No CrossRef data available.
Published online by Cambridge University Press: 26 February 2004
In his letter in the July 2003 issue regarding my article, “The 2002 Midterm Election: A Typical or Atypical Midterm?” (April 2003), Howard L. Reiter claimed that I “erred in repeating the outdated ‘rule’ that the number of House seats that the president's party loses in the off-year is directly proportional to the president's margin of victory two years earlier.” Reiter claims that the dynamic of surge and decline (or “reverse coat-tails”) is no longer in effect because it requires a strongly partisan electorate that no longer exists. He concludes that “it is time to bury the myth of reverse coat-tails.”
The record should be set straight. Reiter is wrong in his premise that the electorate is not strongly partisan and in his conclusion that the process of surge and decline is no longer operative. There is now abundant evidence and a growing literature that has found that the decline of partisanship in the 1970s was exaggerated and that partisanship in the electorate is now quite strong (Miller 1991; Keith et al. 1992; Bartels 2000; Campbell 2000; Hetherington 2001; Lawrence 2001; Green, Palmquist, and Schickler 2002; and Abramowitz 2004). In the 2000 election, once you adjust NES data to make it representative of the actual national vote distribution, 89% of Democrats voted for Gore and 91% of Republicans voted for Bush. Moreover, fewer than 7% of reported voters were pure independents. Fewer than 20% of voters split their tickets between their presidential and House votes in the 1996 and 2000 elections, not much more than the amount of ticket-splitting in the 1950s. The American electorate is strongly partisan.
Reiter is also wrong in his conclusion that surge and decline is dead, though his error is understandable. As I documented in my article, there was a substantial compression in the amount of seat change in the last quarter of the twentieth century, probably owing to the decline of competition and the rise of a substantial incumbency advantage. There is also solid evidence now that there was a realignment in congressional politics with the Republican breakthrough election of 1994 (Campbell 1997; Abramowitz and Saunders 1998). Once you control for the compression in seat change and the 1994 congressional realigning election, the effects of surge and decline are unmistakable. A regression accounting for midterm seat change for a party (adjusted for the compression of seat change) with the presidential vote margin (the party's vote percentage minus 50%), presidential approval, the party's initial number of seats, and the 1994 realignment accounts for 94% of the variance in seat change. It expected a five seat Republican gain in 2002 (only two fewer than what occurred). Moreover, the surge and decline variable is the strongest variable in the equation (a beta of .60) and easily surpasses conventional levels of statistical significance (t=8.01). The expected seat changes are also as accurate in recent years as they were in the 1950s and 1960s.
Does this mean that surge and decline is not weaker? No. Surge and decline is weaker than it once was, but so are all of the other influences on interelection seat change. This is because of the compression of seat change. When the range of seat changes are highly constrained, as they have become in recent decades, the amplitude of surge and decline effects is also limited as are the effects of referendum-like evaluations of the president and the economy. The myth that needs to be buried is the myth that surge and decline is no longer a force in congressional elections.