Many on the Right view the 2014 midterm elections as a sign that Americans are increasingly fed up with government and that the public is now (finally!) turning in our direction. What happened in the midterms is certainly consistent with that story IF we assume that voters elect Republicans to take on government.* But the election results are also consistent with a story about a national electorate undergoing absolutely no ideological change whatsoever.
Political scientists who study the history of electoral outcomes have built models to predict how Americans will vote in midterm elections given presidential approval ratings, the number of seats being defended by the president’s party, changes in GDP during the first half of the election year, and other variables unmoored to the ideological sentiments of the public. The assumption here is that past is prologue and elections are less about campaign strategy, charismatic candidates, or politically salient issues than most of us here in Washington believe.
The potential hole in that argument, however, relates to presidential approval ratings. Shouldn’t a president who reflects the political sentiments of voters gain public approval and, likewise, earn disapproval when the two diverge? But the numbers don’t bear that out. Prof. James Stimson’s empirical investigations find that “Even when presidential policy affects voters greatly, it is unlikely to affect aggregate Approval,” which is instead driven by the state of the economy and perceived competence.
It turns out that the non-ideological fundamentals underlying the models did a reasonable job of predicting the outcome last November. The median estimate from the more prominent models was a 5-6 seat Republican gain in the Senate (4-5 seats short of the actual result) and a 14 seat gain in the House (which was spot-on).
Those embracing ideological causation can argue that the election results comported with trends in public sentiment about government. And that’s true. But as I’ll discuss at greater length tomorrow, those trends may not signal anything of long-term consequence.
A stronger argument might be that the modelers are making very strong claims based on extremely limited evidence. After all, how many data points do we really have to go on? From 1900-November 2014, there have only been seven midterm elections in the 6th year of a presidential term (eight if you count the wartime midterm in FDR’s 10th year in office), and all but four of those elections occurred before reliable data about presidential approval ratings was available. While midterms during a president’s second year may be relevant data points, they are not ideal.
I’d like to think that voters are increasingly libertarian and that we saw that play out last November. Alas, the midterm elections provide only limited evidence of that.
* The last time the nation gave Republicans total control of the federal government, we got the largest increase in federal spending since LBJ, the largest expansion of the social safety net since the Great Society, and a de facto federal takeover of education, one of the last bastions of state autonomy.