'Disparate impact' is the idea that, to assess fairness, you treat individuals not as individuals, but as members of identity groups (e.g., grouped by ethnicity, sex, sexual orientation). Then, if any group doesn't get a proportion of some desirable outcome (e.g., employment at a particular company) matching its proportion in the relevant population, this is presumed to be illegal discrimination.
The burden of proof then falls on the decision-maker (e.g. employer), i.e. guilty until proven innocent.
This seems crazy when you consider processes with any randomness.
There's no reason why every single company, for every role, would naturally have demographic proportions matching the relevant population for every single identity group.
If you toss a coin ten times and get 6 heads and 4 tails, that doesn't mean the coin is biased.
The EO is right: disparate impact doctrine has forced employers to discriminate based on race, so they can maintain proportions.
You’re right in principle, but taking this out of the context it should be viewed in. The only reason such policies were created is because there was already bias against certain groups. So we can claim that disparate impact policies create another bias, but seems like we’re just addressing one source of bias, not the original offender. What is the solution to that?
Politics should not be handled in a stateless manner without understanding root cause. You pick an issue, with 10 imperfect solutions over 50 years and decide to remove only the things you don’t like, you will leave it in a terrible state that will need to be fixed by your replacements.
"You pick an issue, with 10 imperfect solutions over 50 years"
Are you suggesting that disparate impact is part of a delicate system that reduces or eliminates unfair discrimination? And that its elimination would make the situation worse than it is today?
If so, why?
I've given you many reasons why disparate impact is bad and has bad effects. You have given no reasons why, with the state of the country as it is today, it is helpful. Your argument seems to boil down to (i) well, there must have been some reason for its introduction, and (ii) status quo is a fragile system that must be changed all at once and not piecemeal.
I don’t think the example you applied is applicable, you’re using the randomness of a coin flip to a deterministic issue, biasses. Why would it be an issue if 10 homes were sold to a specific identity group and to give preference of the 11th home to a different identity group? I don’t understand why there’s an aversion to promote a heterogenous effect.
It's an issue because people are individuals and we should treat them fairly as individuals, not discriminate based on the colour of their skin or other irrelevant attributes, including membership in identity groups.
Should each NBA team be required to make sure the next player they hire has what you call a 'heterogenous effect' on the team composition with respect to identity groups?
Ok, great, we want equality of opportunity. How do we measure that? Disparate impact seems like a guide, if an imperfect one, but if you can prove to me that everyone has equal opportunity, I’ll accept that, except I think “equal opportunity” is often just shorthand for “we’ll do nothing”. Do I have an equal opportunity to the guy who’s dad’s name is on one of the buildings?
- It imposes liability for neutral policies based solely on statistical outcomes, ignoring discriminatory motive. Example: Griggs v. Duke Power invalidated a diploma requirement due to impact alone.
- It prioritizes group statistical parity over individual qualifications, potentially invalidating useful criteria. Again: Griggs invalidated tests because of their statistical impact.
- It incentivizes racial quotas: fear of liability drives companies (and college admissions offices) toward managing numbers by race, undermining the principle of equal treatment. Example: New Haven (Ricci) discarded firefighter exam results specifically to manage racial outcomes.
- Trying to avoid disparate impact liability can itself constitute intentional discrimination against another group. (Ricci paradox, same case as above.)
- It enables lawsuits against essential, neutral practices based purely on statistical correlations. Example: Challenges to voter ID laws based on disparate ID possession rates. Challenges to 'rat control' regulations based on disparate impact on rents.
- The "business necessity" defense is often hard to prove and subject to judicial second-guessing. Example: Duke Power failed to prove its hiring requirements were essential.
- It conflicts with the Equal Protection Clause's requirement of discriminatory intent for government liability.
- It creates the potential to challenge almost any neutral policy with unequal outcomes. Example: suspensions and other school discipline must be meted out 'fairly' between different races.
Another restricts the use of the so-called disparate impact rule, which civil rights groups have long said is an important tool for showing discrimination against minorities.
If your whole raison d'etre is to 'show discrimination' then obviously you may like a rule that allows you to 'show discrimination' even where none exists.
'Disparate impact' is the idea that, to assess fairness, you treat individuals not as individuals, but as members of identity groups (e.g., grouped by ethnicity, sex, sexual orientation). Then, if any group doesn't get a proportion of some desirable outcome (e.g., employment at a particular company) matching its proportion in the relevant population, this is presumed to be illegal discrimination.
The burden of proof then falls on the decision-maker (e.g. employer), i.e. guilty until proven innocent.
This seems crazy when you consider processes with any randomness.
There's no reason why every single company, for every role, would naturally have demographic proportions matching the relevant population for every single identity group.
If you toss a coin ten times and get 6 heads and 4 tails, that doesn't mean the coin is biased.
The EO is right: disparate impact doctrine has forced employers to discriminate based on race, so they can maintain proportions.
You’re right in principle, but taking this out of the context it should be viewed in. The only reason such policies were created is because there was already bias against certain groups. So we can claim that disparate impact policies create another bias, but seems like we’re just addressing one source of bias, not the original offender. What is the solution to that?
Politics should not be handled in a stateless manner without understanding root cause. You pick an issue, with 10 imperfect solutions over 50 years and decide to remove only the things you don’t like, you will leave it in a terrible state that will need to be fixed by your replacements.
"You pick an issue, with 10 imperfect solutions over 50 years"
Are you suggesting that disparate impact is part of a delicate system that reduces or eliminates unfair discrimination? And that its elimination would make the situation worse than it is today?
If so, why?
I've given you many reasons why disparate impact is bad and has bad effects. You have given no reasons why, with the state of the country as it is today, it is helpful. Your argument seems to boil down to (i) well, there must have been some reason for its introduction, and (ii) status quo is a fragile system that must be changed all at once and not piecemeal.
I don’t think the example you applied is applicable, you’re using the randomness of a coin flip to a deterministic issue, biasses. Why would it be an issue if 10 homes were sold to a specific identity group and to give preference of the 11th home to a different identity group? I don’t understand why there’s an aversion to promote a heterogenous effect.
It's an issue because people are individuals and we should treat them fairly as individuals, not discriminate based on the colour of their skin or other irrelevant attributes, including membership in identity groups.
Should each NBA team be required to make sure the next player they hire has what you call a 'heterogenous effect' on the team composition with respect to identity groups?
Ok, great, we want equality of opportunity. How do we measure that? Disparate impact seems like a guide, if an imperfect one, but if you can prove to me that everyone has equal opportunity, I’ll accept that, except I think “equal opportunity” is often just shorthand for “we’ll do nothing”. Do I have an equal opportunity to the guy who’s dad’s name is on one of the buildings?
Disparate impact is flawed:
- It imposes liability for neutral policies based solely on statistical outcomes, ignoring discriminatory motive. Example: Griggs v. Duke Power invalidated a diploma requirement due to impact alone.
- It prioritizes group statistical parity over individual qualifications, potentially invalidating useful criteria. Again: Griggs invalidated tests because of their statistical impact.
- It incentivizes racial quotas: fear of liability drives companies (and college admissions offices) toward managing numbers by race, undermining the principle of equal treatment. Example: New Haven (Ricci) discarded firefighter exam results specifically to manage racial outcomes.
- Trying to avoid disparate impact liability can itself constitute intentional discrimination against another group. (Ricci paradox, same case as above.)
- It enables lawsuits against essential, neutral practices based purely on statistical correlations. Example: Challenges to voter ID laws based on disparate ID possession rates. Challenges to 'rat control' regulations based on disparate impact on rents.
- The "business necessity" defense is often hard to prove and subject to judicial second-guessing. Example: Duke Power failed to prove its hiring requirements were essential.
- It conflicts with the Equal Protection Clause's requirement of discriminatory intent for government liability.
- It creates the potential to challenge almost any neutral policy with unequal outcomes. Example: suspensions and other school discipline must be meted out 'fairly' between different races.
If you're trying to understand the context of this and other EOs, see here:
https://www.nytimes.com/2025/04/23/us/politics/trump-executi...
That article is pretty one-sided:
If your whole raison d'etre is to 'show discrimination' then obviously you may like a rule that allows you to 'show discrimination' even where none exists.