Peaceful coexistence

I don't mean to make a habit out of rehashing old school papers, but I recently came up with some kind of explanation for a phenomenon I once noted but couldn't adequately address.

The political organization Americans for Democratic Action provides a useful proxy for how liberal or conservative individual congressmen and congresswomen are. Basically, the ADA picks 20 votes in a given year that it considers important and assigns a score to each representative or senator, the score being the percentage of the votes with which the individual voted how the ADA would prefer; because the ADA is a liberal group, a higher score denotes a more liberal individual, whereas a lower score denotes a more conservative.*

We might expect a representative to largely align with the median voter in his or her district, in order to maximize the share of the vote; the same thought process would apply to senators, where their districts are the entire state (or, as a proxy for individual voters' preferences, the median or mean of the ADA scores of a state's representatives could be compared with the mean of the senators' scores). In layman's terms, an elected official should be somewhere in the middle of the voters in terms of how liberal or conservative.

Yet when I was looking for a economics paper topic in fall 2008, I noticed that several states had senators at opposite ends of the political spectrum. In Missouri, for instance, the mean of the representatives’ ADA scores was 52.2 in 2007, while the median representative, Jo Ann Emerson, had an ADA score of 50. Missouri Senators Kit Bond and Claire McCaskill had ADA scores that same year of 25 and 90, respectively. 

Louisiana offered a particularly striking example: the mean and median scores of its representatives were 29.3 and 10 in 2007, while its senators’ ADA scores were 10 and 80 — the latter being an extreme outlier.

This was not a rare situation, either; other examples of wide deviation included Ohio, Virginia, Oregon, New Mexico, Nevada, Minnesota, Indiana, Florida, Colorado, Iowa and Pennsylvania. Significant ideological differences between a senator and a state's broader persuasion were, if not the norm, certainly common. I also found that of those 13 states, only four (Minnesota, Indiana, Pennsylvania and Oregon) had one senator who appeared to be moderate; the other nine had one staunch conservative and one staunch liberal senator.

Recognizing that this was a fairly fascinating discovery but not having any great method to explain it, I cobbled together an argument based on geography (and population density) that, frankly, was hogwash. I believe that recently I was able to come up with a more satisfying conclusion.

Indulge me, though, as I offer a bit more background. Incumbency provides an enormous advantage in an election; several studies have found that it can be extremely difficult to oust a politician from office once elected. We can largely ignore, then, how these strange setups have persisted; once a liberal and a conservative are elected, they can go on coexisting for some time. The key is to uncover how such a situation came about in the first place.

I must also update my data set here with more recent results because, with the states from my paper, I didn't save all the information I need. For simplicity's sake, I'll ignore the scores of a state's representatives and merely focus on states whose senators are divergent. Relying on information from 2010, the latest year available, I will include the following states and senators in my analysis**: Alaska (Begich and Murkowski), Florida (LeMieux*** and Nelson), Illinois**** (Durbin and Kirk), Indiana (Bayh***** and Lugar), Iowa (Grassley and Harkin), Louisiana (Landrieu and Vitter), Massachusetts (Brown and Kerry), Missouri (Bond and McCaskill), Nevada (Ensign***** and Reid), New Hampshire (Gregg and Shaheen), North Carolina (Burr and Hagan), Ohio (Brown and Voinovich****) and South Dakota (Johnson and Thune). That gives us 13 states, and while many of the senators are probably the same as the ones in my 2007 data set******, it seems — to me at least — that this phenomenon isn't a one-time fluke.

I hope the variable I didn't take into account in my college paper jumps out here.

Alaska: R (2002), D (2009)

Florida: R (2009), D (2001)

Illinois: R (2010), D (1997)

Indiana: R (1977), D (1999)

Iowa: R (1981), D (1985)

Louisiana: R (2005), D (1997)

Massachusetts: R (2010), D (1985)

Missouri: R (2011), D (2007)

Nevada: R (2001), D (1987)

New Hampshire:  R (2011), D (2009)

North Carolina: R (2005), D (2009)

Ohio: R (1999), D (2007)

South Dakota: R (2005), D (1997)

The numbers in parentheses, of course, refer to the years each Republican and each Democrat assumed office. The set doesn't make this as clear as I might have hoped, but there's some evidence to support what I would argue is a perfectly viable explanation for the phenomenon: the passage of time.

It is fairly easy to imagine a liberal's being elected in a previously (or even predominantly) conservative state in a period of widespread anticonservative sentiment (say, the tail end of George W. Bush's presidency; you'll notice two Democratic senators above took office in 2007 and four more in 2009). The same could go for a conservative in a time of antiliberal sentiment (you'll also notice five Republican senators in the 2009-2011 range, when there was frustration with the response to the recession and the Tea Party was ascendant).

Once a senator is elected, he or she will (or can — scandals or higher offices may intercede) likely stay in office for some time because of incumbency effects, even if the partisan fervor that the candidate rode to election subsides. Incumbency effects might be so strong, in fact, that the senator could withstand ensuing partisan ripples until another massive change in sentiment came about. So we can have a case like Ohio's, where a Republican comes in at the end of a Democratic presidency and a Democrat arrives at the end of a Republican presidency and the situation persists.

This isn't a perfect explanation: all of these senators are near the edges of the range outlined by the ADA. If we expected senators to locate close to the median voter, we still have some explaining to do; it's unlikely that the median changes so drastically in two to four years (Iowa, Missouri, New Hampshire, North Carolina). But that explanation isn't an impossible task if we reach out to other theories*******, and certainly on the basis of explaining a seeming paradox in the coexistence of Republican and Democratic senators in a given state, this argument has some appeal. Should the median be swayed just enough to the other side of moderation, a candidate could ride that to election.

This year's election will provide some interesting anecdotes for the theory I've outlined. Of the 100 seats in the Senate, 33 are being contested; says Wikipedia, "Democrats are expected to have 21 seats up for election ... while Republicans are expected to have only 10 seats up for election." If and when some of those change hands, it would be helpful to try to assess the change in the states' voter makeups — at the least, the change in voter share from previous elections, with the assumption that the party matters more than the individual candidate.

* - The caveat is that this is a small sample and can't begin to include the nuance of political persuasion. But when one senator agrees 19 of 20 times and the other senator posts a goose egg, there seems to be some real difference in their attitudes.

** - Nebraska has a Republican senator (Johanns) and a Democratic senator (Nelson), but because Nelson appeared to be such a moderate (ADA score of 50), it didn't seem worth including the state in the analysis.

*** - No longer in office. The result is less relevant to our discussion because he was appointed.

**** - After an election, a senator changed midway through the year, but despite the small sample of votes, the results seem clear.

***** - No longer in office, but the result is relevant.

****** - Virginia, Oregon, New Mexico, Minnesota, Colorado and Pennsylvania now have senators fairly close on the political spectrum and have dropped out of the set. Alaska, Illinois, Massachusetts, New Hampshire, North Carolina and South Dakota have been added as their senators have become more polarized. What may be worth noting: since the paper in college, Alaska and North Carolina added a Democratic senator (both in 2009); Illinois, Massachusetts and New Hampshire added a Republican one (in 2010 or 2011). This seems to more or less fit with what we'd expect given the analysis I've undertaken. (South Dakota's senators were elected before the previous paper; that suggests to me, without looking at their old ADA scores to confirm, that they're more divergent than they were then.)

******* - It isn't a perfect comparison, but this is a start.