districts quite directly without relying on a competitive local campaign for maximum
effect. In fact, despite all the emphasis on spending and experience as important
variables for challenger success, I find that a challenger is extraordinarily unlikely to
defeat an incumbent without the aid of a national tide. Independent national forces not
only exist, they are central to understanding congressional campaigns.
The analysis that leads to these conclusions proceeds in three parts. First, I
examine the current emphasis in the literature on local district factors as most worthy of
attention. I then construct a multi-level regression model that incorporates both local
competitive factors and national tides, and use this model to explain challenger votes. I
also compare these results to a analysis of each election year separately. Finally, I use a
multi-level logit model to explain challenger victories, demonstrating that the dynamics
of votes described earlier translates into a very different dynamic for wins and losses. At
the end, I tease out the larger implications the study bears for the understanding of
congressional elections.
The Two Faces of Congressional Elections
Perhaps because it simplified measurement and modeling, early research on
congressional elections emphasized their national character and downplayed local district
variation. The first academic survey research suggested that House voting was strongly
partisan and unaffected by individual candidates (Stokes and Miller 1962), and research
into electoral returns implied that House elections, to the extent that they had ever
reflected a wide range of local interests, were becoming more national in character over
time (Stokes 1967). Indeed, Campbell (1960) argued that House elections followed
presidential elections mechanically in a “surge and decline” pattern: higher turnout
among weakly committed voters in presidential years would be followed by a reversion
to the underlying partisan distribution in midterms. Congress was left with almost no
role to play at all. Furthermore, initial responses to this research contested not the
emphasis on national-level mechanics, but on the choice of national explanatory
variables, emphasizing instead the role of the national economy and presidential approval
(Kramer 1971, Tufte 1975, 1978, Kernell 1977).
McGhee
2
Convention