The new Pew report on trust in government is out, and, as we would expect amidst an economic downturn, trust is low. One question is what this means for the 2010 election. Pew notes “record anti-incumbent sentiment.” This leads Chris Cilizza to write:
..unrest…is at the foundation of what is shaping up to be a strongly anti-incumbent political year…Those soaring levels of dissatisfaction have to worry incumbents of both parties—although the electoral pain will almost certainly be felt more by Democrats since, well, they have a lot more incumbents in the House and Senate.
This suggests a testable hypothesis: more congressional incumbents lose elections when trust in government declines. Conveniently, Pew put some data on-line. It has the number of House incumbents who lost their elections (1958-2008) as well as a database of trust questions (from which I take the American National Election Study data, which I charted previously, combined with one CBS/NYT poll from October 2006, a year when the ANES did not field its standard election survey). Here is the graph:
Hmm, so much for the anti-incumbent effects of trust. Thanks to a few outliers, the relationship between trust and incumbent losses is actually positive, although it is not statistically significant. Thus, I call this a null relationship.
To be fair, both Pew and Cilizza make a second, more qualified claim: this distrust of government is really worse for Democratic incumbents. Here is the key Pew tabulation:
Here is Cilizza:
More potentially problematic for Democrats is that the Pew poll also shows that the discontent toward the federal government runs far stronger among Republicans and independents and appears to be directly correlated with voter intensity.
But Cilizza’s choice of the word “correlated” is important. We really don’t know, based on this tabulation, whether it’s distrust of the government that driving anticipated voting behavior or something else that is correlated with distrust—namely, dissatisfaction with the economy. The distinction is important. If the economy is the key, the Democrats may not have to fix the government to mitigate their losses in 2010, simply preside over an improving economy. To that end, I note some new Gallup data, which show increasing satisfaction with the direction of the country among both Democrats and independents. Levels of satisfaction are still low in absolute terms, but if these trends continue, the Democratic majorities might yet survive 2010.
{ 4 comments }
I’m willing to bet that the trend looks slightly different if you replace “incumbent loss” with “party switches” and separate the graphs by whether its an open seat or not. Open seats, of course, are always more liable to turnover over. But are they more likely to turn over when there is less trust? I’m not sure I’ve seen this published anywhere, but it might lead to the perception that trust matters for turnover if incumbents are responsive to such a trend and strategically retire and/or open seats turnover more.
Since 1970 it looks like the Pew hypothesis holds up pretty well. Might the 1960s be distinct, both in terms of trust and turnover? What is the correlation from 1970-present? That said, since trust is highly correlated with congressional approval across time I doubt a definitive statement can be made on the matter with the aggregate NES data.
Eyeballing the scatter plot, I don’t see a null relationship, I see a parabolic one: incumbents get tossed out when trust in government is very low or very high.
I got a much more interesting relationship between presidential approval/disapproval and change in the composition of the house. You can see my graph here. It takes a bit of explaining: the x-axis is log(approval/disapproval) and the y-axis is the log-odds of the change in the President’s party in the house compared to the previous congress. Red indicates the President was a Republican, black indicates the President was a Democrat.
You can see that only in 1998 and 2002 did the President’s party pick up seats in the House. You can also see that 1974 was an anomaly since Nixon was still president six months before the mid-term but Ford was in office two months before the mid-term.
As a methodological note, I look at log odds for the change in House composition because the number of seats you lose (or win) should depend in part on how many you have to defend. That makes more sense to me than looking at the absolute number of seats won or lost, as you use for your y-axis.
Comments on this entry are closed.