Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

Final Projections 2016

November 8th, 2016, 12:45am by Sam Wang

(Updates: 6:06am data for Presidential and Senate, and added confidence intervals. 9:00 am: more description, also variance minimization.)

Here are the final snapshots. Four Senate races are within one percentage point: Indiana, Missouri, New Hampshire, and North Carolina. Partisans there may want to lawyer up for possible recount battles.

Soon I’ll put out a brief Geek’s Guide to the Election. Also, live blogging starting around 8:00 pm.


President: Hillary Clinton (D).

The Presidential estimates are based on the current snapshot in the right sidebar, except for the most-probable-single-outcome map, where variance minimization was done to give a more stable snapshot for North Carolina, Clinton +1.0 ± 1.0% (N=8 polls).

Most probable single outcome (shown on map below): Clinton 323 EV, Trump 215 EV. This is also the mode of the NC-adjusted histogram.

Median: Clinton 307 EV, Trump 231 EV. Meta-Margin: 2.2%. One-sigma range: Clinton 281-326 EV. The win probability is 93% using the revised assumption of polling error, +/- 1.1%.

(Why doesn’t this probability necessarily match the probability in the snapshot histogram?)

National popular vote: Clinton +4.0 ± 0.6%.


Where possible, variance minimization was used to identify a time window that gave lower variance than the standard time window.

Mode: 51 Democratic/Independent seats, 49 Republican seats; the most likely single combination is shown in the table below.

Median: 50 Democratic/Independent seats, 50 Republican seats. (average=50.4 ± 1.1 ; the 1-sigma range rounds to 49 to 51 seats)



Generic Congressional ballot: Democratic +1%, about the same as 2012.

Cook Political Report-based expectation: 239 R, 196 D, an 8-seat gain for Democrats.

Tags: 2016 Election · House · Senate

168 Comments so far ↓

  • Matt McIrvin

    You’re picking the 323 mode over the 308? Gutsy.

    • Evan

      I think there must be two different outcomes that produce the 308 total.

    • Mårten Fjällström

      There is a couple of scenearios leading to 308.

      Limiting myself to flipping states has in the “barely GOP/barely Dem” piles, I find as the 323-map but Trump wins North Carolina or Trump wins Colorado and Nevada or Trump wins New Hampshire, New Mexico and Nevada.

      Then there is even less likely scenarios where Clinton wins an additional red state like South Carolina and loses North Carolina and Colorado. And so on.

      So the 323 map presented is the “Most probable single outcome”, while 323 not being the most probable number of electors. Or at least that is my understanding of it.

  • Anthony Shanks

    Thanks for your final numbers Sam. Will you be doing any live blogging or podcasts tomorrow?

  • cvermar

    Hoping likely voter screens underrepresented Hispanics and disaffected Rs (esp. women)… but I could live with this. Would love to flip the House but not holding my breath on that.

  • Matt McIrvin

    I’m thinking NC is more likely to go to Trump; I’m giving him the margin thanks to the state government’s vote-suppression measures.

    Also, at least one disgruntled WA elector is going to vote for Jill Stein or write in Bernie or something. So, remaining optimistic about ME-2, that leaves me with Clinton 307, Trump 230, 1 other when the final count happens.

  • Edwin

    Long time follower, first time commenter. Excellent site, beautiful mathematics/statistics, and a cool, cold approach. Much respect from CA-10, a close race due to Trump.

    P.S. To think what you would do with more than a “shoestring” for a budget.

  • Duff

    Have appreciated you efforts since 2012 cycle. Notwithstanding your statistical models, your level headed sensibility has my profound respect. Give a heads up when your next at Caltech in Pasadena and we’ll drum up a group to cheer your social science achievements.

  • George H

    Did I miss IL Senate Kirk/Duckworth?

    • Edwin

      I think the above table only considers the so-called, knife-edge, or close races. It seems as if Rep. Duckworth will coast to victory.

  • JeffE

    Ugh… wish Senate picture looked better. I think Bayh is toast in IN, so to me looks like a real coin flip. Hopefully Clinton can drag Ross across the finish line in NC too.

    • Vatnos

      Obama was able to drag Purdue across the finish line in 2008 in NC, despite her polling slightly under McCrory for much of the cycle. Hoping the same thing happens this time for Ross.

  • David Elk

    Thanks for everything, Prof. This site is such a great resource. Being able to learn so much, and intelligently donate downballot, were high points to this stressful race.

    It’s not incredibly useful, but here’s one final animated median EV:

    Perhaps crazy, but: looking forward to tomorrow.

  • shma

    Please post your 2-sigma intervals for the EV count and the senate as well, if you can.

    How are you assigning NC as blue in the modal result when your margin is Trump +0.5% and your probability (as reported on the Upshot) is 57% Rep?

    • shma

      NC, at 57% R, was one of the few states where your model really stood out from some of the other models (particularly Kos and Huffpo).

    • anonymous

      Is the MM still running? Because it just lost .3 and about 6 EVs.

  • Jeff W

    Thank you for everything you do.
    If you had to pick one state where maybe the numbers might not match the result for whatever reason (poor polling, late breaking undecided voters, herding in polling, who knows), which state would you pick?

    • Sam Wang

      President: Clinton may win the near-tie in North Carolina. A little nationwide polling error in her favor would be enough to do it.

      Senate: Indiana, I wonder if Young will win. Bayh was leading, but has been headed downward. Not much data there.

    • anonymous

      Thank you for making that clearer. I think NC is going to be razor thin due to the electioneering by the state R’s. Bayh was done in by his own mini scandals it seems. My actual concern is NH right now. Seems to really be a coin flip.

    • Tom_b

      NC will be tight.
      In our favor:
      1) vigorous GOTV. Bee hive busy field offices. They had me drive an hour out of Orange County to canvass a smaller county Saturday. MANY of my neighbors have been out in the field.
      2) deeply unpopular Governor (HB2 is 50 unfavorable/ 35 favorable and the law cost the state hundreds of millions)
      3) mostly a bit ahead in the polls. Excited populace; lots of signs around. Turnout favors Democrats.
      4) Over 1000 federal election observers.
      Against us:
      1) residual voter suppression, though a lot was overturned recently by court order.
      2) still a purple state.
      3) some reports black early vote down from 2012, which was already a bad year.
      4) Ross (Senate) consistently underperforming Burr in the polls.

  • Edward

    How likely is it that a massive (D) turnout could just barely tilt Ohio into the Clinton column, if early voting reports about Hillary having a EV lead in OH mainly due to the women vote? BTW, thanks for staying through the course and not settling for click-baiting. Happy Election Day!

  • amk

    Thanks Sam for all your work.

    I am happy that my other go-to guy, Drew Linzer, also has the same 323 EV projection.

    Fingers crossed.

  • Walter Manny

    New Hampshire is interesting. Tiny-sample midnight voting just about replicated primary voting by party. Back in that primary, Trump beat his final RCP avg. by 4.1%, and Sanders beat his final RCP spread by 11.1%. Not sure what those differentials were at PEC, but that state should worry anyone.

    • The Liberal Crab

      You are joking right? You are actually going there trying to do an analysis on a non-random tiny population.. I mean, I guess I’m not surprised, someone was going to do it. Just surprised someone on sam’s site would.

    • Walter Manny

      Not at all. The first sentence was simply a lame kickoff, sorry. The non-predictive polling was what worried me.

  • Veronica

    Dr. Wang, considering what we’ve been through this election season, what with all the worries, guesses, nail-biting, and excitement, you should write a WHOLE “Geek’s Book to Elections”!

  • Paul Stein

    Thank you Prof. Wang for providing a safe port in this polling storm. The insightful comments backed up with mathematics is a wonderful change from this fact-less pundits.

    At this point in the election, data science will give way to the will of American population.

  • Some Body

    If NH Senate is exactly 50/50 (according to the table), what was your “tie-breaker” to pick Hassan?

  • Jared Mounts

    Writing in for the first time just to thank you for all you have done in this election cycle Dr. Wang. In what is basically your free time you have explained the data more concisely than anyone else. I know myself and many others are interested in seeing your work for years to come.

  • NickT

    I am putting my virtual credibility on the line and saying Clinton gets 353 electoral votes, the Democrats take the Senate 51-49 and pick up 12 seats in the House.

    • Paul Browning

      So in-line with the LATimes prediction you think Clinton can pick up Ohio and Arizona ?

    • Greg Gross

      Unlike NickT, as an English and History double-major with one “Poet’s math” class under my belt (and no numbers whatsoever in law school), I HAVE NO virtual credibility….

      So: I predict Hillary wins. That’s it.

      MM makes me squirm though. 2.2!

  • Hannah

    Sam, I’m currently debating someone who believes it is impossible to predict an election with a probability upwards of 99%, or even 90-95% and I really, really want to prove them wrong.

    Can you explain why you are able to come up with a probability that high?

  • A New Jersey Farmer

    The question about whether Hillary’s GOTV effort will enable her to overperform the polls will be answered tonight.

    • Josh

      Yes, but I suspect if Hillary does outperform her polls it will be more because the polls were off than because of her GOTV operation.

      Where we’ll see if the GOTV pays off or not is in whether or not she wins states that are on the knife-edge: North Carolina, Ohio, Florida, Iowa.

    • BrianTH


      I think that could be a tough question to answer. If there is a big surprise in Clinton’s favor, it looks to me like it might be driven by lower-propensity women and Hispanics ending up with higher voting shares than the pollsters allowed for in their demographic models. So on the one hand that would be a polling error, but on the other we could ask WHY those low propensity voters were turning out. And the answer to that question could include GOTV efforts, although of course it could also include more fundamental forces relating to the choice presented in this election.

    • Josh

      Sure. In my admittedly limited experience, my understanding is that GOTV is most useful and effective when deployed in states where electoral votes can be won by motivating a relatively small number of extra people to vote. In a tossup state like North Carolina, just getting an extra 30,000 or 40,000 people to the polls can win you 15 electoral votes. That’s where GOTV matters most.

      I definitely agree that a more generalized approach to increasing turnout–concerts with pop stars, celeb endorsements, targeted advertising, etc.–can also bear fruit. I’m not sure I’d call this GOTV, more like just savvy campaigning–although maybe we can agree to disagree?

    • BrianTH


      I’m actually not sure we are disagreeing at all. My point is more that it could be difficult to figure out what role GOTV efforts are playing, given a reasonably narrow definition of what that means, in light of all the other things going on.

      What would be really interesting to see is some sort of study where you looked at low-propensity voters in states with less of the narrowly-defined GOTV efforts versus states with more, to see if you could detect a difference. And maybe people will do that!

  • William

    Congratulations! Wired Magazine: “2016’s Election Data Hero Isn’t Nate Silver. It’s Sam Wang.”

    • Matt McIrvin

      We shall see, eh? Even if Hillary wins, a large polling miss, especially if Trump does relatively well, would probably lead to a general rehabilitation of Nate Silver in the media.

      Personally I’ve come to suspect that Drew Linzer’s or The Upshot’s uncertainty estimates were probably the best this time around. The way the story we got from the polls fuzzed out at the very last minute is frankly unnerving; that didn’t happen in 2012 or 2008. It’s more like 2004.

  • Rahman


    I sure hope you’re right today! I do not want to have to live in a world where a blatant racist, xenophobic, misogynistic, man child becomes the president of this great nation.

  • Stuart Levine

    I know that this is a question that you cannot empirically answer, but do you think that Comey’s Friday Surprise cost HRC Ohio and the D’s North Carolina and/or Missouri and/or (possibly) Indiana?

    • Andrew

      No, IIRC the current trend in polling started just prior to Comey & the FBI’s election interference.

    • Matt McIrvin

      HRC never led in Missouri or Indiana, or really came close to doing so.

      I think the Comey business depressed her numbers a bit, but it was mostly just that it had been long enough since Trump did anything sufficiently embarrassing to dominate the news. As Andrew said, the slide actually started a few days before the FBI letter dropped.

  • Stuart Levine

    One thing that I forgot–I know that I speak for all of your regular readers: Thank you very much for you efforts. This is truly a remarkable project.

    Can I extract a promise from you to come back in 2 years and then 2 years after that?

    • NickT

      I’m hoping Sam will return in 2017 for the Virginia special election to replace Vice President Tim Kaine (and whomever McAuliffe appoints pro tem) in the Senate.

  • Froggy

    If you’re going to predict, predict big. (Wayne Allyn Root taught me that one.)

    Clinton 368, Trump 170. (Sam’s map, with Ohio, Arizona, and Georgia also for Clinton.)

    Democrats win the Senate 52-48. Maggie Hassan, Katie McGinty, Deborah Ross, Jason Kander, Catherine Cortez Masto, welcome to the Senate.

    Sheriff Joe will finally go.

    Everyone get out the vote!

    • Josh

      I’m on board with this. I think Ohio and Iowa go blue despite the map. Not sure about Arizona, though.

      52-48 in the Senate sounds right.

      I have Clinton with 347 EV. Feel free to roundly mock me for this tomorrow if I’m off.

    • Jim Wright

      Ohio is my “upset” state tonight. Lots of good polling and a superior GOTV effort. Early vote totals came in higher than 2012 in Cuyahoga and Franklin counties, I think HRC wins Ohio

    • BrianTH

      I think if there is a surprise today, this is the more likely direction. In particular, I think it is looking like the pollsters may have assumed away a surge in relative female voting shares that could systematically shift things.

    • Indomitable Ted

      Dems in Arizona may get boost from having cannabis on the ballot. As I recall, a 2 point Dem bounce above the polling median was particularly noticeable in the midterms in 2014 for states with a cannabis initiative on the ballot. Wasn’t enough to save Dem’s chances at the time, but this little core of hard to reach otherwise unlikely voters may help pull off an upset in Arizona.

    • RonL

      I was at 340 ten days ago and I’m sticking. That’s the 323 you see on Sam’s map plus Ohio’s 18. Minus the one EV from Maine. Didn’t check the Maine EV polling, if any, at all. Just liked the rounded number!

    • David N

      I think RonL has nailed it with Ohio going to Hillary. Ohio’s polls were clustered around Comey’s “reopening” comment, with very little polling data taken more recently. There were very many people voting early this weekend and heavy volume at my prototypical Republicans-for-Hillary precinct.

    • Susan Sullivan

      Agree with Jim Wright that Ohio will go blue. Local news is reporting very heavy turnout and long lines. And early voting was smashing.

  • Bruce


    I’m a long-time lurker and admirer. As a developmental neurophysiologist by training who morphed into a code geek, your technique and spirit are superb. Thank you so much for your (and your colleagues) hard work, insight, and cleverness.

  • ravilyn sanders

    Dr Wang would you approve of some GoFundMe fund raising to support one or two grad students to seed a open source poll analysis research ? How much it takes to fund a grad student in Princeton?

  • Ed Wittens Cat

    merci mille fois Dr Wang
    I went from not even going to vote to working on Ivy GOTV in the space of a couple months because of PEC.
    You and Drew Linzer are isomorphic on EVs– the reproducibility of rigorous maths.
    Your simple elegant approach has at least partly restored my hope for the future.
    gratitude also to the best commentariat on the interwebs– i learned a lot from you guys!

  • Matt McIrvin

    So, some scary polls dropped overnight, I guess? The Meta-Margin and the histogram seem to have suddenly plunged at the last minute.

    • Matt McIrvin

      (and I can’t see what changed, really–Michigan tightening up?)

    • Josh

      SurveyMonkey had polls with Michigan and Pennsylvania tied. I think that’s what it was–nothing else really changed.

    • Jon

      Seeing (worrying about) that sudden change in the MM. It’s been stable for days within 0.1% daily fluctuations. Wassup?

    • Matt McIrvin

      Well, it’s all about GOTV now.

    • Matt McIrvin’s model plunged too, but in their case, it was because of large Trump outliers in Nevada and Florida pulling their mean-based averages down. Here’s hoping these aren’t all fragmentary indicators of a nationwide shift toward Trump at the 11th hour.

    • BrianTH

      Michigan hasn’t been polled much. I am not presently worried about it.

    • Paul Stein

      Interesting, I thought the enumeration based approach to modeling was less sensitive to potential outliers. Any ideas on how many paths to 270 are used in the model?

    • Sam Wang

      The convolution approach here includes all permutations that lead to each EV outcome.

    • Andrew

      I have been mystified by the polls over the last week or so. The house race polling was stable for about a month and then it nose dived right at the end with three polls showing Republicans in the lead. Similarly, the Upshot has been estimating early voting in North Carolina with their prior poll as a base which showed Clinton +6, but their final poll shows the race as tied. A similar nose dive happened in 2014 for the senate, but then the polling was really mixed through the pre-election period (& we know how that turned out…).

      So what gives? Why are the polls just before the election so much less favorable to the Democrats? Is this just the effect of having more polls close to the election so their average becomes a better estimate of the actual voter preference?

      On the other hand, more than 50% of the NC voters had already voted by the time the final upshot poll was collected. Since Democrats have a greater tendency to vote early, maybe they aren’t so interested in answering phone calls from pollsters any more after they voted and it is skewing the final polls. Time will tell what the right answer is- but I note that the most likely EV result Prof. Wang shows above resembles the polls from a few weeks ago, but not the more recent ones.

    • Paul Stein

      Convolution approach, cool, more to read,

      Thanks Dr. Wang

    • BrianTH


      This is just a preliminary notion at the moment, but I am noticing more and more that the likely demographics based on EV patterns are not necessarily being reflected in the demographic models being used by pollsters. And I wonder if that could be causing a growing possibility of error if various factors are also leading pollsters to miss a bit on the margins in some of what would end up as underrepresented groups.

      But we shall see.

    • Joel W.

      The number of state polls (191) is down from what it was by a lot, no? Or am I misremembering?

    • Matt McIrvin

      @Joel W.: No, you’re right. There are fewer outfits paying to do polls this year. I read somewhere that poll aggregators like 538 and PEC may be killing their own source material by reducing the demand for individual polls–people are wising up and not hanging on individual poll numbers, so there’s less incentive to produce and publish them.

    • RonL

      Upshot has Trump with a 16% chance of victory. That happens to be the same percentage chance that Clinton has of winning Mississippi.

      Puts the percentage chances into context.

    • Joel Wapnick

      @Matt McIrvin, no, I meant this election season. I thought I saw a number well over 200 not long ago. I may have misremembered, though.

    • Andrew

      Just noticed that pollster posted some new polls running through yesterday with range of Trump +1 to Clinton +8. That gives us five polls ending Nov. 6 or 7 at C+8, C+1, C+0, T+1, T+1. Median C+0, but oh wow look at the spread!

      I think BrianTH is right and some pollsters likely voter models are wrong, but whose?

    • BrianTH


      I haven’t done a comprehensive survey, but someone else tipped me off to the fact a lot of pollsters seem to have a 51-49 female/male ratio, sometimes even less. I spot-checked a few odd-looking polls and that tended to be true.

      So I’d guess all of those polls are a bit off, meaning I suspect the female/male ratio in most (maybe all) states and nationally is going to end up higher than 51/49.

  • Runner

    Obama’s net approval chart shows his approval at its historical high of 8.5%. Yet, HRC’s current MM at 2.2 is much lower than the 4 to 6 MM she was at in May through Sept., and HRC’s current EV estimator of 307 is lower than it has been nearly all year except for her September low point. What correlation is there, if any, between these numbers? And what effect might they have on who will control the Senate after today’s election?

    It seems curious to me that a D President’s approval would increase significantly while a D Presidential candidate’s MM and EV would fall during the same period of time?

  • Aaron

    Sam is there a possibility you’ll do a post-election podcast discussion with Nate Silver to discuss the differences in accuracy of your models and methods?

    • Indomitable Ted

      Unlikely Aaron. Those two have a bit of an unpleasant history (hence Sam referring to him as the Argental Satan in a number of posts this year). It would take a lot for them to sit down and break bread considering that Sam has no incentive to.

    • Aaron

      I only discovered Sam’s prediction on the recommendation of another statistician in the middle of October so I’ve got plenty of back-reading to do. As a fellow stats-enthusiast I always appreciate the details and the little camps that emerge as a result of differences of opinion. I’m a convert!

    • truedson

      In 2012 the two had an amiable discussion on NPR before the election. But 538 seems much different from the NYTimes days. I went to it regularly then.

    • Indomitable Ted

      The issue was primarily Silver attacking PEC’s model out of blue some time in 2012 or 2014 as I recall. He set off this weird Twitter war and didn’t even accurately critique Sam’s model. It seemed like a petty publicity stunt on Silver’s part so far as I can tell, but it kinda pissed off Sam. As I said, I don’t think he’s forgiven him to this day.

    • Sam Wang

      That was an escalation on his part. I had said critical things of him beforehand. It is a long story…best ended peacefully somehow.

    • Indomitable Ted

      Ah yes, I remember now, Sam. You gave basically an academic critique of his model and then he kind of exploded on you. That’s why I remember it feeling like a publicity stunt–the vitriol seemed disproportionate to what you’d actually said.

  • Robert Truax

    It appears in the EV histogram that only 95% of outcomes show a Clinton win. Why is the final headline probability still at 99%?

  • Psych

    This site’s transparent presentation of the stats behind the interpretation of the polling data has been a beacon, for both this election and back to my stat & research methodology courses for my doctorate. Thank you not only for cutting through the data but reassuring me that I still have enough of a grasp on those courses to recommend this site.

    I have 2 questions:

    1. How would unexpected major events that occur within 1-3 weeks before election day be incorporated?

    2. Today for North Carolina in the right column there’s “Trump +0.5%” but the state is predicted to go to Hilary… please explain.

    AGAIN, THANK YOU & your team (research lab) for taking the time & energy to run this site!

    • Sam Wang

      1. Not incorporated, except in the error parameter, much discussed the other day. To be totally honest, I will probably set that parameter higher for the next election – assuming opinion polling is recognizable at that time.

      2. That’s the auto-updated margin. We will soon manually update it to reflect the different time window.

    • Psych

      Thank you, Sam, for responding.

      re: “assuming opinion polling is recognizable at that time”

      I’ve wondered a lot about sampling .. how to approach it “in this day and age” seems like a great dissertation topic

      Again, hats off to you and your team for clear presentation of methods

      PS: I like your lab’s day job research & willingness to publish for the lay person.
      Warm Regards to all!

  • Steven

    Gotta go with the median, straight-up, no choices: 307 EV for Clinton.

    Sorry to say it, as in the last month or so 332 was my number, with an upper limit of 356 lower bound of 308. That was before I knew about Maine. But today 307 is the aggregated wisdom of 191 state polls.

    Same for Senate. Gotta go with the median: 50-50.

    • Richard

      My prediction is 308. News out of NC suggests to me that vote suppression there has probably made it very hard for Clinton to pull that one off despite polling. And I have a gut feeling that Maine is not going to split.

  • truedson

    Interesting discussion on New York Magazine today. I don’t think they know much on methodology or mathematics….

    Jebediah Reed: Poll-watching and election prediction have gotten even more heated than usual this week. There is a rift between Nate Silver, who sees the race as pretty wide open (Trump with a one-in-three chance of winning), and some other generally left-leaning statisticians who see the race as over, a Clinton win basically guaranteed. How certain are all of you that Clinton will prevail? What has you most concerned that the race might be more competitive than the consensus suggests?

    Jonathan Chait: Silver’s right that the size of her lead is within the margin of a non-shocking polling miss.

    Ed Kilgore: There are actual polls run by reputable people showing Trump ahead nationally and in most of the hotly contested states. If I knew more about polling methodology, I could maybe tell you whether Nate’s critics are engaging in a nerd fight or are convinced Trump’s odds increase whenever someone says he might win.

    • Ed Wittens Cat

      Trust the Math, Padawan.
      remember, Data over Drama.

    • Pat C

      Great material. BTW, I will definitely be stealing EWC’s mantra “Data over Drama.”

    • Olav Grinde

      Intentionally or unintentionally, “left-leaning statistician” seems like a slur. It almost implies that the political preferences of professionals such as Sam Wang, Drew Linzer and the NYT Upshot team make their statistical computations “lean”.

      That is hardly the case!

    • Ed Wittens Cat

      Pat !
      its Dr Wang’s mantra, not mine.
      Nates mantra was “demographics is destiny”– seems like he should have at least considered demographics in this election.

    • Ed Wittens Cat

      its my hypoth that all mathematicians and physicists have bluebrain genetic tendency.
      Consider the scientists of the AltRight– genetics and polysci only– only the soft sciences.

    • Catalie

      “its my hypoth that all mathematicians and physicists have bluebrain genetic tendency.”

      It’s my belief that what you’re seeing is caused by reality’s well-known left-wing bias.

    • Sam Wang

      And yet engineers often tilt right.

    • counsellorben

      Must be some gyroscopic effect ;^)

    • Amitabh Lath

      Yes math and physics people are mostly Democrats but that is more a statement about how anti-intellectual the Republican party has become.

      The recent rejection of climate science was preceded by the the Reagan era push for star wars (hit a bullet with a bullet nonsense).

      My thesis advisor, the late Nobel Laureate Henry Kendall, said the Nixon administration asked academic scientists to find a way to win in Vietnam. There were echoes of the Manhattan project. When the answer came back negative, rather than accept it, they started up their own group of scientists, who would give them the answers they wanted to hear. That was the beginning of the cleavage.

  • CBC

    I’m sure this has been addressed before. Apologies. How is it that Clinton’s win probability is 99% when the chart of possible EV outcomes clearly has at least 5% of the outcomes as a Trump win?

    • Sam Wang

      The histogram shows the standard deviation of all possible outcomes. The win probability is calculated separately, using the Meta-Margin. See the 99% post.

    • William

      And with the meta margin now at 2.2% what would be the “adjusted” win probability IF you applied the “reasonable uncertainty” factor of the Meta-Margin to 1.1% that you speculated recently to be the better value?

    • DanH

      Sam, understand that win prob is based on 2sigma EV median. What do you think of tying the “reasonable uncertainty” factor in some way to the size of the outcome tail cross over rather than using a constant. Surely at some point the number of potential losing outcomes must matter? As it is many of us experience cognitive disconsonance trying to reconcile the two.

  • brian

    Long time reader first time commenter. Love the site Sam! But in my household we are freaking out to see the meta margin drop overnight! Even as it seems a batch of last minute polls were good news for Clinton.

    We would GREATLY appreciate it if you took a second to explain to us why the margin would drop so much so fast! And why we should not be freaking out about it. Thanks!

  • Mike Martin

    Thanks Sam! Not only for your hard work on the polls, but for your ‘punditry’. You have supplied sanity where many others seem to speak just to be heard.

  • Ruth Rothschild

    Thanks for being my island of sanity in this crazy election. Your logical, objective, and scientific analyses have really helped to make alot of sense of the many polls. Thanks also for explaining your statistical methods in a way that a non-statistician like me has been able to understand. Much appreciated. And, thank you for the many interesting articles, podcasts, etc, and most of all for the time and effort that you’ve put in to running this site. I’ve recommended this site to a number of people who have also found it to be great. Thanks also to the many people who have commented on the various topics presented at this site, for your insightful discussions and questions, and for having civil discourse in a campaign and election that has been anything but civil. I look forward to coming back here not only tonight for the blogging and tomorrow for the post-election discussions but also again for future election cycles. PEC will now be my go-to site in future elections (along with HuffPost).

    Thanks again, Sam, for all you’ve done in hosting this site. Happy Election Day!!

  • Charles Stanton

    I have been saying for 6 months HRC will win Arizona. Darn it, I’m sitcking with it and counting on the Sam Wang/Beyonce’ bounce :)

    • SJWangsness

      Me too — HRC’s GOTV, Latino surge, Mormon dislike for DJT, plus fall-off in late GOP voters dispirited after early returns from the East show HRC winning FL and NC, essentially ending the contest.

  • Alex Wilkinson

    Interesting that for all the differences in methodology, Sam and Nate settled this morning on exactly the same electoral map as their final prediction. Same states going for Clinton. Of course, they differ a lot on the probability of that outcome!

  • Rob M

    My wild hot take prediction is that Metcalfe will win the Alaska Senate seat.

  • gumnaam

    I learned a lot of new things from this blog this election season, with the Google Correlate and neighboring county approaches probably topping the list of my favorites. Thinking beyond this election though, I hope that President Obama reaches out to Sam when he starts his work on combating gerrymandering and tackling the problems faced by Democrats in state legislatures. Thanks for sharing your insights and analysis, Sam. Best wishes.

    • Mike Martin

      Here Here!

    • marcos

      If Obama did not want gerrymandering, then he would not have screwed the pooch with the ACA in 2010, a census year, when the backlash combined with demoralized base threw the state houses to the Republicans.

      That was an unforced error in a climate of asymmetrical political warfare when one side is playing for keeps and the other side could not be bothered.

    • gumnaam


      I think Obama was not thinking politics as much as policy solutions in 2009-2010. This is admittedly a little bit of an Achilles heel for him, as the two are not completely separate. Nevertheless, I would say that his decision worked out well for the 20 million people who now have health insurance, even if we ignore the other stuff like the real bend downwards in the health costs curve.

  • Rob M

    For all the people looking at polls that were dumped last night or today, stop. This is equivalent to checking the weather forecast when you’re in the eye of a hurricane. I am a bit of an amateur meteorology geek and the term used is nowcasting, stop looking at what the predictions and models said and start looking at what is actually going on.

    Appropos of nothing wishcasting is another great weather geek term that can be applied to politics. You see this in the people who unskew polls or point to rally size as favoring a Trump landslide.

  • Aravind

    My small family has found PEC to be a tranquil oasis amidst all the chaotic news. I stopped watching TV many months ago and have been getting all the election information I need from PEC.
    Thanks for your excellent work, Sam.
    Just Brilliant.

  • Michael B

    Sam, thanks for being an island of sanity. However, I think you have counted NV senate seat projection as a gain for Dems, when it would be a hold. Therefore Senate projection should be 50/50 not 51/49, no?

  • Ed Wittens Cat

    Right on cue– Wired
    Dr Wang is woke…and lit :)

    • Ravilyn Sanders

      Dont listen to that idiot. Dr Wang’s web site is beautiful. It is clean, it is fast, and it works. It does not have stupid eye-candy that slows down. The animated menu and dynamic zoom on the cursor look nice, for about three seconds. Then they are nothing but irritations.

  • 538 Refugee

    I’m busy babysitting the granddaughter today and nothing is more important. Just doing a quick check before I get her outside to take advantage of what will probably be one of the last nice days we have her for this year and ran across this.

    Let’s label this “Thought of the day”.

    Susan B. Anthony died without the right to vote. Now people are covering her tombstone in ‘I voted’ stickers.

    • 538 Refugee

      OK. Granddaughter is taking her nap. I’m skipping mine for today. I’m gonna wimp out on predictions this year. Generally I look at polling in close states and decide by the reputation of the pollsters. This cycle they seem to be on different pages. The Columbus Dispatch does a mail poll and has a good track record. They have Hillary +1 so Ohio might be closer than we think. Last night Chuck Tod punched Georgia and Arizona as blue in what seemed like a ‘think this could happen moment’.

      I’ll ‘wishcast’ a bit hoping for a wave that probably won’t happen but probably NEEDS to happen for the sake of the country. There is still a record backlog of judicial seats open as Republicans stall. Expect a couple retirements on the left of the Supreme Court once Hillary is Nominated.

      Overall, PEC’s map is reasonable and I’ll accept it as a baseline. ;)