sam Wang

      I think Ginsberg is knowledgeable about the mechanics. In person, he appears to be pretty open about these things. I suspect he actually believes what he’s saying.

      However, in my view his belief gets in the way of considering alternate plans that are more likely to escape the fate of a first-ballot Trump nomination.

    • MAT

      The ‘nuclear scenario’ is what I heard Michelle Steele discussing on his POTUS XM show several weeks back. Petey’s right, it’s legal, or at least the rules committee guy they were interviewing said it was.

      Sam’s point below about keeping Trump under 50% notwithstanding, it’ll be very interesting to see who percolates up thru the delegate selection processes in all these county/district/state conventions. It all comes down to who is the best organized and paying attention and the people who have been doing this for years are the party regulars.

    • Petey

      “Sam’s point below about keeping Trump under 50% notwithstanding, it’ll be very interesting to see who percolates up thru the delegate selection processes”

      The fact no one has been paying any attention to is that the GOP completely revamped their delegate selection procedures after 2012 to stop a Ron Paul insurgency. (Ron Paul was able to use superior organization to grab more than his fair share of delegates.)

      This obviously comes in very handy for the GOP Establishment this year.

      In prior years, organization really mattered, and the stop-Trump efforts would have a much harder time. But with the new delegate selection procedures, the deck is stacked against organization, and in favor of the institutional Party.

      It’s almost as if the GOP has 3,000 Superdelegates this year…

      (And all due respect to Sam, but in my view, Ginsberg sees his current minor role as simply casting aspersions on the legitimacy of a 1st ballot Trump win, which lays groundwork for the real mission to come. Stopping Trump from 1,237 is a job for other folks.)

  • W.Dow Rieder

    Glad I was able to contribute. Game theory can lead you to some counter-intuitive but still potentially optimal solutions in multi-sided contests like this.

  • Mark F.

    I can see Kasich picking up a fair number of delegates in upcoming states if he wins Ohio and stays in the race. And then having a significant role at the convention, maybe even being a compromise nominee or at least getting on as Vice President. But the risk is, as you note, that this strategy gets Trump too close to 50% or gets him over…

  • Mark F.

    Good analysis. I assume you think Clinton has the race locked up on the other side, don’t you?

  • 538 Refugee

    Just last night I was going through some of the states yet to vote and wondering how this would shake out. New York and California are probably the two largest outstanding states. NY is the winner take all (Trump?) and California is district level so I really couldn’t make any assumptions on how that shakes out. Trump and company are still working through learning this primary process. The Kasich people got one of their ads blocked in Ohio because the disclaimer was in the ‘wrong place’. Cruz has more of the traditional groundwork in place but I’d think it has to be spread thin in many of the states left.

    That said I think Trump didn’t know from the beginning how he would fare so probably didn’t invest a lot of his money in places. I expect this to change as his chance of winning increases. Losing would hurt the Trump brand/ego so expect the purse strings to loosen when needed. Kind of like “just in time” manufacturing processes. After Trump is done in the Midwest look for him to tack to the middle, something Cruz just absolutely cannot do in any shape nor form. Kasich could hurt him if still in. I think Amit also makes a good point about Republicans just sitting this one out if their only choice is Trump or Cruz. It’s a long step from voting Kasich/Rubio/Carson to voting Trump/Cruz. Ordinal order on a survey is a long way from the ballot box.

    • Froggy

      I think New York is winner-take-all (by district and for at-large delegates) only if someone gets 50%, and otherwise is proportional with a 20% threshold.

      As far as Republicans sitting the general out, it may be a long step from voting Kasich/Rubio/Carson to voting Trump/Cruz, but it’s a reasonably-sized step from viscerally disliking the Democratic nominee, and not wanting that person to control the presidency for the next four (or eight) years, to voting for that person’s opponent.

    • 538 Refugee

      Froggy, I did not mean sitting out the general, just talking the primary here.

  • bks

    Excellent article! This will give the talking heads lockjaw.

  • SoddingJunkMail

    Given this new analysis do you have a revised overall percentage chance Trump wins the nomination?
    As recently as March 9th you had him at 90%, but today’s post would suggest the odds are lower.

  • Amitabh Lath

    The difference between Fig 2 and Fig 3 (Trump v Cruz v Kasich, and Trump v Cruz) is huge. I suppose this is due to the 4/1 reassignment of Rubio and Kasich voters to Trump.

    This makes me wonder, would some fraction of these Rubio/Kasich establishment voters simply stay home rather than choose between Cruz and Trump? Could this be a significant effect?

  • Petey

    “The only candidate who should not want Kasich to win Ohio is Ted Cruz.”

    Oddly enough, I think Kasich should want Trump to win Ohio by a point or two, even if it means him dropping out. (Stipulating we accept the conclusions of the rest of your model.)

    Assuming it’s not a Trump 1st ballot victory, I think Kasich is one of a very small number of the strongest favorites for the nomination. (I’ve got Ryan/Kasich/Romney as my list.)

    He’s acceptable to most factions of the institutional Party, who’ll be responsible for picking most of the delegates. (Only downside is that the Kochs hate him with a passion over Medicaid expansion.)

    He’s viewed as electable. And, of course, perhaps most importantly, he’s from Ohio.

    So whether he knows it or not, he wins by losing. (Stipulating we accept the conclusions of the rest of your model.)

    • Sam Wang

      I currently think the optimal course for Kasich is to win Ohio, then stand back and let Cruz pummel Trump into getting <50% of delegates. This is basically the Romney plan, more or less.

    • Petey

      “I currently think the optimal course for Kasich is to win Ohio, then stand back and let Cruz pummel Trump into getting <50% of delegates. This is basically the Romney plan, more or less."

      Well, sure. Winning OH denies Trump delegates, and it gives Kasich the aura of a winner who gets narrative props for performing in a highly crucial swing state.

      Unfortunately for Kasich, the Romney plan is insane going forward, IMHO. The vast majority of voters don’t go to the polls to pull the lever for that charismatic and inspiring candidate, Strategic Voting.

      And even more unfortunately for Kasich, there’s a Reuters story this morning that indicates Team Kasich fully buys into the strategic voting plan that’ll divide the anti-Trump vote two ways, and will campaign hard going forward if he wins.

      The truly optimal Kasich course would be to win OH, and in his victory speech that night, announce he is suspending his campaign to avoid splitting the anti-Trump vote, and that he plans to win in Cleveland in July. But that ain’t gonna happen…

    • Kevin

      Sam, I see your theoretical point about Kasich’s strategy, but I don’t see how Kasich can execute the “stand back” maneuver in the real world short of winning Ohio and then dropping out. The problem is that Kasich/Rubio supporters will not support Cruz in the numbers described unless placed in a bind where the only alternative is Trump, right?

      I have not gone so far as to parse polling data, but I find this paragraph from Nate Silver concerning(

      The exit polls in Michigan and Mississippi asked voters who they’d pick in a two-way race between Trump and Cruz, also giving them the option to say they’d sit out the race. Among Rubio voters, on average between the two states, about 75 percent said they’d still vote in a Trump-Cruz race, and of those, 80 percent would prefer Cruz to Trump. Kasich voters were somewhat more equivocal; 55 percent said they’d still vote, and of those, two-thirds would go to Cruz over Trump. (/end Nate Silver)

      Did your model incorporate the “I won’t vote” alternative? That option makes the landscape more favorable to Trump, but without running simulations I couldn’t say by how much.

      This analysis is excellent, however, and very clarifying in terms how to think about the strategical consequences of future campaign decisions, even if it has limitations as an actual prediction of final election outcomes.

    • JayBoy2k

      Great Analysis, Sam If anything, Trump is trending up.
      If we just go with the latest polls, Trump is going to win 4 states and lose one. Rubio will then drop out and Kasich will do everything he can to beat Trump in the Northern/Western states. The lone surprise may be Missouri to Cruz.
      The idea that either Kasich or Cruz will hold back is almost as foolish as the idea that Rubio would drop out to make things easier for Cruz.
      I think we go into a contested convention with Trump & Cruz holding 70-80% of the delegates.

  • Andrew

    I live in Florida and want to stop Trump. Was planning to vote for Rubio (who is still technically ahead in the polls), but wonder if I should swap to Cruz. Basically I need to coordinate my vote with the rest of the stop-trump bloc. Any advice? (I recognize it’s a long shot in any case)

  • RodCrosby

    I found in 2008, 2012 and so far in 2016, a cubic relationship holds pretty good for WTA CDs, and a square law for the 2:1 CDs…

    • Sam Wang

      These relationships can basically be derived from first principles by knowing the district-to-district SD. Each relationship corresponds to a particular SD. Footnoted somewhere in my gerrymandering article. Also see Gudgin and Taylor’s book, back in the day.

  • Amitabh Lath

    Nice analysis. I am really impressed you went to all the trouble to Monte Carlo all the delegate selection rules for every state.

    Are there any large unpolled states where you can tweak the estimate by looking at neighboring states (or other demographic info) rather than national average? A first order correction, it would give a sense of what the sys. uncertainty is (smaller than Fig 3 RMS I would guess).

    PS: They fixed Fig 3, but now Fig 2 (Trump vs. Kasich vs. Cruz) is wrong (it is a repeat of Fig 3).

    • Sam Wang

      Delegate rules: I used an approximation that works pretty well, now that winner-take-all races are dominant. Assume district-to-district SD within a state is 5 percentage points, which combined with the statewide candidate percentage gives a distribution of outcomes. Then calculate the winner-take-all Congressional districts using a cumulative probability distribution. A shortcut, but it should give a result that is not far from the eventual outcome. It is much more compact than what I did pre-Super-Tuesday.

      The code looks like this:

      % Convert the CD-level WTA delegates into a seats-votes-like relationship
      % CD WTA states are MO WI CT MD IN WV CA; include loophole states IL, PA
      % note that Romney SD in 2012 in California was 3.6%, MAD-based SD 4.3%.
      % Assume SD of 5% across states. Use tcdf(margin/5,1) to make a long tail
      CDWTA=[2 3 9 11 13 14 16 18 21];
      CD_delegates=[54 40 24 15 24 54 27 9 159];