The costliest vote revisited

Steve Greene and I have been working on our analysis of the effect of roll call votes on the electoral fortunes of Democratic House incumbents in 2010. (See this previous blog post for details.) We've put this research together into a paper that we're presenting this weekend at MPSA.

For the paper, we look at the impact of four roll call votes: health reform, the stimulus, cap-and-trade, and the TARP bailout from 2008. We find that a vote for health reform was very costly for Democrats, reducing reelection margins by six to eight percentage points. This cost at least thirteen House Democrats their jobs. We find a smaller, but still statistically significant, effect for supporting TARP. The stimulus has a mixed effect, harming Democrats in more conservative districts but possibly helping them in more liberal ones. We found no overall effect for cap-and-trade.

Here's my favorite graph from our paper, showing the estimated impact of the health reform vote on Democrats' likelihood of reelection. The patterns are generated by a logit equation and show how Democrats from a broad range of districts fared in 2010 based on this one roll call vote. For example, in a district where Obama got 40% of the vote in 2008, a Democratic representative would have a 54 51 percent chance of retaining her seat if she opposed health reform, but only a 19 an 18 percent chance if she supported it.

Steve and I are going to meet with Eric McGhee, John Sides, and Brendan Nyhan about their research on this topic to figure out where we agree and disagree. We'd talked about meeting for beers, but it's a morning meeting, so we'd best stick with wine.

Late update: A few folks noted a slight coding error in the original paper. We've updated the paper accordingly, and the results are substantively identical. I've corrected the above post as needed.

Comparing politicians and their districts

Are elected officials good representatives of their districts? That's actually a difficult question for political scientists to answer, in part because our measures are poor. We have plenty of good measures of the partisanship of elected officials, most of them deriving from roll call votes. We also have good measures of the partisanship of voters, most of them deriving from surveys and election results. But directly comparing voters and politicians is hard when we don't have a common scale.

Hans Noel and I try to remedy this in a new article (gated or ungated, PDF) in Political Research Quarterly. Our approach is to examine votes on legislative referenda in California. That's when state legislators vote on a bill and then send the exact same text onto the voters -- it's one of the only times that voters and politicians are voting on precisely the same thing. We use these referenda votes as "bridges" between legislators and their districts. Then we can put legislators' roll call votes and districts' votes on initiatives onto the same scale and generate comparable ideal points for districts and legislators.

How do they look? Well, it looks like legislators are a lot more polarized than their districts. There are actually plenty of moderate Assembly districts in California; there are basically no moderate Assembly members. Virtually every Democrat in the Assembly is more liberal than her district; virtually every Republican member is more conservative than her district.

We also find that members of the majority party tend to deviate further from their districts than members of the minority party do. Time out of office, we suggest, causes the minority party to try to moderate to win back the majority.

Informal influences on legislators

Recently, I've been plugging my seating paper, in which I find that California legislators sitting next to each other can influence each others' votes.  Well, in their working paper "Friends in High Places," Lauren Cohen and Christopher Malloy take this a step or two further and find a similar effect in the U.S. Senate.  Senators sitting near each other, they find, tend to vote similarly.  (I'm not sure to what degree freshman senators get to pick their seats, so there may be some endogeneity issues there, but still an interesting finding.) They also find that college alumni networks help explain voting patterns among senators.