southerners to return incumbent congressmen to office, outside periods of realignment such as
1994. The South is controlled for using a dichotomous term, coded “1” if a district were in the
eleven state South, “0” otherwise (Key, 1949; Price, 1957).
Finally, to account for the war vote, we controlled for the roll call decision of each
incumbent member, coded “1” for yea and “0” for nay. The House roll call for use of force in
Iraq passed 250-179, taken on October 10 2002 (The partisan division of the roll call appears in
Table 1). If the war vote hypothesis is sustained, we should expect legislators who voted for
war to achieve significantly higher reelection shares, other effects held constant. If the war
vote hypothesis is not proven, as was the case after the 1991 Gulf War, then no significant
effect will be in evidence.
The analysis is broken by incumbent party, in order to facilitate the comparison of
potential war costs or benefits by incumbents of each party. We expect little if any effect by
the war vote on Republican legislators, who nearly all voted for the resolution (only six
opposed it, resulting in almost no variation in that independent variable among Republicans),
while a war vote effect should be evident among different Democrats, who divided on the issue
within their caucus.
The results of the ordinary least squares regression appear in Table 2. In 2002, all of
the control variables – the Bush presidential vote baseline, the prior incumbent vote share from
2000, the log of incumbent and challenger spending, and the southern control – are significant
and in the expected direction for incumbents of both parties. The Iraq vote, coming less than a
month before the November midterm election, is not significant for either Democratic or
Republican incumbents.
In 2004, two full years after the vote on war, over a year after the end of “major
hostilities,” and after a campaign centered on issues of security and mired in a rhetorical debate
9
Convention