The correlations are not wrong simply because you say so. You cannot eyeball correlations or say the results are flawed because some countries do not fit the pattern. Statistics helps you find those patterns. All the data is publicly available, so download, run your regressions and by all means, submit your rejoinder to Science. Again, the arguments put forward by the Register do not hold any water.

In other words: They did their best to bend the data the way they wanted. If you are as professional in the field as you would like to appear, you know how that is done. Not to mention that you should have heard of a number of papers that were supposedly “peer reviewed” and published in “respectable” journals and later had to be retracted.

Criticizing the Register’s author for looking at data from 2006-2007 is splicing hair. After all he only compared the same variables through comparable sets of data. And if you do look at the data sets from 2003 you will find the same discrepancies.

The data simply does not support Sapienza’s conclusion, as the drawn correlations are fatally wrong.

In the paper (and in the supplementary material, the authors present a number of regressions including not only the indicator of gender equality (not just the WEO, but also results about attitudes from an international survey), and correct for GDP (wealth). In every test they did, the correlation between the gender gaps (in math and reading) and gender equality is statistically significant and high. In their analysis, they also included country fixed effects, to pick up any characteristic of the countries that is not present in the variables included in the analysis (this technique will result in higher standard errors, and, hence, lower statistical significance).

They are also careful to discuss that this does not rule out biological issues. The results also held when they focused on the tail of the distribution (a way to address a criticism about the dispersion of scores), and they also held when they accounted for time spent studying. They examined also sub scores in math, and there are significant differences between genders.

I took away a couple of interesting points: you have the classical debate about nature vs nurture — there is room for both. The issue is that nurture plays an important role and, hence, there is room for better performance. Their results also put in evidence that the reading gender gap needs to be studied: why do female students consistently do better than male students?

It is important to actually read the paper and distinguish what the authors say — they simply document a fact — with the way it is being reported in the press, sensational titles and all. I was following the debate in The Economist web site as well. One person (not me) put very nicely: “Ultimately it is in the best interests of society to promote the potential of all its citizens: male, female, and minorities alike. This speaks to the beauty of this new study.”

Anyway, the ramble in the blog entry about corrupt science is just that — it reflects the author’s prejudices, might find some anecdotes (from blogs?), but he offers no proof. As in any profession, there are egregious cases of downright fraud or interested interpretations, but the scientific peer review system works by and large.

If you had actually read the accompanying piece “Corrupt Science” right here, you would have a glimpse how questionable research makes it into journals being thought of as “respectable”.