ABSTRACT
A central measure of the incumbency effect over time is the estimation developed by Gelman and King. The Gelman-King estimates are a result of open-seat changes. The political changes occurring in open seats beginning in the 1960s was very different from prior years. The year-to-year changes are a puzzle, given the Gelman-King estimates. The Gelman-King estimation may seem logically valid, but it does not end up measuring the impact of incumbents. Indicators of incumbent fortunes have fundamental validity problems. The actual vote of incumbents can be compared to the normal vote to measure how much an incumbent has moved the vote away from the party vote. The problem with the Gelman-King estimation is that they ignore the consequences of how their selection of independent variables dramatically reduces the variance in scores for vote percentages and the incumbency variable.
