Why Voters Sometimes Choose Divided Government

(Bloomberg Opinion) -- For the past 158 years, the president’s party has lost seats in every single midterm election for the House of Representatives with only three exceptions: 1934, 1998 and 2002. It is for good reason that this tendency has been described as one of the few “iron laws” of American politics.

But the extent of the midterm losses can vary considerably. Why that’s the case is a conundrum that has obsessed political scientists for decades. In this year’s election, two very different forces are in play. On the one hand, many voters detest President Donald Trump in deeply personal terms. On the other, the economy is booming. Which will prove more decisive?

In 1975, the political scientist Edward Tufte tried to explain the magnitude of the midterm losses. He parted ways with standard explanations of the phenomenon, most of which portrayed the results as nothing more than a return to partisan equilibrium. In this formulation, midterm losses are more or less unrelated to current events. Rather, they demonstrate that the presidential coattails are no longer operative.

Tufte disagreed with this thesis and a related explanation dubbed “surge and decline.” He correctly pointed out that it was not so much the losses that needed explaining, but the highly variable magnitude of those losses. Tufte speculated that two interrelated factors might explain the magnitude of the losses in midterm election: the public approval of the president at the time of the election and the state of the economy. 

Using both polling and economic data, Tufte found that these two variables — presidential approval ratings and economic conditions — explained “91 percent of the variation in the partisan division of the vote.” He concluded: “From a statistical point of view, the model constitutes a virtually complete explanation of the aggregate vote in midterm elections.”

At first, this finding seemed rather compelling, especially because presidential approval ratings often track with the health of the larger economy. Taken together, they seemed to offer a rather tidy explanation for the extent of midterm losses. But as other scholars started to pick apart this explanation, it seemed less satisfactory than when first proposed.

The first problem was that Tufte’s explanatory formula only covered the years from 1938 through 1974, and skipped over the admittedly anomalous wartime election of 1942. This wasn’t his fault: The polling data he used only went back to the late 1930s. Still, this raised questions about the applicability of the model on a larger scale.

Moreover, Tufte’s findings notwithstanding, researchers found that it was seemingly impossible to explain the nuts and bolts of how national economic performance actually affected votes in local races for the House of Representatives. Indeed, when they sifted through local data, scholars found scant evidence that voters cast ballots according to changes in family finances.

If voters didn’t vote their pocketbooks, was there a connection between midterm losses and the economy? In 1990, the political scientist Robert Erikson looked more closely at the underlying models used by previous researchers and found that they had fallen victim to “a series of unforeseen and devilish problems of model misspecification.”

In an elegant takedown, Erikson highlighted these errors and systematically dismantled much of Tufte’s arguments. Erikson concluded: “The estimated effect of the economy on the congressional vote fails even to achieve statistical significance.” Subsequent research suggested that if there was an effect, it likely died out by the early 20th century, a result echoed in a far more broad-ranging inquiry by Alberto Alesina and Howard Rosenthal.

But what role might presidential popularity play? Here related work by Erikson and two other researchers — Joseph Bafumi and Christopher Wlezien — is illuminating. In a separate article published in 2010, they found that presidential popularity is relevant, but not in the way that everyone imagines. 

While acknowledging that presidential popularity (or unpopularity) does play a role, they discovered that the electorate has already factored this into their likely votes at the very beginning of the midterm year. For example, Trump may be a drag on House candidates, but that’s unlikely to worsen between now and Nov. 6. 

But what does matter, they found, is the dominant political party’s reputation. “The one change in every midterm,” the authors observed, “is the growing negative impact of the president’s party.” Simply put, as the election approaches, voters increasingly contemplate voting for divided government in order to “balance the policies of the president’s party.”

If correct, this theory — that midterm losses reflect a collective attempt to restore balance to the political universe — bodes well for Democrats’ chances this fall. Moreover, midterm losses are most dramatic when a single party controls the White House and Capitol Hill, much as Republicans do today.

Democrats have a history of snatching defeat from the jaws of victory, and this year may be no exception. But the conditions are in place for a watershed midterm election, with dramatic losses for the majority party in the House increasingly likely – so long as Democrats make the midterms a referendum not just on Trump, but on the entire Republican Party.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

Stephen Mihm, an associate professor of history at the University of Georgia, is a contributor to Bloomberg Opinion.

©2018 Bloomberg L.P.