Selective Survival: Consequences and Solutions in Demographic Research

Scott M. Lynch, Princeton University

Despite its ubiquity, selective survival receives little attention in contemporary research outside of the areas of mortality and health demography. Yet, selective survival may affect substantive conclusions in a much broader array of substantive areas. Often, in analyses using panel data, missingness due to mortality is handled using missing data methods that simply do not resolve the biases introduced by mortality. In analyses using cross-sectional data, selective survival is generally not even considered, despite the fact that it affects the pool of potential respondents. In this research, I (1) discuss when selective survival may be a problem and when it can be safely ignored; (2) use simulated data to show the consequences of ignoring selective survival, in terms of biased estimates of means and regression coefficients; (3) show how popular missing data techniques fail to correct for selective survival; and (4) present methods that can partially correct for selective survival.