Yesterday John pointed out that, in his words, “incumbents still dominate congressional elections,” noting that, in the vast majority of congressional elections, incumbents ran for reelection and won.
I just wanted to elaborate on this a bit. Here are the Republican vote shares in House elections that were contested in 2008 and 2010:
As John noted, most of these had incumbents running (the black dots).
Now look at the red dots. In general, open-seat candidates did not do as well as incumbents in comparable districts. On the left half of the plot (districts where Democrats won in 2008), Democratic incumbents did better, on average, than Democratic open-seat candidates. On the right half (districts where Republicans were the incumbent party), Republican incumbents did better, on average, than Republican open-seat candidates.
I don’t have the time now to do a full analysis (following my 2008 paper with Huang) so instead I’ll estimate the average incumbency advantage using a quick regression (as in my 1990 paper with King). Here’s what comes out:
lm(formula = rvote2010 ~ rvote2008 + inc2010 + party2008, subset = ok)
coef.est coef.se
(Intercept) 0.09 0.02
rvote2008 0.97 0.04
inc2010 0.06 0.01
party2008 -0.07 0.01
---
n = 330, k = 4
residual sd = 0.05, R-Squared = 0.88
So the estimated incumbency advantage is 6 percentage points. That ain’t nothin, but it’s a bit lower than the estimates of 8-10% that we were getting from the 1980s and 1990s. So I do think incumbency advantage is a bit lower than before, although it certainly appears to be higher than zero.
Here’s the scatterplot showing the data and fitted regression line. (Statistics tip: You should always do this!) The black lines show the model fit to the elections with incumbents, the red line is fit to the open seats.
So the model ain’t perfect (in particular, it’s not fitting at the low end of the graph). Also, I haven’t checked the data, nor have I tried to make any adjustments for the uncontested elections. That said, the estimated incumbency advantage of 6 percentage points seems reasonably well supported by the data.
The bottom line
1. We estimate incumbency to be worth about 6 percentage points of the vote, which can make a difference in some places.
2. The incumbency advantage does seem to have gone down in recent years, although it’s not clear that 2010 is particularly special.
3. Sometimes it’s ok to run a non-incumbent, even a weak candidate, if the district is highly partisan: take a look at those red dots on the right side of the above graph.
4. In almost all the districts, incumbents ran for reelection. This is the biggest incumbency effect, by far: Once you’re in, you generally get to run again and represent your party in the general election. This is the most important way in which incumbents “dominate”: they don’t need to compete for the nomination on a level playing field each time.