Much research on nonresponse and attrition is concerned with estimating the effect nonresponse has on point estimates. Most substantive research, however, aims to draw conclusions on relationships between variables. How nonresponse and attrition affect relationships between variables has received only marginal attention in the literature. Using data from the Swiss Household Panel (SHP), this study examines the extent to which attrition affects the parameters of multiple regression equations. By comparing models run on the full sample from the first wave of the SHP to the subsample that will continue to participate and the subsample that will drop out in later waves, we assess how coefficients from OLS regressions are affected by the selectiveness of the sample. Our models relating
to two well-researched topics, time spent on housework and political participation, produce comparable coefficients for the full sample and the two subsamples in terms of signs and significance, although one coefficient in the model on housework differed significantly in size. Implications for cross-sectional analyses of longitudinal data are discussed.