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.
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.
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.
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.
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.
> 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.
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.
Generally, looking at the margin better predicts real world behavior. If you add one employee, how much extra revenue would you expect to gain? A reasonable salary = (extra revenue - HR/overhead expenses - the minimum profit the company needs to justify the trouble of adding one employee). If the last term was zero, then there would be no reason for the company to exist, and so it wouldn't exist and neither would the job.
But in reality, hiring managers are probably setting this internal rate on what the job market is showing along with how badly they need to fill a position in their particular case.
> As a result of the findings, which were first reported by the New York Times, Google provided an additional $9.7 million to 10,677 employees, the majority of whom are men, to compensate them for unexplained discrepancies between their 2018 pay and that of co-workers who have the same responsibilities.
So this sounds like some men and women were earning less than others. If pay has a large variance (e.g. because some people are better at selling themselves), and the workforce is overwhelming male, you would expect exactly this. Even if women on average were to earn less, the majority of those being underpaid would be men.
This is the exact quote from the article, it clearly states that men received more relative to their size:
> In response to the study, Google gave $9.7 million in additional compensation to 10,677 employees for this year. Men account for about 69 percent of the company’s work force, but they received a higher percentage of the money. The exact number of men who got raises is unclear.
Having the "same responsibilities" does not necessarily mean you are entitled to the same compensation. I am not why this would ever be the case. If employee A is 50% more efficient than employee B -- who both are making widget X -- why would employee B be paid the same as A?
The article also states "At the end of 2017, it adjusted 228 employees’ salaries by a combined total of about $270,000", which is a very similar figure.
You can assume there's a wide variance, but all that would mean is that most employees are paid basically correctly with a few outliers. If anything, that would diminish the value of the story.
This article is a few years old so I hope it has evened itself out by now.
I wonder why an 'algorithmic' approach to salaries is not more common.. To achieve this there must be a structured career ladder with standardised job descriptions and (most likely) a way to determine a maturity within the role itself, but then it's a case of having a base salary (often heavily influenced by where the employee is) and some multipliers. Out pops a number (or a tight range) and the employee can choose if they take it based on their valuation of their skills. No risk of gender bias and it takes away a lot of the pain of negotiating salaries felt by both management and subordinates.
Regularly reviewing the base numbers to adapt for inflation and market trends will ensure that both new and existing employees are treated fairly.
At my previous position (head of engineering at an old school sales house trying to become more engineering friendly) we had great success bringing in a structured approach to growth and a corresponding salary calculator. Unfortunately the salary numbers were reserved for management as it wasn't rolled out across the whole organisation but it inspired other, non-engineering, departments to adopt a similar pattern. My vision would be to have these figures open company-wide however I understand that in most medium-large companies this is a time-consuming process.
As someone starting a business today I want to ensure that salaries are open (definitely within the company but likely to the world) from the beginning. Employees should feel comfortable knowing that they are compensated fairly for both the work they do and in comparison to others.
If anybody is looking for some open inspiration about how this is done, check out Buffer[0] and PostHog[1] (Gitlab[2] have one too but the numbers are not open).
Because you give up a lot of flexibility when hiring people.
If you want to hire someone asking even just a tiny bit more than what the algorithm says, you have to adjust the algorithm and increase everyone else, play with the role level to get what you want or just say sorry not possible.
My experience is that it's often the second one. "Yeah, he's more like level X than Y, but we need someone quickly and he's the only one in the pipeline, so let's bump his level so that it matches his salary demand and sign him." It kinda devalues the whole thing if you tweak the input to get the output you want.
True - there is no flexibility but that's kind of the point. The situation you describe is exactly how salary discrepencies and potential gender inequality happen.
Realistically if you have a candidate you want to hire but they are not happy with what the algorithm spits out either your baseline isn't correct for the current market situation (and therefore should be adjusted, affecting every employee), or it's not the right candidate.
The algorithm should be an accurate representation of what your company is willing to pay for an individual at a certain level.
If the candidate is not happy, fine. Can't have everything and they may be overvaluing their skillset. If the company is not happy, a single candidate should not get preferential treatment over the other employees.
"Critics said the results of the pay study could give a false impression. Company officials acknowledged that it did not address whether women were hired at a lower pay grade than men with similar qualifications."
These HN topics are almost verging on dogwhistle politics, where people can't take their head out of the sand for a second to see that the compensation system overall in tech needs to be badly overhauled. Even though this is an old article, it addresses something which should be of interest to most here - namely that whatever color or stripe or creed you are yourself, the odds are that you're personally getting less than you should in terms of compensation.
I know people almost single-handedly keeping valuable back-end systems running, that accrue millions in revenue for a company, but yet the company pays them less than the industry average for their role. Conversely, I have worked with corporate ladder-climbers who contribute nothing beyond endless catch-up meetings that earn more. If you do good work, that directly supports or creates revenue for your company, you deserve a fair slice of that, regardless of how your role is perceived by often technically illiterate HR and management layers.
While I agree in principle. We mostly enjoy benefits of those sins. The money/land/power gained by ancestors by means which we believe unethical/illegal now continues to provide value for the descendants.
The overwhelming majority of people live in cities on land that has no relationship whatsoever to their distant family and have no power whatsoever. As for money, look at any table of income by ethnicity/race [1] in the US, and the top earners are overwhelmingly Asians. Similarly for religion [2] where the highest earners are Jews and Hindus.
One can only speculate about this, but it isn't surprising. In a culture where equity and leftist ideals are rampant, and people get their feelings hurt for having a discussion, you'd expect people to err on the side of caution and make everything seem more equal and thus overcompensate, therefore paying men less.
> I come here for technical discussions, but at this rate soon reddit will have subs with more substance and less political fights than HN
I've been reading HN since 2016. It started decreasing in quality around the COVID pandemic's start, and quickly tanked after most of the YC community moved onto Bookface.
In DEI discussions, equity is typically shorthand for "representation" or "equality of outcome" where if x% of the population is from one group, they should be x% of executives, employees, etc.
Contrast that with equality (now verboten), where everyone gets an equal opportunity at success, but the results are up to individual performance.
> equality (now verboten), where everyone gets an equal opportunity at success
I wound strongly contest that this is because, on the one side of the implementation spectrum, this turned out to impossible in a practical sense, and on the other side, it's far too easy to have the appearance of equality but not the actuality.
On the other side of things - (IMO) equity is also just a better concept - give people what they need to do reach their excellence, and you have far more people being excellent. (Although it sounds like we might have different ideas as to what "equity" even is)
> I wound strongly contest that this is because, on the one side of the implementation spectrum, this turned out to impossible in a practical sense, and on the other side, it's far too easy to have the appearance of equality but not the actuality.
Perfect equal opportunity is the goal, not necessarily the reality. It's not like we can go around handing out genes to make people NBA basketball players or genius mathematicians. Sometimes we might undershoot, other times overshoot, but it's a great ideal to aim for.
> equity is also just a better concept - give people what they need to do reach their excellence
What does that actually mean, and how does your definition differ from the one that I described?
AFAIK+IMO, perfect equal opportunity turns out is a terrible goal, and is unreasonably impossible - you can't even get close, and getting close wouldn't end up being a good thing anyway - for pretty much the reasons you point out. Birth circumstances (genes, parents, etc) mean that the opportunities would never be equal, and (IMO) you wouldn't want them to be - people deserve to be able to be different.
AFAIK, "equity" is what you actually want. [Here's](https://interactioninstitute.org/illustrating-equality-vs-eq...) the usual explanation I see. Interestingly - I think "fairness" here is going to use the same definition as used in AI/ML, which (AFAIK) is "these inputs should not affect the outputs". Or, as it usually get encoded into English inside my head - find out what each person needs to succeed, and then give it to them. Now you have a lot more people succeeding than you did when their circumstances had to be a good match for the opportunity.
I do think there's a bit of legitimacy in the reactions against all this, in one very specific way - I think a lot of times it's assumed that this or that member of "privileged" life circumstances has had some specific need met - aka "the white guys don't need any help" - when a specific individual might not have. Personally, I really would have benefited from any mentor at certain crucial times of my life - college in particular comes to mind - and I think people assume I had them, or had access. I didn't, even if people "like me" did.
(side note - that's also why rule #1 is: don't make statements about other people's lives, only ever ask questions)
> differ
Near as I can tell, your definition limits "equity" to only concerning itself with outcome - and specifically equality of outcome - instead of (overly simplifying) methods plus outcome of maximized and/or sufficient success. That is - "equity" as I understand it doesn't need everyone to reach the same degree of success, it's more about making sure everyone reaches enough success ("to see over the fence") and beyond that to reach as much success as they each person can.
> That is - "equity" as I understand it doesn't need everyone to reach the same degree of success, it's more about making sure everyone reaches enough success ("to see over the fence") and beyond that to reach as much success as they each person can.
Okay, but what is sufficient success and who gets to define it? A CEO who says "25% of our employees are women, mission accomplished" would be tarred and feathered by the media and DEI activists. Even if that is the natural outcome of the choices individuals make in a society with excellent gender equality.
For example, Scandinavian women are less likely to go into STEM than women in less equal countries. If that is in fact a matter of preference rather than coercion, shouldn't the "equitable" level of employment in STEM be based on the level of demonstrated interest?
But in virtually every DEI training I have experienced, anything other than perfect representation is viewed as failure, and the agency of minority groups is ironically dismissed. It's quite depressing really. Maybe you prefer the equity model in theory, but from what I have seen in practice, it is often implemented in a toxic and divisive way.
> Okay, but what is sufficient success and who gets to define it?
As best as I "get it" - that's like asking "okay but there's the object?" in functional programming. This stuff is "means"-oriented rather than "ends"-oriented. You find the people most in need of help, then you help them; rinse, repeat. Trying to base it off of ends is a code smell, indicative of other problems and ills that'll either prevent or undermine your progress, even if measuring those ends is useful to calibrating your means.
> But in virtually every DEI training I have experienced, anything other than perfect representation is viewed as failure, and the agency of minority groups is ironically dismissed. It's quite depressing really. Maybe you prefer the equity model in theory, but from what I have seen in practice, it is often implemented in a toxic and divisive way.
Well shit, that sucks. Sorry you're having to go through that. It's pretty alien to my experience; at worst, the DEI stuff I've experienced has felt vapid - The best has been amazing tho, and felt like it gave agency to many minority groups, and greater connection (less division) between people of different demographics. Equality stuff on the other hand, generally feels like finding ways to avoid the issue, and absolve someone of doing more - at least, now that I've seen "under the hood" more. Maybe you prefer the equality model in theory, but what I have seen in practice, it is often implemented in a shallow and avoidant way.
Equity wasn't a typo. It refer to the policy of quotas, whether it be official or unofficial. And isn't it a valid social phenomenon to be studied, proved, or disproved? Surely, if equity or pay policies are being gamed, then perhaps it might be useful to look into it.
And obviously, if you are here for the technical discussions, why did you click on a thread about salaries of men vs. women?
Finally...this place is ruled by emotion just like every other place. It's only when everyone agrees that there is rational discussion.
> if you are here for the technical discussions, why did you click on a thread about salaries of men vs. women?
Because maybe I would learn that certain positions or areas are better/worse paid than others? Or that managers at Google are really bad at math? Or who knows, maybe the CEO hires really expensive prostitutes and pay them with company money disguised as salaries??
Did you have to make it into a political discussion by shouting your political position? I can't see any outcome of doing so that isn't bad for the quality of discussion.
How do you know my political position? I am not a conservative and never vote for them if that's what you mean...if anything, I dislike all political parties.
But anyway, I was merely stating a possible root cause of the problem, hyper lip-service to a political ideology.
Meh, not surprising. Google also has a bunch of these racial support groups whose participation in can literally help with promotion and performance reviews for literally every racial group except one.
I don't like them - I've seen them literally help racial based college clubs get projects to work on at Google.
If you belong to a culture or country that was founded by a particular ethnic group, I would expect that to be true, so long as the majority population is that ethnicity. That being said, averages are absolutely terrible for informing decisions such as whether or not such a support group should exist. Plenty of white men slip through the cracks and end up in depression, drug abuse, poverty, and all the other ailments. Even if they're statistically doing better on average than some other groups in some cases, it could mean that tens of millions are suffering with no help and no support.
If I were to go to China, of course, as a white guy, I would never presume to be able to even achieve any type of leadership or government position there. And it's not just due to communism. I would not be able to achieve that in Japan either. And I struggle to conjure examples of native born U.S. ex-pats who emigrated from U.S. and then came to prominence in another country but I can think of many examples of the reverse.
> If you belong to a culture or country that was founded by a particular ethnic group, I would expect that to be true, so long as the majority population is that ethnicity.
And what about women from your particular ethnic group? Just because things went like they did historically doesn’t mean we cannot acknowledge that some people have an edge over others.
> Even if they're statistically doing better on average than some other groups in some cases, it could mean that tens of millions are suffering with no help and no support.
That’s a fair point, if we’re talking about general care and crisis response. The topic has been career advancement and employee support however, and a white male in the USA or Europe already has all the odds in their favour.
A support group for white men to help each other with the burden of being a white man is therefore, frankly, pretty ridiculous.
It’s called “the company” or “the university”. That’s my point. Instead of looking at what you don’t have, try appreciating what you already do and use it to help others.
I’ll ask you a similar question: when is white history month?
Never. History months are marketing gimmicks though so it's not a thing I would ever seek to support. I do think people should be more aware of history in general and the long chain of complex events that led to the current situation we find ourselves in. But that would require real nuance and our modern society just doesn't have time for that.
White people still are the most overrepresented in leadership https://about.google/belonging/diversity-annual-report/2023/ - you can deny reality all you want - there's only one group that is overrepresented. Let me guess the next reply to this will be some form of "White people are better"
The commenter was talking about promotions so I was looking at leadership % vs staff - nice change of topic though. Why is the US population a good source of comparison? Especially when these companies are full of immigrants? You need to look into your own privilege instead of blaming society friend.
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.