3 ms·

I've thought about the math related to this a bit. Let's say you have a certain (role,level,rating) tuple with 2 women and 8 men. You desire gender equality, a

I've thought about the math related to this a bit. Let's say you have a certain (role,level,rating) tuple with 2 women and 8 men. You desire gender equality, and thus say that the average pay for women and men must be equal within a (role,level,rating) tuple, and to do this, you'll raise the salary of the less-paid gender. The women, due to there being less of them, will have an average pay with higher variance than the average pay of the men. If the women's average pay varies down due to random variance, then the solution is not too bad for Google: pay those 2 women more. If the women's average pay varies up due to random variance, the solution is 4x more painful for Google: pay the 8 men more. Overall, each individual person will on average benefit the same amount from this scheme. But on average, more money will be directed to men due to this scheme than will be directed to women, just due to the fact that more men exist in the company. Similarly, on average, more men will will benefit than women due to more men existing in the company. Disclosure: I work at Google but not on anything related to this.

pseudo0 3y ago
Or the employer shifts to a less flexible compensation system, and everyone loses. Some employers will outright refuse to negotiate pay for a given position because HR gives them a grid or a standard formula.
- Certhas 3y ago
  How does everyone lose? It seems to me that people worse at negotiating are set to win in this case...
  - pseudo0 3y ago
    Pay transparency measures were touted to reduce the "negotiation gap" but ended up undermining the bargaining power of individual workers, so I would be wary of unintended consequences. > Our model predicts that transparency reduces the individual bargaining power of workers, leading to lower average wages. A key insight is that employers credibly refuse to pay high wages to any one worker to avoid costly renegotiations with others under transparency. https://www.nber.org/papers/w28903 https://www.nber.org/papers/w28903
    - Certhas 3y ago
      If only there was a way to address the structural power imbalance in the negotiations between employers and employees....
bitshiftfaced 3y ago
Low sample size has the potential to increase variance, but it isn't necessarily true. Also, you're comparing each group to the average of both groups, but we could instead compare them to an objectively determined value of what that job/role/level is worth to the company, regardless of how it compares to the employee average.
- Thorrez 3y ago
  Comparing salary to the role's worth has 2 problems: * If you set employees' salaries to how much benefit they give the company, the company's profit will be 0, which will make investors mad. * If the worth to the company is somewhere between the average pays of the genders, this doesn't achieve equality between the genders, which is the whole goal of this, unless you're willing to decrease employees' salaries, which will make employees mad. The current system is the minimum change necessary to achieve gender equality without lowering anyone's salary. Any other design either costs the company more, lowers people's salary, or doesn't achieve equality.
  - bitshiftfaced 3y ago
    > If you set employees' salaries to how much benefit they give the company, the company's profit will be 0, which will make investors mad. I don't see how this follows. > If the worth to the company is somewhere between the average pays of the genders, this doesn't achieve equality between the genders, which is the whole goal of this, unless you're willing to decrease employees' salaries, which will make employees mad. In this scenario, all employees within this tuple should theoretically be receiving the same pay. Since it's not related to the average of the group, the company would give raises to those under this figure and would withhold raises from those overpaid until inflation eats up the surplus. You'd do this on an individual basis, so gender doesn't factor in. But it would have the same outcome of equalizing pay among genders.
    - Thorrez 3y ago
      >I don't see how this follows. Yeah, it depends on how "benefit to the company" is measured. One way would be to take the company's current revenue minus equipment expenses, and say that was generated by current employees, and thus that's how much the current employees are worth, and divide that between the current employees. So that gives a profit of 0. That ignores a lot of things though, such as value of past employees, value of investments provided by investors, and value that the current employees are producing that will only be realized by future revenue. It's hard to measure. >the company would give raises to those under this figure and would withhold raises from those overpaid until inflation eats up the surplus Yeah, that works, but it's not instantaneous. If you want instantaneous equality you need something else.

pseudo0 3y ago

Or the employer shifts to a less flexible compensation system, and everyone loses. Some employers will outright refuse to negotiate pay for a given position because HR gives them a grid or a standard formula.

Certhas 3y ago
How does everyone lose? It seems to me that people worse at negotiating are set to win in this case...
- pseudo0 3y ago
  Pay transparency measures were touted to reduce the "negotiation gap" but ended up undermining the bargaining power of individual workers, so I would be wary of unintended consequences. > Our model predicts that transparency reduces the individual bargaining power of workers, leading to lower average wages. A key insight is that employers credibly refuse to pay high wages to any one worker to avoid costly renegotiations with others under transparency. https://www.nber.org/papers/w28903 https://www.nber.org/papers/w28903
  - Certhas 3y ago
    If only there was a way to address the structural power imbalance in the negotiations between employers and employees....

bitshiftfaced 3y ago

Low sample size has the potential to increase variance, but it isn't necessarily true. Also, you're comparing each group to the average of both groups, but we could instead compare them to an objectively determined value of what that job/role/level is worth to the company, regardless of how it compares to the employee average.

Thorrez 3y ago
Comparing salary to the role's worth has 2 problems: * If you set employees' salaries to how much benefit they give the company, the company's profit will be 0, which will make investors mad. * If the worth to the company is somewhere between the average pays of the genders, this doesn't achieve equality between the genders, which is the whole goal of this, unless you're willing to decrease employees' salaries, which will make employees mad. The current system is the minimum change necessary to achieve gender equality without lowering anyone's salary. Any other design either costs the company more, lowers people's salary, or doesn't achieve equality.
- bitshiftfaced 3y ago
  > If you set employees' salaries to how much benefit they give the company, the company's profit will be 0, which will make investors mad. I don't see how this follows. > If the worth to the company is somewhere between the average pays of the genders, this doesn't achieve equality between the genders, which is the whole goal of this, unless you're willing to decrease employees' salaries, which will make employees mad. In this scenario, all employees within this tuple should theoretically be receiving the same pay. Since it's not related to the average of the group, the company would give raises to those under this figure and would withhold raises from those overpaid until inflation eats up the surplus. You'd do this on an individual basis, so gender doesn't factor in. But it would have the same outcome of equalizing pay among genders.
  - Thorrez 3y ago
    >I don't see how this follows. Yeah, it depends on how "benefit to the company" is measured. One way would be to take the company's current revenue minus equipment expenses, and say that was generated by current employees, and thus that's how much the current employees are worth, and divide that between the current employees. So that gives a profit of 0. That ignores a lot of things though, such as value of past employees, value of investments provided by investors, and value that the current employees are producing that will only be realized by future revenue. It's hard to measure. >the company would give raises to those under this figure and would withhold raises from those overpaid until inflation eats up the surplus Yeah, that works, but it's not instantaneous. If you want instantaneous equality you need something else.