Princeton Election Consortium

A first draft of electoral history. Since 2004

Weekend At Bernie’s (no, not that Bernie)

October 12th, 2016, 4:55am by Sam Wang


To the extent that the most statistically stable Presidential race in 65 years can be made to look alive, here it is. Right-click to adjust the looping playback.

Animated Median EV

I have a question for readers. To me, it is plain that this year’s race is statistically highly stable, i.e. it hasn’t moved up and down very much. I am occasionally met with incomprehension or disbelief, even after showing a graph like the one able. Is the difficulty simply the emotional nature of this year’s race? Or is something else at work?

Thanks to PEC reader David Elk!

Tags: 2016 Election · President

98 Comments so far ↓

  • Ken L

    I am looking to quantify the risk of the election in some way as it appears that the emotional content of this election entangles risk with volatility.

    Thanks to Scott J. Tepper for Dune quote.

  • Bob McConnaughey

    well..when the Upshot said a day or so back (paraphrase) that Trump has the same chance of a field goal kicker being successful from 36 yrds out..then minds can get more and more unsettled. And perceptions get moderated by the flood of social and media cues. Basically I very rarely go outside of PEC to obtain my election fix; to some extent PEC (happily) certainly reinforces my set of beliefs in re US politics though i’m likely more paranoid about the “american prospect” than many..Kristallnacht has lurked in the back of my mind since the Republican convention where, unintentionally or otherwise, the iconography on display during Trump’s acceptance was eerily similar to that in the Reichstag when Germany declared war on the USA.

    • JP

      I am a long time follower of PEC. Understand; I am not a statistics guy. But I do appreciate thoughtful and careful analysis of the statistical measures related to the election process. Because of that; I have made a herculean effort to learn the basics of the statistical processes used on the site. So in that regard; Dr. Wang has inspired an old dog to at least attempt to learn some new tricks! Back to Bob’s comments. I too attempt to steer clear of the twenty four hour cable chatter and social media nonsense. Thank you.

    • (((CassandraLeo)))

      As a person with Jewish ancestry, I’m finding the historical parallels unsettling enough that I’m actually experiencing anxiety issues caused by the election. (The fact that I’m also LGBT and Trump’s administration would not exactly be friendly to people like me does not help either.) I’m not the only one; one of my other Jewish friends has said that she finds the Hitler parallels so disturbing she has difficulty even seeing Trump on television. It’s probably worth pointing out that Mike Godwin of Godwin’s Law has also said that he does not find the comparisons to be frivolous, for whatever that’s worth.

      PEC has definitely helped my anxiety from growing too out of control, but even the 3% change that Trump is still being given is far too high for to calm my nerves completely.

    • Olav Grinde

      Bob, NFL kickers are better than that.

      Actually, Trump’s chances were stated as being equivalent to that of a field goal kicker missing the field goal from 36 yards. In other words: rather unlikely.

  • Brian MacDougall

    Well, if my physician tells me I have a 75-85% chance of beating cancer, I’m still freaking out, statistics be damned! This election, for both sides, is like life and death, and the rhetoric and the media coverage have followed suit. We are now down to the image of a titanic battle between good and evil. Statistics pretty much goes by the boards with that kind of emotional gravity.

    And I understand; I want the electorate to crush Trump. Not just win by single digits, but crush in the sense that he will be a broken pariah for the remainder of his days, living in a cardboard box in Queens, or perhaps a mental institution on Long Island. I’m pretty sure that Trump supporters feel the same way about HRC. When you have that kind of emotion, steady-state realities are pretty chimerical.

    • SoddingJunkMail

      I’m afraid that your hopes of a long term Trump crushing are going to go unfulfilled.

      He’s really in a no lose situation at this point. He either (a) pulls off the hail mary long shot and wins, or (b) has 25M raving fans of his brand, complete with their contact information through email or social media.

      It’s a direct marketer’s dream. When he loses in November, you can bet he’ll be crying all the way to the bank.

    • Suvro

      In Hindu mythology, there is an equivalent of good over evil. Long time ago, one demon had become very powerful, and was worrying the gods and goddesses. But he was all too powerful. So they combined their powers and created super Goddess Durga. She had 10 hands, each armed with 10 different weapons. She had an epic battle for 10 days and nights, chasing this demon – Mahishashur (the buffalo demon – half buffalo, half human – like a centaur). Finally on the 10th day of battle, she vanquished the demon. In India we celebrate this victory right around now. In Los Angeles we are celebrating this next weekend. It reminds me of this epic battle, where I am sure good will win over evil!

      Here is the Wikipedia entry on Goddess Durga and the socio religious festival.
      https://en.wikipedia.org/wiki/Durga_Puja

  • AP

    I think Prof. Wang is fixated with the sd of the meta-margin, which is a reasonable notion of volatility but not the only one. Other analysts with a more than decent reputation called the race volatile (I disagree with including the primary season, but who am I). People have suggested sd is the wrong notion as it is order-insensitive. People in this thread have suggested that swings in probability are more important, I would add as they are closer to the payoff for regular citizens (my payoff is expected number of nuclear wars in the next 8 years + alpha times expected concentration of CO2 in the air in the next 8 years, not the metamargin). There are unique, objective factors that are unique to this race that add uncertainty, such as the atypical candidate from one side, the interference from a foreign power, the comparatively large number of undecided and third party votes, none of which is captured in the SD of the meta-margin. I think we shouldn’t get too enamored with mathematical definitions to the point that we complain that reality doesn’t fit them: they are just models, the capture some aspects nicely, the leave some others out.

    • AwK

      In every election there are unique factors and it is easy to argue that there is additional uncertainty in this one. But there does seem to be low variation this year across time (not just overall sd).

      If you are going to redefine volatility as something other than variation across time then you should pick a concrete definition.

    • Sam Wang

      I do not believe there is any convincing evidence for short-time-scale swings. Even if there were, it requires far more analysis than visual inspection of a graph sampled at 2-week intervals.

      Again, if we are going to go there…it is possible to come up with some other definition…but then do it with evidence.

  • Daniel Barkalow

    I think many people assume that races are random walks rather than values selected from a distribution. The fact that we’ve seen a one-week swing that would bring the race to nearly a tie makes people nervous, even though (a) we’ve only seen one large one-week swing, so they’re probably not common and (b) another large swing could go the other way.

    I guess the other thing about that graph is that there aren’t a lot of different values in the range, which suggests there’s a lot of correlation between subpopulations, which means we don’t have the law of large numbers making outliers extremely unlikely.

    Between these, it feels like the model isn’t capturing something, and that something may be important to the outcome, which makes it harder to trust that the election-day polls will be like the polls we’ve seen so far.

  • Ken L

    The beauty of statistics is that it gives us a precise language to discus an uncertain world. It demands that we define our terms in such a way as to be measurable so that we may come to an informed consensus (and focus our attention on the house and Senate.)

    Analogy: based on this test we are 95% certain your cancer is in remission. Now let’s work on that cholesterol.

  • Mike Beers

    People confuse or conflate process with outcome. The events in the process of the both primaries were very surprising indeed and, at the outset, were not predicted. For example, no one predicted that Trump would publicly, coarsely insult, then systematically dismantle, the favored candidates of the Republican party leaders nor that Bernie (who?) Sanders would have gotten such an enormous response (primaries won, money raised).

    Many people then conclude something along the lines of “You couldn’t predict anything of this! Now you’re saying we won’t be surprised by the election outcome?”

  • Sean

    I would say that most people aren’t even aware of this site and that this kind of analysis has been done. They get their news on who is winning by watching CNN or FOX.

    I talk to people every day who refuse to believe that Trump is even behind much less that he will lose in November.

    Full disclosure: I am in West Virginia and the political environment here is truly insane. We had our potential governors debate last night and both were trying to win the “who is the biggest trump supporter” contest.

  • George

    I think Amitabh and “anonymous” both touched on potential explanations that resonated with me – and that I have addressed in a couple of posts. If stability and volatility are measured in a binary “who wins” – yes, very stable and not volatile. But if measured against the actual Meta-Margin, you see a long stable period from May through June in the 3-ish range, then after that you have had a big swing up to about 6.7 or so and down to about 1.1 IIRC. Top to bottom, that is a 5.6 point swing, and just looking at the two top and bottom numbers, it is a five-fold swing in the MM. THAT looks volatile – even though the predicted outcome never changed. In addition, with those swings, the “red zone” did in fact cross the 270 threshold, even if just barely. Finally – to quote Sam, at one point he said (more or less) “if the meta-margin gets below 2.5, it could get/be interesting,” and within about a week, that is exactly what happened.

    • Micah

      If that’s a five-fold swing in the meta-margin, then any election where there’s ever a lead change at all has a literally infinite swing in the meta-margin. You’re arguing that the election is swingy; you probably shouldn’t choose a measure which makes it infinitely less swingy than 2004 or 2008…

  • Richard

    Yes, the difficulty is emotion, which isn’t a bad thing. It is rightfully extremely scary to think that Trump has a nonzero probability of becoming president. That makes even small changes in polls in his direction seem very, very big. It is a great relief that the race is now effectively over, but the Republic is not going to heal anytime soon. Statistical analysis is comforting in trying times, but the reality that a fascist had a real chance of becoming president feels very unstable, even if polls are not.

  • Jeff Alworth

    Why does it seem volatile? I think about this a lot, and have since the primaries, when my liberal friends felt Bernie was doing a lot better than the numbers said he was. I’d identify three factors, the third of which is the biggie.

    1) Uncertainty. Even if a trend has been very stable, we don’t trust that things will remain that way. If circumstances seem volatile (and this is the most volatile election–measured by polatization, violence, and general weirdness–since 1968), that creates an expectation that “anything can happen.” Connected to this is a sense among many people that polls don’t capture what’s really going on.

    2) Media horse-race coverage. I discount this relative to the importance many people give it. But the coverage nevertheless frames the contest as active and undecided, and that can’t help but create an atmosphere of seeming volatility.

    3) Emotions overwhelm rational thought. The Trump campaign is a wonderful metaphor for this phenomenon. Trump is running on the fiction that the US and world are descending into chaos. It’s absurdly out of step with reality, but it’s why he has 35% of the popular vote locked up: what he says *feels* right. Studies have shown time and again that people trust their gut (their emotional reality) more than data. This descibes the state of the election. From the primaries, with Bernie and Trump, through the events of the general, nothing seems stable and expected. Emotionally, the idea that this year is “stable” or “predictable” feels wrong. It may be the case, but it’s not where people live.

  • Ryan

    I’ve read PEC for several cycles but I’ve never written a comment so I’m going to start off thanking Dr. Wang and everyone else who has contributed to the site.

    There are lots of thoughts I agree with in the comments but one aspect of the race that I sometimes struggle with is that the MM is very stable, but small changes in the MM can make the EV count jump as states move from one column to another.

    I admit I’m a partisan. I struggle with trying to understand the people that support Trump. I can’t put myself in their shoes well enough to understand what they are thinking. When I hear the things Trump says I’m incredulous that so many support him. That unsettles me enough that emotions do get in the way of focusing on the numbers.

    As for something that PEC has control over, I do have one thought. If you want to show off the stability of the race I believe it would help to create a Median EV estimator graph with a full range X axis from 0 to 538 and a History of the Meta Margin graph with a full range X axis of 0 – 100. The smooth lines those graphs would show will highlight just how stable the race really has been.

  • Jinchi

    To me, it is plain that this year’s race is statistically highly stable. I am occasionally met with incomprehension or disbelief, even after showing a graph like the one able. Is the difficulty simply the emotional nature of this year’s race? Or is something else at work?

    On the left, I think it’s because most people can’t believe that Trump has any support at all. Every week he says a half-dozen things that would kill a normal campaign. Pundits seem to think this is proof Donald Trump defies the norms. But I think it’s because he did kill off his campaign right from the start. If you’ve stuck with him through Mexican rapists, Muslim bans, inner-city hellscapes, and his decades long record of defending sexual predators like Mike Tyson against their victims, then you’re probably willing to believe that his latest comments are just typical guy-talk. If you abandoned him long ago, you aren’t coming back, now. In either case, you’re no more or less likely to vote for him than you were the day before you heard about it.

  • Kevin

    I think it’s that people don’t care about volatility per se as much as they care about how far the race is from tied. A swing from Clinton +4 to Clinton +1 may not be large in mathematical terms, but people see win probabilities cratering by 20 points and start to crap their pants. Fluctuations in the range of Clinton +11 and Clinton +17 would be less noted, as they would have a comparatively small impact on win probability.

  • anonymous

    I think the perception of stability versus instability is intricately tied to the perception of how even or close-to-even the election is. For example, in the movie, imagine that the median EV was not hovering around 320 but instead around 270. In that case, the exact same time series, which would have the exact same standard deviation, would be perceived to be a wildly volatile contest. Perhaps a complete separation of volatility and the set point of the contest is not psychologically possible?.

  • Amitabh Lath

    The answer might just be in the plot itself. Look at May to the end of Aug, where it barely moves. This conditions people into thinking of this as normal.

    Then a few ticks down and up again, and everyone’s popping Xanax.

    If summer had had the normal amount of jitter, what happened in September would not have seemed so catastrophic.

  • Ken L

    I think if we could determine an “emotional expected value” for the race using some kind of polling to determine the perceived costs/benefits of the possible outcomes and plot this over time we might have a more objective measure of the perceived volatility of the race.

  • AA

    Perception beats the hard cold statistics.

    [1] There is a lack of clarity on the email saga, and Clinton hasn’t helped either. Her fumbled responses have feed into the perception she is disingenuous.

    [2] There is a pseudo theory of hidden Trump voters, bolstered by the Brexit and Colombia vote, that makes people question the efficacy of modern polling methods. I sometimes do fear about it.

    [3] The variation in polls taken around the same time period is used as an argument against their efficacy.

  • TeddyVienna

    Several factors:

    1. Brexit polling.

    2. The 2014 midterms.

    3. The meta-margin and the EV projection have been stable, but Clinton’s win probability here has veered between 97% and … what, about 65%? To a statistician, maybe that’s not a big difference. But when you’re talking about a 35% chance, that worries people more than a 3% chance does. (And Nate Silver had Clinton all the way down in the 50s at one point.)

    4. Unusual variables. Will Sanders supporters show up to vote for Clinton? Will Trump bring new voters out of the woodwork, particularly people who don’t necessarily answer the phone for pollsters? Are Millennials really too lazy to go to the polls?

    5. We didn’t think Trump would win the primaries, either.

    So it’s a mix of emotion and some serious questions about whether we’ve accounted for every variable. I think Nate Silver has turned very cautious and is throwing every conceivable twist into his models. He’s probably overcorrecting after missing the Trump bandwagon in the primaries. But by how much?

  • Jim Higgins

    My perception from news coverage and highly anecdotal discussions with anti-Trump others is that his sheer unpredictability is creating the perception of volatility in polls and the election’s outcome. I hear and see a lot of talk about a negative Bradley effect that reminds me in substance if not origin of Republicans’ delusion about polling accuracy in the 2012 presidential election.

  • Rochelle

    I think there are a number of contributing factors.

    You are presenting an aggregate of polls; the media present each new poll as if it alone will predict the outcome. So the variability of individual polls is magnified by the media.

    People don’t think about the difference between a national poll and your aggregation of state polls. They hear a national poll result right after the Republican convention and think that Trump will win.

    The media cover Trump far more than they cover Hillary. Every tweet, every statement at a rally – he is there, in our face, all the time. And I don’t even watch TV!

    And, of course, there is the emotional factor – fear on the Democratic (and rational Republican?) side – and anticipatory triumph from Trump supporters.

  • Michael

    Is this really new? No later than 1935 Will Rogers said, “I only know what I read in the newspapers, and that is my alibi for ignorance.” Then, in the 1941 newspaper story, Charles Foster Kane’s race for governor ends with a sex scandal. What is most scary now is the way one side is injecting baseless doubt into the process, not just about the opinion polls, but the voting itself. When you can suggest there really is no such thing as objective fact, you can gaslight almost anybody.

  • Sheldon Rampton

    I think the sense of volatility stems in part from other polling aggregators like fivethirtyeight presenting a much greater impression of volatility than this website presents. If you watch their various metrics, in particular their presidential now-cast, the picture seems to be: Clinton has generally been ahead, but Trump took a brief lead after the Republican national convention; then she took back the lead, but Trump fought back to a virtual dead heat until the debate, when she reclaimed the lead. And of course the graphics are accompanied by commentaries every day or two that heighten the drama. What people are getting from fivethirtyeight (and even more so from the TV networks and other news media) is a _narrative_ of turbulence and a race in which the lead keeps changing or threatening to change. And in the case of fivethirtyeight, that narrative comes from someone who appears to be pretty good with the math stuff too.

    I think the human mind is also designed in such a way that we respond more viscerally to narratives and drama than we do to math and statistics. I wish someone would write a book about the human brain that would help us understand that part better. :-)

  • Kenny

    For me it’s the high number of undecideds + 3rd party voters that still add a bit of unpredictability. However, the polls right now suggest that even if those undecideds/3rd party broke heavily (80%) for Trump, it still wouldn’t be enough to catch Clinton.

  • bks

    If every election reached a point where one candidate had a 95% chance of winning on the day before the vote, then about once in a lifetime the candidate with 5% chance would win. I hope that’s not this time.

  • Sharad R

    Perhaps it is because of the popularity of Nate Silver. If you look at his analysis, things look a lot more dramatic…the two candidates were almost even end of September according to him.
    Add to the variability in the prediction of the different websites, a lack of clear understandable explanations of what priors mean, and what the various terms in the analysis are…suddenly the analysis business starts to look like the speculations of the talking heads on TV since the results from one to the other are so different.

    Perhaps an interactive where people can choose their prior to replicate the analysis of the different websites easily to see whats so different about all of them would help.

    By the way, for geeks like me who know about your science, Sam: your website is the ‘go to’ place.

  • JPI

    People are watching and listening to the news media outlets that behave as if this whole thing is one big high wire act that could go any way. This helps them to sell advertisements. The Princeton Electoral Consortium has no ads ( which is what I told a friend who was nervous about the election when I referred him to this site).
    The drama is most of the fun. I admit that I am relishing the thought of not just a huge loss by The Mouth, but a really nasty down ballot wave on election night. I’ve got a feast and margaritas lined up to accompany my schadenfreude. My fun is knowing that Sam is right about this election, and all of the hand wringing is a pointless exercise. Though I admit after that last debate’s nasty tone the fun is starting to diminish substantially…

  • Barney

    It may be the drop in Median EV in early September that has people saying “it doesn’t look stable to me”. The black line did just about get outside the 95% ‘cone’ from late August. It’s not easy to think how often you’d expect that to happen in a continuously updating graph with uncertainties.

  • MikeW

    On the one hand you provide a serious mathematical explanation for what is going on. It answers a straightforward question in probabilistic terms.

    There is a very different industry that is motivated by different objectives — namely selling ad time. I would assert that in every national race I have watched the same phenomena of “stories” repeats itself to keep people paying attention.

    The sports analogy is apt, less competitive games, or games that we know are effectively over (using an intuitive sense of probabilities) have lower viewership or lose viewers.

  • E.J. Chichilnisky

    Your methods have some time-averaging built-in (the time it takes for fresh polls to show up from every state, though I don’t know the details of that). So, it’s not surprising that your graphs look pretty stable compared to media buzz. I’m not sure whether your time-smoothed estimates are more accurate than what we hear other places (we would have to hold a lot of elections every year to test this idea rigorously!).

    • Sam Wang

      Probably the effective turnover is 2 weeks or so during the season, more like 1-2 weeks now. Not all states have to turn over – mainly the closely contested ones. You can get a feel for this by examining dramatic events from past years, such as the first Obama-Romney debate. In that case, a large change was visible within 1 week, and it appeared to be near-equilibrated.

  • JD

    I think it is a combination of things.
    1. Emotion. Definitely a scary prospect of having DJT as POTUS.
    2. Unknowns. With Russia and WikiLeaks involved
    it is easy to think things could turn on a dime.
    3. Clinton fatigue. It is easy to imagine, especially
    if you listen to the media, that HRC has huge flaws that will
    be exposed and the population is tired of the Clintons.

  • Ken L

    Our media, which should be one of the pillars of our democracy, fails to convey the both the value and the pitfalls of polling and statistics, choosing instead to focus on the “horse race” and to amplify every inflammatory statement and event.

    If our mainstream print and broadcast media were to seek to objectively report the facts ( as opposed to a false sense of balance,) provide clearly identified thoughtful analysis, and provide easy to understand explanations of concepts that may be unfamiliar to their audience, we would all be better served.

  • Ebenezer Scrooge

    The race is stable, given Sam’s poll metric. And Sam’s metric is the one to use if the only thing you care about is who wins next month.

    But there are a lot of other things going on in this race that are not stable. They are mostly related to the long-term future of the Republican party. Professional Republican politicians are running around like headless chickens–something inconsistent with any kind of stability in the race.

    If you are a hammer, everything looks like a nail.

  • Scott J. Tepper

    “I must not fear.
    Fear is the mind-killer.
    Fear is the little-death that brings total obliteration.
    I will face my fear.
    I will permit it to pass over me and through me.
    And when it has gone past I will turn the inner eye to see its path.
    Where the fear has gone there will be nothing. Only I will remain.”

    • Ed Wittens Cat

      ok, a Dune-off it is
      on Dr Wang’s polarization:

      When a creature has developed into one thing, he will choose death rather than change into his opposite.

      Scytale, Tleilaxu Face-dancer

  • Andrew

    One potential factor that hasn’t been mentioned is the media narrative. Up until just recently, Clinton has received more negative media coverage than any other candidate, and not just wikileaks but these trumped up stories (please forgive the pun!) by the AP about her meetings as SoS and Clinton Foundation activities, etc. Plus you have all the political attacks (Benghazi). Examined closely there’s nothing substantive in those but cumulative effect is pretty powerful, too much so. To judge the race by media coverage it paints a much closer race than it actually is. That’s my guess: we’re still living in a pundit’s world!

    P.S. Don’t give your blog after this election Prof. Wang, it seems there is still need for an “unbiased estimator,” or as close to that as we can get. I realize that your h-index may suffer for doing this work instead, but it is useful.

    • (((CassandraLeo)))

      I agree. This blog is an invaluable resource and the country would be far worse off without it. It’s not just an unparalleled factual resource; it’s also been indispensable in keeping my anxiety issues in check. I don’t even want to know how much of a wreck I’d be if I’d had to rely on sites like 538 for my election updates this cycle.

  • Christian

    It’s emotion for me, yes. And my emotional reaction can be explained by an equation that’s been tossed about here: the odds of a Trump Presidency (3%) multiplied by the consequences of a Trump Presidency (Fascism + War + Financial Ruin + International Embarrassment) still leads to a significant #.

    .03 * 98598643287543275360695 == !!!!!!

  • Matt McIrvin

    Personally, I am deeply nervous about the Senate: that is pretty stable too, in numerical terms, but it’s poised on a knife edge, unlike the Presidential race. And so far, Trump doesn’t seem to have powerful anti-coattails, though there hasn’t been a lot of polling since he started his weird bout of slamming Republican members of Congress.

    If Hillary Clinton doesn’t have a Democratic Senate, it’s entirely possible that for at least the next four years we will lack not just a full Supreme Court, but a Cabinet.

  • Sophia

    I think it is difficult for people to equate the statistically stable numbers of this race with an unstable candidate like Trump. So in a sense there is instability. (The fact that a person like Trump was elected as the Republican nominee is the unstable part) Just not with the statistics of the presidential race itself which you keep trying to drive home. Also, I can imagine being a Trump supporter and genuinely having an unstable home and unstable life so everything looks unstable. That feeling of the entire world is unstable and the idea that this election is unstable is amplified by the media. You know Sam, so many people are being spoken to by Trump and the media right to their amygdala. I actively rewired my brain by reading your site and staying away from the ones that feed the part of my brain that could easily be hypnotized into fear.

  • bobby earle o'brien

    I think it’s somewhat emotional. For me (and for someone who lives in Boston), I often bring up the 2004 ALCS with the Red Sox & Yankees. You can never completely count out the underdog.

  • Mark Schmaling

    You just found “Backfire.” People judge/filter new information based on their pre-existing beliefs. Facts don’t change minds. Actually, the opposite happens. The made up mind will dig in and defend pre-existing incorrect beliefs when confronted with new facts that undermine those pre-existing beliefs. This may be part of human nature. It’s deeper than just plain old ignorance. http://archive.boston.com/bostonglobe/ideas/articles/2010/07/11/how_facts_backfire/

  • Amit

    For me, even though I understand and believe the math, the problems are:
    1. Amnesia regarding past races, especially with respect to how volatile (or not) they were. There is simply no reference point for volatility.
    2. The sheer terror at even a delta possibility of a DT presidency.
    These 2 effects magnify even the smallest changes that in my mind I know are statistically insignificant.

  • Edge Oforever

    The horse race narrative has taken in many paying attention to pundits (same one predicting Trump would not win primaries). I just went by polls and predicted both primaries and GE. Got a bit nervous a few weeks ago, but that’s all.

  • Stefan Krieger

    Shouldn’t those forecast uncertainty shadings look more like sqrt(t) and not linear in t?

  • Matthew

    It’s definitely emotion, and I say so as someone who has been (emotionally) highly susceptible to even minor shifts in polling this year.

    Like many liberals I feel Trump is an existential threat to the United States, and I have been ashamed and dismayed at the number of my fellow citizens who have been duped by him. Countenancing that he has even a REMOTE chance at the presidency has caused me to lose sleep multiple times this year, going all the way back to March when it became clear he was going to be the nominee.

    The only comparable time in 2012 was after the first debate when the left had a collective concern about Obama blowing it (A Liberal Tradition during presidential election years).

    I am also unsure if the media is even MORE in horserace mode this year than previous, or if it’s just coupled with the disbelief that anyone is willing to treat Trump as a serious candidate. I’d be inclined to the latter, but the more I learn about Jeff Zucker peddling Trump at CNN and NBC’s reticence to release the GTBTP tape the more it genuinely feels like powerful forces have campaigned to keep the “close race” narrative alive.

    TL;DR definitely high degrees of emotion

    • Froggy

      This, although hanging around here long enough has made me less susceptible to panic as polling changes.

    • Charles

      “I am also unsure if the media is even MORE in horserace mode this year than previous”

      They are! And they are going to become worse and worse about this kind of stuff until TV news dies, for the sole reason that they are being made slowly irrelevant by the internet, and thus they are forced to make outrageous claims and pander to the lowest common denominator. This effect is totally unrelated to the election, they will exaggerate any topic they are given if it helps attract viewers.

  • Fritz

    Math isn’t intuitive and doesn’t ‘fit’ anecdotal experience. Between fear and visceral anger that Trump might win, fluctuations in polls reported by the media feed that anxiety. Even now “Trump is one Wikileak reveal away from flip flopping the race” is beginning to take hold in the punditry. Liberals/Democrats are also predisposed to panic (why is that I wonder?)

    For me personally, PEC is my sanity anchor – so thank you for doing this.

    • Charles

      “Math isn’t intuitive” is just about the best answer. The very first poll after pussy-grab-gate showed Trump down 11 points. The media narrative was, “Trump collapses after video controversy.” Yesterday there was a poll showing that he was 9 points behind, and the CNN headline said, “Trump makes gains following debate.” These are two different polls conducted by two different media outlets. To suggest that either one alone can be used to draw a conclusion is ridiculous; to suggest that a two percent difference between them suggests a “gain” is absolute insanity. But for people who don’t read blogs like this, don’t follow how polling ACTUALLY works, they’re just going to read that headline and assume it’s true.

    • dn

      This, this, this. Math just goes over the heads of millions of people. You cite the numbers to them, or show them a graph, and their eyes just glaze over. And there is no poll fluctuation so inconsequential that it can’t be spun by horse-race journalists.

      Plus, I think there’s a certain factor of disbelief that goes along with the well-known unpopularity of the candidates. In an election characterized by extreme polarization and where so many people despise both the mainstream candidates, people find it hard to believe that the overwhelming majority could be firmly in favor of one or the other. Especially the other (how many times have you had a conversation to the effect of “I just can’t believe how many people will vote for Donald Trump!”?).

      More thanks to Sam and PEC for helping to make the decision to avoid the MSM easy this year.

    • alurin

      Yeah, liberals are really quick to panic. I wish I knew why.

    • Kevin King

      I agree with you, Charles, about the coverage of the polls. Trump’s support hasn’t collapsed at all. At best, Clinton appears to have gained maybe a point. What the tape and the debate seem to have done is solidify the candidates’ respective support.

    • Rob

      Yes but Trump hasn’t gained. He’s found 40% who are either racists, nationalists or are simply going to vote for the person with an “R” next to their name no matter who that is. And his current strategy is about stoking up those people rather than trying to broaden support amongst more moderate voters.

      There could be several factors here in that decision:

      A) He’s concluded the moderates are lost to him and so hopes to maximize turnout with his base and hope the Dems suffer depressed turnout.

      B) He’s deluded himself to the size of his following and thinks his base represents a larger proportion of the population than any reputable poll suggests.

      C) He’s never quite understood the difference of competing in a primary election rather than a general campaign and doesn’t understand how to operate.

      All in all, us liberals need to take a breath. We still need to vote and be vocal but this is going our way and strongly.

    • (((CassandraLeo)))

      “Yeah, liberals are really quick to panic. I wish I knew why.”

      I think years and years of incompetent Democratic campaigns hasn’t helped. The DNC still hasn’t gotten its act together despite this year’s Clinton campaign being one of the most professional and workmanlike political campaigns in history. I think most of us were freaking out in September when Clinton’s lead began narrowing, but in hindsight it looks a lot more like she was simply running out the clock so she could dump the opposition research she had on Trump after it was too late for the Republicans to replace him. I don’t think her campaign ever anticipated that she’d be able to create this much disarray in the party, but I’m absolutely certain she intended to throw him off his game with psychological warfare, and she absolutely succeeded. A few more presidential campaigns like this and I wouldn’t worry about them going forward, apart from the facts that Trump is such an existential threat to the nation in a way previous Republican candidates haven’t been, and that I can’t see the 2020 or 2024 nominees being any less so.

      Regarding Trump himself: he’s a malignant narcissist. He is entirely incapable of viewing other people as anything other than tools to be used and then discarded when he no longer needs them, and he honestly believes that he has no significant flaws, so he is incapable of learning from anyone else. All his actions need to be viewed through this prism, and Leah McElrath’s Twitter feed (google it; it’s worth it) breaks them down in a way that provides a deeper understanding of his candidacy than I’ve found anywhere else on the Internet. Because of this, I think the truth is a mixture of Rob’s B and C.

      Democrats still have a lot of work to do – even the 3% chance of Trump’s presidency is far higher than it should be. It shouldn’t even be 1%. This needs to be a 400+ EV blowout. With the latest poll showing a statistical tie in Utah, that looks like it may actually be possible. However, if we don’t GOTV, it won’t happen. And much more work still needs to be done on the House and Senate levels. And then we need to do a much, much better job generating turnout in midterm elections. If we hadn’t screwed the pooch so badly in 2010, we wouldn’t be in this mess with the House in the first place. (Hopefully the DNC becomes more competent with DWS no longer at the helm.)

  • Greg Gross

    It is emotion, whipped up wildly by the MSM in order to garner higher ratings for the entire election cycle. Each MSM camp (e.g. Fox and MSNBC) keeps its tribe in as near a frenzy as possible for that entire cycle. And each tribe is in an echo-chamber of concern, punctuated here and there with genuine panic. Stability is bad for MSM ratings, and so it must be distorted.

  • Paul Benjamin

    I think you’re right; it’s the emotion. There’s so much at stake and so much incomprehension. How could such a buffoon come so far? When Hilary went into that swoon, I feared for the worst.

  • Jay Ulfelder

    Great visualization. I’ll offer three speculations in response to your question:

    1. Availability bias. The people you’re talking to are constantly being bombarded with information about this election, virtually all of which contradicts your analysis. So, even if they have principled reasons to think your analysis (and similar ones) are more credible, it’s really hard to shut out the other stuff.

    2. Substitution effect. Peoples’ strong feelings about the candidate they oppose lead them to conflate their strong fear and anger with a high probability of a hated candidate winning.

    3. Distrust of statistics. Many people still just don’t believe statistical analysis, period. Maybe I should have put this first, because in my experience it’s pervasive.

    • alurin

      All good points, Jay. I would add that most people are looking at national polls rather than the EV totals. And Clinton’s lead, while consistent, has been narrow. I suspect that polling shifts look larger when they are closer to parity. If Clinton’s lead was bouncing around 10 points, it would feel a lot less volatile than when it’s bouncing around 3-4 points.

  • Mike Martin

    I understand that Trump’s polling history has been fairly close to his results, but there is a fear(emotion) that Trump will outperform his poll numbers. Brexit would be cited as a similar example, even though I suspect that the Brexit results were within the margin of error. Unrelated to the actual results, one side is attacking the process, with advanced calls of ‘rigged’ and ‘fixed’, as well as threatening to jail his opponent if he wins. So, yes, emotional, but justifiably so.

    • Matt McIrvin

      There does genuinely seem to have been a systematic polling bias toward Remain that showed up in the poll aggregates. Since it shows up in aggregates of several polls, it was a systematic error, not random sampling error as expressed in the MOE. And it was a disturbingly large error by the standards of PEC or 538, something like 5% in the Leave vs. Remain margin. A systematic bias that big would render the polling in most recent US Presidential elections nearly useless.

      A lot of people have been making facile Trump = Brexit comparisons and arguing that our polls are biased against Trump by analogy, that there’s some reservoir of shy Trumpsters that the polls are missing. One thing that’s different is that the UK had never had anything like a Brexit referendum before, so there was no prior model for turnout. To analogize this to the US you’d have to argue that a Presidential election with Trump in it is such an anomalous thing that all prior turnout models are useless, which seems unlikely.

      Another thing I notice, looking at the historical polling, is that the polling just prior to the shooting of MP Jo Cox favored Leave and wasn’t that far from the actual election results. After Cox’s death, campaigning was suspended for a while, and when it resumed, the polling narrowly favored Remain. The murder might have created a powerful social-desirability bias that wasn’t there before, and made Leavers temporarily shy without changing their votes. Or it might have just created a short-term fluctuation that had dissipated by referendum day.

    • Matt McIrvin

      I’ve been looking at the historical Brexit polling, and there does seem to have been a systematic error there of something like 5% in the Leave-Remain lead, which persisted even in polling aggregates that ought to have cut down on pure sampling MOE. Almost everyone was off in the same direction. That’s a big enough error to be scary in the context of a US Presidential election.

      I’d thought it was a low-turnout election, but turnout was actually much higher than in US Presidential elections. On the other hand, it was a one-off with no prior examples to construct a turnout model.

      One thing I notice looking at the polling history according to Wikipedia is that the campaign was interrupted by the killing of MP Jo Cox by a Leave proponent about a week before the referendum. All campaigning was suspended for a few days. The polling just before Cox was killed seems to match the actual results pretty closely. After campaigning resumed, the results have this systematic bias toward Remain.

      Maybe the shooting created a short-lived social-desirability bias in the referendum-eve polling that wasn’t there before.

      (My previous comment on this seems to have been eaten; not sure why.)

    • Matt McIrvin

      …Also, just the fact that the polls instantly swung that much as a result of the Cox killing should have made people nervous about the volatility of the results. Leave had been slowly gaining for months and the polling was neck and neck with Leave narrowly favored, then there was this sudden swing toward Remain in the aftermath of a shocking event at the 11th hour. I suppose a possible lesson is that those kinds of things don’t necessarily translate to votes.