Hacker News new | past | comments | ask | show | jobs | submit login
[flagged] Gender pay gap shrinks when companies disclose wages (cornell.edu)
68 points by LogicRiver 22 days ago | hide | past | web | favorite | 73 comments

>From 2003 to 2008, the gender pay gap at the reporting companies shrank 7 percent, from 18.9 percent down to 17.5 percent

Is this really a large enough effect size to justify such a headline?

>Our findings suggest that regulatory mandates on pay transparency, as a means to overcome biases against women in the workforce, may be effective in closing the gender pay gap.

No, they really do not. A potential reduction from 18.9% to 17.5% does not 'close the gender pay gap', and attempting to close such a gap in this manner to begin with just demonstrates a misunderstanding of what confounding variables cause it to begin with.

I once read an interesting thought experiment about this, adapted to this story it goes like this:

We assume:

- Greedy capitalist bosses (wanting to pay less for more work)

- Same productivity of men and women

Now if there's an 18.9% pay gap, then women are on average 18.9% cheaper to employ than men.

Greedy capitalist bosses should therefore only employ women, closing the gender pay gap by driving womens' wages higher (and those of men lower).

Yet this doesn't happen, so one of the assumptions is wrong or the gender pay gap difference (the 18.9%) is not correct.

Another assumption that could be wrong is that greedy bosses can accurately determine productivity. If bosses have biases that cause them to underevaluate women's productivity this would also explain the pay difference.

That would still give a huge benefit to the greedy boss who can "accurately determine productivity" and hires all women for cheaper, which means it would inevitably be copied in the end.

Not sure.

Even if you had only 1 out of 10 bosses able to determine productivity, that guy'd only employ women, so he'd cover some of the "losses" of the other 9 bosses.

Also, you'd probably see a bimodal pay gap distribution per company.

Can we assume greedy capitalist bosses aren't to stupid to run (or get someone else to run) A/B tests?

Yes I admit. It is leading question. And personally I guess greedy capitalist bosses would happily ditch more than a few biases if that meant they became richer, quicker.

Given you agree you have three options that I am aware of (interested to know if there exist more):

1. oppressing women on a large scale is somehow more valuable to greedy bosses than getting a huge competitive advantage over other greedy bosses, or

2. for some reason it wasn't such a huge competitive advantage after all.

3. for some reason it isn't possible to test it. Maybe because women chose another career path from the start. Or maybe women aren't as easy to fool into working for greedy capitalist bosses. Or something else.

Your assumption in your model is wrong. You aren't assuming the same productivity of men and women, you're assuming the same observed productivity of men and women.

If you're in a widget factory and both a man and a woman are producing 10 widgets per hour then your effect would be expected - because the boss would be able to evaluate true performance. In the real world however, no job is that simple. The problem with the gender pay gap partly is that women are assumed to be less productive than men in certain jobs. You don't hire a woman engineer because women are interested in people not things - so your expectation is that a women is 20% less productive than a man because you've decided being interested in people is bad for the job and that influences your expectations when hiring.

Indeed - and we observe that quite a lot of factory line work that doesn't specifically require hefting large parts around is done by women. Especially in clothing and electronics.

The point of the model is that there will be many bosses that are not affected by such biases, who will then employ only women because they recognize that the other bosses have overlooked women due to "liking people not things".

The incorrect assumption is an implicit one: the fungibility of workers.

Your argument would work (-ish) if you were arguing against pay inequality. And, in fact, pay inequality is usually a lot lower than the pay gap.

But for many large organisations, the pay gap is due to women being more likely to be in lower paid roles while men are more likely to be in higher paid roles. Replacing men with women and vice versa would reduce the pay gap but wouldn't dent the overall wage bill significantly.

You can't just "replace men with women" though, as it's not clear that the reason men are in higher paid positions is because of discrimination. For example, men on average work longer hours than women, which obviously contributes to the pay gap. Men also take less vacation than women, which generally makes them "more reliable". If I have two SysOps people, and they have the same skill level, I'm probably going to promote the one that took less vacation time.

The priority should be in curbing the behavior of men who are overzealous in their work to the detriment of their own health. That would probably bring down the pay gap significantly.

> You can't just "replace men with women" though

Indeed. I was just pointing out why the argument was broken. The lack of symmetry in the workforce would also mean that that's not a way to reduce the pay gap even without the assumption of a reducing wage bill.

> Men also take less vacation than women

Are you sure? Men are statistically likely to take more of their vacation time (in the US at least) than women. If you are correct that implies that men get more vacation time than women but I'm not sure that that's correct.

Taking the first hit on google, men work 42 min more per day when taking into consideration aspects like vacation and sick days. Every second week that almost translate to an whole additional day in difference.

"among full-time workers (those usually working 35 hours or more per week), men worked more per day than women—8.4 hours, compared with 7.9 hours. (See table 4.)"


The questions we should ask our self is what are the incentives. How does wages, promotions, social status, health, happiness, children and social support look for men who is in the 50% that work longest vs those men who are in the 50% fewest hours at work. Then we do the same question for women.

Yes, but about 3 hours of each day that man is reading HN, so I think they're actually working less time. I think they might be posturing longer hours maybe?

In any organisation it's only the owners who are ultimately concerned with overall cost - managers can have all sorts of incentives for larger or less efficient staff.

> In any organisation it's only the owners who are ultimately concerned with overall cost - managers can have all sorts of incentives for larger or less efficient staff.

I wonder if that could be accounted for by performance/productivity related bonuses or other perks.

>- Greedy capitalist bosses (wanting to pay less for more work)

Well, that, for one, is a given.

I found this to be really interesting:

"The researchers were especially surprised by the way in which the companies adapted to the regulation – by slowing the growth of men’s wages, not increasing women’s."

It seems like a given outcome, to me.

Suppose that half of your workforce makes 60% of the wages, and the other half make 40%. Suddenly, PR dictates that you need to even that out.

Your budget for salaries hasn't increased. Your other costs haven't decreased. Your projected revenues haven't increased. The money to bump the bottom half up has to come from somewhere, and with a fixed salary budget, it comes from the other half's expected raises and new hires.

Unless, of course, you have the flexibility to start rewriting all of your budgets, but that still hasn't changed the fact that both revenues and all other costs remain fixed.

Seems like a self-fulfilling prophecy in that scenario. Equal wages is supposed to result in improved production... isn't it?

>Equal wages is supposed to result in improved production... isn't it?

No, why would they?

Attracts better qualified personnel? More motivation? Some macroeconomic optimization where price and value mismatch results in inflation?

I don't know! \o/ Just what my intuition says.

I experienced this first hand and as a result I departed for greener pastures. Isn't diversity a great thing when it suppresses wages, no wonder all the big boys are pushing for it.

This is happening at my place right now. Leadership has dictated that priority is now to even out gender imbalances rather than giving raises. Since the women in our company are heavily skewed towards lower-paying roles (QA, customer support), this means the company will avoid giving raises to the roles that actually are in demand and have strong salary growth elsewhere. Needless to say, those of us with options have started looking.

Of course, everyone who gets a competitive offer elsewhere will get a counteroffer, so this policy only applies where it’s convenient.

I don't get this kind of destructive attitude. Surely the sensible thing to do when you want more equality and women are skewed towards lower paying jobs, is to hire more women in higher paying jobs?

I'm all for paying people in lower-end jobs better, mind you, but using this as an excuse to hurt salary growth seems like it's intended to discredit the whole idea of gender equality. Well, unless the people involved are provably overpaid, like CEOs tend to be. But somehow I suspect the CEO is still getting his excessive raises.

I don't think the real intention is equality at all. I think it's scoring PR points via virtue signaling, and reducing expenses as a bonus. (Who would dare oppose a move that appears to increase equality, even if it hurts their wallet?)

> When a measure becomes a target, it ceases to be a good measure

Very simply, the measure of women being paid less is useful on wide scale in demonstrating women's position in society, but it doesn't go into the complexities of the issue, so aiming to get your company more equal on that measure can mean so many things as to make it a useless target unless you understand the situation.

Excellent point. Reminds me of the Bechdel Test, a simple test to show how poorly women are treated in movies. To pass the test, the movie has to have at least two named women talking to each other about something other than a man. It should be trivial to pass, yet a surprising number of movies don't.

But that test doesn't mean that movies that pass the test are better or more feminist than movies that don't. Terribly sexist/misogynist movies can still easy pass it, whereas Gravity, a movie starring only Sandra Bullock for most of its duration, doesn't pass it because there's nobody else in the movie to talk to.

Similarly, the point of calling attention to the gender wage gap should be to pay women better and give them better access to higher paying jobs, not to deny men raises.

Equality should lift people up, not push them down.

The article cites a 2.5% reduction in overall wages, and a 2.5% reduction in productivity. (Or one was 2.8, but close enough). Which sounds exactly like what you suggest: the most productive employees used to get other offers, and use them to ask for raises, and now they leave, or just slack off.

If you know any employees of the big boys, you'll know that management is not supportive of diversity until they face worker revolts.

Your model makes no sense. If powerful forces wanted diversity out of self interest all along, we wouldn't even have a diversity movement, because the environment would be diverse already due to natural locally optimizing market forces, so there'd be nothing to advocate for.

The other way to look at it was that men were artificially overvalued and the market is now correcting itself.

This is entirely unsurprising given the mechanism. Disclosure makes companies not want to give anyone a raise as an individual because then everyone else will know about it and want one too.

Then mean wages go down, but if more of the raises had been going to men then wages for men go down even more.

Did the researchers measure employee departures? If wages stagnated in general, then one would expect that employees who are more desirable/mobile to depart. This would make it seem like wage equality went up, when in reality the people who were previously making more are still doing so — they're just doing it at a different company.

They say "we find no statistically significant change in departure rates of males or females across firm hierarchies."

Have not read it closely, but "no statistically significant" need not rule out the scenario that departures drive the whole effect. It's just saying that the error bar from counting departures is large enough to include zero, while the error bars on mean salary are smaller, because everyone has a salary but only a few leave.

Many companies award salaries in fuzzy way where few good negotiators people negotiate themselves larger salaries then similarly performing rest. One would think that salary disclosure helps to close such gap among men themselves bringing average lower.

A pair of scissors labeled "cynicism and greed" poised to shear the head off a crustacean labeled "meaningful societal advancement."

I'm not sure how else they would behave. Increasing the salaries of all the women would be like admitting to discrimination, wouldn't it?

Edit: I mean that instantly increasing the salaries would be like an admission.

How does depressing men's wages not do the same?

There is a material difference between "paying too much" and "not paying enough". Clearly they want to be on the former side and not the latter.

They're not "depressing men's wages" -- just giving them lower raises. No one's salary suddenly went down.

If you've been underpaying women in a discriminatory way while pretending to be a meritocracy, the real fix is to pay women more.

But if you suddenly raise salaries for all the women, it proves you were never a meritocracy.

To keep up your pretense while also fixing your mistakes, your only option is to slow raises in order to restore fairness over time -- which is what they did.

The third option would be to reduce men's salaries, which of course wouldn't be possible.

Based on what I have experienced in corporate America, this is completely expected behavior.

Here's the study: https://wpcarey.asu.edu/sites/default/files/daniel_wolfenzon...

To describe a 7% change as "shrinks" is a bit of a stretch.

And to attribute it to solely to the requirement to disclose may be incorrect as there was only a 1.4% difference vs companies that were not required to disclose anything (the 7% figure is vs pre-regulation).

So wages rose for women post-regulation, regardless of whether the company they worked at disclosed wages or not. Meaning other provisions in the law also likely played a role in the decrease of the gender gap.

The study also references another study that finds:

> that firms with higher pay inequality exhibit larger equity returns suggesting that differences in pay inequality across firms are a reflection of differences in managerial talent.

All just food for thought. Regulation and transparency can produce interesting outcomes and side-effects without making us any wiser as to the dynamics driving the differences in outcomes.

Not just gender I think. I witness over and over how employees one day realize they were underpaid and get paid immediately. Including myself. Sometimes the company asks the employee to keep it silent so the other ones won't do the same. This is ofcourse unhealthy situation, and too common. You got to win your rights manually.

The problem of gender pay gap is a really hard one to tackle. You have to be a fool to pretend there is no issue, but the number of causes is incredible at the gender, education, industry, company and individual levels.

Every time there is a public debate around that topic you end up with two major groups, one saying that women work less or are less productive because of XYZ, the other saying something close to "it's the patriarchy".

Striving for equality of outcome is a good first step, but it's attacking the symptoms and not the causes.

Just like for politics or employees quotas, you can bend the end of the process to make it look fair and balanced but if the 10 steps before are biased against women (and other groups) you end up with an illusion of parity while the only thing that changed is you now make "positive" discrimination.

Making the wages public is a good thing no matter the reason, after all the first step of making a decision is to have access to data. If it helps with pay gaps all the better.

How about option 3: men choose to work on jobs that pay more but are more stressful while women chose to work on more social jobs that pay less and value other things more than just pay.

You'll have to be a bit more precise because right now it's hard to tell if you're trolling or trying to make a point. In any case, the pay gap isn't a thing that can be addressed with a single sentence such as yours.

It's a really simplistic point of view to say social jobs are valued other than with a salary. And I'm fairly certain that social workers, teachers and nurses get their share of stress.

If anything the "men do manly jobs and women do social jobs" is another symptom of the underlying causes. Sure you can say "that's the way it is" and move on, but if you want to take part in the debate you have to show a minimum of intellectual honesty.

Plus it doesn't work for people having the same position in the same company.

I agree stress was a bad metric, I'm sure there are other metrics some sociologists would be able to enumerate, I can just speculate based on my experience: In my country when I started studying computer science -- which was a niche area, I had 90% male colleagues; after 15 years with all the media coverage of high salary in the industry, the proportion shifted to 60-40, because the profession has become main-stream.

So another factor could be risk -- women are more risk averse than men; hence they choose more established professions where the pay is less.

Then, there's the drive and single-mindedness some high paying jobs require -- men are built that way, women are less likely to have obsessions on obscure-technical hobbies; heck, most men aren't built that way, it just happens that for those people that have such drive and single-mindedness, most of them are men.

And that's the way it is -- unless there's some moat protecting your job, due to the laws of supply and demand the higher paying jobs are the ones nobody wants to do because of risk, discomfort and so on, and those that most people cannot do because of the dedication and single-mindedness they require, and for both categories men are the likely candidates.

Yes it is a very simplistic explanation to why genders tend to gravitate to one job or another. And yes, I'm saying that most, but not all of the disparities are biological and while it's a great idea to encourage women to pick higher paying jobs, we're at a point in time where we'll see diminishing returns in those attempts.

For the same position with the same experience, in roughly the same location, I'd be very much surprised if the AVERAGE/MEAN pay gap is more than a single digit percent across an industry.

By all means go after those X% numbers, teach women to negotiate more, assert themselves and so on, but I think you'll end up with numbers similar to the ones in this study -- a couple percent improvement.

What I'm trying to say is there's no conspiracy against women earning more in the western world.

Since gender only seems to play a very minor role in wage differences between men and women, as opposed to say, willingness to work full-time/overtime, personality, and job choice, it is not at all clear that the 'gender' pay gap is a problem that needs tackling.

I believe systems that strive for equality of outcome and put group identity above individual qualities will only create more injustices, not fewer. If the current situation we have is mostly the result of men and women using their freedom to make different choices in life (on average), then forcing different outcomes with these systems seems like an even worse idea.

The assumption that equal work deserves equal pay is fine. But who decides what equal work is? There are a lot of factors that can make one employee more valuable than another employee in the same role, and therefore it only seems reasonable that companies get the freedom to reward those employees differently. All these subtle factors will not show up in the data, and so I do not believe it will help with anything, except for making people see problems where none exist.

> Since gender only seems to play a very minor role in wage differences between men and women, as opposed to say, willingness to work full-time/overtime, personality, and job choice

But isn't personality heavily influenced by gender, job choice by education and stereotypes, will to do over time by lifestyle (getting the kids from school, ...)

I agree with your point of view on identity politics, but you seem to assume that the current "system"* is neutral and isn't favoriting any demographics right now. Last time I checked, studies tend to show that the "gender pay gap" isn't entirely based on gender, as you said, but still, a non negligible part of it is.

* I don't like that word but can't find better

Which data are the fools ignoring, more exactly?

1) The average income for every individual in society?

2) The average income for every working individual?

3) The average income for every full time individual?

4) The average income normalized to work hours and work position?

5) the average income normalized to work hours, work position and experience?

6) the average income normalized to work hours, work position, experience, city of occupation?

7) the average income normalized to work hours, work position, experience, city of occupation, volunteering for less desirable works like overtime and night work?

8) all the above with additional quasi-legal normalization for aspects that may effect productivity, like number of sick days?

The first one that make them feel good about the situation.

You can rationalise a lot of things by interpreting statistical analysis to support your point of view, you just have to cover it with a "duh that's the real _facts_" and move on.

That illustrate the issue between those who advocating against pay gap and those they are trying to convince. The data is not relevant, it is the fight, the struggle, the cause! It is great if there is data to support it but if not, then that is just rationalization of fools.

Each and every one of those create a different data, with different conclusion about what the issue is and what the solution are.

The first ones highlight among other things that fewer women are working in the first place. Men don't have a choice and will end up homeless and at the bottom of the social heriarcy if they don't. Women are also social expected to take care of children, and because of breastfeeding are likely to spend the time at home for the first few years. Men also die earlier, both from work accidents and from a health care that focus more on women then on men. This result in more women at the end of life not working compare to men who will instead be dead.

For social and cultural reasons, women are more likely to work part time while men are more likely to have task wage. Women are more likely stay in education for longer and look for a job until they find the right one, while men start working quickly and with more emergency. Mens social status is tied to income, while women social status is disconnected from income and benefit from social activities. Men will on average volunteer to undesirable high pay works like high risk work, night work and overtime, while women will not. Women will have more sick days and spend more time on work hours with the health care system, while men either skip going to the doctor or only do so if it doesn't conflict with work. Men are more willing to change city of occupation and move out of a social environment than women.

And after all of that we may have bias, which is the exclusive focus point of the discussion. All the above is usually ignored.

  Striving for equality of outcome is a good first step
Did you you mean equality of opportunity?

Equality of outcome is evil. Equality of outcome is a goal that stomps all over individual liberty and completely ignores the fact that different people have different abilities and different personal goals. Outcome should only be equal in cases where ability/desire is identical and opportunity was equal. Anything else is fundamentally wrong.

I couldn't find an N of the study.

It seems important to me in order to ascertain the statistical significance of this. Were there 10 companies in one group? 100? 1000? Did anyone else see this in the study?

Based on table 1, it was probably about 4000 companies and 66000 employees total.

These articles tend to attract comments with dire predictions for US society if people knew each others salaries, so just a reminder Norway always does nationwide salary transparency


> These articles tend to attract comments with dire predictions for US society if people knew each others salaries, so just a reminder Norway always does nationwide salary transparency

True. But I think it is worth pointing out that the US is not Norway. There is a significant culture difference for one. Secondly the policy may simply not scale. The US population, workforce, economy is orders of magnitude greater than that of Norway.

Been working on a tool called http://levels.fyi to help increase career ladder and salary transparency in the workplace (mainly for tech right now). Would love suggestions on things to add to make it more powerful!

Shouldn't remuneration depend on the amount of work done? Gender issue shouldn't come into it.

Should the price of something depend on the amount of work done? On how much the buyer values it? Or on the amount the buyer has to offer to outbid others?

Does an employer just pay for work done, or possibly also for availability, potential, loyalty, goodwill?

Edit: That said, I absolutely agree that gender isn't something that should figure in those considerations. At best it's a poor proxy for much more specific differences between individuals.

I agree its not easy to quantify wage rates. It is highly context dependent. The valuation depends on many things. Still gender should not be the parameter since valuation depends on an individuals performance.

How do you remove biases from such a process though.

Given two employees, same education, same title, same working hours, same productivity (based on managers reports), nothing guarantee that the HR department won't "like" one more than another. Or one might ask for raises more often or with better arguments (having kids, longer commute, getting offers from other companies all the time).

Office politics are what they are, politics, as long as you have human in the process you'll have biases. Now if these biases are stronger against certain group it probably should be addressed.

In addition, the more fuzzy the nature of the work the more room for variations between remuneration. I suspect that the gender pay gap is lower workers in warehouses, factories or fast food restaurants.

The gender pay gap is large in restaurants -- women servers get bigger tips.

Conclusions that the companies hired more women because of it are a bit weird to draw since every company that would be a comparison in the country also has to follow the same rules?

And ofc it's solved by lowering the growth of male wages :D Business

No, they were comparing companies that didn't need to follow the regulation (companies with 24-34 employees) with companies that did (35+ employees).

Of course, the problem with this is twofold:

1. Companies with <35 employees is a tiny slice of the population and should generally not be expected to follow normal population distributions.

2. Companies with <35 employees are already behaving differently than larger companies with >35 employees. I do not think that variable was accounted for in any way by the study and that is problematic.

And, from skimming the paper, they have literally divided companies into two groups, <35 and >35. It could be that one giant employer (like a hospital system) is the whole story.

A more subtle study would be to look for a discontinuity: expect that firms with 10, 30, 100, 1000 employees will behave differently, but anything which happens sharply at 35 (and changes in this time) can be more confidently attributed to the law. But this is obviously harder to do.

Over 80% of all companies in the USA have fewer than 35 employees


i agree with your point #2

the differences between large and small companies are significant enough to outweigh the causal relationship that the OP implies

I think he was referring to the relative number of workers at companies with more/less than 35 employees. Clearly the average worker, by a large margin, works at companies with 50 or more employees. Even more so when you don't include whatever proportion of businesses counted that are essentially self-employment.

"The researchers compared those companies to firms with 25 to 34 employees that didn’t have to release their data"

So they are comparing two different types of firms (with > 35 employees and between 25 and 34 employees): Doesn't that invalidate their study right there?

They did not directly compare the two types of companies, they compared evolution of each type.

Of course that's not perfect, but it's good enough

Very logical. Well, it gets to 0 if the pay is fixed.

You don't have much companies using that in the West

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact