8
We control for the effect of quality challengers and increased vulnerability of freshman
incumbents by including dichotomous variables indicating the presence of a quality challenger,
one who has held previous elective office, and the presence of a first-term incumbent. To
address the effect of candidate spending, we include a measure of the difference in challenger
and incumbent spending. Based on prior work (Canes-Wrone, Brady, and Cogan 2002; Erikson
and Wright 1993) we measure this spending difference as the natural log of incumbent spending
subtracted from the natural log of challenger spending.
In predicting whether or not incumbents are contested, we also include a measure of
district safety to capture characteristics of the district that are likely to attract or deter potential
challengers. Using CQ Weekly ratings of district competitiveness, we measure the relative
safety or competitiveness of the district based on local political forces. The measure ranges from
a low of zero, indicting an extremely competitive district where there is no Republican and
Democratic advantage, to a high of three indicating a safe district and a decided advantage for
the party of the incumbent. We expect that incumbents in safer districts are less likely to draw
challengers in the general election.
The additional independent variables in our models are gender and an interaction term
capturing the conditional effect of gender and ideological distance. Gender is coded one for
women and zero for men. Consequently the coefficient of the multiplicative interaction term
represents the additional effect of ideological distance on women’s electoral fortunes. If the
coefficient is close to zero, then men and women members of Congress experience the same
electoral effect of ideological distance. If the coefficient of the interaction term is significant,
then ideological distance affects men and women differently.
Convention