I’m starting to once again hear about the “anti-incumbent attitude” of voters, and the effect this will have in 2012. As I wrote back in Spring 2010, I think too many people confuse anti-incumbency with ideological landslide. But I was just fiddling (for other reasons) with some election data, and saw something mildly interesting.
I think it’s important to differentiate the two ideas — anti-incumbency and partisan landslide — as collapsing them reduces our ability to describe elections. For example, there may have an anti-incumbent attitude in the 2010 election, but I don’t see a lot of evidence that it electorally manifested itself. Yes, the proportion of incumbents who returned to the House (78.2%) was the second lowest since 1948, but of the 90 seats that were vacated by an incumbent, 66 were former Democratic seats won by Republicans, while only 3 were former Republican seats won by Democrats (21 retiree seats were won by the occupying party). If very few incumbents of the winning party lose re-election, it hard to call it an anti-incumbency situation, per se. It’s just a garden-variety ass-whippin’, in the form of a partisan/ideological landslide.
And so while I certainly can see the anti-incumbent attitude in the current polling data, I’m not inclined to put too much stock into it having a large electoral effect. Yes, approval ratings of both congressional parties are at historical lows, such that we are definitely in uncharted waters, but I think the safe assumption, following the historical evidence, is that anti-incumbency only really manifests itself when you have bi-partisan support for truly unwanted policy (like the Kansas-Nebraska Act in 1854***) or bi-partisan blame for corruption (like 1992). Anything else, and it’s more likely that the aggregate electorate is just going to punish one party or another (like 1890 or 1930 or 1994).
What would evidence for actual anti-incumbency look like, differentiated from ideological drift? Well, one scenario that would be a low incumbency retention rate paired with a low partisan seat swing. Imagine if every incumbent in the House lost, but the net partisan seat gain was zero. That’s probably good evidence. I say probably because it’s not airtight — voters of both parties could have become more radical and remove their incumbents for ideological reasons, for example.
Anyway, here’s a simple plot of incumbent return rate vs. absolute partisan seat swing, 1952-2010:
The trend line, not surprisingly, provides a pretty strong correlation between incumbency return % and absolute seat swing — aggregate incumbent defeats tend to reflect partisan or ideological drift . But check out the outlier. That’s 1992, when only 75% of Members returned to Congress, but the total seat swing was only 11 seats. Again, this isn’t air-tight evidence of anti-incumbency; it’s possible that many otherwise-electable Members retired and were replaced by partisan comrades. But a lot of what we know about 1992 says otherwise: the House banking and post office scandals were symbols of an institution in decline, many of the Member retirements were to preempt losses, and both parties suffered double-digit defeats in seats they were trying to defend.
***For the unfamiliar, 1854 bears a second look, because that’s a true anti-incumbency result: going into the election, the partisan breakdown in the House was approximately 150 Dems and 75 Whigs. When Congress reconvened after the election, the breakdown was 84 Democrats, 60 Whigs, 62 American Party (“know-nothings”), and 46 Republicans. I can’t see how that will ever be topped — the majority party loses almost half its seats, the minority party loses a fifth of its seats, and they all go to a minor party that previously held no seats and a new party that didn’t exist a year earlier!