Constructing Fertility Measures with the Own-Children Technique: Old Methods for New Data

Michael Sobczak, University of Illinois at Urbana-Champaign
Gray Swicegood, University of Illinois at Urbana-Champaign
Carl P. Schmertmann, Florida State University

Own-children measures can be used when direct information on fertility is not available. Although a wealth of fertility information is collected in the United States, many existing data sources are not suitable for addressing important questions because the sample sizes are too small (e.g. origin-specific immigrant subgroups) or they lack information on key covariates. Although the decennial Census no longer collects fertility information, own-children techniques applied to the 2000 PUMS files offer a potential remedy to this fundamental problem. Here we outline a protocol for constructing own-children measures for PUMS data and examine the reliability of the measures using several external comparisons. Our initial assessments suggest that the approach yields reasonable aggregate level estimates for many subgroups. We also investigate instances where the estimates appear to be less reliable. Lastly we examine the extent to which the measures can be employed to estimate individual level models of fertility behavior.