Princeton Election Consortium

A first draft of electoral history. Since 2004

The post-Iowa bounce goes to…Hillary Clinton

February 6th, 2016, 12:54pm by Sam Wang

Update, post-primary: Sanders outperformed his final New Hampshire polling margin by about four percentage points. That is interesting, but not out of bounds. It may reflect the difficulty of polling an open primary. It remains obviously true that Iowa did not have effects of a size to affect the likelihood of a Sanders win in New Hampshire.

It is probably hopeless to counteract reporters who weave the tale that Sanders is surging. Based on actual data, it might be exactly the opposite.

  • In aggregated data, Hillary Clinton has gotten approximately a 6-point bounce in New Hampshire. The median margin was Sanders +21.5% in 4 surveys conducted January 26-30. This narrowed to Sanders +15.5% in 6 surveys conducted February 2-5.
  • A daily tracking poll from U.Mass. Lowell shows even more narrowing. On February 1 it showed Sanders +31%, which by February 6th narrowed to Sanders +14%, a 17 percentage point change in Clinton’s favor.
  • In national surveys, Clinton went from a median of Clinton +12% (4 polls, January 22-February 1) to Clinton +16% (3 polls, February 2-4). This is noisy data, but the median change is a national 4-point bounce for Clinton. It is possible there was little change in either direction (see confidence intervals below).

It is likely that Hillary Clinton has remained level or risen in her national standings. This may be counterintuitive, considering the tone of the coverage by the press corps. Since Sanders is still in the lead in New Hampshire, my guess is that few reporters will get static for portraying an imagined surge for him.

Josh Marshall is all up in arms about the Quinnipiac poll showing a near-tie. He doesn’t highlight the fact that three national polls done at nearly the same time show a wide range of results, including a result showing Clinton leading Sanders, 53% to 32%.

Then again, the Quinnipiac data point is interesting. it shows Sanders at 42%, which along with an online Reuters surveys showing him at 43%, are his highest numbers to date. These are likely-voter surveys and they may be measuring increased enthusiasm for Sanders. If true, we would have the interesting phenomenon that his movement in New Hampshire and nationally are in opposite directions. That could be because people in New Hampshire know him already, and noticed the Clinton win; while nationally, some are newly cueing in to him.

As Matt McIrvin puts it in comments: “I think [movement in both directions is] entirely plausible. In both cases, it’d be a shrinking lead as the respective demographics start considering the other candidate.”

[Confidence intervals: In New Hampshire, I estimate that Clinton's bounce was between -0.3 and +12.3% (one-sigma confidence interval). Nationally, the post-Iowa shift was between 2.4% toward Sanders and 10.4% toward Clinton. Overall, it seems that she gained from her narrow win in Iowa.]

P.S. Please, no comments about individual polls being biased in their likely-voter screens and so on. It is nearly impossible to do a good job of evaluating such claims without falling prey to one’s own motivated reasoning. This is why one takes the median of all legitimate polls – to reduce the problems contributed by extreme individual measurements.

Tags: 2016 Election

14 Comments so far ↓

  • James Hotze

    Sam, what are your thoughts now that New Hampshire shows that the polling bounce did not represent the electorate? Surely a 21-point win for Sanders shows that Iowa did not improve Clinton’s position at all…

  • John

    Hey Sam,

    Instead of trying to look too deeply in to the biases and reliability of polls, it can be helpful to compare them to themselves when looking for breakout trends.

    This gives us a different range of changes:

    Rasmussen = Hillary +2
    Quinnipiac = Sanders +29
    Morning Consult = Hillary +2
    PPP = Sanders +7

    It also shouldn’t be surprising that Hillary is closing the gap in NH. She had been campaigning pretty actively in Iowa and has now switched her entire focus to the next state. She’s also been really turning up her ad spending during the last week before the vote, like she did in Iowa. Early reports suggested Bernie was spending more on ads, but the final results showed Hillary pulling ahead at the last minute.

    Also, 538 has come up with the ingenious idea of using national trends as a counter-weight to state trends. If a candidate pulls ahead in a certain state while falling behind in the national, it just means there support has become more geographically condensed.

  • Doctor Science


    I don’t know if you meant to, but comments are closed on the Symposium post.

  • John McDonald

    What do you mean by LV and RV filters?

  • Matt McIrvin

    …However, both of those results show motion toward Sanders, not Clinton. But it’s not a huge 20-point jump after Iowa or anything like that, it’s gradual.

    I think it’s entirely plausible that the national primary polling is moving toward Sanders while the NH polling moves toward Clinton. In both cases, it’d be a shrinking lead as the respective demographics start considering the other candidate.

  • Matt McIrvin

    Of course, it’s possible that these interpretations could become self-fulfilling if they drive media coverage in a way that affects general perceptions of who is a realistic candidate.

  • Matt McIrvin

    The Bernie fans of my acquaintance are all going bananas over the Quinnipiac poll. Lots of discussion of Hillary’s support collapsing. On the Republican side they’re saying the same thing about Trump on the basis of one PPP poll.

    • Brad

      I personally felt the Quinnipiac poll was an outlier (as a Sanders supporter), but then the Ipsos/Reuters poll showed almost exactly the same results. I still believe they’re a bit exaggerated, but to say Clinton has gained nationally post Iowa is absurd.

    • Sam Wang

      I agree that the suggestion that national polls moved toward Hillary Clinton is counterintuitive. However, these are the concrete measurements we have. If you have other measurements, share them.

    • Matt McIrvin

      Ipsos/Reuters depends strong on the RV/LV filter. Their tracking poll has Hillary hugely ahead if you apply their LV filter, and they’re neck and neck exactly like the Quinnipiac result otherwise.

      (The Quinnipiac result claims to be LV-filtered, though, so that isn’t the entire story, though of course these filters are a black art.)

    • Kevan Love

      The PPP poll showed Bernie ahead in Iowa the entire month of January. By as much as 5 pts.