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

Can Trump bring undecideds home?

July 7th, 2016, 9:05am by Sam Wang

July 20, 11:44pm: In light of tonight’s RNC speech by Ted Cruz, in which he pointedly did not endorse Donald Trump, it seems appropriate to revive this post from last week. Trump’s biggest deficit is lack of support from disaffected red-state Republicans. If they come back into the fold after tonight, it won’t be because of Ted Cruz.

As I wrote yesterday, there are more undecideds in the Clinton-Trump race than in the Obama-Romney race four years ago. The difference is 5.5 percentage points: in early July 2012, approximately 90.7 ± 1.1% (average ± SEM) of voters reported a preference for Obama or Romney. In 2016, total Clinton+Trump support is currently 85.2 ± 2.0%.

In practice, undecided voters tend to break approximately equally. However, these “extra undecideds” may not be the same as the run-of-the-mill voters who cannot yet express a preference to a pollster. Instead, Donald Trump is underperforming in a pattern that suggests that many Republicans in red states are disaffected by his candidacy. In 35 states where polls conducted in 2016 are available, the following pattern appears:

As I have written before, 2016 is shaping up to be notably similar to 2012 in its overall alignment. If Obama won a state in 2012 (points to the right of the vertical axis), then Clinton leads today. If Romney won (points that are left of the vertical axis), then Trump leads today. The sole exception is Nevada, where one NBC/Marist poll shows a Trump lead [this was a data entry error at HuffPollster].

If undecided voters simply break according to their state’s overall partisan leaning, we would expect both Clinton and Trump to be currently underperforming. Qualitatively, that is the case: In 21 Obama states, Clinton’s lead is smaller by 1.9 ± 5.6% (average ± SD). In 14 Romney states, Trump’s lead is smaller by 9.3 ± 5.8%.

However, there is a quantitative difference: Trump’s underperformance is substantially worse than Clinton’s, in a way that differs from random expectations (two-sample t-test, p=0.0004). Electorally speaking, Donald Trump is Romney Lite. His weakness is spread across many states, most notably Utah and Kansas, states that voted for Ted Cruz in the primaries. Together with the increased number of undecided voters, this suggests that a lot of Republican voters have gone missing.

If Trump does not bring enough Republican voters home, there could be unexpected wins for Clinton. If all states within 5% went Democratic, the electoral total would be Clinton 381 EV, Trump 157 EV. This is the downside risk for Republicans. On the flip side, Clinton’s biggest major weakness is Pennsylvania, where she leads by a median of only 1%. If all states within 5% went Republican, the electoral total would be Trump 318 EV, Clinton 220 EV. So it is not crazy to imagine a Trump victory…if he could somehow become a candidate that did not repel members of his own party.

The November win probabilities in the banner quantify how likely these outcomes are. The probabilities are based on evaluating how likely it is that today’s Meta-Margin, which is Clinton +3.6%, would move to zero between now and November. Recall that the Meta-Margin is defined as how much state polls would have to narrow, on average, to create a perfect toss-up. So if Trump makes up 3.6% on average (for instance by making a net gain as undecideds become committed), then he would have an even-odds shot of winning the Presidency.

That doesn’t sound like very much change. However, the last five Presidential elections (1996, 2000, 2004, 2008, 2012) have not shown much movement, a standard deviation (SD) of only 2 or 3 percentage points. This is a feature of intense political polarization, in which opinions are entrenched and don’t move much. If the SD is only 2 percentage points, then Hillary Clinton’s November win probability would be 90%.

The win probabilities in the banner are more cautious. They assume greater uncertainty based on historical standards: from 1952 to 2012, the average July-to-November movement has been 7 percentage points, which I use as the SD. That gives a random-drift probability of 70%. For the Bayesian prediction, I assume that this year’s up-and-down poll movement sets a range for the final likely outcome. This re-sharpens the prediction a bit, to give a November win probability of 85%.

If there is movement, it will probably occur after the back-to-back party conventions (July 18-21 for Republicans, July 25-28 for Democrats), and after the President/vice-presidential tickets are established. I am particularly interested in the Republican VP pick. Some undecided Republicans, like former Minnesota Senator Norm Coleman, who calls Trump a bigot and misogynist, still won’t vote for Clinton. A “normal” pick such as Newt Gingrich might bring them home.

There’s been talk of national and state polls diverging. In 10 national surveys started since June 24th, the median margin is Clinton +4.5 +/- 0.5% (median +/- estimated SEM). The current Meta-Margin of 3.6% is a bit smaller, reflecting the fact that Trump’s weakness is “hidden” in deep-red states where it doesn’t matter that much, for now anyway.


I want to end on a broader note. What will happen downticket? It’s too early to tell for Senate races, where many candidates are not yet established. But we have two other ways to take the national temperature: (a) President Obama’s approval rating, and (b) the generic Congressional ballot. Obama approval is moving upward, and the generic Congressional preference is moving toward Democrats. Obama’s net job approval rating has been moving upward, and is currently bouncing around Approve +3%. The generic Congressional ballot is at Democratic +5.5%. These indicators match the Clinton-Trump contest data.

If national partisan preferences are maintained to November, Democrats would be headed for a fairly convincing win – with the possibility of even retaking the House. This has seemed like a pipe dream for Democrats, but a 5% House popular vote win could be enough to do it. It depends on how strong the effects of gerrymandering and incumbency will be.

Tags: 2016 Election · House · President

57 Comments so far ↓

  • Olav Grinde

    Sam, is there any overlap between so-called undecided responders and those who actually prefer third-party candidates?

    I ask, because many Trump-Clinton polls do not seem to take into account Jill Stein and Gary Johnson at all. Might that not give a misreading as to the number of “undecideds”?

    Another question: Are your Meta-Margin and probabilities based on Trump-Clinton polls only, or on both types of polls? Does including/excluding polls with third-party candidates have any influence on the Meta-Margin…?

    Seems to me that the two types of polls cannot be put “in the same basket”.

    • Zach


      From the polling aggregates I’ve looked at, the margin between Trump and Clinton is very similar in two-way, three-way, and four-way contests.

      For instance take the RCP aggregates:

      2 way: Clinton +5.0%
      3 way: Clinton +4.5%
      4 way: Clinton +4.5%

      And the difference between the 2 way versus the other two is that it includes the Rasmussen Poll released today that shows Trump winning. So that is an effect of how Rasmussen models the likely electorate (which is a bit more Republican leaning than other polls) rather than a result of excluding minor party candidates.

    • Matt McIrvin

      I think Sam just uses them all.

    • Matt McIrvin

      Here’s Martin Longman worrying that Gary Johnson takes more votes from Clinton than from Trump… but it’s on the basis of one poll out of California, a very blue state. The general pattern across states suggests to me that this could be a blue-state thing.

    • Kari Q

      If I recall correctly, polls that include third party candidates as options always show greater support for that third party than they actually receive in votes. The further away from the election, the greater the disparity.

      Of course, this year could prove to be unique, but I’m guessing that as the election approaches, support for Stein and Johnson will decline. They may get a higher percentage of votes than in past elections, but I would anticipate it being less than the polling would indicate.

    • Daniel Barkalow

      I expect that people who have actually decided on Stein or Johnson would poll in a Trump-Clinton poll as “undecided”, acting like they’d decided not to decide the question the poll was asking. On the other hand, I think it might be useful in estimating whether the “undecided” responders will ultimately vote to know whether they’re backing a third-party candidate; however, I doubt we have the necessary historical data to estimate the size of the drift based on how people who aren’t backing either major party candidate report themselves. I haven’t even seen anything firm about how the total size of these two groups together affects drift historically.

  • Richard Vance


    Thank you for the time spent to provide some sense of this. I see PA as the true Keystone state this year. As PA goes so goes the nation.

    • anonymous

      Pennsylvania is supposedly Philadelphia and Pittsburgh with Alabama sandwiched in between them (excepting State College and other college towns, I suppose). I guess the main factor will be turnout in the cities versus the Alabama part. I guess when everything is said and done, the smart money is probably on it ending up on the Democratic side.

    • Matt McIrvin

      Clinton could lose Pennsylvania and still win (Sam’s model generates just such a map if Trump does 2% better than today).

      If Trump loses Pennsylvania, he’s almost certainly lost the election.

    • Matt McIrvin

      …I suppose if you want a national bellwether for 2016, the thing to do is to look for the states whose current margin is closest to the Meta-Margin. That would be Virginia and Iowa.

      People who have watched politics for a while are accustomed to thinking of Virginia as redder than the country as a whole, but it’s not any more.

    • Matt McIrvin

      …But if you’re watching on Election Night, the race in Virginia is often deceptive, since Fairfax County (the most Democratic part of the state) doesn’t fully report in until very late.

      Sam likes to look at New Hampshire, which reports very early, and is not far off the Meta-Margin this year. Comparing that to model results should give a good idea of how things will go.

    • Sam Wang

      Agree, Virginia is a terrible state to watch on Election Night. Polls tell you more about that state’s outcome than live coverage.

    • Matt McIrvin

      I like to watch Virginia, personally, just because (a) I’m from there, and (b) I can look like a prophet by telling my worried friends that Fairfax is going to flip it when it looks Republican early in the evening.

    • Josh

      I see your logic here but the problem is that Hillary doesn’t need PA to win.

    • Bill G.

      I’m chiming in late the party, but I think Florida is the key state to this election (daring opinion, I know). When you look at the states that each candidate is pretty safely expected to win and add Florida to Hillary’s column Trump pretty much has to run the rest of the swing states. I’m not saying that that is impossible, but it would be a tall order.

  • Some Body

    On the House–I remember reading that Republicans, being on average more likely to turn out, regularly outperform the generic ballot polls by about two points on average. This time may be different, of course.

    Also, I couldn’t find the generic ballot aggregate on HuffPost. RCP have it as Dem +2.0, nowhere near 5. In short, we’re probably not there yet with the House.

    • Some Body

      Now I did find the HuffPost chart, which they’ve hidden well, and it has Dems up by almost 8 points, with many polls that don’t show on RCP. Based on their list, the median should be Dem +7.5. The median based on the RCP list would be an exact tie. Weird stuff…

    • Sam Wang

      In the past, RCP has been slightly incomplete, which is why we use HuffPollster. The PEC chart should show the median of the last 3 polls or 21 days, whichever gives more data; and dropping duplicate pollsters. See this description. If we deviate from that, let me know so I can troubleshoot.

    • Sam Wang

      Could be, but I think the error goes in both directions. I agree that it would be unwise to make a prediction based on this – the uncertainties are too large. In the past, detailed district approaches have worked. I can’t recall if it was David Wasserman or others ( in 2008 – worth following, I think. Wasserman has flagged the issue that Democrats have not recruited candidates in all competitive races, which is basically political malpractice.

    • Some Body

      I forgot to take out duplicates. It is 5 points indeed, not 7.5.

      I suspect RCP don’t register a generic congressional ballot poll if the relevant question is burried inside a poll with a different headline, or they don’t include web polls (IBD, YouGov). There are just too many polls missing there.

  • Leading Edge Boomer

    On what planet are we where Newt Gingrich is considered a “normal” choice for a VP nominee?

    • MPP

      It just means he’s a politician with some experience.

      Compared to Trump picking someone like Ivanka, Omarosa or Carl Icahn, or even Sarah Palin, Gingrich is relatively tame.

    • Sam Wang

      He led the GOP takeover of Congress in 1994. A case can be made that he is the most pivotal figure in U.S. politics in the last 20 years.

    • Michael Coppola

      Except that 1994 was more than twenty years ago, Sam. Damn, I’m old.

  • Amitabh Lath

    Very interesting scatter plot. Would be nice to weight by electoral vote. Maybe if size or color of dots could indicate #EV?

    Because Florida (29 EV) was too close to call in 2012 (Polls had Romney leading, it went Obama 0.88%). Currently polls show Clinton ahead by around 4%.

    If this holds then mathematically I don’t see how Trump prevails, even if I give him PA, VA, NC.

  • Josh Burton

    Hi Sam – This is really fascinating, but I have a quibble. You’ve taken two (correct) observations:

    1) Trump is underperforming in red states
    2) There are a lot of undecideds

    And from these you’ve drawn the conclusion that the undecideds disproportionately live in red states and/or are potential Trump voters waiting to be “brought home”. But that’s not necessarily the case. In fact, I’ve just looked through 538’s collection of polls and found that there’s no correlation between a state’s Trump vote share and its proportion of non-Clinton, non-Trump poll respondents. So I don’t know how much we can say about the undecideds before they start deciding.

    • Sam Wang

      That sounds like a very high-noise way to approach the question. Perhaps restrict yourself to red states, defined as places where (Romney2012-Trump2016) is greater than 10%. Then compare the undecided-voter share there with other states, and do a t-test. I think this is a pretty indirect measure, but it could work. Whatever you do, don’t do it by eye.

      The test that persuaded me is the fact that (margin2016-margin2012)_Romneystates > (margin2016-margin2012)_Obamastates by a t-test, p=0.0004.

      Anyway, I appreciate the criticism of the logic, which is now cleaned up a bit.

  • Jewish Steel

    Your discussion about VP picks with your podcast co-host was wide ranging and very interesting. But I was surprised that the pair of you didn’t hit upon a VP pick that targets a particular constituency rather than a state or region.

    An Hispanic VP pick from Clinton could mobilize a segment of the electorate that under-performs its numbers.

    • Dexter Yanagisawa

      Given Trump’s inflammatory Hispanic rhetoric, doesn’t Clinton have enough ammunition to mobilize Hispanics?

      I can’t find a aggregate for Trumps approval ratings among Hispanics, but Gallup has Trump has a 77% unfavorable rating (12% favorable). This was conducted between January 2nd and March 8th, so this doesn’t even include Trump’s Curiel comments.

    • Josh

      I’ve seen polls recently with 10% of Hispanic voters supporting Trump. This is probably lower than whatever the actual number will be on election day, but I find it funny that your prescription for Clinton is to find a VP who explicitly attracts Hispanic voters. Hasn’t Trump made a good enough case for Clinton in this regard already?

  • Samuel

    This race wasn’t even close from the start. It all boils down to demographics and Republicans know it. What ever happened to the minority outreach they discussed in the post-Romney defeat of 2012? That is why level headed Republicans regret that someone like Trump has ended up as their nominee.

    • 538 Refugee

      I think the “minority outreach” Republicans envisioned was for the minorities to ‘wake up’ to some reality were they realized their personal belief system was wrong. It was never about the Republican party shifting on the issues important to minorities.

      “Level headed Republicans” are reaping what they sowed with their tactics. Claiming they didn’t need to compromise because they were just so damned sure they were on the correct side of the issue they have attracted a bunch of voters that reflect them perfectly.

  • John

    Just a lurker piping up to ask a question: any theories on why Nevada has gone from blue to red? It seems to be an anomaly.

  • Deb

    Why is pew on your pollster feed but not on the huffpolster site? Curious.

    • Deb

      Never mind. Though new it’s actually older date rise but further down. So is it common for Rasmussen to put out daily poll?

    • Matt McIrvin

      The HuffPo site now has two different Rasmussen polls dated July 5 (both with Trump ahead). Another data entry error, or is Rasmussen just shamelessly spamming weird pro-Republican poll results now?

  • Stu W

    Just a small thing about the plot. It only looks like 33 or so of the states are being plotted. Maybe the font size, or axes scales could be adjusted somewhat, in order to better visualize all the states on the chart.

    Other than that, this is excellent stuff that I’ve been checking on since about 2010. As a stats professional, I really appreciate the analysis and overall election picture.

    • Matt McIrvin

      There is still little or no polling in some states. That should fix itself over time.

  • Jakob Boman

    Have you made any calculations on the outcome of the election based on what VP the candidates select? It must be one of the biggest jokers, particularly in regards to Trump. It could help him to unite the party so he doesn’t have to run against two parties.

  • William Ockham

    The danger for Republicans is that when we get to September, if it becomes obvious that Trump will lose, they will be in very serious trouble. I think much of Trump’s support is personal, that is, it comes from voters who are only weakly affiliated with the Republican party. If those voters, whose only real motivation to vote is a Trump win, stay home and the normally safe Republican voters Trump has alienated stay home, the Republicans will have a disaster on their hands in the down ballot races.

    • Matt McIrvin

      There’s been some speculation that the opposite might happen: Clinton could win but with relatively short coattails, because a lot of people don’t really associate Trump with the Republican brand, and they might do a lot of ticket-splitting. Democrats running for down-ballot offices would probably have preferred Ted Cruz, a guy much more identified with where the Republican Party is today, to be the nominee.

    • Sam Wang

      Okay, but why is the generic Congressional preference at D+7% right now?

    • William Ockham

      @Matt McIrvin – That’s an interesting theory of voter behavior. I’m trying to figure out if the alternate scenarios (based on who the Republicans nominate) assume two different voter pools or if the notion is that voters are just brand consumers whose behavior is affected by marketing.

      My theory is based on my assumption that we basically know how every potential voter is going to vote (even when they don’t). What we don’t know is which voters will show up at the polls.

    • Matt McIrvin

      It’s true… right now Trump actually seems to be gaining while the Congressional Republicans lose ground. Though the big dip in the Meta-Margin could be spurious (it seems to be driven by a new clump of polls showing Trump ahead in Florida, all from firms that are either Republican-affiliated or have a pro-Trump lean).

  • Olav Grinde

    Even more than Ted Cruz’s glaring “non-endorsement” of Trump, I think many GOP voters are reacting in horror to the angry, intolerant reaction from the convention floor to his speech. They no longer recognize the transformed party they see.

    Likewise, many Republican voters are decidedly uncomfortable with the chants of “lock her up!” – as Chris Christie basically prosecuted Hillary Clinton for normal political views/acts with which Republicans might disagree.

    Strikingly, Senator Jeff Flake was one of very few top-level politicians that reacted publicly to what strikingly calls “the criminalization of political disagreement”.

    The deafening silence from the Republican elite, including Mitch McConnell and Paul Ryan, is encouraging this slide into extremist rhetoric. But then again, before Hillary, from Day One of his presidency they have looked on with encouraging silence as President Obama has been similarly attacked.

    The nomination of Donald Trump as the Republican Presidential candidate is a logical consequence of this, and a harbinger of what is to come.

    We are hearing a growing high-pitched, scream of rage that is unnerving far more than just liberal and moderate Americans – and the world at large that is viewing the spectacle of this election with growing incredulity and unease.

    It is a high-pitched scream that is increasingly making meaningful political debate impossible, that is making compromise unthinkable and in fact treasonous, and which is making tolerance of those with other political views anathema.

    What is the observant conservative, conscientious voter to do when they see strong forces in their own political party cheering this crumbling of the very foundations of democracy?

    How are they to vote this November – if at all?

    • A New Jersey Farmer

      I agree, and the biggest disappointment is that the party leaders are tolerating, at least in silence, what has become of the Republican Party. Trump has no intention of moderating himself and Pence will find in extremely difficult to mollify those who are uncomfortable or outraged by the tone of the debate.

    • Amitabh Lath

      Olav, Farmer, my working assumption is that this disaffection will not matter. It may be a small fourth-order effect in some edge congressional district but as Sam said, these are voters in deep-Red states. Even if Utah goes from -50 to -10 it does not change the basic calculus.

    • Michael Hahn

      Olav: Robert Heinlein once had wrote a pithy, but very true one sentence reply to your final question,
      He basically wrote:

      Always vote. There may not be anyone to vote FOR, but there is always someone to vote AGAINST.

      I will be voting FOR Hilary Clinton, despite some of her flaws. AND I will be voting AGAINST Donald Trump. I have my parents’ experiences from growing up in Germany in the 1920s in the forefront of my mind as well; they have heard and seen what is happening in the Republican campaign before, and it did not end well for Germany!!

      I have rarely voted FOR a candidate for President in my voting lifetime (since 1972), but I have voted, and voted against a candidate many a time. So my answer to your last question is: cast your vote for Hillary. The country WILL survive four years of her governance; I am far less sure that we will survive four years of Trump.

      That said, I am left very angry with the Republican Party, because they have drifted so far to one extreme that I no longer have any choice when I enter the voting booth. And that, as you point out, is NOT good for democracy!!

    • Commentor

      Carefully managed, scripted, conventions are a relatively new phenomenon. In the past, they’ve been much much much uglier than this.

    • Mark F.

      They can vote for Gary Johnson.

    • alurin

      @Commentator: Managed conventions are a hallmark of the primary era, in which the nominee is decided beforehand, by the voters. When the conventions actually decided things, they could get nasty.

  • mediaglyphic

    Does the regression look the same if we plot, todays polls vs. late July 2012 (rather than 2012 outcomes), perhaps we don’t have the data.

  • 538 Refugee

    The local news had a piece that basically said Kasich indicated the Cruz reception is why he is staying away from the convention. He is showing up for delegate events around town though.

    Is it a testament to Trump’s ‘judgment’ for not realizing Cruz is laying the groundwork to run again in 4 years and that couldn’t happen if Trump got elected this year?

  • ottovbvs

    There’s a lot of “moderate” Republican discomfort around with Trump particularly among women. It isn’t going to make any difference in places like OK but in PA, GA or FL we could be surprised. There’s a lot of ground to cover between now and November. This week has been a train wreck so we’ll see what happens in the post convention polls. It’s hard to believe Democrats aren’t going into to next week with their tales up. Then how does Donald get through the debates. One on one, no teleprompters, and a different electorate.

    • Matt McIrvin

      I’m thinking he might just find some excuse to cancel at the last minute, and not debate Clinton at all. He’s done it before, and the move was generally regarded as a humiliation for his opponent.

Leave a Comment