In much of North Eastern India, there are 915 or fewer girls for every thousand boys - a sex ratio of more than 1.09 boys per girl. The persistence of such high sex ratios is puzzling, because it violates Fisher's Principle.
In most species, the ratio of males to females is approximately one to one. Fisher's principle explains why. In a society with fewer females than males, females achieve, on average, greater reproductive success than a typical male. Accordingly, any inheritable tendency that predisposes parents to producing females spreads. The number of females rises until, eventually, equality between males and females is restored.
So how can son preference persist?
One answer is population growth, combined with human institutions. Typically, men marry women who are younger than they are. Suppose that, every year, there are 9 girls born for every 10 boys. Given males' greater vulnerability to disease, this translates to say 9.5 women for every 10 men, by the time that people reach reproductive age. Now suppose the population is growing at a rate of 2 percent a year, men at age 28 at women marry at age 20.
The relevant gender ratio is not between men and women of the same age, but between men and women 8 years younger than they are. In our example, taking into account the 2 percent growth rate, there are 9.5*(1.02)8 or eleven marriageable 20-year old women for every ten marriageable 28-year-old men. By varying the age gap between men and women, a growing population can readily accommodate imbalanced gender ratios.
This is not merely a theoretical possibility. In a recent working paper, Hippolyte d’Albis and David de la Croix calculate a "missing brides" index. Taking into account age at marriage, fertility rates and sex ratios (all of which they take as given), they find that, in the high-growth Asian countries, groom to bride ratios are significantly lower than the ratios of boys to girls in the population. For example, in India, the marriage index is 1.043 grooms per bride, while in Pakistan the marriage market slightly favours males. Son preference only starts having an extreme effect on marriage markets when it is combined with low population growth and a preference for young brides: the most severe marriage market imbalances these authors document are in South Korea (1.194), China (1.139), Singapore (1.134) and Japan (1.092).
Population growth and marriage institutions make son preference sustainable in some parts of South Asia. But what explains its persistence in East Asia, where it is leading to such large marriage market imbalances?
Until recently, son preference did not have disastrous consequences for reproductive success. Human cravings for sugar and fat did us no harm as long as we lived in an environment where these commodities were hard to obtain. These tastes only became a problem with the development of the modern food industry, with its Cheetos and potato chips, cokes and frappuccinos. In the same way, son preference made life miserable for millions of girls and women. Girls named Nakushi or "Unwanted." Girls who grew up being told "Raising a daughter is like watering your neighbour's garden." But it did not prevent girls from being born, or most of them from growing up and having children of their own.
For ideas about son preference to change, they will must begin to have serious negative consequences for those who hold them. Even then may evolve only slowly.
A fascinating paper by Lena Edlund explores how such a change might happen, and its potentially adverse consequences for women. Edlund makes four eminently reasonable assumptions. First, parents prefer having unmarried sons to having an unmarried daughter. This does not require any resort to "culture". Men, on average, earn more than women, so an unmarried son brings more income into the household. Second, parents prefer to have a married son to a married daughter. If married children reside close to the husband's family, this is reasonable enough - and most societies, to avoid consanguinity (in-breeding), adopt either a "with husband's family" or "with wife's family" rule. Third, she assumes that parents prefer a married daughter to an unmarried son - people want to be grandparents. Finally, she assumes that when a woman is choosing a mate, "social position and wealth are important components of attractiveness" - but her model allows for the possibility that men, too, find status attractive.
Edlund's model predicts social stratification - low status parents will choose to have daughers, high status ones will have sons. Why? All else equal, parents prefer sons - married sons are better than daughters, unmarried sons are better than unmarried daughters. This creates a surplus of men in the marriage market. When men are in excess supply, the low status, less desirable ones will remain unmarried. Low status parents start to realize that any son they produce would not be good husband material, but observe that brides are hard to find. To avoid being burdened with a "leftover male", low status parents will, in Edlund's model, choose to have daughters.
When there are excess males, the higher a man's status, the better his probability of marriage. High status parents, knowing that any child they produce will be a hot catch in the marriage market, are free to produce sons, and will do so - after all, a married son is better than a married daughter. Eventually an equilibrium is established, with low status parents producing daughters, and high status ones producing sons.
Edlund's dystopic vision is of a world where girls are raised in poor, low-status families, with less access to education and other resources, while boys are raised in plenty. She leaves it up to the reader to imagine what this would mean the relationships between husbands and wives, for girls' educational opportunities, for the status of women in society as a whole.
Her model avoids making strong assumptions about "culture". In her world, men have no objection to marrying higher status women. This fails to capture one of the sadder aspects of the Chinese marriage market: the poor prospects of highly educated, successful women. A story in a recent New York Times article particularly hit home to me:
Finding a Chinese spouse can be even more challenging for so-called leftover women, even if they often have precisely what the shengnan [leftover men] lack: money, education and social and professional standing. One day in the Temple of Heaven park, I met a 70-year-old pensioner from Anhui Province who was seeking a husband for his eldest daughter, a 36-year-old economics professor in Beijing.
“My daughter is an outstanding girl,” he said, pulling from his satchel an academic book she had published. “She’s been introduced to about 15 men over the past two years, but they all rejected her because her degree is too high.”
(It should be noted, however, that a male aversion to approaching more educated females is not a uniquely Chinese phenomenon - Hitsch, Hortaçsu and Ariely report something similar in this AER paper).
Edlund images a world where, if "sex choice were freely available and the technology perfect, sex ratios might balance" - but the balancing would come through low-status parents choosing to have girls and high-status parents choosing to have boys.
What Edlund does not account for, however, is psychology. Overconfidence bias means that people mis-perceive their actual social status. In particular, people who are towards the bottom of the income distribution still perceive themselves as middle-class. Every parent's firm (but, in some cases, utterly misguided) belief that their son is not that loser who will end up alone and unmarried is another possible explanation of the persistence of son preference.
Modern sex selection technology has only been available for a relatively short period of time. It is only in the past few years that it has begun to have a noticable impact on marriage markets. If something can't go on forever, it won't. Let's hope that this is the case for son preference, or daughter preference, or any other kind of discrimination against children.
Note: the discussion of Lena Edlund's work here has been revised and updated.