There’s been a lot of back and forth about whether cap-and-trade hurt the Democrats this past Tuesday. It’s mostly spin, with people focusing on whatever results they think bolster their case. But the debate matters: as Chris Horner of CEI points out, the fact that people attributed the 1994 bloodbath in part to Democrats’ attempt to pass a Btu tax is a big part of what killed anything along those lines for a generation.
A couple neutral statistical analyses are finally out, and they’re worth reading. Both conclude that cap-and-trade didn’t matter, and I think they’re both largely on solid ground. But they both miss some important dynamics.
To dig deeper, I put together my own statistical analysis. I assembled data on the 334 House races that were contested by both parties in both 2008 and 2010, and in which the same candidate ran in both races.
Before I get deep into the weeds, here are my bottom lines:
– The district-by-district vote can be explained almost entirely by the partisan tilt of each district; the historical popularity of the incumbent; and whether the incumbent was a Democrat. Alternatively, it can be explained by partisan tilt; historical popularity; and whether the incumbent voted for health care reform. Votes on Waxman-Markey predicted pretty much nothing, by themselves, about how individual candidates fared.
– There’s some chance that voting for Waxman-Markey hurt Democrats in Republican-leaning districts. Specifically, there’s a roughly 75% chance that voting for Waxman-Markey in districts that tilted toward John McCain and George W. Bush hurt Democrats. That said, there’s a similar chance that voting for Waxman-Markey helped in Democratic-leaning districts.
– The story looks a bit different if we focus only on challenging districts with Democratic incumbents, which is what most of the news has been about. (I define what I mean by that below.) Voting for Waxman-Markey is highly likely to have hurt Democrats in Republican-leaning districts, though it may have also helped Democrats in marginal but Democratic-leaning districts. The effect, though, appears to have been relatively small.
– Four defeated Democrats could have gained a 10% or more chance of saving their seats had they voted against cap-and-trade, though most of them were in dire straights anyhow. Four more who won while voting against Waxman-Markey would probably have lost had they voted for the bill.
– There’s one defeated Democrat who may have been able to eke out a win had he voted for the bill, but it doesn’t look there are others who fit that pattern.
Now for the technical details. I should note that the analysis is pretty crude: I used simple multivariable linear regressions for all of it, so please take the results as suggestive, rather than definitive. That said, the basic contours are pretty clear.
Most of the vote in each House district can be explained by three variables: the partisan tilt of the district, measured by PVI, which reflects how the district voted in the last two presidential elections; the popularity of the individual incumbent, which I measure by the degree to which they outperformed their party’s generic candidate in their district in 2008 (the performance of the generic candidate is given by PVI); and whether or not the incumbent was a Democrat. These factors explain 95% of the variance in vote counts. First, partisan tendencies intensified, with each point of PVI translating into 1.5 points of vote. For example, if Republican presidential candidates have won by an average of 10 points in the last two elections, that translated into an edge of 15 points for the Republican candidate on Tuesday. Second, personal strength counted for less than in 2008, with each point of personal edge from that election translating into only half a point this time around. For example, if a Democrat won by 25 points in 2008 in a district where Democratic presidential candidates beat Republicans by an average of 5 points in the last two elections (a difference of 20 points), he should only have expected to have a 15 point (5+20/2) advantage this time. Basically, this means that voters cared less about individuals, and more about parties. Third, being a Democrat sucked, costing the unlucky incumbent about 4.8 points. This one is a little less certain – the 95% confidence interval is from 1.4 to 8.2 points — but it’s still robustly positive.
Once you’ve accounted for these three factors, votes on various bills, including Waxman-Markey (ACES), health care (HCR), and the stimulus (ARRA), don’t add much. If you add those three variables but get rid of “Democrat” (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 it’s 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.
I also looked to see whether voting for ACES, HCR, or ARRA might have hurt the incumbent more in more heavily Republican districts, which makes intuitive sense. (I also played around with a few other variables that turned out not to matter, including freshman status and per capita greenhouse gas emissions.) It turns out that there’s a roughly 75% chance that voting for Waxman-Markey in hurt Democrats in Republican-leaning districts, and a roughly 85% chance that voting for the stimulus did the same. But including this effect doesn’t really change the predictive power of the model. The best way to understand this that, as far as voters in Republican-leaning districts were concerned, voting for Waxman-Markey and the stimulus was a significant part of what constituted being a Democrat – and those who did so were punished.
Things look quite different, though, if we focus on the candidates that everyone’s been talking about: Democrats in relatively tough districts. I took a focused look at districts with Democratic incumbents but PVIs of D+10 or worse. In English, that means that Obama and Kerry beat McCain and Bush by an average of 10 points or less in these districts; in many of them, the Democratic presidential candidate lost. This captures all but one of the Democrats who voted against cap-and-trade (Pete Stark is the one exception) and a total of 122 candidates.
This time, personal popularity and partisan tilt (PVI) are only enough to explain 68% of the variation in vote counts. Adding in all the other variables – votes on HCR, ARRA, and ACES, as well as variables that accounted for partisan preferences on those bills – boosts the model’s explanatory power to 74%. In particular, two variables stand out as mattering far more than the others: voting for health care reform, and voting for Waxman-Markey in a Republican-leaning district. These two alone boost the explanatory power of the model to 73%. 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. That’s probably pretty confusing, so take a specific example. Imagine that a Democrat was running in a district that preferred George Bush and John McCain over John Kerry and Barack Obama by an average of ten points in 2004 and 2008. Then voting for Waxman-Markey would have cost that candidate 5 points (+/-4) this year. Conversely, though, this suggests that a vote for Waxman-Markey may have actually helped candidates in Democratic-leaning districts.
What all this means, for the most part, is that Waxman-Markey did the most damage to candidates who were already in deep trouble (recall that PVI was a great predictor of the vote this year). The model suggests, though, that there are a few candidates who might have benefited from voting against the bill. In particular, Bob Etheridge (NC-2), who is currently asking for a recount, appears to have lost by less than one point in a fairly neutral but slightly Republican district (where he used to be a popular incumbent, before he was caught on video hassling two student activists.) The model gives him a 60% chance of having won had he voted against Waxman-Markey. Michael McMahon (NY-13), who lost by three points in a slightly Republican district, would have had a 20% chance of winning had he voted against ACES, according to the model, while Tom Perriello (VA-5), who lost by 5 points, would have had a 10% chance of winning had he flipped. (It’s worth noting that the latter two both won their 2008 races in Republican districts after Republican incumbents disgraced themselves, suggesting that they were fighting uphill battles in general. It’s also worth noting that Rick Boucher isn’t part of our sample, because he ran unopposed in 2008; he should probably be part of this category, and I’ve included him in the summary up top.) Jerry McNerney (CA-11) might also not be fighting for his life right now had he voted against the bill, but his race would have been very close regardless. Finally, Jim Costa (CA-20), who appears to have lost by a couple points in a slightly Democratic district, might have benefited had he voted for the bill: the model suggests he’d have had about a 75% chance of winning had he done that.
On the other hand, several Democrats probably saved their seats by voting against the bill. The model identifies four members who probably would have lost had they voted for Waxman-Markey: Joe Donnelly (IN-2; 60%); Jason Altmire (PA-4; 85%); Jim Matheson (UT-2; 85%); and Dan Boren (OK-2; 60%). The percentages in parenthesis are the odds that the Congressman would have lost had he voted for cap-and-trade.
Again, this is all very crude, so please take it with a generous grain of salt. But the bottom line? Cap-and-trade mattered in individual races, but not much, and when it did, it was because it exacerbated partisan tensions. Neither story being peddled – that cap-and-trade killed the Democrats, or that it could have been their salvation – holds up.