Multiple Imputation for Demographic Hazard Models with Left-Censored Predictor Variables

Michael S. Rendall, University of Maryland
Angela Greulich Luci, Université Paris 1 Panthéon-Sorbonne

A common problem when using panel data is that an individual’s history is incompletely known at the first wave. We show that multiple imputation, the method commonly used for data that are missing due to non-response, may also be used for data that are “missing by design.” Our application is to woman’s risk of first birth as a function of how long she has been fulltime-employed. Using “complete cases” with two years of employment history before the birth interval, we multiply impute employment status two years earlier to “incomplete cases” for which employment status is observed only in the year immediately before the birth interval. We find that, relative to having not been fulltime-employed in the year before having been exposed to a birth, having been employed fulltime for two consecutive years is statistically associated with a higher propensity to give birth, whereas having just entered fulltime employment is not.

  See paper

Presented in Session 228: Missing Data and Bayesian Models in Demography