Princeton Election Consortium

A first draft of electoral history. Since 2004

Politics & Polls #19: Election Night

November 3rd, 2016, 2:48pm by Sam Wang


In this episode, Julian Zelizer and I talk about what we’ll be watching for on Election Night.

Tags: 2016 Election · House · President · Senate

36 Comments so far ↓

  • Melissa

    Sam, are you going to be posting here on Tuesday- commenting as the data appears?

  • Meg

    I, for one, am in a pretty big freak out over the polling out of New Hampshire.
    Anybody who wants to talk me down, is welcome to it!

  • Paul Ruston

    Sam-
    Are you going to produce your “Geek’s Guide” as an Election Day reference like you did in 2012?

  • Jeff

    Just saw Steve Schmidt on MSNBC say that Hillary will win with 300+ e.v. implying that he has access to campaign level polling data. He said the campaigns have far more accurate data than public pollsters and that they have much better voter models. Given that he worked on the McCain campaign, and others, he would seem to possibly have credible access to such data. Doesnt seem like the kind of person to risk his reputation pumping Hillays numbers just because he is on MSNBC.

  • Richard

    Sam’s two models assume either random drift of or a Bayesian prior for the meta-margin. The prior is that the average of the meta-margin sets an expectation for the outcome of the election. Under these assumptions, according to the models, the likelihood of Clinton winning is 97% and 99%. But these are empirical assumptions and they could be wrong. The Bayesian prior might be chosen incorrectly and drift we are seeing might not be random, in which case the models are wrong. Also, bear in mind that average of the polls can be off; Obama outperformed averages by a couple percent in 2012. I believe this is why Sam includes the Clinton +2 and Trump +2 maps, so folks get a sense of the most likely bounds on the outcome. Unfortunately, we might find the meta-margin less than Clinton +2 by the eve of the election, which puts the outcome as more susceptible to a miss in the average of the polls. All models can fail if assumptions they are built on are wrong. I hope Sam’s assumptions are correct, but models aren’t reality. November 8 is.

    • 538 Refugee

      They are also built on polling in a year that pollsters with good track records aren’t in good agreement from time to time.

  • DLane

    Can someone tell why the Clinton campaign isn’t getting her voters in Utah to vote for McMullin? Is there no scenario where denying 6 EVs to Trump is beneficial? Thanks.

    • Anthony

      There is no benefit to Clinton if either Trump or McMullin wins UT.

    • Richard

      Clinton has to get 270 otherwise it goes to the House which is controlled by the GOP and they will elect Trump even if Paul Ryan doesn’t like him. Whether Trump or McMullin win Utah does not affect Clinton’s total.

    • CW

      As others have said, Clinton has to get to 270 EVs or the election goes to the House. It’s not enough to get most EVs. And once it goes to the House, each state gets one vote and can choose among the top three EV winners. And they keep voting until one candidate gets 26 votes (a majority).

      Any plausible Clinton-Trump tie will leave Trump with the majority of states. See this for example: http://www.270towin.com/maps/NKERV

      Trump has a significant advantage in this scenario… State delegations don’t have to vote the way their state did, but it’s probably the most likely option. Add in McMullin and it just adds to the ways Clinton can lose.

    • Jack Chin

      I’d say that if HRC his 270 or more, a McMullin win in Utah is better than a Trump win because it is a rejection of DJT. But it would of course be better for HRC to win.

    • DLane

      Thanks all.

    • CW

      I guess the one scenario where it would make sense for Clinton is the map here:
      http://www.270towin.com/maps/xPZeN

      In this case, McMullin would keep Trump from getting to 270, giving Clinton one last hope in the House. In this case, I think that hope is sufficiently distant that it’s not even worth Clinton thinking about as a strategy

    • Ravilyn Sanders

      It is true, Trump losing Utah to McMullin does not help Clinton. But for small investment on her side, and she is flush with cash, she can force Trump to defend Utah. Any resources diverted from battleground states is good for her, right?

    • Tafsyk

      McMullin with 6 EVs could be useful to the USA.
      If (Randomness forbid) HRC gets only 269, according to Amend. XII the House chooses from the top THREE vote getters in the Electoral College, which would put McMullin the contest. Given the general polarization, hatred, and disgust, along with the possibility of an almost even House in January, a three-way race might lead to rejection of both HRC and DJT in favor of an apparently squeaky clean rational person and a tacit agreement by a bipartisan working majority to try the novel idea of governing the country.
      James Monroe could envy President McMullin in such a Moment of Good Feelings (if not an Era) and also understand the collapse of one of the two major parties.
      OK, perhaps waking up now would be best, even if it involves looking at reality.

  • Fat Bob

    Let’s say that it’s true that Clinton’s GOTV effort is leagues ahead of Trump’s, would that be reflected in the public polls or just in the actual election?

    • BrianTH

      In theory if they are getting a properly representative sample and they are properly coding likely voters, it should be reflected in the polls. However, if, say, there was a big surge in tough-to-poll sub-groups, it is conceivable some polls could struggle to pick it up.

      I wouldn’t want to bank on that for the 270th EV if I didn’t have to. But if, say, a state like Nevada ends up a bit more favorable for Clinton than the polls suggested, that could be a significant part of the reason.

    • Josh

      GOTV efforts are deployed at the margins of very close states, where getting an extra 20,000 voters to the polls can win you a bunch of electoral votes. Campaigns won’t waste their time turning out voters in California or Texas or Illinois; you’ll see robust GOTV operations in Iowa and Ohio and Florida and North Carolina. A very good GOTV operation may get you a couple hundred thousand extra votes across the country. Out of 140 million total votes, this would be noise in polling.

  • Rachel Findley

    PS enjoyed the podcast, too. Good comments on the media overcoverage of Comey/FBI and undercoverage of climate, immigration, taxes, security, civil rights, and other issues of substance.

    Will be listening to Senate returns. And looking toward post-election life in the body politic.

  • Rachel Findley

    Looking for advice on Senate contributions:
    How much does one dollar in campaign contributions affect each vote in small states? In small states with huge pools of money sloshing around, is there a decreasing marginal return per dollar? Or is a dollar really better spent in New Hampshire and Nevada, as compared with Indiana, Missouri, North Carolina, and Wisconsin?
    Does anybody know?

  • Anthony

    Just finished listening. I know it seems like you don’t like the horse race talk but I was hoping you would address one issue that seems very different than 2012. As you mentioned as far as the MM is concerned it has been stable over the long run, however currently we are at a very sharp slope in the decline. What precedence is it for this extremely sharp slope right at the end of the race (compared to 2012)? I’m just a bit scared this very sharp decline is going to end up missing in the polling data on election day.

    • Matt McIrvin

      In 2004 it seemed like the state polls were gyrating around wildly in the last couple of weeks, near 268-268, with the lead changing sides repeatedly. It wasn’t a stable situation. Yet both PEC and electoral-vote.com almost nailed the entire map on Election Eve with simple state-poll aggregation (discounting bad special-sauce adjustments). The result wasn’t pretty for Democrats, but they got it.

    • Matt McIrvin

      …eh, I guess that looks stabler in PEC’s version than electoral-vote.com’s, so maybe it’s not the best example.

  • A

    Great podcast. Sam, you said you expect to see NH be a +5 for Hillary on election night.

    Do you still see that, even with a lot of polls seeming to show the contest in NH tightening up?

    Also, your thoughts on the senate were a bit chilling. Hate to think we’re in for 4 years of these shenanigans, even if Hillary manages to win but Dems don’t pick up the senate or make a meaningful dent in the congressional races…

    • Adam

      My question on that for Dr Wang: is New Hampshire a hard state to poll? It was pretty off in 2012. I remember that Latino groups can be easily missed but could it apply to largely rural states? Just wonder why…if there is a why.

    • Sam Wang

      No, New Hampshire should be easy.

    • Rob M

      Regarding NH, Sam, could you update us on Monday night as to what the new normal would be? That is, what the margin is from which you would look for the over-under, as you put it in the podcast ? I say this in the light of apparent tightening in the NH race that may lower your 5% figure (hopefully not, but still…)

    • Josh

      As Sam has pointed out elsewhere, when you have lots of polling, you’ll naturally get more outliers. There have been a dozen polls in NH in the last two weeks, far more than before, which means you’re seeing results like Clinton +14 and Trump +5. But just because you’re seeing wacky results doesn’t mean the race has fundamentally shifted. The median of all the polls in NH over the previous ten days (when the Meta-Margin began to shrink): Clinton +4. A one-point difference between two aggregated medians, especially when it comes to public polls, is probably noise.

      There will probably be half a dozen more polls in NH before Election Day. If all of those half-dozen polls have Donald Trump winning or close to winning, I’d be concerned that the race had narrowed. That doesn’t seem to be the case today, and in fact, with only 4 days to go until Election Day, there isn’t much time for things to change anyhow.

      Look at it this way: if Clinton wins NH by something in the realm of 5-6 points, then it basically shows that polling this year was dead on. If she wins by less, or by more, then polling was slightly off. And if she wins by much less or much more, then polling was significantly off.

  • Bill

    A very good listen.

  • Xavier

    You say on the podcast if NH is closer than 5–6% things will be close for the election. It seems pretty clear from polls that NH is guaranteed to be much closer than 5–6%.

  • SP

    The link appears to be broken.