Since I posted my statistical analysis showing that voting for a climate bill didn't have a statistically significant effect on Democratic electoral success, two other independent analyses have come to roughly the same conclusion. My model found a statistically insignificant effect from climate legislation (half a point to a point, at best), and a significant effect from Affordable Care (on the order of 5-6 points).
RealClearPolitics, a politically conservative site with a solid reputation in terms of data analysis (e.g., their polling averages are widely cited and reliable) did an analysis with slightly different variables than mine, and found nonetheless that "Cap-and-trade cost an incumbent a point, but we're only about 63 percent sure of that." Meanwhile "a flip from 'no' to 'yes' on the health care bill cost those incumbents 5.5 points." That's almost exactly what I found (a somewhat larger role for the climate bill, but within a standard deviation of my model's estimate).
And the Council for Foreign Relations' Michael Levi got a similar result when he analyzed all of the House races:
Once youâve accounted for these three factors [PVI, popularity of the incumbent, and whether the Rep. is a Democrat], votes on various bills, including Waxman-Markey (ACES) [climate], health care (HCR), and the stimulus (ARRA), donât add much. If you add those three variables but get rid of âDemocratâ (on the guess that those bills are what people hated about Democrats), the predictive power of the model stays pretty much the same (to be precise, it increases by a tiny amount). But itâs health care, not Waxman-Markey or the stimulus, which does the explaining. Doing a regression just on partisan tilt (PVI), personal popularity, and HCR suggests that a Democratic vote for HCR cost 7.6+/-3.5 points in the final count.
Levi then does something interesting, but which I'm not sure is really right. He restricted his analysis to competitive races, to seats with PVI < D+10 (meaning a Democratic presidential candidate does less than 10 points better in that district than the nationwide result). In that model:
Voting for health care reform cost a candidate 7.3+/-3.9 points. Voting for Waxman-Markey cost a candidate 0.5+/-0.4 points for every point the district tilted Republican in general. ⦠Imagine ⦠a Democrat ⦠running in a district that preferred [Republican presidential candidates] over [Democrats] by an average of ten points in 2004 and 2008. Then voting for Waxman-Markey would have cost that candidate 5 points (+/-4).
When I use the same subset of the data, I still do get a significant interaction between PVI and support for the climate bill (but only weak support), and the effect in a D+10 district is about 3.5 points (+/- 3). The effect of voting for Affordable Care comes to around 6 points. But I worry about taking this subset, because it feels artificial. Certainly it's true that we should expect a different dynamic in contested races than in races without a credible challenge, but I'd prefer to have variables in the model that capture the differences between contested races and uncontested ones. I get at that partly by adding a quadratic term for PVI, allowing its influence to change somewhat across the model (which does have a significant effect). Others add information on total amount spent by the challenger (no significant effect), personal popularity of the incumbent, etc. But excluding data on 60% of Democrats (and only 1% of Republicans) hardly seems like the best solution.
Methodological geekery aside, the fact that these different analyses all wind up with nearly identical conclusions is important. We all seem to have looked at different subsets of the data (RCP seems to have used all races, I excluded those without challengers as basically uninformative; Levi looks at seats contested by both parties in 2008 and 2010), and using different variables and different assumptions about the nature of the interactions. Despite those differences, we're finding similar effects on electoral outcomes, with a negligible to nonexistent effect from the climate vote, and a significant effect from Affordable Care.
Climate change legislation didn't hurt Democrats in 2010.
- Log in to post comments