Anna Rybinska, University of North Carolina at Chapel Hill
The intended and the desired number of children are widely used measures in family studies yet, despite their popularity, some key questions about these measures remain unanswered. First, little is known about their statistical reliability. Measurement error in either, or both, constructs can cause biased estimation, blurring the relation between these concepts and observed childbearing behavior. Secondly, while extant analyses provide evidence for a distinction between childbearing intentions and desires, these studies do not consider their reliability. Given their wide use, it is necessary to confirm the dimensionality of these concepts and to do so while accounting for measurement error. In this study, the link between childbearing desires, intentions, and behavior is revisited using a structural equation modeling (SEM) approach in which I test if childbearing desires and intentions are distinct constructs while accounting for measurement error. Using data from the 1979 National Longitudinal Survey of Youth I estimate a SEM of latent intentions and desires and then use the results to estimate the odds of having a(nother) child within the next three years. The results indicate that measurement error causes major bias in the relationship between childbearing intentions, desires and behavior. In models that account for measurement error, the effects of childbearing intentions and desires on childbearing behavior are twice as large as in models that assume perfect measurement. In addition, I find that while childbearing intentions and desires are distinct constructs, when used independently they might predict childbearing behavior with similar precision. Researchers should be aware of this nuance when including either, or both measures in their analyses. Combined these results suggest that researchers interested in childbearing behaviors need to account for both measurement error and the distinction between childbearing intentions and desires in their models or risk severe bias in their results.
Presented in Session 84. Fertility preferences 2