Markov Variation and Sex Ratio Dependence in the United States

Michael S. Pollard, University of North Carolina at Chapel Hill

The measurement of an individual's sex is often assumed to be among the most precise in social science, and if the sex of children born at parity x is exogenous with respect to a variety of factors linked to choice behaviors we may use it as powerful tool for identifying causal relations. The proposed paper examines U.S. data for evidence of Markov variation, which suggests that the probability of having a boy varies within couples according to the sex of previous births. Prior research in other national contexts has generally failed to identify this form of sex ratio dependence. Using the Current Population Survey, Markov variation is identified. Additional sociodemographic factors that may be associated with sex ratio dependence are also examined. Finally, in light of surprising levels of incomplete information in the CPS regarding sex of children, an examination of the levels and correlates of missing information is conducted.