This is the third in our series of posts by students on the job market this year.
Gender disparities in educational attainment, labor market opportunities, political representation, and many other areas of economic and social activity generally favor men around the world (Schwab et al. 2014). Gender also matters for perhaps the most fundamental activity of all: reproduction. In China, India, and several other countries in Asia, parents with daughters often keep having children until they have sons. Where parents want sons, girls tend to belong to larger families, leaving them at a disadvantage when family resources are spread thinly across many children (Jensen 2002). Girls in many of these countries are also more likely than boys to be aborted or die in infancy (Guilmoto 2012). Imbalanced sex ratios distort marriage markets and can have other harmful economic and social consequences.
In the rest of the world, the story is more mixed. Parents in some countries generally want sons, in others daughters, and in some countries (like the United States) parents tend to want a balance of sons and daughters (Angrist and Evans 1998, Bongaarts 2013, Milazzo 2014). However, in much of Africa and Latin America there is no clear evidence of sex preferences during childbearing (Arnold 1997). Sex-selective abortion is unavailable or uncommon and parents are just as likely to keep having children after having a daughter as after having a son.
A New Framework for Measuring Sex Preferences
In my job market paper, I show that these standard signals of sex preferences mask substantial heterogeneity: around the world, many parents want sons and many want daughters. I reach this conclusion by developing a new framework for measuring sex preferences. I model the decision to have children as governed by three preferences: ideal share of children that are sons, ideal number of total children, and the importance of the sex of children relative to their number. The key contribution of this model is that it separates the direction of sex preferences (whether parents wants sons, daughters, or a mix of the two) from the intensity of sex preferences (the degree to which the sex of children actually influences parents’ childbearing decisions). For example, a couple may want a single son, but whether the couple has a second child after having a daughter depends on the relative importance of having a son versus having just one child.
When parents cannot control the sex of each child, parents that want sons may end up with only daughters and vice versa. However, patterns across a group of families can indicate population-level sex preferences. From a large set of possible preferences over the sex and number of children, I identify the combinations of preferences that best explain an observed population. There are generally many such combinations. Although sex preferences cannot be point identified – for example, I cannot say that a specific share of parents want sons – the bounds on estimated preferences are meaningfully narrow. This approach contributes to a growing body of work among economists that allows for partial identification (Tamer 2010). Obtaining point estimates of sex preferences requires placing severe restrictions on possible preferences. I show that, when these restrictions are eased, the set estimates are still informative about the direction and intensity of sex preferences.
Sex Preferences are Widespread
I compile hundreds of thousands of birth history records from Demographic and Health Surveys, China’s Two-per-thousand Survey, and several other surveys in Asia, Africa, and the Americas. I estimate that, for between 40 percent and 90 percent of parents in each of these regions, the choice to have additional children depends on the sex of previous children. Even at the lower bound, these estimates indicate that more than one-third of parents have and act upon sex preferences. In Asia, most parents prefer sons. In Africa and the Americas, I estimate that at least 15 percent of parents prefer sons and at least 10 percent prefer daughters.
The uneven distribution of children across families generates these estimates. In Asia, most families have a last-born son, indicating that parents are generally satisfied only once they have had a son. Elsewhere, there is more variety. For example, parents in Ghana are more likely to stop childbearing after having a first-born son than after having a first-born daughter, but less likely to stop after having two sons than after having two daughters. These patterns suggest that son preference is most common in Ghana among parents that have very few children, while daughter preference is more common among parents that are willing to have many children. Comparable variation at different birth orders across Africa and the Americas cannot be explained by sampling error and similarly indicates that, even though roughly half of last-born children in these regions are sons, some parents want sons and others want daughters.
What drives this heterogeneity in sex preferences within regions? I show that pockets of son and daughter preference in Africa are associated with agricultural traditions: son preference appears most common where inheritance of land traditionally follows patrilineal ties or men perform most agricultural tasks, while daughter preference appears more common where inheritance traditionally follows matrilineal ties or women perform most agricultural tasks. However, these patterns are correlations and I cannot yet conclude that cultural incentives for having sons and daughters trigger sex preferences. I plan to further explore the institutional channels that connect historical cultural traditions with contemporary childbearing decisions.
Widespread sex preferences are commonly thought to inflate fertility levels by leading parents to have more children than they otherwise would want (Mutharayappa et al. 1997). For example, if all parents have a minimum of two children but then keep going until they have a son, then some parents would end up with many children; if sex preferences are eliminated, then all parents would have two children. The model I develop more flexibly allows parents with sex preferences to also stop early upon reaching a particularly desirable combination of sons and daughters. Under counterfactual scenarios in which sex preferences are eliminated, I estimate that some parents would have more children and some fewer, but aggregate fertility levels around the world would change little. These estimates suggest that sex preferences, although widespread, are not a leading cause of high fertility.
The natural likelihood that each child is a boy, access to contraception, infecundity, and other aspects of childbearing may vary across populations, across parents within a population, and even over time for the same parent. The main finding of my job market paper is that sex preferences are widespread. Although this finding is robust to variation in assumed values of many of these aspects of childbearing, particular estimated bounds on sex preferences vary across different assumed values. Additionally, although the choice to abort a daughter clearly indicates son preference, the new framework I develop requires the sex of each child to be random, and I focus on settings in which sex-selective abortion is uncommon (for countries with imbalanced sex ratios at birth, I limit attention to women who completed their childbearing careers before ultrasound technology became available in the 1980s). Finally, I study the relationship between the sex of previous children and whether, but not when, parents have additional children. Sex preferences likely also influence the timing of childbearing and I plan to incorporate spacing between births into the new framework.
Conclusion and Research Agenda
The findings in my job market paper – that son preference and daughter preference during childbearing are widespread – appear to conflict with consistent advantages enjoyed by boys and men around the world. Parents may invest more heavily in their sons even when they wanted daughters. Why do the reasons for wanting sons or daughters differ from the reasons for differentially investing in boys and girls? I don’t yet know but I want to find out, and my job market paper is the first step in a broader research agenda in which I will address this and other questions about gender, fertility, and economic opportunities.
Johannes Norling is a Ph.D. candidate at the University of Michigan and is on the job market this year.