3 ms·

You're listing off a bunch of factors that would be very hard to measure exactly and then concluding they don't add up to 100%. I think it's much more likely t

You're listing off a bunch of factors that would be very hard to measure exactly and then concluding they don't add up to 100%. I think it's much more likely they don't add up to perfect measurement. It's standard to consider the null hypothesis as the answer if it lies within the margin of error of what you're measuring. Last I knew comparing exact jobs (not merely professions), actual hours worked (no "full time" category) and years of experience (not age) explained more than 90% of the gap. Are all these other causes you mention measured accurately enough to ensure the null isn't within the error range?

Spivak 5y ago
Other than the tone you have it exactly right. There is, depending on the study, a 3-10% pay yet unexplained pay gap — it always shows up. If it was actually 0% +/- error you would expect to see women as a group making more sometimes once all the explained parts are controlled for. However, people take “explained” to mean “it’s fine, we don’t have to do anything” when instead it’s actually the stuff we know we should be targeting. * Men on average work higher paying jobs, male dominated fields are on average higher paying. * Men on average hold higher titled positions in a given field, the glass ceiling is a thing, and women have a harder time becoming managers and directors. * Women have fewer years of of experience on average because of childcare. This is all stuff we need to work on because “controlling” for all this stuff and then forgetting about it makes it seem like it’s totally fine that 99 men and 1 woman are senior SWEs as long as the 1 woman makes the same pay.
- s1artibartfast 5y ago
  I would question if we still need to work on some of these topics. Particularly #1 and #3 that you bulleted, and your example. Sometimes you can find the explanation and it is "totally fine", if your goal is to eliminate gender discrimination. This is only a problem if your goal is equality of outcome, independent of preference. >Women have fewer years of of experience on average because of childcare. It has been long shown that women desire to raise children more than men. Fixing this difference would require making both men and women do something other than what they want. For what purpose, so that the numbers are symmetrical? >Men on average work higher paying jobs, male dominated fields are on average higher paying. As long as these fields are open and non-discriminatory to women, there is no problem! Some fields do have discrimination issues, and we should focus on that, not try to fight the market demand for occupations. >all this stuff and then forgetting about it makes it seem like it’s totally fine that 99 men and 1 woman are senior SWEs as long as the 1 woman makes the same pay. If women don't want to be SWEs, this isn't a problem. Last I checked, there was a massive imbalance in computer science application and graduation rates. Last I checked, SW firms were tripping over themselves to hire and promote women. At some point, you have to question the premise that numbers would be equal if you removed unequal barriers and discrimination. At some point, if your sole metric is equal numbers, you have to force men and women to do jobs that they don't want to.
  - Spivak 5y ago
    It's weird reading a post that I aggressively agree with except for the conclusion. You have it exactly right in the cases where we should ignore pay gaps due to preference differences and what we should do about it. Because you're right, it would be silly to try and force the numbers to be equal, even if you could it wouldn't fix the more fundamental issues that caused the numbers to differ by such a wide margin in the first place. > At some point, you have to question the premise that numbers would be equal if you removed unequal barriers and discrimination. At some point, if your sole metric is equal numbers, you have to force men and women to do jobs that they don't want to. I feel like we're 99% in agreement, the difference is that I see a 20-30% difference among groups that have more variance within them than between them as evidence of those barriers and discrimination and a system that continuously reinforces the same outcomes. And it goes both ways, the barriers that push men out nursing, child care, and art are just as important, the difference is what data you use to demonstrate it.
    - s1artibartfast 5y ago
      Sounds like we are mostly in agreement so I would add two parting thought around differences. #First, the simple numeric argument: Small innate differences in mean preference can have large effects at the margin for normal distributions. If you take two identical distributions, and move the mean of one 20-30%, the number that falls above or below a threshold changes a lot more than 20-30%. For example, decrease the mean of one curve a single standard deviation (~35%), and the tail drops a lot more. A value that is in the top 2% of the first curve is now the top 0.1% of the second curve. That is a 2000% difference in samples above the value. #The second is the feedback-loop argument: People have preferences, and lets assume the innate preferences between genders is small. Over time, positive feedback loops can amplify these differences. This doesn't have to be sinister, but can be as simple as becoming interested in what are exposed to via same gendered friends and what your social network does. You see it often in certain communities or families that specialize in a given trade (eg, my uncle was a carpenter, or my aunt was a nurse). The only way to stop these influences is to raise people in a bubble or gender randomize their friends and role models. What we DO want to avoid negative feedback loops where people are actively pushed away from a trade. (e.g. inhospitable workplaces for male nurses or female carpenters)
- godelski 5y ago
  > Men on average work Just to be clear, the above discussion is not about median wages between genders, it is about median wages between genders for similar job experience and job titles. The former is a 17% difference in wages and is related to men and women choosing different types of jobs that have differences in pays while the latter is a discussion about differences in pay for similar work. There are different decisions that women make in a job, but that doesn't explain everything. Hence the "no smoking gun" argument that keeps being said.
godelski 5y ago
> You're listing off a bunch of factors *that would be very hard to measure *exactly and then concluding they don't add up to 100% Which is why it is unexplained. i.e. We don't have strong confidence in the causal structure. No one is really disagreeing with this. But to create causal graphs you have to have... causes. If you can't measure something well then you can't form that graph well.

Spivak 5y ago

Other than the tone you have it exactly right. There is, depending on the study, a 3-10% pay yet unexplained pay gap — it always shows up. If it was actually 0% +/- error you would expect to see women as a group making more sometimes once all the explained parts are controlled for. However, people take “explained” to mean “it’s fine, we don’t have to do anything” when instead it’s actually the stuff we know we should be targeting. * Men on average work higher paying jobs, male dominated fields are on average higher paying. * Men on average hold higher titled positions in a given field, the glass ceiling is a thing, and women have a harder time becoming managers and directors. * Women have fewer years of of experience on average because of childcare. This is all stuff we need to work on because “controlling” for all this stuff and then forgetting about it makes it seem like it’s totally fine that 99 men and 1 woman are senior SWEs as long as the 1 woman makes the same pay.

s1artibartfast 5y ago
I would question if we still need to work on some of these topics. Particularly #1 and #3 that you bulleted, and your example. Sometimes you can find the explanation and it is "totally fine", if your goal is to eliminate gender discrimination. This is only a problem if your goal is equality of outcome, independent of preference. >Women have fewer years of of experience on average because of childcare. It has been long shown that women desire to raise children more than men. Fixing this difference would require making both men and women do something other than what they want. For what purpose, so that the numbers are symmetrical? >Men on average work higher paying jobs, male dominated fields are on average higher paying. As long as these fields are open and non-discriminatory to women, there is no problem! Some fields do have discrimination issues, and we should focus on that, not try to fight the market demand for occupations. >all this stuff and then forgetting about it makes it seem like it’s totally fine that 99 men and 1 woman are senior SWEs as long as the 1 woman makes the same pay. If women don't want to be SWEs, this isn't a problem. Last I checked, there was a massive imbalance in computer science application and graduation rates. Last I checked, SW firms were tripping over themselves to hire and promote women. At some point, you have to question the premise that numbers would be equal if you removed unequal barriers and discrimination. At some point, if your sole metric is equal numbers, you have to force men and women to do jobs that they don't want to.
- Spivak 5y ago
  It's weird reading a post that I aggressively agree with except for the conclusion. You have it exactly right in the cases where we should ignore pay gaps due to preference differences and what we should do about it. Because you're right, it would be silly to try and force the numbers to be equal, even if you could it wouldn't fix the more fundamental issues that caused the numbers to differ by such a wide margin in the first place. > At some point, you have to question the premise that numbers would be equal if you removed unequal barriers and discrimination. At some point, if your sole metric is equal numbers, you have to force men and women to do jobs that they don't want to. I feel like we're 99% in agreement, the difference is that I see a 20-30% difference among groups that have more variance within them than between them as evidence of those barriers and discrimination and a system that continuously reinforces the same outcomes. And it goes both ways, the barriers that push men out nursing, child care, and art are just as important, the difference is what data you use to demonstrate it.
  - s1artibartfast 5y ago
    Sounds like we are mostly in agreement so I would add two parting thought around differences. #First, the simple numeric argument: Small innate differences in mean preference can have large effects at the margin for normal distributions. If you take two identical distributions, and move the mean of one 20-30%, the number that falls above or below a threshold changes a lot more than 20-30%. For example, decrease the mean of one curve a single standard deviation (~35%), and the tail drops a lot more. A value that is in the top 2% of the first curve is now the top 0.1% of the second curve. That is a 2000% difference in samples above the value. #The second is the feedback-loop argument: People have preferences, and lets assume the innate preferences between genders is small. Over time, positive feedback loops can amplify these differences. This doesn't have to be sinister, but can be as simple as becoming interested in what are exposed to via same gendered friends and what your social network does. You see it often in certain communities or families that specialize in a given trade (eg, my uncle was a carpenter, or my aunt was a nurse). The only way to stop these influences is to raise people in a bubble or gender randomize their friends and role models. What we DO want to avoid negative feedback loops where people are actively pushed away from a trade. (e.g. inhospitable workplaces for male nurses or female carpenters)
godelski 5y ago
> Men on average work Just to be clear, the above discussion is not about median wages between genders, it is about median wages between genders for similar job experience and job titles. The former is a 17% difference in wages and is related to men and women choosing different types of jobs that have differences in pays while the latter is a discussion about differences in pay for similar work. There are different decisions that women make in a job, but that doesn't explain everything. Hence the "no smoking gun" argument that keeps being said.

godelski 5y ago

> You're listing off a bunch of factors *that would be very hard to measure *exactly and then concluding they don't add up to 100% Which is why it is unexplained. i.e. We don't have strong confidence in the causal structure. No one is really disagreeing with this. But to create causal graphs you have to have... causes. If you can't measure something well then you can't form that graph well.