The 111th Senate

A while back, Nate Silver over at fivethirtyeight.com had a post up about how various Senators voted on the auto bailout*, and how that might be a preview of which Senators were likely to defect in the upcoming congress. Nate's post was interesting, but I was a bit disappointed at the lack of analysis of the effect, especially given that Nate was writing this. So, naturally, being the numbers geek I am, I decided to do it myself.

I started by compiling a list of all 78 Senators who:

• Voted on the bailout bill
• Were not lame ducks
• Had not announced their planned retirement before their next election
  
• Were, at the time of the vote, not leaving the Senate for a cabinet post
  

Nate's basic hypothesis was that a Senator's vote (V) would be roughly determined by their political party (P), Obama's margin of victory in their state (O), and by how many years they had until re-election. Accordingly, for each Senator I collected their party (assigning Lieberman and Sanders as Democrats), margin of Obama victory, class (i.e., how many years until re-election), and how they voted for the bailout bill. I also colelcted their state, for a counter hypothesis I had which I will discuss below. Votes were coded as 1 for Yea and 0 for Nea. Parties were coded 1 for Democrat and 0 for Republican.

I ran a linear regression estimation for

V = a*P + b*O + c

where a, b, and c are constants esitmated by the linear regression. I did a separate regression for each class of senators, and came up with results that looked pretty good for Nate's hypothesis:

Class III (election in 2010): V = .44*P + 1.14*O + .286
Class I (election in 2012): V = .53*P + .81*O + .303
Class II (election in 2014) V = .64*P + .78*O .231

So, the longer a senator has until his next election, the less he cares about Obama's margin of victory, and the more he is likely to care about his own party - exactly what Nate hypothesized.

The value of V at which a senator should be voting Yea is 0.5, as a Yea was coded as a 1, and a Nay a 0. However, it didn't quite work out that way. Instead, 0.3 seemed to be the more accurate predictor. Only 6 Senators (3 from each party) with a vote-score above 0.3 voted Nay, while only 1 with a vote-score below 0.3 voted Yea. Based on this model, the senators most likely to defect from each party are, in order of likelihood:

Republicans:

• Olympia Snowe (ME)
• John Ensign (NV)
• Arlen Specter (AP)
• Judd Gregg (NH)
• Charles Grassley (IA)

Democrats:

• Blanche Lincoln (AR)
• Byron Dorgan (ND)

Note that every single Republican on that list had a vote-score closer to .5 than did Byron Dorgan.

Now, as I mentioned above, I had a counter-hypothesis. I thought that, instead of being about Obama's popularity, the auto bailout was more about regional voting. My hypothesis was that Rust Belt Senators would be likely to vote in favor, while Southern Senators (from states which house foreign auto manufacturers) would be more likely to vote against. So I organized the states into regions and found the following results:

• Five of the six Republican Senators from New England (defined as New York, Rhode Island, Connecticut, Massachusetts, Vermont, New Hampshire, and Maine) and the Rust Belt (defined as Pennsylvania, Ohio, Indiana, West Virginia, and Michigan) voted for the bailout bill. The lone defector from that group was Judd Gregg (NH).
• Five of the six Republicans who voted for the bailout were from New England or the Rust Belt - the only other was Kit Bond (R-MO).
• There are three Democratic Senators from the Deep South (defined as Arkansas, Louisiana, Texas, South Carolina, Georgia, Alabama, and Mississippi). Two of them were just re-elected. The third, Blanche Lincoln, is up for election in 2010, and voted against the bailout.

The only other Democrats to vote against the bailout bill were both Montana Senators, Jon Tester and Max Baucus. I have no idea what's going on with them.

At any rate, it would seem that there was, indeed, a regional effect going on. And the effect might be stronger than any effect of Obama's popularity on the vote - both John Ensign and Chuck Grassley had high vote-scores, yet neither voted 'yea' on the bailout bill.

My final conclusion: Nate's hypothesis might very well be right, but the auto bailout bill is a poor vote on which to test it.


* - By "a while back" I mean "when the auto bailout was actually a contemporary and relevant topic. Note to self...