One of the central pieces of evidence of an increased incumbency effect has been the increase in the vote percentage won by incumbents since 1946. Assuming that election results are recorded correctly, the important issue is whether an upward trend actually occurred. As noted before, Mayhew's (1974a) analysis prompted a focus on this indicator. He made four decisions in creating his trend analysis. First, he began the examination with a focus on incumbent vote percentages. Second, he calculated the percentage of the vote received as a percentage of the total major party vote. Third, he presented results in terms of the percentage of incumbent outcomes that were safe or marginal. Fourth, he included contested and uncontested races.

Mayhew's (1974a) first two decisions were quickly accepted for subsequent studies. Almost all the studies conducted since his analysis have focused on incumbents and their vote as a percentage of the major party vote. His latter two decisions, however, were set aside, and that has had significant consequences. Rather than use the percentage of marginal or safe seats, Born (1979) argued that with the presentation of outcomes by categories, small percentage changes could result in a shift of many cases between categories and the impression of a large change. He used the average vote percentage for incumbents (Born, 1979: 812–13), and almost all subsequent efforts to track the trend focused on average vote percentages or changes in them (Payne, 1980: 471–72; Alford and Hibbing, 1981: 1047–51; Garand and Gross, 1984: 21; Gross and Garand, 1984: 230; Jacobson, 1987: 127; Ansolabehere, Snyder, and Stewart, 2000: 24; Jacobson, 2001: 24–26).