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

Which way to November?

September 22nd, 2012, 12:00pm by Sam Wang

I promised you a prediction challenge. Hang on a couple of days – too much discussion arising from the House prediction.

A number of concerns have been expressed about the House calculation, which requires more assumptions than the Senate or Presidential calculations. One, concerning the Monkey Cage model (and others like it) is resolved by pointing out that their model predicts a very broad range of outcomes between R+13% and D+9%, with which my more precise calculation is fully consistent. Another is that district-level polls may be more accurate than the generic ballot question. That will be true…once fresh district polls are available.

Still not fully resolved is the predictive accuracy of September generic Congressional preferences. On Election Eve, it is fairly accurate. But where will it go between now and then?

In my calculation, I assumed that it could move in either direction. Kevin Drum suggests that it might tend to move toward the Republicans. Three of his five examples are midterm elections. During a midterm Congressional election season, opinion moves against the sitting President’s party:

Net Change in generic congressional vote from February to October. Dashed lines, Republican presidents. Solid lines, Democratic presidents (Bafumi et al. 2010).

In midterm years, the amount of movement against the incumbent President is pretty significant post-Labor-Day. The above figure, from a paper by Bafumi, Erikson, and Wlezien does not address on-years like 2012, but it does show the kind of analysis that is possible.

There is a well-known coattail effect in which the President’s party does better than in off-years. But that does not answer the question of movement, which is the relevant concept now that we have a precise reading of current conditions. It would make sense for movement to be toward the Presidential winner’s party. Indeed, Kevin Drum says that there was movement of 3.3% towards Republicans in 2004 (Bush re-election) and 2.5% towards Democrats in 2008 (Obama election). If true, the probability of a Democratic takeover probability would be somewhat higher than the 74% I gave. Although we only have information for two elections at present, the question seems worth pursuing.

With apologies for leaving this issue not fully resolved, for the time being I stay with the assumption that future movement could go in either direction. This leaves the prediction unchanged: a likely (74% probability) Democratic takeover of the House in November.

Thanks to DaveM and others for stimulating comments.

Tags: 2012 Election · House

8 Comments so far ↓

  • wheelers cat

    Could one use RAND and the betting markets as early indicators of movement direction?

  • Olav Grinde

    Dr Wang, you have mentioned the paucity of polling in the House races. Any idea what portion of the races from which we can expect to see fresh polls in, say, the next two weeks?

    Is there a site where all House polls are collated in an orderly fashion? Preferably sorted by state — and perhaps even with a map where statistics are shown when hover your cursor over the state/district?

    In other words, something similar to what does for the Senate races:

    I’m aware that has a nice overview of House races sorted by date — but it’s really difficult to get an overall understanding from that.

    I also know that has some information…

    …but I haven’t really found what I’m looking for.

    Any suggestions?

  • OwlofMinerva

    One point to keep in mind is that with SuperPACs, money can be flexibly allocated. Thus, if it becomes increasingly certain that Obama will be re-elected, we would expect an increasing amount of money redirected into down-ballot races. This might somewhat blunt any Obama coattail effect.

  • Amitabh Lath

    House elections are tough problems. I admire you for tackling it, but so much is hyper-local.

    I suspect there are cases of people voting down the line for one party (which might explain your bias towards the president). Also, there might be people who split their vote, because they want to keep their congressperson who has seniority in some committee, even if he/she is of the “wrong” party.

  • Olav Grinde

    I find it striking that the Rand poll now has Obama at 50 %. Well, 49.86, actually — I shouldn’t exaggerate. That’s more than 6 % ahead of Mr Romney.

    • wheelers cat

      Its also striking that the RAND survey measures change……not conversion of undecideds or independents.

  • Hans

    Dr. Wang, if the EV probability chart looks like it does today on election day, it could be quite difficult to make an actual EV prediction that is meaningful, because of the very spiky nature of the probabilities. I’m guessing the third spike from the left is at 332 EV today, but the other three spikes are very likely and quite different from 332. Your overall analysis could be highly accurate, but your EV guess off by a substantial amount.

    • Sam Wang

      Good point. This is caused by 50-50 states with large numbers of electoral votes. In 2008, it was easier because the two uncertain states, Indiana and Missouri, each had 11 EV. As a result the two middle peaks fused to a single spike at 364 EV.

Leave a Comment