6 ms·

Did the analysis in the NYT report that TechCrunch is cribbing from here account for Simpson's Paradox? Google employs more men than women. Even if a higher pe

Did the analysis in the NYT report that TechCrunch is cribbing from here account for Simpson's Paradox? Google employs more men than women. Even if a higher percentage of women are underpaid, we could see a situation where more men get raises in an equalization.

rjf72 8y ago
This is not how they did their calculations. It's described here [1]. They were not comparing individuals against some set average and making sure each employee was then brought up to that average and then saying e.g. x% of men and y% of women were bumped up. What they did was compare groups against one another and ensure that there was cross-group consistency. E.g. - the regression formula to determine a man's compensation based on various inputs (such as e.g. performance ratings) should yield a mostly identical result for a women of the same inputs. And then they contrasted groups against various groups. They found that if you were a women, you were paid disproportionately more than other employees. This is why in their announcement they speak of groups instead of individuals, "[T]he 2018 analysis flagged one particularly large job code (Level 4 Software Engineer) for adjustments. Within this job code, men were flagged for adjustments because they received less discretionary funds than women." Remember the goal is not to have every individual be identical, but to ensure that there is no specific group bias. Any given man/woman/etc may be paid more than another employee for reasons that cannot be easily quantified, but when the entire group of all e.g. men or all women are disproportionately paid more than other groups of the same quantifiables then you either have inadequate quantifiables (e.g. - you are missing why certain people are paid more), or there is simply no good reason and you have a bias. In this case, they decided it was a bias. [1] - http://services.google.com/fh/files/blogs/pay_equity_info_sheet.pdf http://services.google.com/fh/files/blogs/pay_equity_info_sh...

rjf72 8y ago

This is not how they did their calculations. It's described here [1]. They were not comparing individuals against some set average and making sure each employee was then brought up to that average and then saying e.g. x% of men and y% of women were bumped up. What they did was compare groups against one another and ensure that there was cross-group consistency. E.g. - the regression formula to determine a man's compensation based on various inputs (such as e.g. performance ratings) should yield a mostly identical result for a women of the same inputs. And then they contrasted groups against various groups. They found that if you were a women, you were paid disproportionately more than other employees. This is why in their announcement they speak of groups instead of individuals, "[T]he 2018 analysis flagged one particularly large job code (Level 4 Software Engineer) for adjustments. Within this job code, men were flagged for adjustments because they received less discretionary funds than women." Remember the goal is not to have every individual be identical, but to ensure that there is no specific group bias. Any given man/woman/etc may be paid more than another employee for reasons that cannot be easily quantified, but when the entire group of all e.g. men or all women are disproportionately paid more than other groups of the same quantifiables then you either have inadequate quantifiables (e.g. - you are missing why certain people are paid more), or there is simply no good reason and you have a bias. In this case, they decided it was a bias. [1] - http://services.google.com/fh/files/blogs/pay_equity_info_sheet.pdf http://services.google.com/fh/files/blogs/pay_equity_info_sh...