Political Behavior, Vol. 24, No. 4, December 2002 ( 2002)
PARTISANSHIP AND INCUMBENCY
IN PRESIDENTIAL ELECTIONS
Herbert F. Weisberg
Party identification is a standard part of our understanding of presidential voting, but
the effects of presidential incumbency on presidential voting have not been recog-
nized in most voting models. Democratic candidates in the twentieth century received
10 percent more of the two-party vote when Democratic incumbents were running for
reelection than when Republican incumbents were running. National Election Studies
surveys show that the effect of incumbency varies with individual partisanship, with
the greatest effect, as expected, among independents. Opposition party identifiers de-
fect at a higher rate than incumbent party identifiers when the incumbent is running
for reelection. Even after controlling for retrospective and prospective economic vot-
ing, a 6 percent effect is found for incumbency. Incumbency thus conditions the im-
pact of partisanship on presidential voting.
Key words: partisanship; party identification; incumbency; presidential voting.
Partisanship is a crucial determinant of the vote decision, as has been recog-
nized since the first report on the 1952 Michigan election study (Campbell,
Gurin, and Miller, 1954). There has been debate as to its meaning (see most
recently Green, Palmquist, and Schickler, 2002), and the precise strength of
its effect on the presidential vote has waxed and waned over the years (Bartels,
2000), but it remains an essential element of a vote equation. There is, how-
ever, an important factor that conditions the role of party identification: in-
cumbency. The early literature similarly focused on partisan explanations of
the congressional vote until congressional scholars forced voting specialists to
realize that incumbency plays a key role in conditioning the importance of
party in that vote. Incumbency is now routinely included in models of the
congressional vote, along with partisanship and other variables; the argument
of this article is that it also belongs in our presidential vote models.
Herbert F. Weisberg, Department of Political Science, 154 N. Oval Mall, Ohio State Univer-
sity, Columbus, OH 43210 (email@example.com).
0190-9320/02/1200-0339/0 2002 Plenum Publishing Corporation