THE EFFECTS OF INCUMBENCY IN UNITED-STATES CONGRESSIONAL ELECTIONS, 1950-1988

Using a random coefficients model and regression analysis, the authors examine the effects of incumbency in U.S. House and Senate elections since 1950. The model is based on assumptions about the behavior of vo ters. A strong effect for incumbency is demonstrated, amounting to at least 10 percentage points in the popular vote in elections since 1974 . Part of the effect occurs because incumbents by definition do not ru n against incumbents. A significant time trend emerges, and there is a significant difference in the effect between presidential and nonpres idential years. The effect is robust, being relatively insensitive to changes in the specification of the underlying model. The model is als o used to examine changes over time in the attachment of voters to pol itical parties.