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

A better August for Romney

August 24th, 2012, 8:27am by Sam Wang

(Update: Posting updated at end.) There was a big load of polls yesterday. Our Meta-analysis distills it. The bottom line: the Ryan bounce seems to have peaked at a considerable 3.0-3.6 percentage points, though Obama retains a lead equivalent to 1.8%.

Just looking at individual polls – or even poll averages – makes it hard to get a clear read on the race. Commentators suggest variously that President Obama is still ahead, and may or may not have lost ground (Drum TPM 538).

The following graph, which probabilistically converts all available state polls to a single electoral snapshot, clarifies matters a bit.
The Ryan bounce, pre-convention

That’s in units of electoral votes.

Now, in units of Popular Meta-margin:
Meta-margin Ryan bounce pre-GOP convention

One week after the August 11 VP announcement, I pinpointed the size of the bounce at 1-2 points of swing in opinion. At the two-week mark, the swing has grown to 3-4 points. At the end is an uptick back towards Obama, which suggests that the post-Ryan bounce may have peaked.

As I showed yesterday using autocorrelations, swings in the race are not predictive of the final outcome until the end of September. So the Bain- and Ryan-driven events are of only momentary interest. However, those events do contribute to the average of what’s happened so far. I use this average (and the variability of past Presidential races) to make a long-term prediction. In the long-term outlook, President Obama’s re-elect probability is 88%. The likely electoral outcome is 283-353 EV (+/- 1 sigma) or 251-360 EV (+/-2 sigma).

A note on my use of “Ryan-driven”: it is not possible to say with certainty that the Ryan nomination was the cause of Romney’s improved standing. It could also have been the de-emphasis of Romney’s income taxes or Bain Capital association, because the news was filled with Ryan instead. It could have been the Romney campaign’s ads on welfare. Or something else. All we know is the timing and size of the change.

Update, 12:19pm: Further thoughts on why the bounce has ended.

More recent events are not measured in the average in the graph above: all the stuff about “legitimate” rape and other post-Ryan stories. Now that media attention has shifted toward Todd Akin, the conventions, and whatever comes next, the race will inevitably move again.

My general view is that the race has a “set point” much as a thermostat does, and that opinion tends to be drawn back to the set point. From this perspective, both the Ryan and Palin VP nominations are powerful but risky moves. Now we have a measure of how effective the Ryan nomination was in the short term. I continue to believe the main effect for the November elections is to yoke the Presidential and Congressional elections together more strongly.

Tags: 2012 Election · President

12 Comments so far ↓

  • wheelers cat

    I think you are correct about the Ryan bounce fading.
    The GOP just moved Romney’s nom up to Monday.
    In dressage there is a term we use called “impulsion”….it means forward power.
    I think the GOP is striving for impulsion.
    There is another thing we say in dressage….that which is forced can never be beautiful.

  • wheelers cat

    “I continue to believe the main effect is to yoke the Presidential and Congressional elections together more strongly.”

    And I continue to believe that Ryan was the leastworst pick Romney had left after Rubio turned him down, and it was PRIMARILY motivated by relative telegenicity of the remaining candidates.

    You and Nate should incorporate a telegenicity index into your models. After all, that is the only Palin got as far as she did.

    only REASON Palin got as far as she did. Survival of the Prettiest.

    consider how much conservatives love Christie….for his “policy”.
    But it is common knowledge that he is unelectable because of his weight.

  • Matt McIrvin

    Presumably the effect of how good-looking the candidates are is already baked into opinion polls. Adding a separate index for it would just muddy the waters.

  • wheelers cat

    I think it is underweighted then.
    I think appearance is far more important to the voters than policy.
    Perhaps a poll question about “appearing presidential” would cover it.
    But it really is a relative telegenicity factor.
    People are loathe to admit they vote on looks.

  • xian '86

    it’s not weighted, it’s reflected in the choices people make. adding a weight based on subjectively guessing their reasons would add noise to the data.

  • wheelers cat

    But a question like “do you think candidate x looks presidential?” with a scaled answer could extract additional information.
    If you were to ask respondents if appearance contributed to their choice they would be horrified.
    Everyone wants to believe they are policy wonks.
    But it isnt true.
    Americans dont care about policy.
    consider how much conservatives love Christie….for his “policy”.
    But it is common knowledge that he is unelectable because of his weight.

  • Matt McIrvin

    They may not tell you they vote based on appearance, but will they lie and claim they’re going to vote for the uglier candidate just to keep you from thinking they’re voting based on appearance?

    I doubt it. The stigma associated with appearance-based voting just isn’t that large. I think they’ll just rationalize the choice some other way, in which case you’re already getting all the relevant information by aggregating polls.

  • John Wills Lloyd

    Does “set point” correspond to “regression to the mean?”

    • Sam Wang

      John Wills Lloyd: No, not quite. “Regression to the mean” is the concept that if you sample the same population over and over, an extreme measurement is very likely to be followed by a less extreme measurement — even if nothing about the population has changed. (Generally speaking, my Meta-analysis attempts to remove any such apparent movement.)

      I am saying something about a process that is dynamic, i.e. changing over time. I assert that there is an intrinsic balance of opinion this year, set by external factors. The two campaigns are fighting it out in a manner that looks symmetric. However, because the set point is not a tie, one of them benefits by any move that drives the race towards the natural set point. The other is trying to move the race away from the set point. Since the set point favors Obama, this makes Team Romney’s job very hard.

      Matt McIrvin: This is going to seem odd, but as far as I can tell the RV/LV thing doesn’t matter as much as people think it does.

  • Matt McIrvin

    How much of the pre-Ryan-bounce set point do you think came from the use of registered voters rather than likely voters?

    TPM claims the latest CNN poll shows a gigantic difference between the two, 7 points of difference toward Romney when the LV screen comes in (bringing Obama from 9 points ahead down to 2 points ahead).

    That would swamp all the fluctuations in your chart, which actually makes me a bit suspicious of it, and obviously it’s just one poll. But since a lot of these polls start applying LV screens around now, I’m wondering how much of the Ryan bounce is really just likely-voter screens coming in (which would suggest it’s not going away).

    I suppose part of the explanation could be that the people who say they’re unlikely to vote are overwhelmingly in safe R or D states, where they feel there would be little point, or feel safe staying home as a protest. That would create a big difference in a national poll but have little effect on your aggregations.

  • wheelers cat

    “but will they lie and claim they’re going to vote for the uglier candidate”
    No, but I think that in Christies case conservatives may lie and say they are voting for the telegenically unelectable candidate…a sort of Bradley effect for morbidly obese people.
    My specific interests tend more to game theory and neuropolitics, although I do have some classical math stat background.
    My hypothesis is that Ryan’s sole basis of selection after Rubio was removed as a potential was relative telegenicity. If you did a kepler trigo matrix on electability for the candidate group, physical pulchitrude is the only place where Ryan comes out ahead. Look at all the beefcake photospos that came out after the pick.
    The Ryan plan is not well liked by the greater electorate, and as attention is focused on the details, it will become increasingly unpopular.
    Dr. Krauthammer basically told R&R to run away from the Ryan plan. Romney had to accept that he is now joined at the hip with the Ryan plan, and that is a liability with any demo not already part of his base.
    That is extremely risky. Romney HAS TO increase his share of the white vote to win.
    See Brownstein.
    btw I disagree with Brownstein. The white vote is projected to be between 70 and72% in 2012.
    There is prob(lim 0) that it will be 75%.

  • BillSct

    My MKI eyeball analysis of the “Median EV Estimator” graph puts the “Set-Point” at 310 electoral Votes for Obama +or- 10.

Leave a Comment