One of the more commonly cited approaches to measuring the incumbency effect and its increase is that developed by Gelman and King (1990). The empirical results that their technique generates suggest that the incumbency effect began to increase in the 1960s and has remained relatively high since then. Given that their results indicate an increase and this book argues that there was not an increase, an analysis of their approach and results is in order.

Their concern is that existing measures of the incumbency effect (the sophomore surge and the retirement slump) are biased estimates. By their calculation, the sophomore surge underestimates the incumbency advantage by about 2 percentage points and the retirement slump overestimates the incumbency effect by about 2 percentage points (Gelman and King, 1990: 1147–49):

They define the incumbency advantage for an entire legislature as “the average of the incumbency advantages for all districts in a general election” (Gelman and King, 1990: 1143). To address their concerns of reducing bias and increasing statistical efficiency (Gelman and King, 1990: 1142, 1145–49, 1151), they present the following variable specifications and equation: