Princeton Election Consortium

A first draft of electoral history. Since 2004

Primary discussion thread – All About Kasich edition

March 15th, 2016, 6:13pm by Sam Wang


Tonight, on the GOP side: Florida (99 delegates, Trump by +19%), Illinois (52 delegates, Trump +8.5%), Missouri (52 delegates, Trump +7%), North Carolina (72 delegates, Trump 46%, Cruz 30.5%, Kasich 11.5%, Rubio 7.5%), Ohio (66 delegates, Kasich +5.5%). Except for North Carolina, these states are all either winner-take-all, or nearly so. The leaders are favored to win, with the highest possibility for upsets in Illinois and/or Missouri (124 delegates total) for Ted Cruz. Under these projections, this suggests an evening’s delegate haul of Trump 258 (but as low as 134), Kasich 75, Cruz 23 (but as high as 147), and Rubio 6.

On the Democratic side: based on polls, Hillary Clinton should get approximately 450 delegates, Bernie Sanders 343 delegates. Details: Florida (246 delegates, Clinton +25%), Illinois (182, Clinton +1%, unusually high variability in polls), Missouri (84, Sanders +1%), N.C. (121, Clinton +19.5%), Ohio (160, Clinton +9%). Something’s weird in the Illinois data – maybe pollsters had to adjust their turnout models in response to the surprise result in Michigan last week.

The likely outcomes above are basically more of the same, i.e. Trump and Clinton widening their leads.

In the face of this, is there any possible way to stop Trump? Actually, yes…but only if John Kasich does something totally counterintuitive. Read on…

Yesterday I wrote that if Governor John Kasich wins his home state of Ohio, that’s – paradoxically – good news for Donald Trump. Since many upcoming primaries are winner-take-all, a three-way race between Trump, Kasich, and Senator Ted Cruz would favor Trump, who is in the lead with 30-40% in most states. If these polling conditions persist, Trump has a good shot at getting to 1,237 delegates, enough to win on the first ballot at the convention. In short, Kasich’s win may look good, but in the long run it’s good for Trump.

However, my reasoning followed standard political logic: if Kasich wins Ohio, he should stay in the race. If he loses Ohio, he should get out. But the consequences look like this. Note how good things look for Trump in the left-hand graph.

A lot of that histogram is close to the red line. Even though Trump is unlikely (given current data) to get 1,237 delegates on his own steam, recall that there are over 100 uncommitted and dropped-out-candidate delegates. They would need to be wooed and recruited.

But what if Kasich wins, and then decides to stay out of winner-take-all states?* Then we get this. The right-hand outcome is optimally bad for Trump.

That outcome could hold Trump to about 40% of delegates – and therefore an open convention. Cruz would also have approximately 40% of delegates, with Kasich and Rubio holding most of the remaining 20%.

This strategy is optimal from a game-theoretic standpoint: it maximizes Kasich’s leverage, while minimizing Trump’s delegates. And it gives Kasich quite a lot of leverage at the convention. The main downside of this plan is that it is risky. What if Trump pulls ahead of Cruz? Also, it gives a lot of influence to Cruz – and Kasich wants the nomination for himself.

An alternative plan would involve Kasich staying in the race, and increasing in popularity enough to eclipse Cruz. This doesn’t seem likely.

An open convention provides new opportunities for Kasich (and Rubio, for that matter). Delegates are often more loyal to their state party organization than to their candidate, making them open to persuasion. Delegates are also only bound on the first ballot, and can vote however they like on later ballots. They are also free to vote as they like on rules and procedures. See Sasha Issenberg’s rundown of convention tricks.

Assuming you buy my pretzel logic, now I want to know if Kasich is willing to go down this road. There are members of the Republican Party who say they’re desperate to stop Trump. Are they desperate enough to try this play?

*Update: If we were to really optimize this strategy to the max, Kasich would concentrate all his resources on non-winner-take-all states. They are: Hawaii, Utah, New York, Connecticut, Rhode Island, Oregon, Washington, and New Mexico.

Tags: 2016 Election · President

64 Comments so far ↓

  • counsellorben

    One correction. Pennsylvania is not a winner-take-all state for the Republican primary. Pennsylvania has a “loophole” primary, with delegates on the ballot for each congressional district.

    In fact, most of the remaining primaries are not winner-take-all. However, because the statewide winner in most Republican primaries does receive extra delegates, Sam’s theory proposing Kasich drop out of certain states has merit, especially for the true winner-take-all states (AZ, DE, MT, NE, NJ, SD, WV).

    The calculus is more involved in the other remaining states. Certainly, where Kasich dropping out of a state is highly likely to tip a state to Cruz, it is is Kasich’s best interest at this time to do so. Such states are most likely CO, ND, UT.

    Otherwise, most of the remaining states are likely to be more favorable to Kasich than Cruz (especially a state like MD, and similar states where Kasich can campaign as the most palatable alternative).

    In summary, as the delegate allocation rules for each Republican primary/caucus are byzantine, the optimal strategy for Kasich is to exit only those states where he has little likelihood of winning, and there is a high likelihood his exit tips the state to Cruz.

    • counsellorben

      I will backtrack a bit, as even though DE is winner-take-all, it is a state which likely is more favorable to Kasich, and dropping out is unlikely to tip the state to Cruz. The same applies to NJ as well.

    • Joseph

      The larger question is, what is Mr. Kasich’s long term game plan? I’d imagine he has to be thinking four years out. Mr. Kasich will want to come out of this in the strongest possible position in order to solidify consideration for; (1) a vice presidential spot with Mr. Trump and a possible role as an actual Vice President, or (2) the Republican nominee in 2020.

      So I agree that he will stay in to the bitter end.

      Will he drop out of states where he has zero chance of winning a delegate, but some chance of pulling votes away from Mr. Trump? Not necessarily. He might be seen as alienating Mr. Trump if he does so. OTOH, he might be seen as alienating the GOP leadership if he doesn’t.

      What about states where he can’t win delegates but could pull votes away from Mr. Cruz (assuming those exist)? I’d expect that, if he really is playing the long game, he’d stay in those.

  • Phoenix Woman

    And Missouri’s been officially called for Clinton.

    Clean sweep!

  • Amitabh Lath

    Now that “contested convention” has gone from being a risible hail-mary to the last best hope of stopping Trump, how do proponents plan to persuade Trump delegates to switch to them and prevent theirs from switching to Trump? The former has to happen at much, much higher rates than the latter.

    Perhaps there is a plan. But if all delegates have an equal probability of switching (flat prior) then Trump wins eventually.

    This reminds me of students who blow off homework, flunk quizzes, tank midterms, but continue to believe they have a chance because somehow they will do awesome on the final exam and pass. Mathematically possible, but has that ever worked?
    (Sam, this type may not exist in the Princeton biology dept.).

    • Rieux

      To my understanding, a large proportion of RNC delegates will be committed to “their” candidates by rule only–that is, they will be required by convention rules to vote for a particular candidate on the first ballot, but in many cases (HOW many is obviously a crucial question) will actually not be supporters of that candidate at all.

      Here’s a Bloomberg article documenting attempts by the Cruz campaign and other party actors to install their own loyalists as RNC delegates–including, vitally, delegates who will be bound to cast a first-ballot vote for Trump:
      http://www.bloomberg.com/politics/features/2016-03-14/how-to-steal-a-nomination-from-donald-trump

    • 538 Refugee

      The pull of political pork is strong. That’s why people gravitate to the ‘winners’. The loser can’t give you a plush appointment. A brokered convention would be the first true test of Trump’s self professed ‘awesome’ political deal making abilities. The next Surgeon General, Ben Carson, claims Trump is a very smart, reasonable guy when the cameras aren’t on him. Working the back rooms of a brokered convention should be no problem for Trump.

    • Roke

      As far as I know the delegates are mostly chosen by the local republican parties. So the loyalty towards their nominee is – at least – questionable. And when enough delegates are of questionable loyalty they can choose whomever they want (Ryan, etc).

      But I must admit these kind of plans seem like an Hail Mary for the republican establishment. And I wouldn’t rate the success chances very high.

    • Amitabh Lath

      Local Republican officials might not want to cross local Trump voters just to score brownie points with National Republicans. Ryan is only “local” in WI, after all.

    • MAT

      Amitabh,

      Most delegate types I’ve been around are rank and file party members, those county chairpersons and others who faithfully go year in and year out to monthly district meetings, sit thru endless delegate election contests and argue about party platforms. In other words, exactly the type of people who aren’t likely to be Johnny come lately Trump supporters. Getting delegates elected up thru a multi level process requires knowledge of the rules and organization – something that the Cruz folks have demonstrated.

      Of course, Trump can start promising delegates free Vegas stays, so it’s going to be interesting.

    • Lorem

      I think talking about the probabilities of delegates switching might not be the most useful thing to do, since the probabilities are anything but independent.

      In fact, apparently bribery is allowed, so if you were a billionaire interested in causing one of the biggest scandals in American history, you could walk in there, and brazenly offer each delegate a million dollars to vote for you. I suspect that would secure a majority. Of course, I doubt there’s any way to spin that which would make the general election anything but an automatic loss, but it seems probable that some more subtle bribes could be given.

      Incidentally, in the past, I’ve passed (serious) courses after outright skipping everything before the final exams, though I can’t say it was a terribly good idea.

    • Matt McIrvin

      I just heard that Bernie Sanders’ campaign manager is starting to talk about trying to flip superdelegates. I guess this is what you do when things start to look bleak.

    • Amitabh Lath

      True, bribery would be difficult to model in assigning p-values for delegate switches. But to zeroth order one could estimate the amount of bribery going on by observing how delegate switching differs from that expected of a level playing field.

      Lorem, I suspect the people running the GOP are not nearly as capable as you (or other folks here on the PEC). They all read “The Party Decides” and cheerfully ignored impending Trump.

    • Jim H

      Maddow had a great piece tonight on a bit skullduggery at the last Republican convention in which the Ron Paul camp managed to plant delegates in numbers far exceeding his vote share for several states. The Republican establishment has taken note and things may get crazy this year.

    • Sam Wang

      Many of those loopholes have been closed this year.

  • Kevin

    Do the deltas on the polling in the big March 15 states have anything to tell us about #nevertrump? It seems that Marco Rubio lost about half his support outside of Florida during that time (he should lose the rest now). While this tide lifted the boats of Kasich and Cruz, did Trump share unequally in the benefit?

  • Petey

    Only minor note to an other fine piece of analysis by Sam:

    “Delegates are often more loyal to their state party organization than to their candidate, making them open to persuasion.”

    This is new. This is not usual order of things. This is the first time in the open-primary era in either party where delegates slates are chosen by the state Parties, rather than by the candidates or by state conventions.

    (All due to a rule change in 2012 engineered and rammed through by Ben Ginsberg to prevent a Ron Paul insurgency against President Romney.)

    Thus, while I entirely agree that officially ‘bound delegates’ are likely to have incredibly weak allegiance to their candidate, it’s worth noting this is totally uncharted territory. We’ve never seen anything like this before in the open-primary era.

    (In the crucial ‘free votes’ in the ’76 GOP convention, and ’80 Dem convention, ‘bound delegates’ had very high allegiance to their candidate. But that was under the old normal delegate slate selection rules, not these new ‘everyone is a pseudo-superdelegate’ rules.)

  • Tim in CA

    Kasich could conceivably compete and possibly win Maryland, where I was born and raised, as well. Maryland leans very heavily toward the Democrats, but it also has a history of occasionally electing very moderate Republicans to statewide office. Kasich would seem to fit that bill, and he mentioned in his victory speech tonight something about “heading out to the Eastern Shore [of the Chesapeake Bay]” to campaign in Maryland. Not something you usually hear from POTUS candidates. Maryland also has the other half of the DC suburbs, the MD counterpart to the northern Virginia suburbs where Rubio performed well. Now that he is gone, I would expect most of those voters to migrate to Kasich. Something to keep in mind.

    • Steve

      I question whether Kasich is “very moderate”, although he is probably the most moderate among the Republican candidates. He just defunded Planned Parenthood in Ohio, and in 2011 he signed a 20-week abortion ban that didn’t have any exemptions for rape or incest. He is against unions, wants to cut Social Security, is against regulations. His tax cut proposal takes the top marginal rate from 39.6% down to 28%.

      Anyway, as he heads 1000 miles East to Maryland, I will be heading 6 miles East to NKU this afternoon.

    • Eric

      But Maryland has a closed primary, and Gov. Larry Hogan (R) was elected due to low turnout, and significant numbers of Democrats crossing over in the general election.

    • Matt McIrvin

      When the frontrunner seems completely unhinged, it moves the window of what is considered “moderate” further and further to the extreme. I could imagine a lot of Democratic-leaning centrists voting for Mitt Romney or John Kasich this year just out of relief that he’s not Donald Trump or Ted Cruz.

  • Doctor Science

    Looks like Hills overperformed in OH, under in NC, can’t tell about the others yet. Her delegate count will be *robust*.

    • Matt McIrvin

      With more of the vote in, it looks like she didn’t overperform all that hugely in OH. Sanders did modestly overperform in NC and IL.

    • Matt McIrvin

      …All in all, it’s a night of relative polling successes. Nothing as surprising as Michigan.

    • Adam

      Very cool stuff Sam. I saw your analysis pop up on the prediction market forums. This is very insightful. I have predicted Trump will win it all since July, but that is just a gut feeling haha. Thanks for sharing!

  • Art Brown

    Real Clear Politics shows Trump half way to a majority (without Missouri).

    • Art Brown

      Trump has 47% of the delegates awarded to date, I think. (Just summing up all the candidate’s delegates on RCP.)

    • Art Brown

      Hmm, Trump has only 41% of the total delegates voted to date (plus whatever he wins in Missouri).

      Since no one has a majority of the votes cast “for active candidates” in Missouri, 5 delegates will be awarded for each Congressional district plus 12 for winning the overall vote (per the Southeast Missourian). I can’t find anyone tracking that way.

  • Greg

    Sam, last week you were giving Trump 90 percent odds to get a majority of delegates. These models show something very different. What changed?

    • Sam Wang

      I don’t think anything has changed. If he’s a little below 1,237, he’d have to pick up a few delegates. This is all in the Prospect piece.

    • timothy

      for the non-math experts…Sam please clarify your above statement..so you think if he’s below 1237 its still 90% because he is very likely indeed able to get those “few delegates” ?
      and what are your thoughts on likelihood of trump winning the a contested convention given the likelihood of the number of delegates you think he’d have if thgat came about?

    • Matt McIrvin

      If there’s actually a contested convention, we enter a regime in which opinion poll analysis probably won’t tell you anything directly…

  • James burgess

    Some states lock delegates to the election results for multiple votes at convention. Is Kasich on the ballot in PA?

  • A New Jersey Farmer

    I agree that Kasich’s win helps Trump because it creates a three person race. Where else could Kasich win? He’s lost the states around Ohio already and is trailing in many more. In addition to Trump running an effective campaign, he’s been extraordinarily lucky that the Party has been late to the party.

    • 538 Refugee

      Kasich was born in PA but he is in danger of getting booted from the ballot there. His petitions were certified but challenged by Rubio. Turns out Kasich didn’t have enough valid signatures. Rubio’s people file online. The office closes at 5. They filed at 5:13. They are debating if this is a technicality since the papers were filed within the specified time frame and the time the office closes should not be relevant.

      It is unclear if Rubio can now drop the challenge since he would have no personal interest in seeing it through. Thing is, can the state just ignore it now that it is shown Kasich didn’t submit enough valid signatures? Who knows. It’s a ‘law thing’.

  • Marc

    I’m a Democrat, and I’m not sure who to root for in the Republican primary. Current national polling suggests that Clinton has the widest lead over Trump in the general, compared to running against the other Republicans. Or, @MAT’s third party option would also weaken Trump in the general.

    Maybe let Trump be the nominee because he is more likely than Cruz or Kasich to loose the general.

    • Marc

      Of course there is no state-level polling for the general yet. And, any polls this early are not necessarily reliable.

    • Kevin

      I don’t think your rooting is going to matter. It’s Trump, Trump, Trump.

      Kasich during his “victory” speech: “We’re going to go all the way to Cleveland and secure the Republican nomination.”

    • Matt McIrvin

      There actually is some state-level polling for the general. Wikipedia has it here. It can be a little sparse and stale, though.

    • Sam Wang

      For the general election, Clinton v. Trump data is here. There’s quite a bit of it – six surveys since March 1st.

    • Craigo

      General election polls have no significance at this date.

    • Sam Wang

      Actually, they might be getting close to interesting. Needs analysis.

    • Matt McIrvin

      The ones on Huffington Post are national-popular-vote surveys, though, with all the caveats that apply to those.

      For a while it seemed like the Clinton v. Cruz and Clinton v. Rubio ones had a starkly gigantic split between online and phone polls (Clinton winning by huge margins online, but losing narrowly on the phone) that had me wondering whether I should be looking at them at all.

  • public_editor

    Please fix the second image. It’s presently the same as the first. Very curious!

    • Sam Wang

      Fixed, thanks. Basically the two Trump v. Cruz graphs should be different to account for the fact that in the first one, Kasich does get any Ohio delegates. In the second one (with the red diagonal arrow), the condition is a Kasich win in Ohio.

  • JayBoy2k

    Fox has called NC for Trump and CNN has called Illinois for Trump. If you just bet the favorites, you would be 4 for 4 with Missouri left to go.
    Sam,
    I find it impossible to believe that Kasich would follow the GOP Establishment down the rat hole. He just won his first state and he is going to take a dive and a hit to his integrity with the people who would vote for him.
    Better to just run his race, thinking he and Cruz will win enough to keep Trump from the goal line. He is still on a path to a brokered convention win and gets to stay away from the GOP Establishment -

  • bks

    Ironically, if we trust Sam’s logic, Cruz is telling Kasich to get out, and Kasich is saying he’s staying in to the end.

  • Doctor Science

    So it looks like Hillary did better than Sam predicted in FL, more like 30% than 25%.

  • Sichu Lu

    With Rubio dropping out, what happens to his delegated?

    • Sam Wang

      Depends on the state-by-state detailed rules. In some cases they are still bound to him, in other cases they are liberated to vote as they please on the first ballot. There might even be cases where they have to do what he says (I forget).

    • MAT

      He didn’t drop out, he suspended. His name will remain on ballots, he keeps the delegates. Technically, he’s still running

    • MAT

      On further research, Sam’s right about the Rubio delegates & I’m wrong.

  • Mark F.

    Rubio drops out.

  • Art Brown

    Re the Issenberg article, it is silent on what “tricks” Trump might employ if he is short a majority of delegates. You know he will be looking hard to make a deal.

  • MAT

    In the meantime, while we are waiting on results, one of my longstanding theories shows some signs of life:

    http://www.politico.com/story/2016/03/top-conservatives-gather-to-plot-third-party-run-against-trump-220786?lo=ut_a1

  • amit

    Repeat of issue from previous thread: the voter assignment from Rubio and Kasich to Cruz assumes no abstentions.

    Maybe that’s how it goes down, but given that end of Rubio and Kasich signals the closing of the establishment lane, there may be establishment-minded voters who cannot stomach either Trump or Cruz and stay home.

    At the very least, this adds a little uncertainty, increasing the RMS of the plots.

  • MAT

    Wow, this is nice work.

  • Amitabh Lath

    Quite the opposite: Rubio says that he will stay in even after losing Florida.

    Lots has been written about Republicans not being very good at collective action that requires individual sacrifice for the common good.

  • Art Brown

    How can “Kasich wins” or “Kasich loses” both end up at the same graph? If Kasich loses, Trump has 66 more delegates…

    I also don’t understand the Trump vs. Cruz cases. In the past you’ve written that Trump’s support improves in the remaining primaries relative to Cruz’s. Did I miss an update?

    • Amitabh Lath

      The mean of the lower right-hand graph is presumably smaller, by the number of OH delegates.

      Sam, you should turn on the printing of stats ( and RMS) on these plots. Otherwise to the eye they do look the same.

    • amit

      That should be “mean and RMS”. The top right plot and the bottom right plot should differ by 66, the number of OH delegates. The scatter should be the same.