Son Preference: Statistical, Economic and Policy Significance

My Economy Lab column this week is about Missing women in China – and Canada too?

Recent research by Douglas Almond, Lena Edlund, and WCI regular Kevin Milligan has found evidence of "son preference" - families planning their children to make sure that they have at least one son - within some ethnic groups in Canada. The paper does a number of careful and rigorous tests which are hard to explain in 400 words or less, but this finding is simple and easy to understand:

Their test families were ones with two girls. If those families decided to have a third child, did they have a boy or a girl? Absent sex selection, about 1.05 boys are born for every girl. But when their group of Chinese, Korean and Vietnamese Canadian families with two girls decided to have a third child, they have a boy 1.39 times more often than a girl – a difference outside what can be explained by biological variation. Some parents in Canada do seem to prefer sons.

The results pass the test of statistical significance: more sons are born can be explained by random chance.

The results pass the test of economic significance: son preference has a real effect on people's behaviour. After all, (micro) economics is the study of choice, of trying to explain and predict people's decisions.

But do the results have policy significance? Policy analysis considers, first, the consequences of individuals' choices for society as a whole.

Sex selection in China may have profound consequences reaching out far beyond the borders of the middle kingdom. A recent UN study reports that there are about 120 boys for every 100 girls in the 0 to 4 age group in China. That gender imbalance, multiplied by the size of the Chinese population, implies that there may be millions of men who will never find partners. Predicted social consequences range from increased militarism, higher savings rates, growth in the sex industry, increased sexual violence and human trafficking to (my own personal prediction) more guys staying home and playing video games.

But what about Canada? Sadly, I lack Professor Milligan's econometric skills (and access to the census master files), so I did the standard "missing women" calculation: is the ratio of men to women in the population more or less what one would expect?

In Canada, there are more Chinese girls than boys. Based on the 2006 public use census files, 51 per cent of Chinese children under 15 in Canada were girls. This is surprising. Normally more boys are born than girls – nature’s way of compensating for boys’ higher mortality rate. Perhaps some of China’s missing women can be found in Canada, girls who have been adopted by Canadian families.

To exclude adoptees, I also considered only Chinese children who had been born Canada. In the 2006 census, there were 105.15 Chinese-Canadian boys under 15 for every 100 girls – compared to 105.15 boys for every 100 girls in the non-visible minority population. There is no more evidence of “missing girls” in Canada’s Chinese community than in the non-visible minority population. Yes, there may be some families who prefer sons. But there aren’t very many of them, and they may be offset by families who prefer daughters [or one of each sex]. And the many more families who love and accept their children for who they are.

Son preference exists in Canada. But either sex selection exists in the non-visible minority community as well, or it is not widespread enough to have a measurable impact on population ratios. In any event, there are no obvious signs that sex selection, to the extent that it occurs in Canada, will have large social consequences.

After assessing the consequences, a policy analyst must ask: what are the alternatives? Stopping sex selection from happening does not change the underlying economic realities and social values that make people prefer sons or daughters. Without sex selection, one risks condemning girls to the hard life of the unwanted daughter.  One of the (many) fascinating paradoxes is that despite - or perhaps because of - widespread sex selection, China does very well on the United Nation's "Gender Development Index." Its gender development index as a percentage of its human development index - a rough measure of the level of gender equality - is number 10 in the world.

As an economist, I find the Almond, Edlund, and Milligan research fascinating - if I was refereeing their paper, I would give it the thumbs up and recommend publication.

But there would be no point in even writing a policy analysis of the effects of son preferences -- it would end up being far too speculative. And what policies could one recommend? Restrictions on telling parents the sex of their child when fetal ultrasounds are performed are already in place in British Columbia. Mao Zedong often quoted the proverb "women hold up half the sky," but his cultural revolution was a dismal failure. Preferences for sons or daughters might disappear if gender did not shape a child's economic and social opportunities - and one could write a 2000 page treatise on alternative ways of making that happen.

So policy analysis has little academic respectability, not because it's too easy, but because it's far too hard to do rigorously and well.

Comments

[Kevin Milligan]

Hi Frances,

thanks for your comments on our work. A couple of things:

1) You are quite right that the social implications of son preference in Canada are quite different in Canada than in China
a) proportion of population that exhibits son preference is smaller; intermarriage can relieve marriage market imbalances.
b) As you mentioned at one point (possibly by email) the richer country (Canada) can import the missing sex to redress marriage market imbalances (ie mail order brides)
c) son preference manifests for the most part at 3rd birth in Canada, whereas it happens more often at first birth in China. This means that even among the Asian population there is not that much *overall* sex imbalance, since the great majority of births are first and second ones, and those are roughly balanced.

So, I agree with you there.

Where I might push you a bit is this statement:
"Yes, there may be some families who prefer sons. But there aren’t very many of them, "

What we see in the data is that Asian families with (girl,girl) exhibit son preference likely through sex selective abortion.

What we do not see is what other families would do if they counterfactually had (girl,girl). My contention would be that the son preference pattern is likely widespread, but we only get to observe it among those who happen to end up at (girl,girl).

In fact, this contention must be true if the sex of the first two kids is random. (Randomization balances both observable and unobservable characteristics (in expectation), so the son-preferences in girl,girl families should be the same on average as those in boy,boy families, if the sex of the first two children is random.)

This is important, because if all those other families (say one with boy,girl) still have the same attitudes as we see in the (girl,girl) families, then that girl in the (boy,girl) family, while she was not aborted before birth, is likely to be subject to a lifetime of living in a family with a strong son preference.

That is, I'm saying that a son preference can manifest itself in behaviors that, while falling short of aborting them, might still be detrimental to girls.

[Kevin Milligan]

How do you do sex determination? Well, outside BC you ask the sonologist. That's not hard.

In BC, you could make a short trip to Blaine and go here.

Abortions in Canada are free at both public and private clinics (except New Brunswick which does not pay for private clinic abortions). There is no legal barrier to abortion in Canada up to and including the end of the third trimester. It is a decision between a doctor and the woman. Access in rural areas may be an issue, but most Asian Canadians live in large cities where that is not an issue.

In short, if you were keen on making sure your third child was a boy, it is quite likely you could get that done.

[Frances Woolley]

Kevin "This is important, because if all those other families (say one with boy,girl) still have the same attitudes as we see in the (girl,girl) families, then that girl in the (boy,girl) family, while she was not aborted before birth, is likely to be subject to a lifetime of living in a family with a strong son preference.

That is, I'm saying that a son preference can manifest itself in behaviors that, while falling short of aborting them, might still be detrimental to girls."

Growing up in a family of three girls, this was something my sisters and I discussed frequently.

There's lots of studies from India, based on data collected before sex selection was so widely available, documenting differences in boys' and girls' access to medical care and high-nutrition foods - Amartya Sen's original 100 million missing women article blamed neglect for the gender disparity.

There are studies that use Canadian or US data that look, for example, at the effect of gender of children on the probability of divorce (the study I've linked to doesn't find that having only girls increases the probability of divorce, but other studies have obtained that result). Other studies have found that sons increase the time that fathers spend with children - and actually daughters benefit by having brothers because it means Dad is around more.

On the other hand, it's not obvious that being an adored son is good for a child's development either. This gives me an excuse to quote from a book review of a recent Mo Yan novel "Big breasts and wide hips":

Mo Yan tells the story of Shangguan Lu, born in the last days of the Qing dynasty, raised with bound feet and married off to an infertile Northeast Gaomi blacksmith who is, she concludes, "a useless gob of snot outside the house and totally subservient in front of his mother." Desperate for a son, the girl cultivates a succession of other dolts and losers, who give her seven consecutive daughters named Laidi (Brother Coming), Pandi (Brother Anticipated), Niandi (Brother Wanted), and so on. After a coupling with Pastor Malloy, an ineffectual Swedish missionary, she finally produces a son, yellow-haired Jintong (Golden Boy), who is the book's narrator. Selfish and indulgent, Jintong remains breast-fed until well into his teens. "Do you plan to suck until you reduce me to a piece of dry kindling?" asks his exasperated mother. Read more: http://www.time.com/time/magazine/article/0,9171,1027589,00.html#ixzz12W5KBgcu

Like I said, policy analysis is actually fiendishly difficult.

[Rachel]

I had a abortion because there was a significant chance that the child I was carrying was disabled and even then it was a difficult thing to do. If my family had decided that I should have an abortion because the child I was carrying was a girl, I would have said no fucking way! Women in Canada are much hard to coerce into having abortions than women in rural China or India.

[TGGP]

The "bare branches" theory of surplus males leading to violence is pretty much bunk. Videogames it is.

[Determinant]

Gwynne Dyer had an article on this. Video games and porn may only go so far. The answer is immigration. China may be famous for having a strong inward-looking culture, but that will likely melt when the delights of married life are waiting to step off a plane. Then of course these new immigrants will want family visits, the grandparents will be a regular flow through Beijing and Shanghai Airports, and China will find itself opening up to the world in a way it didn't expect.

[marcel]

A comment and a question.

1) In fact, this contention must be true if the sex of the first two kids is random. (Randomization balances both observable and unobservable characteristics (in expectation), so the son-preferences in girl,girl families should be the same on average as those in boy,boy families, if the sex of the first two children is random.)

I would be surprised if preferences with respect to a 3rd child are the same in BB, GG and BG/GB families. I would think that most BB families looking to have a 3rd child would prefer a G, while most GG families would prefer a B. The BG/GB families would either more likely be indifferent or if they express a preference could well be thinking about things like the family dynamics, etc.

2) In all this anxiety about unmatched males, how much of a mitigating effect is there from differential rates of homosexuality in the two sexes?

[Frances Woolley]

Marcel - on homosexuality - people often distinguish between homosexuality and men having sex with men. It's not uncommon for a man to have sex with another man (desperate times call for desperate measures) without considering himself gay. The evidence seems to suggest that female sexuality is more fluid than male - women "switch teams" more often. I don't know whether the "surplus males" issue will change attitudes towards homosexuality and sexual practices, but it certainly could. Interestingly, within the last year India has decriminalized homosexuality.

Determinant - are you suggesting that immigration will happen as a consequence of gender imbalance? If so, then I agree with you - in fact, I think it's already happening at, e.g., the N. Korea/China border and with Korean farmers marrying Vietnamese brides. Or are you suggesting that immigration will solve the problem of gender imbalance? If so, then I disagree - basically it just moves the problem from one place to another.

Rachel - You're brave to speak of this so openly - my heart goes out to you.

So much comes down to "what are the alternatives?" The complicity of women in the oppression of other women speaks to the dismal alternatives they face.

[Determinant]

As a consequence, yes. I think Gwynne Dyer has the article online on his website.

[Kevin Milligan]

Marcel: given that you are BB people are likely to behave quite differently than given that you are GG. I agree with you. But that is not contrary to my stated contention.

My contention was that if I took those families that are BB and could observe them in the GG state, they would behave the same as the ones who truly were GG. That is, they all have the same preferences and would act the same under the same circumstances--but some of them were dealt BB by nature and others GG.

[Crystal]

About 15 years ago Koreans had a proportion of male/female infants about 220/100 for the 4th child. They succeeded to reverse this problem without any restrictive policy by means of explanatory work only.

[Frances Woolley]

Crystal, I'd like to find out more about the Korean situation. Certainly both Kevin Milligan's work and my own calculations based on census data suggest that son preference exists in the Korean-Canadian population - though the numbers are small so it's hard to say exactly how strong the effects are.

So perhaps cultural change occurred inside Korea but Korean-Canadians retain the values of the old Korea that they left in the 60s, 70s or 80s?

Perhaps there's less sex selection on the fourth child because people are having smaller families and exercising sex selection earlier?

Perhaps the movement of women from N. Korea, Vietnam, etc into S. Korea is creating a more balanced sex ratio. Immigration can have a measurable impact on sex ratios - see, e.g. the relatively large number of ethnically Chinese girls in Canada who have come here through adoption.

Certainly you're correct in saying that the Korean situation is often cited as an example of the power of education etc. to change people's behaviour, but I'm a little skeptical.

[TGGP]

"Then of course these new immigrants will want family visits"
Who says China has to give immigrants what they want? I bet the Chinese themselves didn't want to be restricted to one child.

[David]

Nice post, Frances. I recently wrote a paper on evidence of sex-selective abortion in B.C. for a graduate population economics course, so I read your post with interest.

Your finding of a non-differing sex ratio among Canadian-born youth of Chinese origin is interesting. I wonder what happens if you run the numbers differently, however. With my data, the preference for boys was driven by Cantonese-speaking families, while Mandarin-speaking families exhibited a statistically insignificant preference for girls. I'm not sure what the reason is for this cultural difference between Cantonese and Mandarin people, but if you're looking at Chinese people as a whole, the differing effects may cancel each other out and not show you the whole story.

As for policy relevance, there are several possible policy solutions to address sex-selective abortion, in addition to restricting when parents can be told the sex of their fetus through ultrasounds:

- If certain demographic groups are statistically more likely to engage in sex-selective abortions, immigration policies targeting these groups could affect the quantity of sex-selective abortions.

- Alternatively, public awareness campaigns could be directed at specific cultural groups to discourage the practice of sex-selective abortions.

- I find evidence (albeit weak) that sex-selective abortions may be in part a symptom of women having less power relative to their male partners in decisions; policies that seek to eliminate the male-female wage gap, and increase females’ relative bargaining power in the household as a result, could affect the sex ratio.

- I also found some evidence suggesting that lower abortion costs (both social and monetary) increase the sex ratio. So restricting access to abortion could reduce instances of sex-selective abortion.

- The B.C. government recently announced plans to charge $50 to reveal the sex of a fetus to parents from an ultrasound; this effectively increases the cost of a sex-selective abortion.

- A Vancouver hospital opened a room where women can anonymously abandon their babies without fear of prosecution. If people consider giving away a newborn as at least a partial substitute for an abortion, such a policy could affect a woman’s decision to have a sex-selective abortion at the margin.

Of course, some of the policies would have major impacts on other issues and would be unpalatable as a result. But the point is that, if you think creatively, there should be policies that can affect sex-selective abortion, at least at the margins.

[Frances Woolley]

David - "The B.C. government recently announced plans to charge $50 to reveal the sex of a fetus to parents from an ultrasound; this effectively increases the cost of a sex-selective abortion." I had though that in BC technicians typically refused to tell parents the sex, and so parents had to travel down to the US. Wouldn't a $50 charge for revelation of the sex be a price increase in this context?

- the different results for Cantonese/Mandarin groups are interesting (which data set were you using that allowed you to break down the two?) My totally untested hypothesis is that son preference may be higher for people with family businesses who are looking for a son to run the family firm, and lower for, e.g. doctors, lawyers, dentists, accountants etc, who could easily have a daughter follow in daddy's footsteps. Would this fit with your hypothesis?

[David]

My understanding is that B.C. health authorities have policies as to when they'll tell parents the sex; most jurisdictions won't tell parents the sex early on but some will tell parents at a time when it's potentially feasible (albeit riskier) for the mother to get an abortion. So the $50 fee is an added hurdle, so yes, it's a price increase.

I was using census data on mother tongues. So I can't say definitively that Cantonese mothers have more boys, but increasing the proportion of mothers who speak Cantonese in B.C. neighbourhoods is linked with an increase in the proportion of babies who are boys.

Your hypothesis about wanting sons to run the family business sounds very plausible; it's not one I had thought about, so I didn't include any variables in my analysis that would support or counter your hypothesis. But now you've got me wondering if there is census data on the proportion of people who are self-employed; if there is, I could potentially add that as a variable to see if it's significant.