Looks like a null result that has been p-value hacked. Without the (paywalled) original paper it's tougher to say, but: The effect only occurs in teachers wit

Looks like a null result that has been p-value hacked. Without the (paywalled) original paper it's tougher to say, but: The effect only occurs in teachers with less than 10 years experience, and only teachers in Switzerland and Austria, not in Germany. Population of Germany: 80 million. Population of Austria: 8 million. German speaking population of Switzerland: 5 million. Approximate average career length for a Physics teacher in German speaking countries: 30 years. So about 4.5% of their sample exhibited a bias against girls. The typical p-value to accept a result is .05. If we sliced and diced the sample into 20 groups, we would expect one to show a significant effect at p<.05, which is exactly what they showed. Note that they did a retrospective analysis based on years of teaching experience, gender of teacher, country of teacher, and fake major of student. There are plenty of combinations of subgroups to find something that makes it over the p<.05 hurdle. Can't find a non paywalled full text of the original paper, but it looks strongly like p-value hacking a null result to me. They found no effect in 80 million population Germany, and only an effect among teachers with less than 10 years of experience in 8 million population Austria and 5 million population German-speaking Switzerland. If we assume that their sample is evenly distributed across German speaking Europe (The test question was in German),