Princeton Election Consortium

A first draft of electoral history. Since 2004

Confidence is associated with increased turnout

November 5th, 2016, 10:40pm by Sam Wang


I’m getting mail claiming that when voters are sure their candidate will win, they are less likely to vote. Therefore (these are Democrats writing) I should pipe down. However, this speculation contradicts both human nature and empirical evidence.

First, think about why we vote. Since a single vote basically never swings a race, the rational argument for voting is not strong. Instead, we vote because it is our duty, because we build the habit over time, and because voting makes us feel good. In light of that, the obvious consequence of supporting a winner is increased likelihood of voting – there’s more emotional reward.

Now, some evidence. Consider this post of mine from 2008, The Exuberance of Likelier Voters.

This graph shows outcomes plotted as a function of polling margins. If polls were accurate, the slope of the green line would be 1 with an intercept of 0. But it is not. Indeed, for every 1% of actual margin, only 0.84+/-0.03% is captured in polls*. This underperformance implies that there is a hidden bonus for whoever is ahead. The intercept essentially goes through the origin, so there’s no overall bias toward either candidate.

It seems that this phenomenon could be caused in three ways, not mutually exclusive:

  1. If you live in a state where you are certain that your candidate will win, you are more likely to vote than predicted by likely-voter screens.
  2. In the same state, if you support the losing candidate, you are less likely to vote than predicted by likely-voter screens.

Now, I should say that it’s not clear that this idea applies to a national race. But my point is that we don’t have evidence that confidence is a turnout-killer.

Finally, consider the converse. When Donald Trump claims that the election is “rigged,” that tells his supporters that he is doomed. Based on the state-level data above, this might depress turnout.

Tags: 2016 Election · President

86 Comments so far ↓

  • Newton123

    Could Sam please comment on this perhaps unbelievable item from Upshot’s latest NC poll (Note that likely well over 60% of votes are already in):

    >>Mrs. Clinton leads among voters who have already cast their ballot by a nine-point margin, 49 percent to 40 percent. But Mr. Trump leads among those yet to vote by 17 points, 52 percent to 35 percent.<<

    Full article here:
    http://www.nytimes.com/2016/11/08/upshot/trump-and-clinton-tied-in-final-upshot-poll-of-north-carolina.html

    Also see: http://www.nytimes.com/interactive/2016/upshot/north-carolina-early-vote-tracker.html?_r=2

  • TK1

    This certainly makes intuitive sense. Why else would politicians always claim that they’re going to win right before elections, no matter how much the polling indicates otherwise?

  • Markus Meister

    I’m new to this blog, so pardon me if this thought has appeared many times before. Nate Silver is now a professional forecaster. He has a lot to lose from being spectacularly wrong. So even if he believed that Trump’s chances are 5%, announcing that would have a 1 in 20 chance of being a career-ending move. I suspect that’s why he’s dialing his predictions back close to 50:50. Then he can’t be accused of getting it wrong regardless of the outcome. Sam Wang gets paid for teaching neuroscience classes. He has no such conflict of interest.

  • T Chase

    I study prophecies, and it is clear from a prophecy of Nostradamus, Astrology, and the King James Bible Code, that Hillary will win. I have been saying this for years on my website and videos. A Nostradamus prophecy from 1500s France predicts Hillary will win because “Florida will hold” for her. (Nostradamus spelled it “Flora”).
    A key Astrology sign is that Mars goes into Aquarius on election day. See: http://revelation13.net/calendar.html

    • 538 Refugee

      wow² If New Jerusalem is being towed to Earth by three giant UFO’s won’t they need a President Trump to do the real estate deal? (front page stuff)

    • Arun

      All that fuss about probability generating functions, random drift, and complex interactions could have been avoided if only we had properly interpreted Nostradamus.

  • EDYS

    As a MA voter — *bluest state in the union* — I should be one of these unmotivated voter. Nothing could be further from the truth. I voted early, plus volunteered for Hillary’s GOTV in NH. Never done either before. The threat of a psychopath such as Trump has me motivated to the max. Every vote for Hillary slams the door harder in his sick face. No doubt the brainwashed white folk who adore this creep feel equally motivated. I truely believe this election involves a unique level of motivating partisan rancor.

  • EDYS

    As a MA voter — *bluest state in the union* — I should be one of these unmotivated voter. Nothing could be further from the truth. I voted early, plus volunteered for Hillary’s GOTV in NH. Never done either before. The threat of a psychopath such as Trump has me motivated to the max. Every vote for Hillary slams the door harder in his sick face. No doubt the brainwashed white folk who adore this creep feel equally motivated. I truely believe this election involves a unique level of motivating partisan rancor.

  • Michael

    Sam, For the non-math majors among us, please explain what you mean by a systematic polling error being on the “edge.” Do you mean on the edge of a margin, or that the polling would now be in the “edge,” meaning exceptionally close — a “knife’s edge.” Bottom line: can you say how likely a systematic polling error is, and how it would affect your calculations?

  • jennifer

    just want to say that I have really enjoyed your scientific approach to the election. I literally check this site every day for math/facts vs. clickbait and everything else out there for the election. Thank you for being a beacon of fact in this frustrating slew of misinformation that is election 2016. I think I missed my calling- i should’ve been a statistician!

  • Lorem

    You say “could be caused in three ways” and then only list two, and there’s a “*” link that seems to go nowhere. I am mildly confused.

  • Anonymously

    Nate Silver gets into it with Ryan Grim who said on Huffpo,

    “Silver’s guess that the race is up for grabs might be a completely reasonable assertion ― but it’s the stuff of punditry, not mathematical forecasting.”

    Silver: “This article is so f*cking idiotic and irresponsible.”

    http://www.mediaite.com/online/nate-silver-goes-to-war-with-huffpost-writer-after-highly-critical-column/

    And a shot at some Princeton prof?

    “There’s a reasonable range of disagreement. But a model showing Clinton at 98% or 99% is not defensible based on the empirical evidence.”

    • Ed Wittens Cat

      Nate began his transition to pundit in 2012– with pandering.
      He disappeared a post on why “dem oversampling” in polls was bunk and stood up for climate change denialists in his book.
      but his “no one really knows” schtick is based on how WRONG he was abt Trump securing the GOP nom
      https://fivethirtyeight.com/features/donald-trumps-six-stages-of-doom/
      Dr Wang got it correct, because he was Seeing the Math.
      not his emotions.

    • Jim Wright

      Pretty good barometer of the competing models will be Tuesday. 538 has Trump winning NV, FL, OH, and NC, even this morning. I can’t imagine there will be too many more polls to put into the hopper.

      Most of the other services think NV and NC, at the least, are blue, and then some also include FL and OH. In past years, while the confidence may have differed, the electoral map wasn’t too different between models.

      I think Tuesday will be a lot different.

    • Michael Coppola

      I think that’s more likely a shot at Huffpo. It makes little sense as a shot at Sam. PEC’s model is completely transparent, and clearly based on empirical evidence. Silver’s actual criticism of the PEC model would basically boil down to “there’s more to it than that” and would fit into a tweet just fine. But it really doesn’t fit into a Twitter rant at someone other than Sam.

    • Randy Haugen

      That other guy is under a lot of stress.

    • Ed Wittens Cat

      Just watched Enten and Cohn on Fareed– zero mention that Huffpo and Linzer are much closer to PEC than they are to 538.
      But both agreed that if HRC gets NC or FL the election is over.
      No mention of the core difference– parametrics v nonparametrics– the mean is quite sensitive to changes– the median less so– isnt that simple enough for the average American citizen to grasp?

    • Ed Wittens Cat

      Wow…read the chirpstory.
      Never seen Nate so hostile and defensive.
      Male Pattern Baldness is Bad, so Bad.

    • Ed Wittens Cat

      good grief.
      comparing Nate to Dean Chambers is a pretty low blow.
      headline:
      Nate Silver is Unskewing Polls– All of Them– In Donald Trumps Direction
      http://www.huffingtonpost.com/entry/nate-silver-election-forecast_us_581e1c33e4b0d9ce6fbc6f7f
      ;;ouch;;

  • Ryan Pearman

    99% or 95%

    Sam — I see the election probability at 99% Clinton — even assuming random movement.

    Yet, your simulation / distribution of electoral college results shows a Trump win clearly in the blue, which would suggest a larger than 2.5% probability *today* of a Trump victory.

    That does not seem consistent with the banner result, unless the banner result is the probability of the meta margin reaching 0… which would mean there is a <1% chance that Trump has a 50% chance of winning. (that would feel right to me looking at the data.) But that is not the chance Trump currently has to win, and would in fact under-estimate it.

    What is the difference?

    (ps, good job advocating for the real race in the senate!)

    • Scot Gould

      You are not the only one who has noticed this. However, where it turns green, this is at the 97.5% point. I now suspect it is simply a rounding of the display, in that given that each green bar must possess a minimal width, the bars overstate the location of the 97.5% mark. It was more prominent last night.

    • Sam Wang

      Short answer: the tail of the snapshot is not how the win probability is calculated.

  • Runner

    It always amazes me that any citizen would choose not to vote, regardless of which side you are on.

    What is the estimate of the % of eligible voters who will vote in the Presidential election this year? Is the % turnout estimated to be higher or lower than recent Presidential elections?

  • Justin Jolly

    When I took statistics in college I did miserably. Can someone explain to me, in terms Trump could understand, what the meta-margin is?

    • Michael

      The meta-margin is the amount by which the state polling median would have to change for the race to be a statistical tie.

  • Chas

    What Dr. Wang poses seems to me to be a behavioral economics question. Seems like this this is a question to pose to Prof. Dan Ariely.

  • Betsy Teutsch

    Re: one vote not mattering…. Those Floridians who didn’t bother to vote but leaned Gore — would love to hear about them. Wonder if anyone has tried to find Floridians who take voting more seriously because of the Bush/Gore debacle, complicated by th

  • Bill

    Off topic, Speaking of Nate Silver- As a reader happy to see some state specific divergence between the different methodologies. It’s easy to argue everyone was right when PEC had a 90%+ chance of a Clinton win and 538 had a 65% chance. With 538 currently listing NV, NC, FL and OH as Trump wins, we will soon see when those %’s become binary predictions by state.

    Anyone- as I understand, 538 applies a national trend adjustment to its state level polling results, doesn’t that pose a risk of capturing the same movement twice if the state poll has already captured whatever movement is in the national?

    • Arun

      Your last statement is exactly correct. I do believe he is making the assumption that state polling needs correction, not just from national polls, but also other state polls. All of this results in an intricately correlated and complex system. You’ll see big jumps in win probability from obscure polls conducted in deep blue or red states. In the 2012 presidential election, the average of national polls gave Obama a very small advantage, as they do to Clinton this year. The state polls gave Obama a very clear advantage. Similar story this year, though the state polls are closer.

    • Bill

      Arun-

      Thank you, yes now I see both from 538
      “Polls are adjusted based on this regression. For instance, if Trump led in a North Carolina poll by 1 percentage point in June, but the trend line shows him having gained 3 percentage points nationally since then, the model will treat the poll as showing him up by 4 percentage points. This calculation varies slightly from state to state based on a state’s elasticity score.”

      The state adjustments-

      “The error from state to state is correlated. If Trump significantly beats his polls in Ohio, he’ll probably do so in Pennsylvania also. Figuring out how to account for these correlations is tricky, but you shouldn’t put too much stock in models that don’t attempt to do so. They’ll underestimate the chances for the trailing candidate if they assume that states are independent from one another.”

      Interesting

    • Arun

      Another small point. NC, NV, FL are all in the 60-70% probability range. Losing any of them individually would not be an uncommon event at all. Losing all three of them collectively would indicate a special cause such as the polls under predicting or that the polls do not sufficiently bake in an interdependence.

    • Nick B

      I think you’re right concerning the weight of national polls. While I do believe the national polls are a good indication of momentum, I’m beginning to question their quantitative usefulness to establish state trends.

      It becomes particularly more cumbersome when there’s a possibility the state polls lag the swing polls. We saw this toward the end of August after Clinton’s convention bounce. National polls started to tighten, but state polls showed no movement. Nate even wrote an article about this on August 21. Then, as he predicted, state polls started to tighten about a week later.

      I suspect we’re seeing this same movement now with respect to the new Comey announcement last week. National polls tightened (although they started slightly before the announcement), state polls still strong for Clinton. Now, we’re seeing the national polls go back to a Clinton +3 (call it equilibrium?), with new state polls starting to tighten (especially in New Hampshire, possibly Michigan?).

      The risk, as you’ve pointed out, is double counting movement. If national polls moved the same as state polls, I wouldn’t think of it as a problem. But if there’s a lag, what you’re essentially doing as counting that national + state movement twice, further depressing that candidate. You’re essentially treating it as two separate swings toward or away from a particular candidate.

  • Lee B

    I have a dumb question: when Sam says that the polls would always regress to the mean and that it is natural for Clinton’s numbers to come down from where they were, does that mean that they haven’t really changed at all, but that the numbers being captured were just kinda off from reality slightly in one direction or another?

    I just ask because I really enjoyed when the left side of the red line was barren except for the occasional tumbleweed or buzzard, but now that there are a couple of blue peaks and a little bit of green over there I find myself wondering what it is that causes people to change their minds back and forth like that. Then I think about what prof. Wang has been writing and, not being a numbers guy myself, I wonder if that just means the numbers haven’t ever really changed, it’s just our observations that have.
    Anybody know? Or, barring total knowledge of the minds of all Americans, have a better explanation?

  • Stephen Balzac

    This is consistent with what I’ve seen in organizational behavior: the better things are going, the more excited and engaged people become. Teams that are doing well work harder. Teams that are beating the schedule pull further ahead. Conversely, teams that are behind become more stressed and discouraged and tend to fall further behind.

    Hares may go sleep just before crossing the finish line; people not so much. Maybe that rabbit tortoise a lesson after all :)

  • Bill G.

    Your final point is one that I’ve wondered about ever since Trump started the “rigged” talk. Not that the man thinks these things through, but I don’t see how telling your supporters that the whole this is rigged anyway is supposed to inspire them to turn out.

  • Rick Howard

    ABC News/WaPo Tracking Poll Growing Bullish on Hillary Again

    Shortly after James Comey’s letter to the FBI, Donald Trump briefly took the lead in the ABC News/WaPo tracking poll. At the time, we noted that ABC’s poll is often accurate one week out, but is even more accurate the day before the election. And if that holds, it’s good news for Hillary Clinton, because she’s trending upward and has taken the lead again. To be more precise, Clinton dipped below Donald Trump for one day—October 30, when he was at 46% and she at 45%. They were tied the next day at 46%, and then Clinton pulled ahead again. She’s now up 47% to 43%, which is exactly where things stood a week before Comey’s announcement. So, the effect of the new e-mails, assuming there ever was one, may be dissipating.

    – Electoral-Vote.com

    • BrianTH

      She ticked up another point in their tracker and is now +5.

      This isn’t going to be as infamous as the Gallup miss since they have converged, but it may become Exhibit A in the growing case against the reliability of daily trackers, meaning it probably is not the case actual sentiment is flipping back and forth that much.

    • Arun

      I am coming around more and more to the belief that these are less real moves and more indications of willingness of people to respond to pollsters. People are more likely to respond when the candidate of their choosing is on the positive side of media attention or as has happened more often in this strange election, the opposing candidate is on the negative side.

    • Josh

      I agree with Arun. I think at this stage of the game, virtually everybody has made up their minds–and most of us made up our minds months ago. I doubt it’s coincidental that the Meta-Margin has basically just oscillated between a couple of points above and below ~4 for the entirety of the general campaign.

      My guess is that whatever the popular vote margin is that corresponds to a MM of ~4 (5-6%?) is what we’ll see on Election Night.

  • Scott H

    “The Myth of the Rational Voter: Why Democracies Choose Bad Policies,” by Bryan Caplan has the same argument. The emotional reward outweighs any marginal change in policy your vote is expected to cause. (Caplan writes that independent markets forces are needed to keep democracy from going crazy, and vice versa)

    People want to feel like they are on the winning side, so either switching to the winning side or sitting out if losing can be rational choices.

  • Emigre

    I wonder how peer pressure affects turnout?
    My question was prompted by recent data and graphics on the NYTimes Election 2016 web site.
    “We’re sorting by the way we live, think and — it turns out — every four years or every two years, how we vote.”
    Unfortunately, no data on voter turnout in these counties are listed:
    http://www.nytimes.com/interactive/2016/11/04/us/politics/growing-divide-between-red-and-blue-america.html

  • gumnaam

    I find the claims of herding among pollsters a little weak, since that would involve changing their methods in response to polling trends. Why would pollsters do this?

    If you look at the Huffpo graph, the Trump percentage of the vote has increased more than the Clinton percentage of the vote has fallen. So, perhaps Republican voters coming back to Trump?

    Regarding Sam’s niche, I think his predictions may have secondary effects on at least one other prediction, e.g. the 538 forecasts always seem to be maximally distant from whatever Sam predicts.

  • WDR

    “the rational argument for voting is not strong”

    I disagree. The arguments most commonly used aren’t–but if you treat election voting as a collective iterated Prisoner’s Dilemma, the rational case for being willing to invest some effort in voting is quite strong.
    It’s not easy to understand if you don’t know game theory, but that’s a different problem 8-)

    • Ed Wittens Cat

      yup, an Iterated PD– but not infinite iteration– comes with a demographic time bomb for the GOP.
      some interesting Q & A here–
      https://www.cam.ac.uk/discussion/opinion-decision-time-in-the-us

    • Ravi

      No, the rational argument for voting is quite weak. It starts from the wrong premise – that the possibility of deciding the election is all a voter gets.

      The flaw is that the argument does not look at the rational behavior of politicians. Once you do, you notice that margins of victory matter. They rationally influence the behavior of politicians (positions taken, resources devoted to future races and so on). That is a tiny but *guaranteed* impact of your vote. For that matter, geographical and demographic breakdowns also matter and your vote will move those in your preferred direction as well. And a tiny but guaranteed impact is can rationally overcome some significant cost to voting if you care enough about political issues.

    • ravilyn sanders

      The IPD problems/simulations very quickly showed nice strategies out do nasty strategies, and tit-for-tat did very well, randomly forgiving tit-for-tat did even better.

      But it became very clear tit-for-tat will degenerate to “always-cooperate” once the defectors are reduced a lot. At that point the problem is not nice/nasty, It is the free riders. The lazy non-thinkers.

      It is in that context Sam’s comment should be seen. It is rational and selfish or just plain lazy to skip voting when you are sure your candidate is going to win, and use that time/energy for other activities that will reward you personally.

      But psychology is, when you have a winning candidate you feel more enthused, you vote, you display lawn signs, paint yourself with reflective paint to shine bright on the reflected glory.

    • Ed Wittens Cat

      ravilyn
      the problem with cooperation theory is that its non-intuitive to the masses– its like Hoefstaders SuperRational games– lovely (useless) toys.
      i wasnt even going to vote this election and just dropped in to see what u guyz were doing–
      amazingly now im volunteering for Ivy
      This is what motivated me
      https://www.salon.com/2016/11/06/why-donald-trump-scares-you-so-much-and-why-it-matters/

    • Adam Rosenthal

      No I think they’re intuitive without knowing game theory. I only have a passing layman’s acquaintance with game theory, but I’m well acquainted with performing civic/social acts that are a bit of a hassle because “If everyone cheated the system would break down.”

    • Ed Wittens Cat

      But Adam–
      one side is cheating
      republicans have gone full frontal “sinners”

    • Ed Wittens Cat

      In the podcast where Dr Z and Dr Wang talked about normalization of extremes– that is the simple answer.
      The GOP has gone full-frontal “sinner”.
      The civil norms and taboos, even the Truth, and (mirable dictu)– EVEN THE MATH have no validity in the face of extreme polarization.

    • Ed Wittens Cat

      ravilyn–
      as for TFT (tit-for-tat)
      “Sinners” beat “Saints” in the Iterated Game

    • Lorem

      I think you’d find that an iterated Prisoner’s Dilemma with a huge number of participants in which they neither closely monitor each other nor mete out punishment does not converge to cooperation.

    • Billikin

      The rational argument against voting is not strong, either.

      In particular, to take everyone else’s vote as a given, so that your own vote will not change the outcome, so why vote? is a fallacy. If you treat whether you yourself vote as undecided, the rational thing to do is to treat everyone else’s vote casting as undecided, as well.

      This is Kantian reasoning, as it universalizes the question. But it does not yield a duty to vote. If you don’t care about the results, don’t vote. But if you do, don’t keep from voting because “your vote doesn’t count.”

    • Ed Wittens Cat

      Lorem is correct– im mortally sick of my profs and advisors and committee and math and physics heroes telling me “we are all one”.
      We arent. we really arent.
      an’ this election is absolute incontrovertable proof of that.
      :(

  • Scot Gould

    The headline reads: chance of Clinton win >99%. Yet the top graph on the right hand side shows if the election were held today, Trump has at least 2.5% chance of winning – if I understand your color bars correctly. These two fact appear to be in conflict with each other.

  • jim robson

    No matter what just vote.
    Unless you are voting for the tiger guy that was on Jon Oliver…then yeah I think your vote won’t matter ;)

  • CW

    Not related to this post- but what looks like a bug in the code… The right hand column shows MS as being Trump +3, while the huffpost link (by my calc) has Trump +10, with every poll since Sept having Trump up by 7 or more.

    • Greg

      I just noticed that too. I downloaded the CSV file of polls and didn’t see anything for MS.

      If Hillary comes within three points in MS, I’ll eat an ant farm.

    • Ketan

      The “scraped polls in CSV format” link on the right hand side… gives a file that has 1332 state polls. None of those polls seem to cover a state with initials MS.

      Actually only 43 unique states are in the table. The missing states are:
      HI, MS, MT, NE, ND, RI, SD

    • Ketan

      Sam’s data is assuming 2012 results for those 7 states. Only MS “matters” as the other states are all 0/100% decided but looks like Upshot might have the same bug in their MS estimation.

      Wasn’t able to determine why all polls for these seven states were dropped; of if polls are randomly getting dropped.

      Sam’s Median is pretty robust against bugs, though. :)

    • ravilyn sanders

      @CW // Nov 5, 2016 at 11:58 pm:
      “… but what looks like a bug in the code …”

      Don’t worry about it. Sam has promised to eat that bug. ;-)

    • Ketan

      This got fixed in the last few hours; looks like PEC is now hard-coding a better number for Missisippi.

      Trump + 13

      http://election.princeton.edu/code/python/ev_update_polls.py

      So Ant Farms can rest easy when they see @Greg walking around. :)

    • Sam Wang

      Yeah, that is a glitch in the HuffPost feed, in which pollsters who report 50 results at once get coded in weird ways that were not obvious to us. The hard-coded value is the recent median, including those pollsters. It was a kludge to hard-code the value, but there is no time for anything else.

  • Trevor

    Interesting. I wonder if pollsters don’t adjust because if they used this simple linear equation to adjust polls for the bonus (with the adjusted margin of error) they would be accused of cooking the results beyond base result suggests. I know when I took my first stats class in grad school we were told to just report the results, with the MoE dictated by the sample size (along with weighting to capture the right stratified sample size) and leave the rest to the pundits and aggregators. You were one of the first when I was in grad school in 2004. I lost a bet because I thought challengers would break to Kerry and he would win Ohio.

  • Jay Sheckley

    The percent probabilities for a Hillary win seem to be interpreted as her margin. On Facebook, intelligent friends claim that in most states it’s “safe” to vote for somewhat unvetted 3rd party candidate Jill Stein without harming Hillary’s chances or increasing Trump’s.
    Couldn’t this [combined with binary-polling, widespread gerrymandering and GOP mass unregistering of millions of minority voters] mean all figures could easily be off by at least the typical 2%? Is there closer modeling of that than the 270toWin Trump +2 map linked at right?
    More and more friends are advising others this is a fine time to vote for Jill. They praise one another for this sound advice.
    Aren’t they chancing crowning the likes of Trump?Meanwhile more dirt on him comes out daily. Yesterday we learned Melania did work here in illegally, while simultaneously the state of Georgia filled with Trumpish fliers demonizing “illegals”.
    We also learned that Trump had a two year affair during this current marriage too, the rights to which story were puchased and hushed by the Enquirer… So many of us are no closer to accepting Trump, while at least one of his armed “poll guardians” questioned voters yesterday.

    What info can I give friends regarding the “safety” of voting for Jill Stein? Aren’t more than the usual swing states at risk?

    • Diego de Valera

      GOP mass unregistration is a real problem, but one limited to GOP-controlled states. Looking at the traditionally-defined Blue Wall, I see Democratic Secretaries of State in all but MD, MI, NJ, and WA. Of these, only MI has a poll median closer than C+11, and both campaigns now consider it in play.

      Gerrymandering only affects legislative districts, and has no impact on the EV calculus, unless you consider ME-2 and NE-2 gerrymandered.

      As for binary polling, I calculated a worst-case scenario for Clinton bleeding voters to Stein by taking Clinton’s current median in each state and subtracting Nader’s percentage in 2000. I say “worst-case” because many polls do include Stein, Stein is a much weaker candidate than Nader, Trump 2016 is much scarier than Bush 2000, and the Clinton campaign strongly emphasizes the anti-spoiler message.

      The only bricks in the Blue Wall where Clinton falls behind are ME-2, where the median is C+2 and Nader got 5.56% in 2000, and MN, where the median is C+5 and Nader got 5.20% in 2000. (The purple states where Clinton falls behind are CO, NV, and NH.)

      TL;DR Based on current data, Stein spoilage is not a rational fear in any Blue Wall territory except ME-2 and possibly MN.

    • ravilyn sanders

      I also see lots of people conflating probability of win/loss as polling margins.

      But mostly in rabid red sites who are showing, “look it is only 47-42 but the “media” is misreporting it is 90-10 for Hillary!” They are also against cartograms. They are upset the liberals are drawing a distorted map of USA making red states look small and blue states appear big. Faulty science! Liberal conspiracy!!

    • ADC Wonk

      This may seem silly, but my response would be: “think globally, act locally.” I suppose a slightly more sophisticated argument would be: don’t be a selfish freeloader.

      I.e., what’s the problem with my purchase of a gas-guzzling clunker? Will one person make a difference? Same argument, no? The argument your friends are making is a selfish one: “well, _I_ can do it, but I hope most others do something different.”

      No. The moral way is to act the way you want others to act.

    • DaveM

      “…only MI has a poll median closer than C+11, and both campaigns now consider it in play.”

      It’s not clear whether either campaign considers it “in play,” or whether it’s simply the blue wall state in which a poll-defying Trump surprise win is least improbable. Much has been made of late about which states the candidates are dashing off to at the 11th hour—”Trump sees opportunities” and “Clinton playing defense” fit the media horse race narrative—but the reality is that both candidates have to be somewhere!

    • Sam Wang

      Clinton might be in Michigan because there are three swing Congressional districts. Also, if there *were* a national systematic polling error, it would be on the edge.

    • Matt McIrvin

      My experience with committed third-party supporters is that they’ve already heard every variant of the spoiler argument and dismissed it, and will get very angry and come back with some variant of “you’re not the boss of me” if you even bring it up. If they’re even justifying their move by saying they’re in safe states, that puts them ahead of the pack. Probably not a lot you can do.

      I keep thinking about John Oliver’s masterful hit piece on the third-party candidates, in which he shows a perfect example of Gary Johnson going off on one of his own supporters for daring to mention it. And then Oliver does something very clever: he explicitly sets the spoiler argument aside, takes them seriously as candidates and just tears into them on their own merits. The piece seems to have actually done fatal damage to Jill Stein’s fundraising (her supporters being the ones most likely to pay attention to a John Oliver segment).

    • Ken

      I still curse Ralph Nader every day pretty much. Stein’s case is not fully analogous. Nader had much more support from people who would otherwise have voted for Gore (such as my dewy-eyed 22-year-old daughter; she knows better now). Stein seems to be the candidate of the pure looney who either doesn’t vote or always votes on the fringe. But this impression is based on what I read in the news and not on hard data. I wonder if anybody has tried to measure the extent to which Stein pulls from Clinton. The other question is: WHERE are these deluded people? If they are not in swing states, they can vote for Richard Nixon for all I care. Having said that, the answer to your friends is that even one vote for Stein is too many because you never know. Even if we assume that voting for Stein equals not voting, which it does since she cannot possibly win, that vote may be important. Certainly, if Hillary Clinton were to capture all of the two percent that polls for Stein it would be a very good thing. On the other hand, if your friends might vote for Trump if they don’t vote for Stein, then you should encourage them to record their protest against the evil elites and remind them that both Trump and Clinton regularly consult with Satan’s henchpersons. They’ll probably believe you.

    • Diego de Valera

      ADC Wonk:

      “The moral way is to act the way you want others to act” is a very dangerous argument to make in the context of third-party voting.

      Sure, a Stein/Johnson/McMullin voter in a safe state might not want a *few* swing state comrades to spoil a possible win for their least-disliked major-party candidate. But they would almost certainly prefer it if *many* others acted the way they did, enough to actually elect Stein/Johnson/McMullin.

  • mcorn

    We always seem to have a theory for why people do what they do, regardless. Many people believe in the bandwagon effect, which could be explainable by the desire of people to conform or be associated with a winner.

  • Alan White

    But given that the stats track given here is about a specific event, and that the blanket psychology of the electorate in 2008 and 2016 is so very different given the high level of negatives for both candidates, couldn’t it be that that Trump’s “rigged” comments motivate the collective feeling of disenfranchisement of his emotionally reactionary base more than the “she’s better overall” tepid rational sentiments of hers? That’s a worry to me, even given the added emotions against him to Latinos.

    • Josh

      “the blanket psychology of the electorate in 2008 and 2016 is so very different…”

      I think the question here is: Why would the fundamental factors that motivate people to do [x] be completely different in two events that are almost identical in nature and scale and occur within the same decade?

    • Alan White

      The high level of negatives is my point.

    • Josh

      So you’re arguing that Trump will turn out more voters because an historic number of people don’t like him…but the minority that do like him will be rabid about voting for him?

  • A

    Your work is too niche to affect voter turnout, in my view (no offense, it’s also better in my view than 99 percent of the other analysis).

    I’m more interested to hear about your thoughts on Drew Linzer’s comment about herding in national polls and what that might mean.

    The notion that the race has tightened–is that coming from undecideds and third party voters finally coming back–more of them to Trump then Clinton?

    Or something else?

    • Dharma

      What do you mean too niche. He shared a screen with Beyonce ya know…. ;)

    • Sam Wang

      Your work is too niche to affect voter turnout

      Well, duh. But yes, I should say that too.

    • bks

      But is Nate Silver’s niche big enough to affect the stock market?

    • EDYS

      If Nate’s nitche has gotten big enough to encourge Trump supporters to vote because he has convinced them Trump can win, then I would greatly appreciate Dr. Wang expanding his nitche VERY quickly in an effort to discourage Nate motivated voters with the information that Trump will almost certainly loose. A front page editorial in the NYT tomorrow morning would be nice. Only sorta kidding.

  • SK

    Any idea what this looks like for Obama-Romney?