Do Fathers Prefer Sons?

Parental preference affects divorce, child custody, marriage, shotgun marriage when the sex of the child is known before birth, child support payments, and the decision of parents not to have any more children.

In Asia, the preference of many parents for sons over daughters has led to some 80 million girls "missing" from what should be the normal balance between men and women in a society, perhaps because they have been aborted, neglected, or directly killed. Yet while Americans may read with some horror the fate of female embryos and infants in Asia, they may not realize that American parents, especially fathers, also favor boys over girls. This preference for sons is less severe and subtler than in Asia, but it has consequences nonetheless.

In The Demand for Sons: Evidence from Divorce, Fertility, and Shotgun Marriage (NBER Working Paper No. 10281), authors Gordon Dahl and Enrico Moretti show how this parental preference affects divorce, child custody, marriage, shotgun marriage when the sex of the child is known before birth, child support payments, and the decision of parents not to have any more children. They find that the bias for boys is quantitatively important. Although it manifests itself differently now than it did in the past, it remains significant today.

The statistical evidence based on the 1940 to 2000 U.S. Censuses shows that a first-born daughter is significantly less likely to be living with her father than is a first-born son. Three factors are important in explaining this difference. First, women with only daughters are less likely to marry than are women with only sons. Taking account of the size of families, women with only girls are 2 to 7 percent more likely to have never been married than women with only boys. Strikingly, in terms of "shotgun marriages" - which follow a pregnancy in an unmarried couple - data from California show that for those who have an ultrasound test, first-time mothers carrying a boy are much more likely to be married at delivery. If the gender of the baby is not known, the odds of marriage are no different whether the child turns out to be a boy or a girl. "This evidence suggests that fathers who find out their child will be a boy are more likely to marry their partner before delivery," Dahl and Moretti write.

Second, parents with girls are more likely to be divorced or separated than parents with boys. It may be that fathers like living with sons more than with daughters and, since fathers generally lose day-to-day access to their children in divorce, fathers in marginal marriages may be more likely to want to stay married if a child is a son. The effect is substantial, ranging from a 1 to 7 percent higher probability of divorce, with larger families seeing more divorce. This effect is present in every region of the United States and occurs across race and education levels. But it has declined over the past several decades, so that it seems to have disappeared by the year 2000.

A third possible manifestation of parental bias is that divorced fathers are 11 to 22 percent more likely to have custody of their sons in all-boy versus all-girl families, the authors find. This effect has become quantitatively more important over time as the number of children living with divorced fathers has increased. This difference in custody rates, as well as the difference in marriage rates, has risen over time even as the divorce differential has declined. The result is that the overall gender differential in the probability of living without a father remains large in recent years.

Using a simple model, the authors show that each piece of empirical evidence, taken individually, is not sufficient to establish the existence of parental gender bias. For example, child psychologists and sociologists have found that a father's presence in the household is more important for boys than for girls. So, it is possible that parents of boys avoid or delay divorce because they recognize such possible harmful effects on their sons. Or, it may be that girls are more expensive to rear than boys. The authors therefore turn to revealed and stated preferences on fertility to help sort out parental gender bias from competing explanations for their findings.

In families with at least two children, they find, the probability of parents deciding on having another child is higher for all-girl families than for all-boy families. The magnitude of the effect increases for families with at least three children. Further, among divorced mothers, the probability of receiving child support is lower for those with two girls than for those with two boys.

The preference for boys, the authors find, seems to be largely driven by fathers. At least since 1941, men have told pollsters by more than a two-to-one margin that they would rather have a boy. Women have only a slight preference for daughters. Taking all of this evidence together, the authors conclude that parents in the United States do have a preference for boys over girls.

The authors also examine this question of preference for boys by looking at five developing countries: China, Vietnam, Mexico, Colombia, and Kenya. Overall, they find, all-girl families are more likely to experience divorce and to have additional children than all-boy families. Divorced fathers are more likely to have custody of their sons. Mothers with daughters are more likely to be in a polygamous relationship, at least in Kenya.

The authors note that this preference for boys could matter more in the future. Technology already permits parents to choose a baby's sex, but the methods are now costly and unreliable. "As the cost of procedures falls and their reliability rises, the sex-ratio in the population may slowly become more male," Dahl and Moretti conclude. "More importantly, the bias for boys evidenced by our results may lead to worse outcomes for daughters."

-- David R. Francis