We have predicted a one-seat gain in the House for the Democrats this fall. However, we deliberately limited our analysis to fundamental factors like the economy, presidential approval, and broad-brush district factors like incumbency. Those factors leave a lot unexplained. Some candidates are stronger than others, and the way individual races play out is important. This is true regardless of the competitiveness of the district or whether one of the candidates in an incumbent–the two district-level factors we already include in our model. To get at this idea, we added campaign finance data. Money is a sign of a candidate’s strength: both skills and political prospects. How does this change the results?
The upshot: our estimates become a lot more precise, but the core prediction is basically unchanged. Our model predicts a two-seat loss for the Democrats, and a 10% chance of taking back the House — which is not much different than the model’s original prediction of a one-seat gain.
That said, this prediction comes with some important details and caveats, so keep reading.
Our model is estimated using the amount a campaign actually spent in past elections (specifically, the Democrat’s share of total spending). But when it comes to predicting the 2012 election, the most recent comprehensive report of campaign finance activity is from June 30, and the next one isn’t due until October 15. If we’re stuck with those figures, it makes a lot more sense to use money raised for our prediction. Candidates are only likely to have spent money by June 30 if they had a competitive primary challenge, but they’ll be fundraising early in anticipation of a competitive race in the fall. Of course, candidates can raise a lot of money and change the dynamics of a campaign between June 30 and election day. So what we have should be thought of as a snapshot of candidate strength in mid-summer. Not perfect, but better than nothing.
(One other concern can be put to rest. The amount a candidate has raised by the end of the election cycle is a very good predictor of the amount they will spend, at least when calculated as a share of the total money between the two parties. In 2010, the Democratic share of money raised in each district correlated with money spent at 0.96. So we seem to be on solid ground from that standpoint.)
Here’s our prediction of vote share from this model:
And here’s our prediction of seat share:
In short, the state of campaign finance as of mid-summer suggests little hope for Democrats to take back the House. And the predictions from this model are remarkably accurate for most years, even in terms of seat share.*
Nonetheless, money isn’t everything: for instance, even this model continues to completely miss the 1994 and 2010 Republican wave elections. There was something happening those cycles that campaign finance, and other aspects of the model, couldn’t capture. Moreover, since our numbers come from the end of June, the dynamics of campaign finance could be shifting under our feet. In the next post we’ll try to get at one way this might be happening: Super PAC activity (where our numbers will be far more up-to-date).
*Observant readers might notice that the first graph above suggests Democrats get a slight majority of the vote in certain years–1996, 2000, and 2004–when they didn’t hold a majority of seats. That’s an artifact of the way we calculate the aggregate vote. First we compute the Democratic vote share in each district, and then we average those numbers across all districts. This can produce slightly different numbers, but nothing that alters the seat share in any way. In reality, Democrats won slightly fewer votes than Republicans in each of those years.