Statistics always has noise. I work with a lot of data, I can confirm that there is only so much truth to the results of a multiple regression. But it's easier for me to believe the measurement for the pay gap have merit to it, given all the other structural gender biases in existence.
Unfortuantely I can't access the article about Australia or the CJOE article, but the world of labor article you linked still seems (they didn't give many details on the emperical work they did, which always tends to makes me uneasy) to calculate the gender pay gap based on an aggregate pay averages. Is this really the best way to go about this, as it has some obvious issues in its analysis? As much as I'd like for structural gender bises to give sup-par statistics merit, that's not exactly how science works, and probably is not the best approach when trying to develop an accurate concensus about these issues. I'm not saying we need as complex as a model as the article I linked, but I'd be very interested to see some work that tries to take an approach somewhere in the middle.
Here are a few from 2015 mostly pertaining to OECD countries:
- http://search.informit.com.au/documentSummary;dn=016438411065463;res=IELHSS - http://cje.oxfordjournals.org/content/39/2/441.short - http://wol.iza.org/articles/wage-compression-and-gender-pay-gap
Statistics always has noise. I work with a lot of data, I can confirm that there is only so much truth to the results of a multiple regression. But it's easier for me to believe the measurement for the pay gap have merit to it, given all the other structural gender biases in existence.
Unfortuantely I can't access the article about Australia or the CJOE article, but the world of labor article you linked still seems (they didn't give many details on the emperical work they did, which always tends to makes me uneasy) to calculate the gender pay gap based on an aggregate pay averages. Is this really the best way to go about this, as it has some obvious issues in its analysis? As much as I'd like for structural gender bises to give sup-par statistics merit, that's not exactly how science works, and probably is not the best approach when trying to develop an accurate concensus about these issues. I'm not saying we need as complex as a model as the article I linked, but I'd be very interested to see some work that tries to take an approach somewhere in the middle.