On the Care and Handling of Regression Specifications in Fertility Research

Economists, sociologists and demographers must often attemptto answer important questions with data not well suited to the problemat hand. One example that crops up frequently insocio-economic-demographic literature is the use of samples of women,whose ages span the entire fecund period, to study the effects ofcouples' characteristics on "completed" fertility, or on the demand forchildren. In this case, the usual procedure is to control either for ageor for duration of marriage, and to assume that issues concerned withtiming and spacing of children can be ignored. Under this assumption,differences in the level of the stock of children among families at anypoint in time (any age) bear a one-to-one correspondence to differencesin completed fertility observed at the end of the fertile period. Inthis paper, we explore some of the pitfalls researchers may encounterwhen using data with this characteristic, especially in the absence ofexact knowledge of the functional form of the relationship between ageor duration of marriage and other variables thought to affect actualfertility.