The Lifecycle of Response Propensities in Fertility and Family Demography Surveys

Kristen Olson, University of Nebraska at Lincoln

With declining response rates in sample surveys throughout the Western world, demographers collecting sample survey data are increasingly using post survey adjustment via response propensity models. These models assume each sample element has a nonzero probability of participating in the survey that is both conditional on a fixed design and correctly estimated through a multivariate model. However, not all sample persons in fertility surveys receive the same recruitment protocol. Using data from the Wisconsin Divorce Survey and National Survey of Family Growth, we find evidence that response propensities of individuals vary systematically over the course of the data collection in reaction to different features of the recruitment protocol. We find that mode switches and altered incentives move those propensities. We find that those altered propensities are connected to reduce nonresponse bias in key family demography variable statistics.