Princeton Election Consortium

A first draft of electoral history. Since 2004

Is Donald Trump Really The Frontrunner?

July 20th, 2015, 1:18pm by Sam Wang



Conventional horserace polls put him in first or second place. However, these single-choice polls are a terrible way to tell who is likely to make it in a field of 16 candidates.

Today in The New Republic, I argue that horserace polls are asking the wrong question. Instead, a more elaborate method – instant runoff voting* – may more accurately answer who is likely to survive the coming campaign. Read on!

*Postscript: I’ve been hearing from people who point out the problems with instant-runoff, both as an actual voting system and as a feasible polling technique. One possibility that may address both concerns is Bucklin voting.

Tags: 2016 Election

72 Comments so far ↓

  • Amitabh Lath

    What happened to the comments? This is like when the supermarket reorganizes and you have to walk through five aisles to find seltzer.

    For instance I didn’t see that Craigo had actually done the research on the sore loser law. Awesome!
    And followup: are there states that make it difficult/impossible to run in a party primary and simultaneously appear on the general election ballot? Perhaps by putting the registration deadlines on the same day?

    Maybe bks is right and there are no good commenting utilities. We should all just meet in a coffeehouse some evening and thrash this Trump thing out. We couldn’t do much worse than the commentariat.

    • Sam Wang

      Sorry, perhaps not a great experiment. One more try: reversing time order, keeping threads. If it’s too horrible we can go back to Coke Classic.

  • bks

    USENET software was so, so superior to Web 2.0 interfaces for threading and for keeping track of what was new. No modern commenting software does this well.

    It would be nice to have the ability to edit one’s comments.

    My real pet peeve is that no blogs put a time zone in the date stamp.

    I’d vote to return to the one-level threading from before, but I’m not religious about it.
    –bks

  • 538 Refugee

    Sam: The problem is that as these threads get longer, and in this case older, if someone wanders by and responds to something near the top it will go unnoticed by most of us. On sites I don’t read much I’ve gotten used to reading backwards and mentally rebuilding threads. Now is probably the time to get the input though. I’m sure others will have opinions. I can live with whatever works best for the most. Right now it’s a little jarring because people posted expecting to be threaded so references are somewhat lacking. Oldest first and quoting previous posts might be easiest to follow for most.

  • A New Jersey Farmer

    Fiorina got enough of a bounce to include her in the next debate. I don’t think it will last, but she will have a wider forum to make her case. Carson is far more popular with conservatives and will likely stay near the top of the polls.

    • Amitabh Lath

      Carson is a seventh day Adventist, and their belief is a literal 7 days of creation. No evolution, no big bang cosmology and certainly no 13+ gigayear universe.

      Imagine a guy like that in overall charge of the NSF, NIH, NASA budgets.

      Better Sarah Palin as Secretary of Energy. I wonder if she knows what the DoE actually does.

    • 538 Refugee

      At least she’ll have us all speaking “American” so we won’t have anymore of those metric/imperial type debacles.

    • Amitabh Lath

      Damn straight. Those of us with somewhat foreign sounding names should take extra precautions.

      So before she asks, here it is in proper American units:

      The higgs boson is 4.9 e-25 pounds (or more accurately 1.5e-26 slugs, but if one knows the difference between weight and mass one probably isn’t a real American).

  • A New Jersey Farmer

    I’m thinking that each of the GOP candidates will get their place in the sun this fall and then falter for one reason or another. Fiorina and Carson are there now. One of the debates will feature an all-out attack on Trump by the other nine on the stage and he’ll react badly.

    • bks

      Fiorina? Balderdash! In the last three national polls she was at 3% (YouGov), 4% (Reuters), 4% (Monmouth).
      –bks

    • Craigo

      Yeah, the Carson surge is real, and something that the Village media has ignored. The Fiorina surge is imaginary, and something that the Village media plays up.

    • Sam Wang

      Craigo: Where’s the like button on my blog?

    • 538 Refugee

      You can address that when you get the new software that doesn’t make us go back and look at the dates on a 200+ comment thread to see what’s new. ;)

    • Sam Wang

      538 Refugee: I confess I never notice the clunkiness of comments because I use the backend to read them.

      I have a limited number of options here. I just turned off threading and set “put newest on top.” Do you like that better?

      Downside: this requires you to address one another by name. I can try different versions of threading. Or…new software? Really???

  • Jon

    My hypothesis all along has been that Bernie is simply setting the table for Hillary (whether intentionally or not) by generating early excitement and commitment to the race among Democratic and other progressive activists, and that even as he gets demolished pretty much everywhere but Vermont and New Hampshire, he will have jumpstarted the mobilization of progressive activists (most of whom will stay engaged) and of Democratic messaging in this presidential cycle. Is there any way to check this hypothesis historically by reference to other progressive campaigns, like Jesse Jackson’s, etc?

  • bks

    On Trump signing GOP loyalty oath:

    As a special bonus, Trump also gets to define what “fairly” means. If at any time in coming weeks and months, Trump even so much as hints that Republicans are treating him unfairly, mass panic will again set in, without anyone even knowing by what objective metric “fair treatment of Trump” can even be gauged.

    –bks

    • Amitabh Lath

      I read in the Atlantic that it is difficult if not impossible to run as an independent if you fail in the primary for another party. Several states have “sore loser” laws to prevent just that, although I wasn’t clear how many apply to the presidential race. Also, some states may have the registration date for the general close to that of the primary.

      Anyway, Trump may have seen that the 3rd party threat did not hold up after closer inspection and made the best of it.

    • Craigo

      Amitabh – Last I checked only South Dakota, Texas, Mississippi and Ohio apply their sore-loser laws to presidential elections.

      In Ohio’s case, you can almost certainly access the ballot by gaining a minor party nomination. And none of them prevent a write-in candidacy.

  • bks

    Robert Reich thinks there’s something new under the sun:
    http://robertreich.org/post/125702366950
    –bks

    • Amitabh Lath

      I agree that Trump is sui generis, but Sanders is quite typical. He comes from a long line of populist Democratic challengers: Ted Kennedy, Jesse Jackson, Bill Bradley. Except he poses far less of a threat for the Democratic nomination than any of the above did.

      Recall that Jesse Jackson won 9 states, and had 1023 delegates to Dukakis’s 1792. Jackson posed a real threat. I doubt Sanders will win anywhere but Vermont.

    • Matt McIrvin

      I think there’s actually a good chance Sanders will win New Hampshire. It’ll create a flood of Hillary-in-trouble stories in the news. He’ll be crushed on Super Tuesday.

    • Craigo

      Even before Super Tuesday we have Nevada and South Carolina, neither of which he’ll win.

      But yeah – that looks like it’s going to be a bad day for Bernie:

      Alabama (nope), Arkansas (lol), Colorado (maybe), Georgia (not a chance), Massachusetts (dogfight), Minnesota (decent chance), North Carolina (nope), Oklahoma (are you kidding?), Tennessee (seriously, are you kidding?), Texas (not even funny anymore), Vermont (whoo hoo!), Kansas (shit).

      All but four states are dominated by white moderate Democrats, or Democrats of color. He gets to follow that up with Louisiana, Nebraska, and Kansas (again?!) before the friendlier confines of Maine a week later.

      Ouch.

    • Sam Wang

      It seems extremely unlikely that Sanders is going anywhere. The question is whether he ends up as a neutral or positive influence on the Democrats, for instance by bringing policy discussions into the news.

      In my view he is behaving as if he expects Hillary to be the nominee.

    • Craigo

      Agreed. Keeping her on her toes on policy and politics is a net benefit, and he’s shown no sign of wanting to damage her. None of the Democrats have gone cutthroat yet, in happy contrast to the GOP field.

  • Amitabh Lath

    New Fox News poll of LV: Trump is up 26 to 15 vs. Bush.

    So to answer the question at the title of this post: Trump may not be the frontrunner, but he’ll be playing one on TV.

  • 538 Refugee

    Trump is running first with independent-leaning Republican primary voters, winning 19 percent of their support. And he captures about a quarter of GOP voters who say that they could not see themselves supporting Jeb Bush for president in 2016.

    Those who say they consider themselves conservative are equally split between Trump and Walker, with about 18 percent backing each man.

    And Tea Party supporters most favor Ted Cruz (22 percent), with Trump (20 percent) and Walker (20 percent) close behind.

    http://www.nbcnews.com/meet-the-press/trump-surges-new-nbc-news-wsj-poll-n402036

    • Amitabh Lath

      N=252. So stat error is yoooge, 1/sqrt(N)= 6.3%.

      At the end of the article they quote “The margin of error for 252 interviews among Republican Primary Voters is ±6.17%”. I don’t understand the discrepancy, but nevermind, at least it is likely voters.

  • Amitabh Lath

    Off topic, but it appears the Trump name is an important one in our field. Donald Trump is the nephew of John Trump who, along with Robert Van de Graaff, developed the Van de Graaff generators.

    These were the first MeV electron accelerators, ancestors of modern particle accelerators.

  • bks

    Must read story about confusion at Fox about rules for 6 August debate:
    http://nymag.com/daily/intelligencer/2015/07/roger-ailes-primary.html

    –bks

  • 538 Refugee

    Well, name recognition IS the name of the game this time around it would seem.

    Pew claims that even the low response rate doesn’t indicate a bias between those willing versus those not willing to participate in polls. Not that correlation of events are always correct for the right reasons.

    http://www.pewresearch.org/fact-tank/2015/07/21/the-challenges-of-polling-when-fewer-people-are-available-to-be-polled/

    • Amitabh Lath

      Interesting read, esp the 2012 assessment of non-responders linked within. Pew’s argument seems to be that since non-responders (as determined from high response govt. surveys) match demographically to those that respond to polls, there isn’t any bias.

      This smacks of whistling past the graveyard. The items where there is agreement are broad (homeowner, have children, social security, unemployment). Questions that might go to politics/ideology (contacted public official, volunteered) show big disagreement.

      The basic assumption in this Pew assessment is that demographics == ideology and voting preference. I don’t know about that.

    • 538 Refugee

      Band/circus wagon (clown car?) effect? The interesting thing here is that this poll has him at about the same number for the primary and a general election 3 way. That would indicate he would take as many votes from Hillary as Bush? Of course “Wouldn’t vote” leads them all and the text indicates a 15% undecided.

    • Sam Wang

      No, the arithmetic is a little more involved than that. Voters are approximately split equally between GOP, Democratic, and “independent” (i.e. people with some loyalties but don’t accept being labeled). Trump takes more from Republicans than from Democrats. He is a problem for Republicans.

    • Amitabh Lath

      So either a) the polls are right and Trump ascendant, or b) the polls have Trump specific errors, say due to his celebrity status, or c) the instrument doing the measurement — random phone dialing in most cases — does not work in this era and needs to be discarded like bubble chambers.

      I would guess b), Trump’s wide name recognition is affecting the results.
      Note the large eye-popping leads are in the RV polls. The LV polls also show him leading but single digits. Would tighter LV filtering reduce or eliminate Trump’s lead?

  • bks

    The (not fully revealed) scheme for selecting the top ten GOP contenders for the 6 August Fox debate provides the candidates with a motivation to game the opinion polls, which devalues both the polls and the process.

    And as the mechanism is still obscure, the possiblility arises that Kasich, Graham and Santorum may yet be added to the Fox clown car by appropriate tweaking of the algorithm (margin of error, etc.) to force a “tie”. –bks

  • bks

    I think Nate Silver is wrong here:
    http://fivethirtyeight.com/features/donald-trump-is-the-worlds-greatest-troll/
    I’ve always hated the Internet usage of “troll”. Does Nate think that all the participants in Twitter are having calm reasoned discussions about the issues? Balderdash!

    All the candidates are trying to get media coverage. They know the rules of the game. Trump is doing it better than they are –that’s not trolling, that’s being highly effective.

    –bks

  • bks

    “The polls are skewed!”, say GOP pollsters
    :
    http://www.politico.com/story/2015/07/what-trump-bump-120693.html

    –bks

    • Amitabh Lath

      This is rich. Where was this fabulous analysis when Jeb was leading?

    • pechmerle

      The Republican campaign pollsters quoted in the article make several good points. Not using a likely voter screen at this point does indeed overstate support for some candidates vs others.

      And polling Iowa registered voters re who is likely to win the Iowa caucuses is, as someone in there notes, particularly far off the mark. Only 6% of registered Repubs in Iowa likely to attend & participate in the Iowa caucuses means the outcome of those caucuses is quite possibly not at all being captured by general polls.

      That these campaign pollsters have an ax to grind doesn’t mean that we shouldn’t pay attention when they make valid points.

      Of course, the influence of the Iowa caucuses is itself a blot on the political landscape, given how few people they actually represent, but that’s another story of its own.

  • bks

    Ed Rollins told H. Ross Perot that he could win the 1992 election if he was willing to spend $120,000,000 of his own money. Perot was not willing to go that far. I wonder how much of his personal fortune Trump is willing to risk?

    –bks

  • Amitabh Lath

    Yes, there are a couple of polls that have Trump still ahead. At this point it’s clear he will rack up some delegates even if he does not win any primaries or caucuses , since pre-March 14 they award delegates on a proportional basis.

    Given that he is running 2nd in quite a few states (and there is a different candidate in 1st in each state) Trump could rank pretty high in delegates near the start. Once the winner-take-all states come in, he will fall behind.

    Who does he release his delegates to? Probably not Bush or Walker. They could be a bonanza for a middle-tier candidate.

    The press does not seem to be doing any sort of math on this. They are far too busy doing neurology on Trump and his supporters.

    • Craigo

      I don’t blame the press for not working the delegate counts when no delegates will be awarded for six months. Summer primary polls have little predictive power.

  • pechmerle

    Latest Economist/YouGov poll (July 18-20) has Trump surging to 28%. But this poll did ask, who do think will win the nomination:

    “Despite the [28%] results, Republicans still expect Bush to eventually win the nomination.

    He was chosen as the most likely primary winner by 36 percent of respondents. Walker followed with 11 percent, then Trump with 10 percent.”

    • Craigo

      And as I mentioned above, the prediction number is the most important result, not the intention. One paper found that in over 75% of cases where the two results differed, the prediction number won out.

  • 538 Refugee

    You are going to find some belligerence in any group that will use something like this to let out frustration. This could be similar to a batter slowly charging the mound to give his team mates adequate time to stop him. This doesn’t necessarily translate into votes when it comes right down to making the actual decision.

    What this may be showing is the number of disenfranchised in the Republican party. The Tea Pary wing is still alive and ‘well’?. This could be a very scary number for Republicans.

    I see George Will had the same idea as me. Blame it on the Democrats.

    http://www.realclearpolitics.com/video/2015/07/05/brit_hume_if_donald_trump_was_a_democratic_mole_would_his_behavior_be_any_different.html

    • Amitabh Lath

      Is there any data showing that Tea Party voters are responsible for the Trump surge? That ABC News/Washington Post poll:

      http://www.langerresearch.com/wp-content/uploads/1170a22016Politics.pdf

      shows him at 25% with moderates and 17% with “very conservatives”.

      Also, people who completed college gave him 8%, but he is at 30%+ with less educated.

      I don’t know if “moderate” and “high school education” is congruent with the Tea Party. Too bad they didn’t ask if they are frequent watchers of reality TV.

    • bks

      I’m having trouble convincing myself that Trump is more of a joke than the rest of the GOP candidates despite that being the prevailing wisdom. On social issues Trump is a huge win over the field. That’s not to say that Sam’s argument-from-unfavorables is wrong, just that the rest of the field are pygmies in their own right. Paul, Cruz, Santorum, Huckabee, Kasich, Perry, Carson, and Walker are totally whack.

      –bks

  • Amitabh Lath

    A hypothesis as to the sudden rise in Trump’s poll numbers: A (false) rise in the number of likely voters answering pollster questions.

    I imagine a regular primary poll this far before elections has a very small efficiency in finding respondents, especially ones that pass LV filters. So N (the number of people called) has to be huge, so that N*eff (the number of people willing to talk) is statistically significant.

    Now Trump enters the race. He has a recognition factor orders of magnitude above any of the other candidates (The Apprentice may not be a huge hit by Nielsen standards, but its audience probably dwarfs the number of people who would recognize Jeb on the street).

    So now N can be a lot lower to get at the same N*eff, and if you do not account for the spurious increase in efficiency due to Trump’s celebrity status, you are going to get an artificial bump.

  • bks

    On 21 July 2011 Romney was comfortably in the lead in a WaPo national poll of registered GOP voters. Later Perry and Cain came to the top, but Perry self-destructed and Cain ran out of money. Trump will not have an “Oops!” moment (he lives in front of a camera) and he will not run out of money. Limbaugh has come to Trump’s defense (in 1992 Limbaugh sabotaged H. Ross Perot so keep an eye out for the long knives).

    I’m headed to Costco to make sure I have enough popcorn for the 6 August debate. –bks

  • Craigo

    There’s an even simpler method.

    Ask respondents who they think will win, not who they are going to vote for. Most people, surprisingly, will bet with their heads and not their hearts.

  • pechmerle

    “Using this information, I came up with a simple measure of candidates’ hidden support: a “second-choice ratio,” defined as the number of second-preference responses divided by first-preference responses.”

    An example of this exercise — either hypothetical or using one of the actual candidates — would have been useful for the reader at that point in the article.

    “instant-runoff voting in the 2020 Republican primaries” — overwhelming odds that this doesn’t happen, useful as it is to try to get it into the national conversation.

  • A New Jersey Farmer

    Looks good Sam. It will be interesting to see what Trump’s comments about McCain do to his poll numbers. I expect that Trump will qualify to be in the first GOP debate, and so will Christie.

  • Amitabh Lath

    I’m having a tough time seeing how a series of single-choice polls could all be systematically wrong in one direction (and how instant-runoff polling would mitigate that). Perhaps a Monte Carlo?

    • Sam Wang

      The point is not that the polls have made a bad measurement. The point is that they have made the wrong measurement. Single-choice polls are a bad way to measure who is likely to make it in a field of 16 candidates.

    • Amitabh Lath

      Okay I think I see. If we had better measurements of this system we would see a few giants (Bush, Walker, Rubio), a few smaller rocks (Paul, Cruz) and some safely ignorable Kuiper-belt dwarfs (Fiorina, Pataki).

      Trump was thought to be one of these tiny motes, on a highly eccentric orbit. However he seems to be occluding the Jovians. Therefore the distrust of the instruments doing the measurements. Or maybe we don’t understand the dynamics of this system like we thought we did.