The Journal of Politics
One mystery of U.S. politics is why the president’s party regularly loses congressional seats at midterm. Although presidential coattails and their withdrawal provide a partial explanation, coattails cannot account for the fact that the presidential party typically performs worse than normal at midterm. This paper addresses the midterm vote separate from the presidential year vote, with evidence from generic congressional polls conducted during midterm election years. Polls early in the midterm year project a normal vote result in November. But as the campaign progresses, vote preferences almost always move toward the out party. This shift is not a negative referendum on the president, as midterms do not show a pattern of declining presidential popularity or increasing salience of presidential performance. The shift accords with “balance” theory, where the midterm campaign motivates some to vote against the party of the president in order to achieve policy moderation.
Bafumi, Joseph; Erikson, Robert S.; and Wlezien, Christopher, "Balancing, Generic Polls and Midterm Congressional Elections" (2010). Open Dartmouth: Faculty Open Access Articles. 2359.