Overpaying women in Silicon Valley hasn’t traditionally been much of a problem, but results of a recent Google pay equity study concluded that the tech giant has been doing just that.
Really, Google thinks they’re paying women more than men for the same work. And they are making one-time payouts to the men to make up for the hardship. Needless to say, the study most likely did not capture the true state of gender bias at Google.
The study of Googlers compared the pay for men and women with similar jobs and at similar levels of experience. All in all, about 91% of the Google employees were included in the study. To correct for what Google deemed gender bias, they provided $9.7 million in adjustments to pay to a total of 10,677 Googlers (or an average of $908 per person).
There was one large job category, what Googlers call Level 4 Software Engineer, in which women were paid more than men. Men in this category were given extra compensation to make reparations for the gender bias against them. It's not clear how many of the 10,677 employees who were given extra adjustments were men from this category.
Could it be true that there is such little gender bias at Google, or perhaps the tide is turning in favor of women? It's extremely unlikely. The more likely explanation is the Google study was too simplistic. To give a more accurate picture, the study should also examine gender bias in promotion and how Google determines the employment level an employee is assigned.
Segregating women into lower-level jobs
A lawsuit brought by former Google engineer, Kelly Ellis illustrates this major problem with the Google study. When Ellis was hired in 2010 she had four years of experience. She was hired as a Level 3 employee, the level Google assigns to software engineers who are recent college graduates.
In her lawsuit, Ellis explains that shortly after she joined Google, a man who also had four years of experience was hired on the same team. He was made a Level 4 employee, a higher level. Her lawsuit also states that other men with qualifications similar or less than those of Ellis were hired at Level 4. Because these men were assigned a higher level of employment, they received a higher salary and were given more opportunities for bonuses and raises.
The statistical analysis conducted at Google compared Level 3 employees only to other Level 3 employees. It compared Level 4s only to other Level 4s. If women are assigned to Level 3 and equally or lesser qualified men are assigned to Level 4, Google’s current methodology will think everything is just fine. In other words, Google's statistical test wouldn't catch the problem described in Ellis's lawsuit.
A related problem is that men may be promoted more quickly than women. Google's study would not capture this problem either. Nor would it detect if men were given better assignments which resulted in more promotions.
Performance ratings are subjective
A final problem with the Google study may be related to performance ratings. Performance ratings are subjective, and therefore are subject to bias. Unconscious bias may cause women at Google to receive lower performance evaluations than their male peers, even when their performance exceeds that of their male coworkers. When Google compared the pay of men and women in the same job, they controlled for factors that might influence pay like tenure and location, and they also controlled for performance ratings. If there is bias in the performance review process, and women are paid less as a result of it, the Google methodology would still conclude everything is just fine.
One big clue that Google is barking up the wrong tree is that the size of the gender pay gap they found is only $908 per person per year. Estimates suggest entry-level employees at Google have a starting salary of around $100,000. More experienced employees earn significantly more. $908 would be less than 1% of a Googler’s salary. If the gender pay gap were less than 1%, nobody would be discussing it anymore.
Gender bias is not easy to quantify. But Google has a lot of extremely talented employees, and they could figure it out if they really wanted to. So far, it doesn't seem like they really want to.