People born in China under the one-child policy (OCP) – a policy applied since 1979, restricting urban couples to having only one child – are less trusting, trustworthy, competitive, conscientious, risk-seeking, and optimistic than those born prior to its implementation.
At least, that was the finding of a recently-published article by Australian economists in Science Express.
With this in mind, imagine you are one of the many individuals born under the OCP. According to the research, you are one of the xiao huangdi, a “little emperor”, privileged and pampered by indulgent parents. You have several undesirable psychological characteristics, and the OCP is to blame.
Now, imagine a different scenario. You are a Chinese policy-maker being lobbied to abolish the OCP. You’ve endured years of bleating about the appalling consequences of the policy, such as the gender imbalance resulting from abortions of baby girls and the increasing proportion of elderly people with fewer offspring to care for them.
You’re now told that those born just after the introduction of OCP – who are now in their thirties – perform more poorly on a range of psychological measures than those born pre-OCP, just a few years older than them.
Both individuals would be justified in asking for evidence: these are, after all, strong claims with significant implications.
But is the evidence compelling or merely suggestive? Suggestive data are fine if conclusions drawn from them are suitably restrained. In the Scottish philosopher David Hume’s words: “a wise man […] proportions his belief to the evidence.”
Using a range of economic games involving decisions about giving and investing money, and sophisticated regression analysis, the authors of the Science Express paper – from Monash University, ANU and the University of Melbourne – drew the conclusions referred to above: those born prior to the OCP (1975 or 1978 for the purposes of the study) are more trusting, trustworthy, competitive, conscientious, risk-seeking, and optimistic than those born under the OCP (1980 or 1983).
The individual born under the OCP may want to ask: is there really a difference between those born under OCP versus pre-OCP? The policy-maker will want to know whether the OCP really caused this difference.
Taking the Science Express article at face value, the answers appear to be “yes” and “yes”. Group differences (for example, OCP participants invested 58% of a hypothetical sum versus 66% in the non-OCP group) are adorned with asterisks – the more, it seems, the better.
The term “causal” is used five times and “impact” ten – these are bold terms, describing a relation that is much more precise than mere association. The results, at first glance, could hardly be clearer.
Same but different
But are there really differences between the people studied? In simplified terms, a result in social science is typically called “significant” if the probability of the result occurring merely by chance, rather than reflecting a real difference, is low. The widely-accepted criterion is 5%.
In the Science Express paper, when controlling for relevant variables to rule out alternative explanations, four of the five differences in behavioural outcomes (trust, trustworthiness, risk-taking, competitiveness, but not altruism) are labelled “significant”, but using a criterion of 10%.
Had the widely-accepted 5% cut-off been used, two of the outcomes (trustworthiness and competitiveness) would not have been labelled significant.
Although the cut-off is largely arbitrary and somewhat flexible, the fact remains that for two of the outcomes, a conclusion of “no difference” is admissible by widely-accepted standards.
And invoking effect sizes – an alternative way to compare groups – does little counter this: the effect sizes for trustworthiness and competitiveness are smaller than for trust and risk-taking. In short, some results are marginal at best.
So can the supposed differences in trust, trustworthiness, competitiveness, conscientiousness, risk-seeking, and optimism be attributed to the OCP?
The authors controlled for a range of variables in order to rule out alternatives. They also correctly describe the study as a natural experiment, which is to say their groups are naturally occurring rather than randomly allocated.
The problem with natural experiments is that alternative explanations cannot be easily dismissed.
Tobacco companies understand this all too well: “the deceased did smoke our cigarettes and die from lung cancer, Your Honour, but was also exposed to asbestos …”
Perhaps the OCP is the most likely explanation for the group differences, but are there alternatives?
The bigger picture
China experienced dramatic societal change over the period in question. The Cultural Revolution ended around 1976 and triggered the opening of China’s economy. China established diplomatic ties with the United States in the very year the OCP was introduced. In other words, a lot was going on.
The extent to which the authors statistically control variables purported to assess these alternatives (e.g. attitudes to government intervention as an indicator of the effect of the opening of the economy) is admirable, but involves many assumptions.
The OCP may be the most likely explanation of the group differences, but alternative explanations are not completely precluded by the data.
So, are there really differences in these psychological variables between those born pre- and post-OCP? Yes, but some are marginal. Can these differences be attributed to OCP? Perhaps – although several plausible alternative explanations cannot be dismissed.
The push for the abolition of OCP can gain momentum if strong evidence for its negative implications is discovered. Strong claims based on merely suggestive evidence from a single study on a narrow sample will convince nobody with influence, and perhaps only offend a population of young Chinese-born individuals who must wear the “little emperor” label.
The conclusions, to reiterate Hume, are not yet proportional to the evidence.