This is a guest post by political scientist Eric McGhee, one of the people behind our House and Senate election forecasts.
A recent Gallup poll found that only 17 percent of registered voters think that most members of Congress deserve to be re-elected, a record low. They also note that years when this number was low correspond to larger turnover in Congress and ask, “Can 2014 be a wave election?”
Our forecasting model of House elections, by contrast, has predicted a status quo election, with relatively few seats changing hands. Is it possible that we are missing something big? The easiest way to tell would be to just add the results of this Gallup question to the model. But Gallup only reports results going back to 1992, while our model goes back much further (to 1952). However, there is another, perhaps more speculative, way to ascertain the importance of these Gallup numbers.
No forecasting model can account for every factor that influences elections. Certain aspects of elections will always be “unexplained” (or not accounted for) by the model. For example, no forecasting model will be able to predict that, say, a Senate candidate will suddenly make controversial comments about “legitimate rape.”
But here is what we can do. Our model can tell us whether either party did unexpectedly well or unexpectedly poorly as a whole in any given year. That is, overall, did one party win more votes than our model predicted, likely because of things the model did not account for? If so, did either party’s over- or under-performance correspond to the public’s desire to throw out incumbents? Since the Gallup question focuses specifically on the House of Representatives, we’re doing the same here.
In general, this Gallup question was only weakly related to over- and under-performance for the years that this question has existed (1992-2012). When more voters said that “most members do not deserve reelection,” the majority party in the House did a bit worse than expected. But the relationship was not very strong. This suggests that our model already captures most of what the public is expressing when they say that “most members” don’t “deserve reelection.”
Here’s why that is. People routinely interpret these survey questions in terms of whether it suggests a coming backlash against “incumbents.” The authors of the Gallup post, Andrew Dugan and Brad Hoffman, write:
…self-identified registered voters of both parties largely agree that most members of Congress do not deserve re-election. Eighteen percent of Republicans, the majority party in the House, say most members deserve re-election, identical to the percentage of Democrats saying so. Thus, voters’ wrath appears not to be directed toward one party in particular, as much as toward any incumbent member of Congress.
But as Alan Abramowitz has noted, anti-incumbent elections are largely a myth. High turnover is always anti-party, not anti-incumbent, and the Gallup question is not great at telling us which party is worse off. For example, Democrats did a bit better than the model’s forecast in 1998, and a fair amount worse in 1994. Yet if we consider the Gallup question to reflect feelings about the majority party, both elections should have had about the same outcome.
To the extent the Gallup question does capture backlash against a single party, our model already accounts for a lot of that — in particular, through the president’s approval rating. That’s not to say that congressional performance isn’t separately relevant to House elections: lower Congressional approval ratings appear to penalize the House majority party. It’s more that the Gallup anti-incumbent question just isn’t measuring congressional performance as well as you might think.