Congressional races and Bush support

I was quite intrigued with the scatter plots which compared Bush vote percentage to Kennedy vote percentage in my post about the sixth district race. I've now sat down to look at all of Minnesota's congressional districts. In the left column are plots which show the relative distribution, by precinct, of votes for Bush and the GOP congressional candidate in 2004. On the right are the same plots for 2000 after adjusting for current congressional district boundaries.

Once again, you interpret these plots as follows: precincts which fall above/left of the diagonal line are those where the GOP congressional candidate got more votes than Bush; precincts lying below/right of the line are those where Bush did better than the GOP candidate. The farther away a precinct is from the diagonal line, the higher the proportion of ticket-splitters in the precinct.

In the 2004 plots (left column), I've shaded the dots by total number of presidential votes recorded in the precinct. This technique shows (not suprisingly) that the outliers tend to be precincts with smaller populations.

In the 2000 plots (right column), I've colored the dots according to which congressional race people were voting in (listed by number and 2000 incumbent). Though all the precincts in each plot are currently part of the same congressional district, in 2000 district lines were different. Thus, the current 2nd congressional district boundary lines incorporate portions of the old 1st, 2nd, 3rd, 4th, and 6th districts.

You can click on any plot to see a larger version. In Firefox, a second click gets you to full size; in IE, I had to click on an icon that appeared in the bottom right of the picture to get full size. (BTW, does anyone know a secret for uploading a bunch of images to Blogger simultaneously? Doing it five at a time is a pain for this many images.)

What I'm looking at here is how closely the precincts cluster around the diagonal line. I'm making the following assumptions here: (1) The closer the precincts cluster around that line, the fewer ticket-splitters (folks who voted for one party in the Presidential race and a different party in the Congressional race). (2) When there were few ticket splitters, voters tended to let the presidential race influence their votes for Congresional candidates.

For the record, let me state that neither statement can properly be inferred from this data. It could be true that there were plenty of ticket splitters when the fit is very close to the diagonal but that Rep-Dem and Dem-Rep voters offset each other almost exactly. And it could be true that many voters paid equal or even greater attention to the congressional race than they did to the presidential race. But neither of these possibilities makes much sense intuitively, and I am comfortable discounting them.

The comparison across all congressional districts makes the pattern in the 6th even more striking; in no other district was there as remarkable a shift. In the 1st and 2nd, precincts were tightly clustered around the line in 2004, but those patterns were already evident (if not quite as marked) in 2000. Conversely, 7th and 8th district precincts are separated from the line in both elections, even if they were slightly closer to the line in 2004 than 2000.

Using just this data, Bush's low approval ratings, and incumbency advantage -- and ignoring a lot of important information (like the fundraising strength of each party's nominees), the 6th jumps out as the most likely pickup opportunity for the DFL provided the president's approval ratings remain poor. Next would be the 2nd, followed by the 1st. Ramstad seems unassailable in the 3rd. Looking at the DFL seats, all four incumbents look very safe indeed. The 7th appears to be a potential pickup opportunity for Republicans whenever Peterson leaves office, since a very large portion of his support has been in precincts that have voted for Bush. (Indeed, Bush won the entire 7th district in both elections.)

However, it's not clear how much disapproval of Bush will translate into DFL votes in specific off-year elections. There have been a number of academic studies which examine the role of presidential approval in off-year elections, but all of those with which I am familiar have looked at the performance of the president's party nationwide (overall gain or loss of House seats), not race-by-race.

posted by DK @ 2:35 PM permalink