Alternative Corrections for Sample Truncation: Applications to the 1988, 1990, and 1992 Senate Election Studies

High levels of nonresponse plague all three waves of the NationalElection Studies' Senate Studies. Each of the studies failedto elicit interviews from close to one of every two selectedsample persons, a rate far worse than the NES regular Pre- andPost-Election Studies. This paper addresses three interdependentproblems: Given limited data about the nonrespondents, how canwe model the causes of nonresponse? Using these models, howcan we adjust our analysis of the Senate Study data to compensatefor nonresponse? What difference does nonresponse make for ourunderstanding of the dynamics of Senate elections?