Thanks for this. Some more questions, if you have time:
Is it true, as Hanania claims, that they have to prove a test is nondiscriminatory for each race and site individually?
How easy is it to prove legitimate reason? If I say "I want my schoolteachers to do well on an IQ test, because schoolteachers should be smart" does that pass?
Why can't Sheetz say "We don't want people with histories of violent crime because we think they might be violent or criminal while working for us"?
Why was Duke Power Co decided the way it was, since they asked people to take a mechanical aptitude test for a mechanical job?
1) He may be right about that (I don't know actually) but even if he is right, so what? If a test is relevant to a job, that evidence will apply to each worksite. It's not like there's some affirmative requirement that employers prove the test works before they can implement it--they can do whatever they want and the only check is a lawsuit. A plaintiffs' attorney is not going to bring that case if it doesn't have some evidence the
2) Very easy. You just have to show there is a “manifest relationship to the employment in question" (a more lenient standard added by subsequent more conservative courts) then the burden shifts to the plaintiffs to prove its not legitimate or that the employer could achieve the same goal in a way that doesn't have a disparate impact. In Griggs, there was direct evidence from the employer's own experience that the test they were using was uncorrelated with job performance.
3) That is likely enough. But if, for example, their experience showed that people with a criminal history were no likelier to be violent and criminal than that argument would rightly fail. I think it is also unlikely the EEOC will win this case in the current legal environment.
4) As I said above, if you read the actual case, the facts were that the test did not predict success at the job. This turns out to be very common.
If a plaintiff wins they can (though don't alwasy) get attorneys' fees, which is not true of defendants. But that's still a big risk for an attorney to take -- getting fees can take years, courts often quibble about the amount, and if you lose you are out years of work withiut compensation. Also plaintiffs do have to pay costs (filing fees, deposition transcripts, and some other things) for defendants if they lose and that can be tens of thousands of dollars in aditional losses.
Costs are not that expensive and usually run to hundreds or low thousands. Fees are where it's at. The major plaintiffs' shops run it as a factory business and can easily keep their costs to a minimum if the claim isn't shaping up to be lucrative. Since over 95% of cases settle, the chances of having to fight about a lodestar rate with the court are slim.
My larger point is that Hannania is a deeply (and admittedly!) dishonest person and engaging with his description of factual questions is unwise--as you found when you looked into the facts of the Tesla case, which were awful. He doesn't want anti discrimination laws not because they are counterproductive or over broad but because he thinks racial discrimination is a good thing as his past history of white supremacist posting shows. (Yes he has "apologized" for that but I don't see any reason to believe him).
> Good afternoon. I'm Monique Lillard, from the University of Idaho. I am very pleased to inform you these proceedings will be published by the Employee Rights and Employment Policy Journal. I want to acknowledge that EEOC Commissioner Paul Miller is in the audience. [To Commissioner Miller] It's a pleasure to have you here again.
Indeed. That makes me very confident that the following research will be deeply honest and not at all biased in favor of EEOC and its mission statement.
It was you who suggested that if he was serious he would have engaged with 'the actual empirical literature' and gave that as an example. Would you have chided anti-smoking people in the 50s for refusing to seriously engage with the actual empirical literature paid for by the tobacco companies?
PS: Tu quoque is not an argument, and in any case I did not argue that he has no research bias.
Are there any reasons to oppose disparate impact policies or anti-discrimination laws or affirmative action other than a desire for racial discrimination, in your view?
There are critiques of affirmative action around the idea that it leads to stigma towards the groups who benefit that outweighs the benefits. Also that it causes employers to ignore socioeconomic diversity. I don't find those persuasive but they're coherent.
In terms of bans on disparate impact discrimination, you might just not believe that courts can figure out which hiring qualifications are relevant to predicting job success. But why you should believe employers can do so in that case I don't know.
I would turn the question around and ask what the social purpose is of allowing employers to hire using mechanisms that demonstrably do not relate to employees success and cause discrimination.
More generally, we don't have hypothesize about what employment would look like if we got rid of antidiscrimination law. We only passed those laws in the 1960s. It wasn't true then that the market eliminated discrimination!
> There are critiques of affirmative action around the idea that it leads to stigma towards the groups who benefit that outweighs the benefits.
Yup. First time I ever heard the concept criticized was in my late teens, from a latino friend whose argument was essentially, "if there's affirmative action in place and I get a job somewhere, everyone's going to look at me like I only got it because of the policy and not because I deserved it, and I don't ever want that to happen, therefore affirmative action needs to die." That conversation was quite the eye-opener for me!
It's in some ways a transfer from people who could get in regardless to those who couldn't. There is a question of whether the net benefit justifies that cost. I tend to think it does--and most folks in the affected groups agree--but it's a reasonable question.
I think your statement about disparate impact is slipping in a few assumptions that aren't obviously correct:
You said: "I would turn the question around and ask what the social purpose is of allowing employers to hire using mechanisms that demonstrably do not relate to employees success and cause discrimination."
a. There is a big and important difference between mechanisms that can't be shown in court to relate to employee success and mechanism that demonstrably do not relate to employee success. It's similar to the difference between "not guilty" and "proven innocent."
In fact, all else equal, we should expect employers to be a lot better at figuring out what makes a good employee than a court will be--the judge is an expert on the law, but not an expert on running an electric company or a convenience store chain or a software house or whatever. And the world is absolutely full of things that aren't legible or easy to prove, but are still important. In my field, I think I can look at a young researcher and have a pretty good intuition about whether he or she will be a star in 10 years, even compared to other people who look very similar on paper, and I don't think I am unique here. The court is only involved here because the employers may also be making these decisions for bad reasons that are forbidden by law, like not wanting black employees.
b. It seems like "and cause discrimination" is assuming your conclusion. Having a racial/gender proportion of employees that is not the same as the population is not the same thing as discrimination, at least not in the normal way the word is used. Indeed, the world is absolutely full of fields and workplaces where the race and gender ratio is not much like that of the population, even in places where it seems extremely unlikely to be due to any kind of discrimination. (For example, women are overrepresented among veterinarians and East Asians are overrepresented among academic computer scientists. It's hard to see how either of these would be due to anyone discriminating.)
Many thanks for your detailed and informative answers on all these questions!
One quibble:
>In terms of bans on disparate impact discrimination, you might just not believe that courts can figure out which hiring qualifications are relevant to predicting job success. But why you should believe employers can do so in that case I don't know.
Typically, one would expect employers to be domain experts in their area of business, while one would expect the courts to be domain experts in the law.
While I have never been a manager, I have been one of a team of department members evaluating job candidates, and I asked technical questions closely analogous to situations that arose in the code base we maintained and extended. (e.g. "What is the topology of a CMOS NAND gate? What does the transient waveform in a low-pass RC circuit fed a pulse look like? etc.)
Wow, those are some pretty high standards indeed. When I was in a similar scenario, I asked basic CS questions, like "you have a large list of Person objects, and you want to be able to efficiently look them up by the person's name. Every name is unique. How would you do it?" It was a bit depressing to see just how many applicants for a senior dev role could not come up with "hash table" / "dictionary" / "map," etc.
"allowing employers to hire using mechanisms that demonstrably do not relate to employees success"
The idea that courts can figure out a company's business model, hiring criteria, and employment practices better than the employer itself is one of the foundational philosophical divides between the sides.
Yes, but courts are put in that position all the time; the recent SC case about challenging the FDA approval of mifepristone is a good and timely example. Whether the issue should be left to those who “Know better“ was central to the argument.
"In terms of bans on disparate impact discrimination, you might just not believe that courts can figure out which hiring qualifications are relevant to predicting job success. But why you should believe employers can do so in that case I don't know."
3 reasons:
1. Employers have a financial incentive to figure this out. Court's don't.
2. Employers have more experience with their industry than courts do
3. All Employers have to do is figure out which are LIKELY to predict job success. Courts deal with proof.
The tldr is "likely" just means that the company believes it will predict job success. Because all a probability is, is a degree of belief. If I have a hunch that being smart will make someone a good teacher than the probability that being smart makes someone a good teacher is over 0.5 (at least my probability for this happening). Now I could be wrong, sure. Like maybe being smart is not correlated with being a good teacher. In a fair market companies will make reasonable, educated guesses and will be right more often than they are wrong.
But if being wrong comes with an expensive lawsuit then they will err on the side of assuming their tests doesn't work just in case someone can prove it doesn't work so that even tests that do work won't be used because maybe they dont.
> A plaintiffs' attorney is not going to bring that case if it doesn't have some evidence
Tell me you're completely unfamiliar with the concept of "the process is the punishment" without saying you're completely unfamiliar with the concept of "the process is the punishment"...
Sure there are nuisance settlements--I'm familiar with that concept. But my point was that the "you have to answer job site by job site" is irrelevant to that. Sure, there are some meritless cases, but there are lots of errors in the other direction too--meritorious cases that aren't proved because its too expensive and a lawyer won't take them. One-sided fee-shifting is meant as a means of addressing the imbalance in resources between employers and employees.
Re 3, I had a boss I was on good terms with a decade or so ago who definitely *thought* that he wasn't allowed to (explicitly) exclude people with a history of theft, and that the company would *also* risk being held liable for anything such a hypothetical thief did end up stealing. Sample size of one and all, upstate New York (so salt accordingly for different jurisdictions/local views on law/etc), but it's at least weak evidence of the chilling effect from the double-bind. (We didn't hire any thieves so far as I know, but I was never in a position to see background check results either.)
> 4) As I said above, if you read the actual case, the facts were that the test did not predict success at the job. This turns out to be very common.
This does not mean the test isn't a good test in the sense that it doesn't measure job performance. See how there is no correlation between a players height in the NBA and how well they perform. This is because if there was a correlation then selectors would be leaving money on the table and they could improve their selection for the coming year by increasing the weighting on height (compared to everything else), which would in turn reduce the amount of correlation. Rinse and repeat until there is no correlation left.
The test not predicting job performance could equivalently mean that Duke Power had a very well calibrated way to choose their employees where they were prefectly capturing the information from the apitutde test compared to all the other factors involved in hiring. Indeed the fact that this turns out to be very common suggests to me that this is going on here (and elsewhere).
Excellent. Call it Hartgood's law: a measure used for control purposes is worthless for observing statistical regularities in the phenomenon being measured in proportion to how optimal the control is.
Thanks for this. Some more questions, if you have time:
Is it true, as Hanania claims, that they have to prove a test is nondiscriminatory for each race and site individually?
How easy is it to prove legitimate reason? If I say "I want my schoolteachers to do well on an IQ test, because schoolteachers should be smart" does that pass?
Why can't Sheetz say "We don't want people with histories of violent crime because we think they might be violent or criminal while working for us"?
Why was Duke Power Co decided the way it was, since they asked people to take a mechanical aptitude test for a mechanical job?
1) He may be right about that (I don't know actually) but even if he is right, so what? If a test is relevant to a job, that evidence will apply to each worksite. It's not like there's some affirmative requirement that employers prove the test works before they can implement it--they can do whatever they want and the only check is a lawsuit. A plaintiffs' attorney is not going to bring that case if it doesn't have some evidence the
2) Very easy. You just have to show there is a “manifest relationship to the employment in question" (a more lenient standard added by subsequent more conservative courts) then the burden shifts to the plaintiffs to prove its not legitimate or that the employer could achieve the same goal in a way that doesn't have a disparate impact. In Griggs, there was direct evidence from the employer's own experience that the test they were using was uncorrelated with job performance.
3) That is likely enough. But if, for example, their experience showed that people with a criminal history were no likelier to be violent and criminal than that argument would rightly fail. I think it is also unlikely the EEOC will win this case in the current legal environment.
4) As I said above, if you read the actual case, the facts were that the test did not predict success at the job. This turns out to be very common.
Sorry, didn't finish 1) a plaintiffs' attorney isn't going to bring a case if they don't have some evidence of likelihood of success.
I recall Hanania saying such lawsuits are unusual in that court costs are still covered for losing plaintiffs, but not defendants.
If a plaintiff wins they can (though don't alwasy) get attorneys' fees, which is not true of defendants. But that's still a big risk for an attorney to take -- getting fees can take years, courts often quibble about the amount, and if you lose you are out years of work withiut compensation. Also plaintiffs do have to pay costs (filing fees, deposition transcripts, and some other things) for defendants if they lose and that can be tens of thousands of dollars in aditional losses.
Costs are not that expensive and usually run to hundreds or low thousands. Fees are where it's at. The major plaintiffs' shops run it as a factory business and can easily keep their costs to a minimum if the claim isn't shaping up to be lucrative. Since over 95% of cases settle, the chances of having to fight about a lodestar rate with the court are slim.
My larger point is that Hannania is a deeply (and admittedly!) dishonest person and engaging with his description of factual questions is unwise--as you found when you looked into the facts of the Tesla case, which were awful. He doesn't want anti discrimination laws not because they are counterproductive or over broad but because he thinks racial discrimination is a good thing as his past history of white supremacist posting shows. (Yes he has "apologized" for that but I don't see any reason to believe him).
If he was serious, he would engage with the actual emprical literature on the impact of discrimination cases instead of cherry-picking examples. See, e.g., https://scholarlycommons.law.wlu.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=1181&context=wlulr or igitalcommons.law.uidaho.edu/cgi/viewcontent.cgi?article=1405&context=faculty_scholarship. The literature is far from perfect but he has no evidence at all.
> Good afternoon. I'm Monique Lillard, from the University of Idaho. I am very pleased to inform you these proceedings will be published by the Employee Rights and Employment Policy Journal. I want to acknowledge that EEOC Commissioner Paul Miller is in the audience. [To Commissioner Miller] It's a pleasure to have you here again.
Indeed. That makes me very confident that the following research will be deeply honest and not at all biased in favor of EEOC and its mission statement.
Hanania clearly has no research bias.
It was you who suggested that if he was serious he would have engaged with 'the actual empirical literature' and gave that as an example. Would you have chided anti-smoking people in the 50s for refusing to seriously engage with the actual empirical literature paid for by the tobacco companies?
PS: Tu quoque is not an argument, and in any case I did not argue that he has no research bias.
Are there any reasons to oppose disparate impact policies or anti-discrimination laws or affirmative action other than a desire for racial discrimination, in your view?
There are critiques of affirmative action around the idea that it leads to stigma towards the groups who benefit that outweighs the benefits. Also that it causes employers to ignore socioeconomic diversity. I don't find those persuasive but they're coherent.
In terms of bans on disparate impact discrimination, you might just not believe that courts can figure out which hiring qualifications are relevant to predicting job success. But why you should believe employers can do so in that case I don't know.
I would turn the question around and ask what the social purpose is of allowing employers to hire using mechanisms that demonstrably do not relate to employees success and cause discrimination.
More generally, we don't have hypothesize about what employment would look like if we got rid of antidiscrimination law. We only passed those laws in the 1960s. It wasn't true then that the market eliminated discrimination!
> There are critiques of affirmative action around the idea that it leads to stigma towards the groups who benefit that outweighs the benefits.
Yup. First time I ever heard the concept criticized was in my late teens, from a latino friend whose argument was essentially, "if there's affirmative action in place and I get a job somewhere, everyone's going to look at me like I only got it because of the policy and not because I deserved it, and I don't ever want that to happen, therefore affirmative action needs to die." That conversation was quite the eye-opener for me!
It's in some ways a transfer from people who could get in regardless to those who couldn't. There is a question of whether the net benefit justifies that cost. I tend to think it does--and most folks in the affected groups agree--but it's a reasonable question.
I think your statement about disparate impact is slipping in a few assumptions that aren't obviously correct:
You said: "I would turn the question around and ask what the social purpose is of allowing employers to hire using mechanisms that demonstrably do not relate to employees success and cause discrimination."
a. There is a big and important difference between mechanisms that can't be shown in court to relate to employee success and mechanism that demonstrably do not relate to employee success. It's similar to the difference between "not guilty" and "proven innocent."
In fact, all else equal, we should expect employers to be a lot better at figuring out what makes a good employee than a court will be--the judge is an expert on the law, but not an expert on running an electric company or a convenience store chain or a software house or whatever. And the world is absolutely full of things that aren't legible or easy to prove, but are still important. In my field, I think I can look at a young researcher and have a pretty good intuition about whether he or she will be a star in 10 years, even compared to other people who look very similar on paper, and I don't think I am unique here. The court is only involved here because the employers may also be making these decisions for bad reasons that are forbidden by law, like not wanting black employees.
b. It seems like "and cause discrimination" is assuming your conclusion. Having a racial/gender proportion of employees that is not the same as the population is not the same thing as discrimination, at least not in the normal way the word is used. Indeed, the world is absolutely full of fields and workplaces where the race and gender ratio is not much like that of the population, even in places where it seems extremely unlikely to be due to any kind of discrimination. (For example, women are overrepresented among veterinarians and East Asians are overrepresented among academic computer scientists. It's hard to see how either of these would be due to anyone discriminating.)
Many thanks for your detailed and informative answers on all these questions!
One quibble:
>In terms of bans on disparate impact discrimination, you might just not believe that courts can figure out which hiring qualifications are relevant to predicting job success. But why you should believe employers can do so in that case I don't know.
Typically, one would expect employers to be domain experts in their area of business, while one would expect the courts to be domain experts in the law.
While I have never been a manager, I have been one of a team of department members evaluating job candidates, and I asked technical questions closely analogous to situations that arose in the code base we maintained and extended. (e.g. "What is the topology of a CMOS NAND gate? What does the transient waveform in a low-pass RC circuit fed a pulse look like? etc.)
Wow, those are some pretty high standards indeed. When I was in a similar scenario, I asked basic CS questions, like "you have a large list of Person objects, and you want to be able to efficiently look them up by the person's name. Every name is unique. How would you do it?" It was a bit depressing to see just how many applicants for a senior dev role could not come up with "hash table" / "dictionary" / "map," etc.
"allowing employers to hire using mechanisms that demonstrably do not relate to employees success"
The idea that courts can figure out a company's business model, hiring criteria, and employment practices better than the employer itself is one of the foundational philosophical divides between the sides.
Yes, but courts are put in that position all the time; the recent SC case about challenging the FDA approval of mifepristone is a good and timely example. Whether the issue should be left to those who “Know better“ was central to the argument.
"In terms of bans on disparate impact discrimination, you might just not believe that courts can figure out which hiring qualifications are relevant to predicting job success. But why you should believe employers can do so in that case I don't know."
3 reasons:
1. Employers have a financial incentive to figure this out. Court's don't.
2. Employers have more experience with their industry than courts do
3. All Employers have to do is figure out which are LIKELY to predict job success. Courts deal with proof.
What does the word "likely" mean?
See this article: https://www.astralcodexten.com/p/in-continued-defense-of-non-frequentist
The tldr is "likely" just means that the company believes it will predict job success. Because all a probability is, is a degree of belief. If I have a hunch that being smart will make someone a good teacher than the probability that being smart makes someone a good teacher is over 0.5 (at least my probability for this happening). Now I could be wrong, sure. Like maybe being smart is not correlated with being a good teacher. In a fair market companies will make reasonable, educated guesses and will be right more often than they are wrong.
But if being wrong comes with an expensive lawsuit then they will err on the side of assuming their tests doesn't work just in case someone can prove it doesn't work so that even tests that do work won't be used because maybe they dont.
> A plaintiffs' attorney is not going to bring that case if it doesn't have some evidence
Tell me you're completely unfamiliar with the concept of "the process is the punishment" without saying you're completely unfamiliar with the concept of "the process is the punishment"...
Sure there are nuisance settlements--I'm familiar with that concept. But my point was that the "you have to answer job site by job site" is irrelevant to that. Sure, there are some meritless cases, but there are lots of errors in the other direction too--meritorious cases that aren't proved because its too expensive and a lawyer won't take them. One-sided fee-shifting is meant as a means of addressing the imbalance in resources between employers and employees.
Re 3, I had a boss I was on good terms with a decade or so ago who definitely *thought* that he wasn't allowed to (explicitly) exclude people with a history of theft, and that the company would *also* risk being held liable for anything such a hypothetical thief did end up stealing. Sample size of one and all, upstate New York (so salt accordingly for different jurisdictions/local views on law/etc), but it's at least weak evidence of the chilling effect from the double-bind. (We didn't hire any thieves so far as I know, but I was never in a position to see background check results either.)
> 4) As I said above, if you read the actual case, the facts were that the test did not predict success at the job. This turns out to be very common.
This does not mean the test isn't a good test in the sense that it doesn't measure job performance. See how there is no correlation between a players height in the NBA and how well they perform. This is because if there was a correlation then selectors would be leaving money on the table and they could improve their selection for the coming year by increasing the weighting on height (compared to everything else), which would in turn reduce the amount of correlation. Rinse and repeat until there is no correlation left.
The test not predicting job performance could equivalently mean that Duke Power had a very well calibrated way to choose their employees where they were prefectly capturing the information from the apitutde test compared to all the other factors involved in hiring. Indeed the fact that this turns out to be very common suggests to me that this is going on here (and elsewhere).
Excellent. Call it Hartgood's law: a measure used for control purposes is worthless for observing statistical regularities in the phenomenon being measured in proportion to how optimal the control is.