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

The real news today…downticket

September 13th, 2016, 8:42am by Sam Wang

It is good for a cheap laugh to flay the media for its obsession with horserace. This week’s ongoing ruckus with Phlegm-ghazi confirms that reporters cannot get out of their mental rut of some older storyline. In this case, the storyline is “Clinton is secretive.” Let us pause for a moment. She was concealing her pneumonia because the press would make a big deal out of it. And, wait for it…the press made a big deal out of it.

You, Dear Reader, are complicit in this. I notice that more of you click Presidential links than on the nifty Competitive Congressional District Finder. You like the Presidential horserace. My reason for generating the best prediction I can is to reduce the noise of campaign news. I thought it would clear mental space for thinking about policies, or downticket issues.

The Presidential forecast [methods] takes a low-noise snapshot of state polls, then adds possible drift based on recent elections and this year. Because of intense polarization, few voters are movable. The calculation says that Clinton’s win probability is 90%. The Senate forecast does the same [methods], but also factors in Presidential-year or midterm-year bias. It says that Democrats’+Independents’ probability of taking control is 72%, which is in the 20-80% range, meaning that things could really go either way. Other forecasts tend to count uncertainties twice, or to overestimate how movable voters are. Other forecasts are also under commercial pressure to attract eyeballs.

Still, the comment section is still peppered with anxious questions about Clinton’s chances. Honestly, some liberals can be total ninnies. You don’t see the conservatives in hysterics…though actually, here is their version of a meltdown. I take it back. You go.

Here are some news items that matter more. In Minnesota, the Democratic-Farmer-Labor party (similar to Democrats elsewhere) failed in their lawsuit to keep Donald Trump off the November ballot. At a literal level, this didn’t matter because Minnesota is a safe win for Clinton. The real importance was in the three close House races in Minnesota, as you can see using our Competitive Congressional District Finder. In addition, the DFL is within striking distance of taking back a legislative chamber there. The lawsuit was a tacky move, but it would have been effective.

In a second news item, the National Collegiate Athletics Assocation (N.C.A.A.) decision to take basketball and other championship games out of North Carolina is likely to have repercussions. North Carolina is crazy for basketball. The trigger was North Carolina H.B. 2, an anti-bathroom-access law that is directed at transgendered people. Republican Governor McCrory was the force behind it. However, he is already a median of 7 percentage points behind Democrat Roy Cooper in polls. Any possible anti-Republican backlash is more likely to have downticket effects on Deborah Ross’s effort to unseat incumbent Republican Senator Richard Burr. She is currently behind by a median of 3.5 percentage points. On average, between now and Election Day, Senate polls move close to 4 percentage points toward the winning Presidential candidate. That election should be close.

Ross is one of five Democratic women running in close Senate races. Their outcomes will determine which party controls the Senate after the election. No matter which side you support, these races matter quite a lot. You can contribute at ActBlue if you’re a Democrat, or at the N.R.S.C. if you’re a Republican.

Tags: 2016 Election · President · Senate

91 Comments so far ↓

  • Percysowner

    I clicked on your map! I had to go 150 miles to find a competitive House race and it’s not even in my state. If I go to 250 I get a whopping three. I live in Ohio where, apparently, Strickland decided that his big goal was to be nominated to run for the Senate then do nothing else. What had a chance to be competitive, just isn’t any more.

    I’m still going to canvass and register voters, because I’d like Ohio to go for Clinton, but according to you, there’s not much else for me to do.

    • 538 Refugee

      Strickland is advertising in the Youngstown area. I wouldn’t say effectively, but it is there. Portman’s pac got a jump on him and he’s basically responding. It is a good response but he isn’t aggressively pointing out his strengths at this point. He’s indicated he has more to come though and is just getting started. I don’t like his chances though. I voted for his opponent in the primary.

  • Jack Tenold

    Thank you for this calming missive. I’ll admit it; I’m worrying more than I should. I think.

  • Jeff

    I’m a physician and I find the coverage appalling. I certainly see people in the ICU with pneumonia, but more often, I see them in the office, give them antibiotics, and tell them to take it easy. I certainly wouldn’t send out a press release! Would you expect the campaign to publicize a urinary tract infection? Hemorrhoids? They can all be serious, but do you actually expect them to be newsworthy?

  • Sophia

    When the video of Hillary came out and friends started calling me scared I merely asked are you voting for Trump? Of course they went silent and said good point. Then I point them to Act Blue.

  • Sam

    Thank you as always for your clear-eyed analysis!

  • Anonymous

    Agreed on the obsession with presidential race. As a constructive suggestion of what you can do about it, consider changing your masthead to reflect and direct folks. Some ideas:
    1. balance the prominence of senate and presidential race to two lines each.

    2. post some links to top 3 races to watch with their current probabilities, so folks can click and not have the mental block of 33 races to watch

    3. why not also include two more lines for a house summary similar to senate. I know it is unlikely to switch, but just for completeness and help further distribute attention away from presidential race.

    4. A three-column format for president, senate and house would also be fantastic, if not too complicated to add.

    • Rachel Findley

      Yes, please. Give us some big blinking buttons to click on for the downticket races. We are all easily distracted at this point.

  • bks

    Let’s posit that the optics of Hillary wobbling into a van will have an effect on the election. How long will it take to show up in the state polls? I’d guess two to four weeks at a minimum. I’m taking a nap till October.

    • Kanwaljit Singh

      By then there will be debates happening, so I am not quite sure if there will BE any effects of her health scare.

  • Robin Privett

    Thank you for this information. I come here every day to see what you have to say. :)

  • Michael

    Can what you’re saying be translated to: “Whatever is driving a current narrative about Clinton — the “deplorables” remarks, the pneumonia video — does not change already solid opinions for or against her? And, if so. what do we make of the Trump camp’s emphasis on his apparent lead with independents?

    • Sam Wang

      Well, we will get more evidence on this soon. The whole pneumonia thing seems like it appeals mainly to partisans. Kind of like George H.W. Bush puking on the Japanese Prime Minister back in the 1990s, which was mainly of interest to his opponents. Only pneumonia is less gross.

    • Matt McIrvin

      I’ve been seeing the pneumonia thing get some interest among Bernie Sanders supporters who have rekindled hopes that Clinton might drop out and the DNC pick Sanders as the replacement nominee. I’ve been trying to bring them down gently.

    • Phoenix Woman

      What Matt said. Those of us in the reality-based community and who aren’t interested in relitigating the primaries aren’t the ones driving this story.

  • Lil Sister

    I’m afraid a lot of your readers are using expectation — multiplying the probability of Trump’s winning times the horror of it. Thanks for keeping us almost sane, though.
    Meanwhile, downticket in CA 49 it’s thrilling to see a small, but nationally recognized, chance of dumping Issa from our previously safe seat.

    • Emigre

      Indeed CA49 could be thrilling but once you see the financial advantage Issa has ($3.7M vs $0.13M) reality sets in:
      Time to donate more to Applegate!

    • Bruce Sands

      Probability times consequence is how risk is calculated, so perhaps we aren’t overreacting. If the consequence is high enough we need to pay attend to low probabilities. It would be great to send Issa into retirement, even if for two years.

    • Phoenix Woman

      Issa’s spent millions on a losing battle before. He financed the Gray Davis recall drive with a view to run himself against whoever the Dems picked, only to be told by the national party bigwigs to sit down and shut up while they ran Schwarzenegger, who unlike Idea could actually win a statewide election.

  • Paul

    Agree Sam that Dems get panicky pretty quickly, but as you know it’s all related to how intensely we perceive the ramifications of a “bad” result. While a 90% chance should have us feeling pretty sanguine, the 10% outcome precludes the possibility of a remotely reasonable Supreme Court appointment – which IMHO is as responsible for the dreary state we’re in right now as a country (e.g., Bush-Gore, Voting Rights, Citizens United, gun control laws, etc.) as anything else. So I know I’ll be sweating the top of the ticket and four or five Senate seats for the next eight weeks.

    • Michael

      I think we Democrats are like the homeless man in the old Yiddish story. He finally gets a meal, but goes on about how long he starved, and how the food wasn’t so good, and the portions were so small. The moral: “Some people need to complain more than they need to eat.”

  • Jeff Alworth

    Your point about the presidential horse race is well-taken and undoubtedly true (we know because it happens every four years), but there’s also a non-silly reason for voters to be focused on it more than usual. Donald Trump’s candidacy is extremely unusual and represents a plausible catastrophe for the country. (Since no one knows how he’d govern, it’s far from a certainty–but what reasonable person wouldn’t acknowledge the very real dangers?)

    I worry that in this case, we see familiar trends and fail to acknowledge that they might mean something different. It’s like a false-positive.

  • anonymous

    I think the solution may not be to go against human nature, which loves horse races (literal or metaphorical). The solution might be to turn downticket elections into horse-races. I can even imagine a graphic with horses representing candidates for each district.

    • Rachel Findley

      I like this idea. I wish the wonderful map of competitive House races would include a prominent list of names. And a link to any reasonably current polling data.

      Are there any state legislatures where a change in control is likely?

  • Don Gorton

    I think confidence in prognostications, which all point in a similar direction, is eroded by all the talk of Trump as an unconventional candidate to whom the usual rules do not apply. But the red/ blue divide appears stable, such that the Blue Wall that felled McCain & Romney seems to be Trump’s primary obstacle.

  • Sean

    I live in West Virginia so there aren’t really any competitive races going on here. We are a weird state in that we usually go democrat for governor and republican for president. The other races once people like shelly moore capito get elected they are normally untouchable in future elections.

  • Jaymes Winn

    It’s pure speculation on my part, but I’ve always felt that if the gap between Trump and Clinton’s ground operations holds and is as wide as they say it is, you could add a couple points to Clinton’s vote share in the battleground states.

    • Lance S.

      Jaymes, I agree. I also think Hispanic voters will add something, too, especially in states like Nevada and Florida. Historically traditional pollsters have had a great deal of difficulty capturing their preferences. As a result heavily hispanic states tend to vote a few point more Democratic than the polls suggest. So for me, no panic at all. I seem to recall some folks suggesting both McCain and Romney has taken the lead around this same point in time in 2008 and 2012.

  • Bill Herschel

    NY Congressional District 1, Suffolk County, is definitely close and should be Democratic. Anyone living in Suffolk, take note.

  • Joeff

    The media’s livelihood depends on keeping people glued to the screens and wearing out their clicking fingers. Do not adjust your set–turn it off!

    Question: last time there was a point–early October?–where every poll showing a 3%+ lead for a candidate accurately predicted the winner of that state in November. This was true over multiple elections (hundreds of polls), with only a handful of deviations. My question is: when do we reach that point this time?
    Many thanks for what you do.

  • Jason Bennett

    Sam, why has the map diverged from the EV snapshot in recent days? I thought the map was supposed to reflect the snapshot.

  • Anthony

    Hi Sam,

    I am a avid follower of yours and I read all of your articles and listen to all of your podcast and have been since 2012 and general agree and find them very insightful, however, in at least the way I am interpreting your article, I could not disagree more. Let me explain.

    I accept and trust your model that Trump as of today only has a 10% chance of winning, however I believe that the 10% chance of Trump becoming president is MUCH more important than tossup house or senate races. Of course I believe Democrats should be focusing down ballot is and generally something democrats are weak on (especially midterms) and the party organization really needs to improve on this, however the amount of damage a bad president can do, especially someone like Trump, has national and international ramifications this world has not seen in at least 80 years. It is perfectly legitimate for democrats and in general reasonable people to be VERY nervous at that 10 percent number, especially the TREND of that number of the past 2 weeks. I am already having much more trouble sleeping at night than I did a month ago.

  • Dan Engelberg

    Mea culpa. I’m one of those who regularly check on the horse race. But I also used the competitive congressional district webpage to identify a competitive district within driving range and make the most of my personal efforts. I’ll be going to VA-10 this weekend to knock on doors for LuAnn Bennett. Thanks for this great resource.

  • Stuart Levine

    But shouldn’t we obsess about the Presidential horserace for this reason: The size of the victory matters. That is, a very large victory margin by one party’s candidate will likely pull in some of that party’s senatorial and house candidate. Thus, Phlegm-ghazi, the media’s obsession with the Clinton emails, and the general decline of the Clinton lead in the polls are important not because they indicate that she won’t win (she will), but because the likelihood of Dem. control of the Senate and a significant erosion of the size of GOP majority in the House are reduced.

  • Gary

    So, here’s the drill. Every time you check the horse race (which hasn’t changed much for the past year) force yourself to send $10 to ActBlue or the candidate of your choice. You can avoid paying by checking a local or state race.

  • Emigre

    “Because of intense polarization, few voters are movable.”
    Here is a ton of statistics from PewResearch to chew on this further:

  • Olav Grinde

    Sam, let me see it I understand this correctly:

    1) The current probability (risk) of a Trump victory, assuming random drift, is quite comparable to being forced to play Russian roulette with an old Colt 45 six-shooter.

    2) If we instead assume Bayesian drift, the risk of a Trump victory is slightly worse than if I were forced to play Russian roulette with a choice of two six-shooters … albeit with “only” one bullet in the chamber of one of the guns.

    Right? Forgive me, but I really do not find that very reassuring!

    Given all that Donald J. Trump has said and done, I find it downright depressing that the Meta-Margin is not double-digit. Honestly, I am stunned that Trump is not 20 points behind.

    Politically, it really does seem we’re now living in a post-factual world.

    There has always been a risk of a Black Swan event totally changing this ball game, I get that. But now, with the drastically reduced Meta-Margin, it seems that any number of Light-Grey Swans might be sufficient… And what might that be? Well, 1) the economy tanking, 2) a major terrorist attack on American soil, 3) Julian Assange releasing some deeply compromising material on Hillary acquired by one of his hackers, or 4) Assange, just before Election Day, releasing deeply compromising material that has been faked by Russians (but which there is insufficient time to prove is untrue and fake).

    My point is: Sh*t can happen. And I haven’t even mentioned the 2nd Amendment people that Trump has all fired up.

    Yes, I get that Trump probably will lose. And I believe that most likely. But even if he does, the genie is out of the bottle. The Brown forces in America have been unleashed – and they are not going away!

    Trump may scare me. But it is his followers (and their numbers) that truly terrify me.

    • Than

      By “Brown Forces”, I assume you mean “Brown Shirts”. Not to parse too much, but just to sure.

      By the way, if “2nd Amendment Folks” heed either Trump himself, his adviser who called for Clinton to be executed for treason, and/or Matt Bevin, the Governor (Governor!) of KY, who recently essentially called for blood in the streets if Clinton wins, it will be dark days for sure. But that extreme aspect of US society will not win in the end. In spite of ongoing frustrations and extreme rhetoric and action, I’m convinced we’ve made too much solid, albeit gradual progress in recent years, to fall back.

  • mike paul

    Lance S.,
    Good call. Indeed, there were inaccurately high metrics for Romney right up to polling day on the Real-Clear-As-Mud website fuelled partly by Gallup’s last gasp. However, one unexpected byproduct was the Homeric schadenfreude of watching Karl Rove having his egregious meltdown on FoxNews whilst deluding himself that Fox had prematurely conceded the election to Pres. Obama.

  • Paul Mainwood

    I plead guilty to watching the Presidential race more than the Congressional ones.

    But I’m not a US citizen. And I hear whispers that there are several others on the internet too.

    I believe (rightly or wrongly) that the President has more influence on international policy, while Congress has more on domestic. So this focus seems rational, especially given the approaches that one of the candidates has advanced to NATO, to international trade policy, along with interesting approaches to US oil security.

    So, I’d expect your urging to look at the news of the Minnesota Democratic-Farmer-Labor party to be ignored by a certain proportion of your readership. I don’t know if your traffic stats allow splitting by international IP address, but that might give a cleaner picture of the focus on the horse-race vs the other sears.

  • Mikey

    Hi Sam- in your calculations I see you assume for simplicity’s sake that all the EVs for Maine will go to the same candidate, but that assumption doesn’t look that solid this time- Trump seems to have a good lead in district 2, Clinton handily winning district 1 and well ahead for the 2 overall EVs. To avoid making the calculations more complicated but to allow for the very possible split EV distribution, have you considered just splitting out ME into a 3 EV district 1 and a 1 EV district 2 (so just make the assumption that the district 1 winner takes the 2 overall EV too)?

    • Sam Wang

      If the probabilities get close, one could do that. It does not affect anything at the moment.

    • Mikey

      Assuming Trump gets all the states currently at >= 50% in your model, then his most realistic road to the Whitehouse requires all of the next 3 most likely states (NC,WI, and NV). With one of those ME district EVs, though, Trump would be President with only NC and WI, so no longer need to flip NV to get to 270… I’d have expected that modelling ME differently would have a noticeable effect on overall calculated probability of Trump victory?

    • Mikey

      Ah, apologies, ignore- just remembered that 269 Trump or 270 Trump are both defeats for Clinton on account that the Republican party control the chamber of parliament that decides the president if there’s a tie.

  • Anita Blanchard

    I actually come here when the mainstream media is going crazy over some recent poll for reassurance that the PEC still predicts 90% chance of win for Hillary.

    That said, I’m in NC and your post just encouraged me to contribute (again) to Deborah Ross. The rest of you all should do that, too.

    • Suvro

      Thanks for the reminder – I am putting my money this cycle on two nearby races (getting rid of Darrell Issa in CA, and John McCain in AZ).

      I just put in more money into their campaigns.
      Still long shots, but that is where I will put my money in this cycle, UNLESS the HRC campaign really needs it in October.

  • br

    Hard to believe any Democrat really has a shot in the NJ 5th. I grew up in Wantage township in Sussex county. Middle of nowhere, and quite far from the usual perception of what NJ looks like. I remember supporting Aronsohn when he lost, Wantage is full of the farming type of conservative who eat up the Fox narrative. They once voted down a school board referendum to improve 3 buildings on federal money because they simply hate taxes, despite the fact that their local tax burden would have decreased. You should see the conditions of the schools now-a-days. Garrett usually wins big in Sussex County, with Bergen making up most of the other side of the vote. For more details on the NJ 5th:—new-jersey-s-5th—paul-aronsohn

  • Kevin King

    Even with all that the race could have gone either way. In fact, with a 4-4 Supreme Court Gore wins as Florida’s supreme court’s order would have stood, forcing IIRC correctly s statewide complete recount. Think also how Gore distanced himself from Clinton whereas Hillary has embraced Obama, whom a large swath will turn out for. That result was not preordained, and in fact a lot of things had to go Bush’s way for what happened to have happened. And shame for citing individual polls out of context! Trump’s not nearly as likable as Bush, at the time, Krugman’s complaints about the dishonesty of his campaign notwithstanding.

    • Mark F.

      Why do you assume Gore would have won a full state recount? An analysis of the ballots after the election showed that he would have narrowly lost using the recount standard that he wanted. The race was essentially a dead heat within the margin of error any way you want to look at it.

  • Amitabh Lath

    Downticket is interesting this year. NH Senate is tied but an Ayotte victory would require fighting massive headwinds form the top of ticket. Is there some reasonable way to estimate an upper limit on magnitude of ticket splitting?

    If there is tension between straight up Senate polls and estimate from Presidential polls * max(P(ticket-splitting)) how to resolve?

    As for the nellies, I think a lot of the Clinton panic from the Democrats is unconscious sexism.

    Obama’s meta-margin lead against Romney was much closer for most of the season but aside from a short freakout after debate 1, Romney wasn’t given much hope.

    Clinton is doing better than Obama was in 2012, click on the +2 Trump map: Clinton can give away FL, OH, IA and still win because VA and CO are deeply in the bag.

  • Michael Levinsohn

    Did I read correctly that if the Bayesian probability drops below 80%, and is between 20% and 80%, then anything can happen?

  • Randy

    I’m more interested in the Senate at this point, and the likelihood that they’ll let Hillary fill Scalia’s seat even IF she wins. Of course, if the Dems claim a majority, they’ll have to nuke the filibuster, but would they? I think so. If the GOP retains majority control, I really think they’ll let the seat stay vacant til the end of Hillary’s term. Ok, back to fretting about Hillary losing…and time to donate to ActBlue.

  • A

    I don’t know. I want to believe Sam, that we’re making much ado about nothing.

    But at this point, the trend downward for Clinton is strong and has been moving that way for some time.

    She’s lots a tremendous amount of ground and I don’t see any signs of it stopping. No reason why Trump can’t continue to chip away at her lead and win, from here on out.

    Just a week or two ago, Sam had it at 94%-96% probability of Clinton winning, and now we are down into the 80s.

    The notion that this is all media theatrics seems a bit naive, to be quite honest, at this point.

    I am more than happy to be proven wrong though.

    • David D.

      Along these lines — is it possible to see the history of the Clinton Nov. win probability anywhere? Perhaps it’s a little bit meta but it would be nice to see the changes over time.

    • Sam Wang

      The Upshot just graphed it today – you can find it over there.

    • Jeremiah

      @David that is the EV estimator and Meta Margin history. I see what you are saying about the probability you can get an idea of that from the grey bar which gives one standard deviation I believe.

    • 538 Refugee

      Politico has 5 things that might/should lead to Trump losing ground come November. My favorite?

      “2. The cable and TV networks are going to vet Trump like he might actually be president of the United States. This one is wishful thinking based on zero evidence. Move along.”

      (I think we are actually seeing some real progress on this front now. The Washington Post’s story on his foundation is a good start.)

    • Jeremiah

      @David I realize I was wrong about the grey shaded area – this is the 95% confidence interval not the prediction (I guess akin to the “now probability.”) The history of the Clinton Nov. win probability can be found here: (it is the last 2 columns, random/Bayesian.)

    • George

      Just like there is a portion of the population that would vote for a dead skunk in the middle of the road if it had an (R) after its name, there is also a proportion of the population that would vote for Clinton if every vile lie about her was true. I think we are approaching that number. So what is unknown is how those two numbers compare and whether, as things get more ‘real’ the true undecideds will break for or against HRC. I am still betting on an HRC win, but have accepted that a Trump win might actually happen in the universe I live in.

  • Beth T.

    I was just alerted to this site by Kevin Drum at MotherJones, for which I am very grateful. That being said, one of the reasons that I keep panicking is because I’m having trouble understanding how so many people are being taken in by Trump. That sentence is a much nicer re-write of my actual opinion of the people taken in by Trump. Not the actual, unredeemable deplorables, but the more normal ones. Will click here more often.

  • Heavenly Blue

    The “Clinton is secretive” narrative persists not because the media can’t move on, but because many Americans believe she is secretive. The issue isn’t that she was “concealing her pneumonia.” Only her supporters believe the story that she had pneumonia. Nearly half the country believes that was just another lie/cover-up for her serious health problems.

  • Dissenter

    The great historical advantage that the PEC has had is that this method is not prone to wild swings, and therefore will not fall to the idea of a horserace that has been traditionally promoted by the media.

    The thing is, political climate right now is not the same as it has been for the past two decades. This is not true just in America, but over the entire western world. The British EU referendum has been dismissed as giving any indication of what may be happening in the US, which is relatively sensible from a polls+numbers approach – different votes in different countries with different methods – but from a theorists standpoint, someone who takes the numbers as part of the picture rather than the whole picture, they are very similar. The mood and the voters that swung that referendum towards Leave are the same that Trump is trying to achieve.

    The justification for claiming that this election is like the others this century – polarised with low variability – may have been a miscalcuation, not because the entire electoral map was ready to be shuffled (which was the basis for the decision), as it clearly isn’t, but because Trump’s appeal is to this group that is manifesting itself most obviously in the states he has been aiming for. This ‘low variability’ decision appears to have resulted in a belief that the race was over months ago, and therefore any media speculation today that it IS a horserace is summarily dismissed, even though Clinton’s numbers have been dripping away.

    All this, though, isn’t really a critique of the model, as Clinton’s chances have indeed been creaking downwards with her poll numbers, and I’m sure they will continue to do so given the national-state polling lag. But I suspect our host may have made an error in using the model to justify a ‘race over’ call last month

  • Fred

    When you adjusted your prior, is it at all possible that the switch from RVs to LVs may allow polls to go outside of the range that the model shows? The meta-margin has not been this low since the RNC

  • Nervous Synapse

    I don’t know, Sam. I really, really, really hope you are right and that there is nothing to worry about. But I am still very nervous. I fear that perhaps the reason the meta-margin has been so stable for the last 6 months is because of the unusually large percentage of undecided voters for most of this election. Many of those voters are probably republicans who were not comfortable with Trump but definitely do not want Clinton. They seem to be looking for ANY reason to tell themselves that Trump is actually presidential material. Reading a speech off of a teleprompter without insulting anyone – Presidential! Visiting Mexico without shooting his mouth off while he was there – Presidential! The bar is so low for him it’s unbelievable. Yet I am afraid this is what’s happening in the last few weeks. If you see Trump’s favorability rating it’s been creeping up and now his highest ever, and it correlates well with his improved polling. All Trump has to do is to OCCASIONALLY pretend he is a human being for the next 8 weeks. He can even continue his insults since people are used to it by now. My other worry is the Johnson/Stein factor. I saw a poll that over 40% of young people are going for third party. Clinton seems to always lose ground in a 4-way race. If young voters stop voting third party then Clinton will win. But if young voters are going third party while older republicans are warming up to Trump, I think we have some serious problems here.

  • John Parenteau

    I’ve refrained from commenting for days, but I am weak and cannot resist.

    I, for one, am not complicit. I knew well before the House district finder appeared where the nearest competitive House district was – it is my own CO-6 Coffman v Carroll. I have known for a long time where my House and Senate votes and time and money are going.

    I disagree that the Presidential race is pretty well in the bag. I just can’t help thinking that you, Dear Blogger, are repeating your 2004 Bush/Kerry mistake: making unwarranted assumptions about how Undecideds will break. Looking at the polls since the convention bounces settled, Clinton has remained steady at 42%-43%, while Trump has gained steadily from 36% to 39%. This tells me that the Undecideds are breaking reliably for Trump. I see no reason for that trend to change. Your thinking re Undecided results in assuming that the the meta-margin will be as stable as in the last 5 cycles, but it seems to me that that thinking is based on the math from far too small a sample set. I think the meta-margin will prove a whole lot less stable than you are thinking. Just in the past few weeks it has shrunk dramatically (more than 3% – not only more than the typical 1996-2012 standard deviation, but more than the maximum for that period.)

    Like the Russian Roulette poster whose post has disappeared, I am uncomfortable with this Election Roulette. The best that the Democrats can hope for is the Presidency and a 50/50 or perhaps 51/49 Senate split. They should concentrate on achieving that. It they don’t achieve that, then this country truly is in trouble. Taking State Legislatures, Governorships, and districts like the Colorado 6th would be gravy, but averting catastrophe is, to me, more important than gravy.

  • Scott

    Looking at Sam’s competitive Congressional races and the power of one vote for the Senate. It looks like the New Hampshire first is far and away the place where time is most valuable, especially if volunteers can make face-to-face contact.
    New Hampshire also appears to closely track the Trump-Clinton race as well. So if the race becomes competitive, spending in the district could benefit your preferred party thrice.

  • James Wimberley

    I stick to my non-statistical argument that the drift probabilities are not neutral but weighted heavily against Trump. As the last fortnight has shown, the things that go wrong for Clinton tend to be trivia: pneumonia, the Clinton Foundation, the emails. How many votes and abstentions does this stuff swing? The things that go wrong for Trump tend to be more serious: the Khan freakout, the Russian connection, Hillary’s careful and documented charge of racism, the Bondi bribe.

    This pattern will continue. So far Trump’s standing has been Teflon, but one of the grenades might explode. Clinton will not flunk the debates. Trump might well.

    The other large asymmetrical factor in this election is the ground game. Sam Wang’s poll-based approach was developed against a dataset from a long period when both campaigns were professional and roughly cancelled each other out. This year Trump is running an innovative campaign based entirely on manipulating the media and social media. It’s untested, risky, and could explode in his face.

    • Ken Schulz

      “Innovatiive” is an interesting word to describe Mr. Trump’s campaign. I’ve seen reports that he has far fewer staff and offices than Secretary Clinton – these are quantifiable comparisons, and I would expect them to correlate with numbers of volunteers also. I believe existing voter-turnout models are based heavily on historical data and survey questions – but these have been validated in elections in which the two major-party campaigns were, as you note, more or less evenly matched. We may find out this year how effective GOTV efforts really are, without a ‘restriction of range’ problem. On that topic, a quick Google search turned up this concise review:

  • Larry Guy

    How assumption-dependent is your Bayesian drift model? I ask because it seems to assume that this election cycle’s polling observations follow the usual statistical (stochastic?) rules (estimated from previous elections). But what if they don’t this time round? It’s like “mean reversion” or “momentum” investment strategies in the financial markets. They can work for long stretches, until they lead you astray (example of mean reversion failure: “Enron is down 80%; it’s gotta go back up. Let’s double our position!”). In other words, how mindful are you of possible model failure? When do trends get recognized as such instead of divergences that “should” revert?

    • Sam Wang

      There is no trend projection, only the application of a prior. Generally speaking, whichever distribution (the prior or the random-drift calculation) has a lower standard deviation (SD) will have the strongest influence on the November prediction. Currently random drift has more effect; that will increase even more as the election draws near.

      The prior contains the assumption of mean-reversion: the Meta-Margin’s range so far this year should be an indication of where it will go. That’s consistent with polarization 1996-2012. This year’s Meta-Margin was low-volatility, consistent with those years.

      To deal with weird outlier events, instead of a normal distribution, I use a t-distribution to fatten the tails a bit.

    • Mikey

      I was wondering which answer you use in your calculations when a poll asks both the 4-way question (Clinton vs Trump vs Johnson vs Stein) and the 2-way question (Clinton vs Trump)? I see when both are asked, both get picked up in the StatePolls.csv file.

    • Sam Wang

      Whichever one the pollster reported first in its description of the results.

    • Mikey

      I noticed that generally in the Huffpost ticker, that when the two questions are asked, that the question that gives the best lead for Hillary is usually the one reported first and is made the headline figure.

      In your csv file I had a look at how often there was a difference in margin between the answers to the 2-way or 4-way question, and how often the response that favours Clinton is first on the spreadsheet for a few of the tight states.

      CO 6/6 times the response favouring Clinton is first
      NV 2/4
      IA 2/2
      VA 6/9
      GA 4/5
      SC 1/1
      WI 1/3
      NC 6/6
      AZ 5/5
      FL 7/8

      I suspect Huffpost are consciously or subconsciously picking the response their liberal-leaning readers will prefer to read.

    • Sam Wang

      Even if your concern were true (I am skeptical), it does not matter from PEC’s standpoint. If there are multiple polls from one pollster asking the 2-way and the 4-way, we take the first poll that a pollster puts in its report. This is possible because HuffPollster enters the results in the order that they appear in the pollster’s report. (At least that’s what they have done so far…)

    • Mikey

      Fair enough. Btw, do you have an opinion on whether the 2-way or 4-way question Clinton vs Trump rate differences are the more accurate estimate of the observed Clinton vs Trump rate differences we will see in each state on election day? My hunch is 2 way is more useful when a couple of months away, and 4 way in the closing couple of weeks or so of the race.

  • Rex

    Sam, it seems like there’s been a lot of fluctuation in the meta-margin these past few days. Do you expect a stabilization as we get closer to election day? In 2012 you had the election at win probability 99.9% in November. Do you think that this will happen again?

  • Mark R

    The random drift prediction has fallen into the 20 – 80% range. The idea that the recent fall in Clinton’s meta-margin is just “noise” is no longer a good one. Quoting from above: “things could really go either way.”

  • Anthony

    Hi Sam, like others have already said, I think contrary to your article, I think its legitimately time to panic now.

    With that said, I really want to go on a long rant about our voting electorate, but I don’t think you will allow that so I’ll make this short point.

    We know Romney lost by ~4%, so Trump will need to do better than Romney to win, however Trump is doing WORSE than Romney in every demographic except uneducated white males. It seems like the big tent, very diverse democratic electorate is extremely fragile if Trump can only move the needle in one demographic (uneducated white males) and lose ground in every other demographic and still win. That is 1) extremely disconcerting that this is possible, 2) makes me concerned that Republicans can for the foreseeable future control our federal and local governments with ease and finally 3) Nate Silver might end up being vindicated when he wrote a few articles last year when he made the claim “there is no blue wall” and that demographics won’t save democrats. I remember him getting a lot of heat about those articles by liberal outlets.

  • Bill

    I think it is time to try wrapping our heads around the idea of a President Trump, and all that implies. If the meta-margin continues to drop like it has been the past few days things will be in the toss-up range.

  • TJ

    Would really love to see some analysis and input as to what this sudden drop really could mean.
    Sam did say he expected things would drop back down, but how close was that expectation to get? What’s the comparison of the current numbers to those of where Obama was last election?
    Lots of questions I am curious about before I’m ready to consider accepting a Trump POTUS reality.

  • kenn_sun

    Look at the meta-margin and ev histories. Things are no worse now than they were in July, before the conventions. We’ve just returned to the equilibrium.

Leave a Comment