A New Explanation For the Plight of Winter Babies

Ant passes along a Wall Street Journal report on research that turned up a new explanation for the lifelong challenges experienced by winter babies. "Children born in the winter months already have a few strikes against them. Study after study has shown that they test poorly, don't get as far in school, earn less, are less healthy, and don't live as long as children born at other times of year. Researchers have spent years documenting the effect and trying to understand it... A key assumption of much of that research is that the backgrounds of children born in the winter are the same as the backgrounds of children born at other times of the year. ... [Economist] Mr. Hungerman was doing research on sibling behavior when he noticed that children in the same families tend to be born at the same time of year. Meanwhile, Ms. Buckles was examining the economic factors that lead to multiple births, and coming across what looked like a relationship between mothers' education levels and when children were born." Here's a chart in which the effect — small but significant — jumps out unmistakeably.

Comments

Jumps out?

[ucblockhead]

Of course the difference jumps out. The chart was deliberately designed to make the change jump out by not using 0 as the origin of the Y axis.

This is a very common technique for making a difference look a lot larger than it actually is.

Re:Jumps out?

[wjh31]

Much more important is the lack of error bars, they are what you can use to decide if the difference is greater than noise. However since they seem to be confident enough to include a secondary maximum and minimum, we are led to assume that the error bars are rather small. Since TFA says the study looked at 52 million children over 12 years, it sounds fairly reasonable to suggest that error bars are relatively small w.r.t atleast the primary max an min.

Re:Jumps out?

[selven]

It's well known that children born in Jan/Feb/Mar are much more likely to get ahead because age cutoffs tend to be January 1, so kids born on Jan 1 compete with kids born on Dec 29 in the same year despite having 11 months more experience. Because of this, more attention is given to these "stars", and they perform higher. You should look at the birth months of some professional football teams.

Measured data includes uncertainty

[craklyn]

Any measurement made requires two peices of information: the measurement and the uncertainty associated with that measurement. To present data as though its known with 100% certainty is misleading and incorrect. It seems pendantic to worry about uncertainty, but when you're dealing with small effects on the order of less than one percent, if the error bars are +/-2.5%, then it's absolutely incorrect to refer to the result as "jumping out".

Re:Unwed mothers?

[jonadab]

> Unwed? What is this, 1950?

Statistically, the marital status of the parents is highly relevant to the child's prospects. Children whose parents are married to one another from prior to conception clear through until the child is an adult get on average much better grades in school, are significantly more likely to consistently hold down jobs as adults, make more money on average, are significantly less likely to have a criminal record, are less likely to be smokers, and so on and so forth. These are quite strong correlations.

Now, correlation is not causation. It's possible that the parent's strong marriage does not *cause* the child's good prospects and performance, but rather that both are caused by some of the same socioeconomic factors. But it's still very much relevant in a statistical study like this.

Re:Correllation is Not Causation

[ceoyoyo]

Sigh. Correlation means one of three things with regard to causation. In this case those are:

a) being born in the winter causes increased risk of health and education problems for the baby
b) the baby's increased risk of health and education problems causes him or her to be born in the winter (clearly ridiculous)
c) a third factor causes the baby to both be born in the winter and have increased risk of health and education problems.

The correlation between birth month and risk of health and education problems has been observed. This study is pointing out that the direct causative option (a) is probably not true since they have found possible third factors (c) that appear to influence birth month and are known to have an effect on the risk of health and education problems.

In other words, the study is saying, with actual data and without the childish, misunderstood slogans, the same thing you are - birth month does not cause increased risk of health and education problems.

Showing correlation is required for establishing a causative link between two observations so no, correlation studies do not "need to die." It would be nice if people (including you) understood them a little better though.