The Numbers Don't Lie: The Elections Weren't a Referendum on the Climate Bill Midterm Elections Not Referendum on Climate Policy, Statistics Say

A headline on Politico this morning crowed "Greens desperate to avoid blame." You'd think by reading that that the Democrats' feckless attempt to create climate change legislation directly caused the massive power shift in the House.

Except it didn't.

---

Yesterday I linked to a couple of pieces that explained why the midterms weren't, in fact, a referendum on climate and clean energy policy, nor on the failed cap and trade bill itself.

In the meantime, Josh Rosenau took up the hero's task of actually digging into the numbers to find that, and posted his findings on ScienceBlogs last night. His bottom line findings:

"There's no evidence that voting for or against the climate bill did Democrats any harm at the polls." (Emphasis mine.)

\n

Here's Rosenau's methodology (warning: stats chatter ahead):

To really get at this issue we have to do some statistical analysis. The hypothesis we're testing is straightforward: Representatives who voted for the climate change bill performed worse in this election than they should have.

Using Charlie Cook's PVI as a basic proxy for the underlying politics of the district (a statistic measuring the partisan trend in a district based on previous presidential races), I ran a series of regressions against the fraction of the vote going to the Democrat in each House district, as reported at the New York Times (for statistical reasons, I left out races with no Democratic candidate, no Republican challenger, or where the Democrat got less than 10% or more than 90% of the vote). As you'd expect, the regression is highly predictive, with the regression explaining 85% of the variability. Adding in information on the party of the Representative raises that predictive power to 90% (by accounting for some of the effects of incumbency).

\n

So Rosenau builds a model to predict the results based on the partisan natures of each district, and it's 90 percent accurate on its own. He then adds variables for the health care bill and then the climate bill. The health care vote makes the model even more accurate, but the climate bill vote does not.

And here's the brow-furrowing chart that results:

Basically, what you need to know here is that the higher up a point is, the closer the result was to his model's prediction. Democrats that lost are marked with the "X" and green dots are Democrats who opposed the cap and trade bill. If voting against cap and trade was "safer" in these elections for Democrats, the green dots would be clustered toward the top. But they're not.

But do yourself a favor and read Rosenau's post to better understand. He's a scientist that doesn't write like one.