The new antiracism is not, as its etymology suggests, opposition to racial discrimination. Ibram X. Kendi demonstrates this in his 2019 bestseller, “How to Be an Antiracist.” He defines “racism” as a combination of policies and ideas that “produces and normalizes racial inequities.” This racism has nothing to do with individual discrimination. Rather, it is support for institutions that yield disparities. Lest there be confusion, Mr. Kendi emphasizes that “focusing on ‘racial discrimination’ takes our eyes off” the policy goals he and other self-proclaimed antiracists support.
How urgent is it that we alter institutions in order to remove racial disparities?
Institutions are rules and practices, such as laws against using drugs or the practice of requiring SAT tests for college admissions. Racial disparities are outcomes that are on average worse for African-Americans, such as under-representation among the very wealthy or over-representation in the prison population.
Consider two extreme views:
a) We should get rid of any institution that might cause such disparities until the disparities disappear.
b) Unless an institution explicitly uses race or skin color as a criterion for discriminating against African-Americans, that institution should be preserved, assuming that it serves a good purpose.
My guess is that a lot of people nowadays would position themselves somewhere between (a) and (b). I would hold up my hand for (b), with no in-between.
For example, consider the use of credit scoring to screen loan applicants. I believe that credit scoring is non-discriminatory with disparate impact. That is, a black borrower and a white borrower who each have a credit score of 650 will have the same probability of defaulting on the loan. But if you set 650 as a cutoff for approving loans, then the proportion of loan approvals that are for blacks will be less than their share in the population.
Someone committed to (a) would want to remedy the disparate impact. Either explicitly or implicitly, they would lower the cutoff for black applicants until they receive a proportionate share of loans. As someone committed to (b), I would advocate using the same cutoff for blacks as for whites.
Some further thoughts:
1. This would get me accused by the religion that persecutes heretics of being a white supremacist. But by most people’s standards, I am not.
2. There are plenty of institutions that might not qualify as “assuming that it serves a good purpose.” For example, drug laws may not serve a good purpose, but not because of any disparate racial impact. I am inclined to get rid of the college admissions process and replace it with open admissions rationed by a lottery system. Again, that is not because of any disparate racial impact. [UPDATE: See the essay on college admissions by Jeffrey Selingo. It appears to me that the main purpose of college admissions office is to perpetuate itself.]
Marcus concludes,
To defeat racism, we must turn away from the new antiracism.
I could never really fathom variant a. Every institution or process will lead to some disparities and the causes can be quite variable. We would open a Pandora’s box of what is considered an unacceptable disparity and we would have to change our institution with every new disparity. I guess many people just *feel* that this is a good solution without thinking how to actually implement it.
If you believe that all groups are inherently equal, it is easy to see any disparities as the result of injustice, either now or in the past. Then, it may seem to be a matter of simple justice to “make up” for that injustice by privileging members of the group that is not doing as well.
There is, of course, a logically simple way of “fixing” the unjust inequality: setting quotas on results.
The highest earning demographic group in the United States, last I checked, is Indian American. What injustices did Indian Americans commit against everyone else?
This point of the absurd logical implication of absolutist egalitarianism was explored by Vonnegut in his “Harrison Bergeron” short story. However, that criticism is based in the idea of “abstract principles applying equally to all”, which is very admirable and liberal, but not actually how it is going to work.
The key mechanism to keep in mind in our democratic system is the feedback between ideology and power, when optimizing for electoral efficiency (i.e., minimizing what it takes to get past the important posts).
The left’s political formula is inherently clientalist. It is the promise (tacit and explicit) to use the coalition’s power and influence (state and non-governmental) to give to those who envy by taking from the enviable. A traditionally vurtuous statesman is interested in creating as much solidarity, cohesion, fraternity, and harmony as possible in the governed population. But in a democracy there is much hay to be made by defecting from this goal and whipping up as much resentment, bitterness, envy, and tension as possible. This is like running Soviet “Active Measures” on one’s own countrymen, which is what gives sense to the combination “cold civil war”.
If we were going to prohibit some category of speech, this kind of subversive agitation propaganda should have been the top priority. Instead of being suppressed, even in a normative manner, it is perversely glorified, precisely because of its political utility. This is our major Social Failure Mode, leading to another “national nervous breakdown”, and the bridge is now wobbling so wildly that small pieces are starting to fly off.
Now, it’s also important to keep in mind that it’s not just about money – it’s the whole complicated function that goes into the core emotional feelings driving all of this – of social rank, esteem, prestige, and status. The left already redistributes tremendous amounts of money and material goods and services, but that is not enough. Once people thought it might be enough, “Old Left”, but the problem is that *nothing is ever enough*, because status is zero sum, and you are always going to have losers who feel bad about being losers, who can be quite easily provoked into feeling even worse and violently enraged about it especially if there is a designated bad guy to blame for it, and who suffer real social disability from being losers. As regards social strength, health, and stability, this is a big. But for that political formula, it is a feature, because the formula will *always* work, which is precisely why it should be politically incorrect if anything is politically incorrect, and suppressed if anything is going to be suppressed, and it was in fact suppressed in most places for most human history. A society can build up social capital like it can build it physical capital of, say, a large and beautiful wooden bridge. The person who wants to profit from running a ferry boat business will prosper if they can burn the bridge down and get away with it. Heck, even someone who is just a little chilly can get the tiny benefit from the temporary heat for the huge cost of torching a bridge. Whipping up tensions is social-arson, which is in many ways much worse than actual arson, because it is easier to rebuild a bridge than to make bitter rivals love each other.
What keeps the lid to Pandora’s Box partially closed (but also partially open for intentional but selective demon release) is the constraint of optimizing for efficiency. People talk about “single-issue voters”, but what’s *really* valuable are “zero-issue voters”, who are going to vote for your side no matter what, with just the absolute minimum investment in campaigning time, effort, and money to motivate them to vote.
I think it’s possible to give a more precise technical definition to “polarizaion”, a “Handle Polarization Index”, by counting up the proportions of 0-issue voters, 1-issue voters, etc. If you have a lot of >5-issue voters, the overall distribution will be normal and narrow, with a lot of people bunching near the mean by virtue of the central limit theorem. If you have a lot of 0s, then you are going to have a lot of dispersion and even a bimodal, bifurcated, polarized distribution.
The point is that the left doesn’t need everybody, just enough for the right majorities, and it wants the cheapest possible set. That means that the membership of their coalition is fluid and dynamic, especially given rapid demographic change, and they are always looking to throw the highest- maintenence allies under the bus, since they have the lowest marginal rate of transformation from costly burdens – especially internal rivalry and contradiction – into power.
The major practical consequence is that there need only be a few salient identity group characteristics for which any disparities must be leveled. Race, and sex-based disparities are probably enough, with some nods to generic wealth-class disparity anxiety to get it over the top. But all the other kinds of ‘x-privilege’ and ‘y-shaming’ are not strictly necessary except to heighten the symbolic contrast between teams, and so don’t have to be taken more seriously after any legal distinctions are erased and prohibited.
Kendi’s definition of racism describes a condition – outcome inequality – that must always be true of any institution. Because people can be categorized by countless traits, one can always identify inequalities between identifiable groups – if for no other reason than “identifiability” implies differences or “inequalities.”
Kendi is demanding the impossible – that we devise a system for which his definition of racism, which by definition is always true, is not true.
“Kendi is demanding the impossible”
I’m thinking that this is basically the whole point – Kendi and his disciples in the inclusion and diversity industry get to decide when, where and how much “racism” exists so that they can decide unanimously on the appropriate remedies.
(They are just hoping that you don’t see the contradictions, so please stop pointing them out :))
Lastly, I’m not statistics savvy, but isn’t most of this disparate impact stuff just a causation/correlation fallacy?
It really isn’t impossible at all. It just means racial quotas for everything where blacks are at a lower than representative proportion (i.e., not sports), and including income and wealth in the form of redistribution. Many other countries deal with tensions and jealousies between rival sub-populations by simply giving certain portions of things exclusively to members of certain groups. This is especially common for multi-ethnic, mutli-cultural amalgamation countries, but of course every country is multi-gendered. Often times this is purely for state-related positions, representation in power-sharing arrangements (e.g., number of seats in a legislature or executive council), university admissions, or government military and civil service positions, and I would recommend reading up on how Lebanon does it, or affirmative action for Malays in Malaysia, or especially the quota system in Pakistan, keeping in mind that while by coincidence meaning “pure”, the “PAK” is actually an acronym for Muslims living in regions “Punjab Afghan Kashmir”
Now, in Pakistan, 50% of civil service spots go to Punjabis, 19% to Sindhs, 11.5% to Kybers, 6% to Balochis, 3% to Tribals, 2% to Kashmiris, 1% to Baltistanis, and only 7.5 to “Merit”. That is, it is explicitly a mutli-ethnocracy, not a meritocracy, or only marginally so. Also, at least 10% from the share of each must be women.
And why stick to only government matters? One would expect plenty of expansion into areas of the market where there are clearly market-dominant identities and thus potential sources of tension and desirable resources which one can grab and pay to one’s clients.
As an example, consider the recent trend of laws requiring a gender quota – often a minimum of one or a proportion of 30-50% – for women on corporate boards. It’s just another step to require such things at every organization, at every level. Individualistic ideas like merit, worthiness, and identity-blind fairness are not the priority. “Deserve’s got nothing to do with it.” Which is why “Social Justice” is not “Actual Justice”.
So what Kendi is proposing is not only possible, it is the actual de jure or de facto policy in many other places and contexts (even in the US, even when it is technically against the law), and the only reason it is not already the case in the US is becaues of the temporary stumbling block of a dwindling opposition and some legal constraints that will inevitably be overcome, one way or another.
Kendi is not proposing the impossible, he is showing us the future. Boots and human faces and everything. Have a nice day.
“It really isn’t impossible at all. It just means racial quotas for everything…”
The key word is “everything”: Positions up and down every institution’s (corporations, universities, hospitals, governmental bodies, churches, chess clubs) hierarchy, equal representation in every type of employment from janitor to brain surgeon, ownership of luxury goods from BMWs to beach houses, investments, level of education in every college major, geographic location, demographics (marriage rate, number of children, average and median age), etc. Equality across all possible points of difference isn’t even theoretically possible.
First, there is a long, long way to go in the direction of the possible before you have to worry about the impossible, about as much ruin there is in a nation.
But second, a problem you can solve is not a politically useful problem, because someone might do it, probably your opponent who doesn’t want it hanging over their head any more. But an unsolvable problem is better; it never stops being useful.
That’s the thing about the politics of status envy. Once you start down that road, there is no end. The only way to win the game is not to play. But we started playing, and so there is go going back.
Sometimes people that are treated unfairly long enough and can’t do anything about it turn to make increasingly strained arguments and ask the government for protection, even though it probably isn’t going to produce a very healthy solution.
Just like you do with cancellations.
The only problem with this, at least here in the U.S., is that the aggrieved have basically no evidence to support the supposed unfair treatment that they have experienced. Or, at the least, they aren’t willing to share the empirical data.
And, there are so many counter examples in the U.S. to such accusations of unfair treatment (gosh darn those Asians, Indians, West Indians, Sri Lankans, Filipinos, Pakistanis, Nigerians, etc.).
That’s precisely you are seeing so much push back and disbelief.
(And, intentionally ignoring one ethnic minority that will remain unnamed because, for some odd and unjustifiable reason, they tend to drive the aggrieved class into fits of overt racism and conspiracy theories)
https://www.npr.org/sections/live-updates-protests-for-racial-justice/2020/07/15/891422809/fired-over-anti-semitic-comments-nick-cannon-wants-wild-n-out-ownership
My position is, you can’t get rid of racism by constantly obsessing about race.
We’re not (primarily) talking about racism here. We’re talking about “systemic racism” (= racial disparities).
I think b) is a good ideal, but it does imply some kind of algorithm transparency. For instance, I would not be surprised if someone could create a machine learning algorithm that does a pretty good job at predicting someone’s race from their name and interests.
I’d support b) if you got rid of the word “explicitly.” It is possible for an institution to choose formally neutral rules for racist purposes – to deliberately seek a disparate impact. This was done under Jim Crow (e.g. literacy tests) and affirmative action (e.g. the Texas 10% rule).
I think this would be a big objection to b: that nominally neutral standards were selected to marginalize minorities. The SAT is race neutral, but many contend that the questions are (subconsciously if not deliberately) selected so that black students will do more poorly. As you note with voting literacy tests, this has been done before. An issue with this point though is it’s probably pretty much impossible to disprove a claim that a standard was selected for non-racist reasons.
The problem with the claim that the SAT test is biased against Black students is that the test scores track very well with how well students do in college. In other words, the test does what it’s supposed to do.
The response would be that the measures that determine success outside of college are also designed to marginalize black people. This all of course largely ends up being rather circular, where we can only believe a measure of success that is truly race-neutral is if it doesn’t correlate with race.
If that’s true, then you have a huge profit-making opportunity. All you need to do is hire black people who are not being paid what their worth and you’ll make a fortune.
Elegant in principle but demonstrates lack of two things: 1) Empathy for a large groups of people 2) A sufficient level of understanding of how institutions perpetuate discrimination
Your take is also elegant in principle, but also demonstrates a lack of two things:
1) empathy for individuals (as part from artificial group distinctions).
2) lack of any actual evidence of how institutions are perpetuating racism
there might be courses of action between a and b. affirmative action is one (instead of getting rid of the institution entirely). in the case of loan, the government could subsidize more loans for black borrowers a certain percentage below the 650 cutoff.
the idea being to try to equalize an outcome when institutions and history and culture up to the point of a loan are not equal. when opportunity for individuals are not equal.
is there a point between a and b you would be comfortable with? because if we are past the point of racism on individual level, we still have racism baked into the American way and how do we deal with that?
“we still have racism baked into the American way”
I’m not sure that’s very important. “Male superiority” is also “baked into the American way”. For the longest time, women couldn’t own property and were largely kept either at home or in certain fairly dead-end jobs. Today, women are all over, and getting the same pay for the same job. Women make up 60% of college students, and about half of law school and medical school students.
What if the problem isn’t “racism baked into the American way” but blacks not taking advantage of the opportunities available? I know it’s terribly rude to ask such a question.
Paying the Danegeld doesn’t get rid of the Dane…
What’s the limiting principle on affirmative action? Why stop at a particularly point between A and B. Won’t B+1, B+2, etc always be in someones interest? Won’t there always be people agitating to move further down the road from B to A.
I think you have in mind an idea whose time has passed. Namely, a sort of decades ago consensus that blacks, only 13% of an overwhelmingly white population, could be given some token giveaways but otherwise not disturb the running of society.
But of course if blacks deserve B+1, why not B+2. And if blacks deserve it, why not Mexicans. Why shouldn’t people from India get special minority loan programs? Why shouldn’t women or gays or trannies get their own giveaways?
Let’s get blunt about it, in the past the people doing the looting would always have wanted more, but the overwhelming demographic might of middle class bourgeois whites (especially men and their wives) kept things from progressing from B+1 to B+2 all the way to A.
As they lose their demographic power, as their children are indoctrinated to hate themselves, why shouldn’t things go further than the good old days? Whose going to stop them.
Handle is right to point out the state of affairs in a place like Malaysia, which I’m familiar with. All multi-cultural/multi-ethnic societies end up like that.
Why would we assume problem with pursuing option (a) (eliminating institutions that are alleged to cause racial disparaties) will improve racial disparaties?
Firing all the police officers in Seattle could improve the racial disparity in use of police force, while vastly increasing the racial disparity in impact of crime. So, does policing stay or go?
Because black lives matter, we should be against Black Lives Matter.
At some point, we can simply eliminate all crimes, so we won’t have any more criminals
https://news.gallup.com/poll/316571/black-americans-police-retain-local-presence.aspx
81% of black Americans want a greater or equal police presence in their neighborhoods as in the status quo. So it should probably stay. Black Americans are also much less likely to believe the police will treat them with courtesy and respect. “Stay or go” is a false dichotomy. There is no reason policing in black neighborhoods inherently requires protection of pointless hostility and racist officers, it’s not a choice between “defunding” and “no reform at all.”
“There is no reason policing in black neighborhoods inherently requires protection of pointless hostility and racist officers”
Maybe not POINTLESSLY hostile and racist.
But what about REASONABLY hostile and racist. I don’t think there is a lot of pointless hostility, but perhaps lots of rational wariness based on lived experience and reasonable evidence based assumptions.
Nobody has found a way to get around the tradeoff between getting harsh on high crime minority communities and keeping crime under control. Nobody. Anyone that could would become the hero of the whole world, yet nobody has found the $100 bill on the sidewalk.
Instead, we have a Null Hypothesis in policing. Nearly all common sense police reforms and methods have already been implemented in most places. It would be better if people would simply accept that we aren’t going to get any better outcomes by going further down the Laffer curve of police restraint.
Turns out, as Micheal Bloomberg pointed out, you need to have officers pat down lots of brown youths and get their guns combined with lots of broken windows policing. There just isn’t a better way. Whenever people think there is a better way we get the BLM 2015 Ferguson effect or the BLM 2020 murder spree.
> Nearly all common sense police reforms and methods have already been implemented in most places.
This is complete nonsense. There is no national police decertification database — officers that are fired with cause in one place are hired in other precincts. It is still onerously difficult to fire bad officers in the first place. Civil asset forfeiture is still a massive conflict of interest and expensive to contest.
https://www.baltimoresun.com/news/crime/bs-md-ci-fop-response-20200812-3fqa5nsr7raaxdcenmpf6fqqju-story.html
Richard Pinheiro was found guilty of the crime of fabricating evidence but cannot be fired by the department because the internal investigation took too long. Is it not a “common sense” reform that an officer found guilty of fabricating evidence should be fired?
In DC, half of officers rehired due to arbitration had none of the facts of their firing contested, but won solely based on failure to adhere to deadlines in their investigations. There’s a great breakdown of this issue here
https://www.washingtonpost.com/graphics/2017/investigations/police-fired-rehired/
including the hilarious case of an officer who was caught on video telling a handcuffed and compliant man, who was arrested seemingly for pointing out that the officer had missed something, to fight him for his freedom. This officer, after previously having been fired and rehired, was again fired and again rehired.
It’s also still ordinary to retaliate against internal whistleblowers. Prohibiting retaliation against whistleblowers seems like a common sense reform.
https://www.buzzfeednews.com/article/albertsamaha/breaking-baltimores-blue-wall-of-silence
Officers are still rewarded for putting up high numbers of arrests and tickets
https://theappeal.org/nypd-commanders-text-messages-show-how-the-quota-system-persists/
I don’t know about you, but I think firing officers that are found guilty in court beyond a reasonable doubt of fabricating evidence should be fired and firing officers that challenge compliant citizens to fight them is common sense. But the power of the police union prohibits meaningful accountability, and the need to look busy coupled with cities’ reliance on the police to raise revenues incentivizes putting up high arrest and fine numbers. Perhaps what you mean is “all common sense police reforms and methods which are politically easy to implement have already been implemented.” Great, but sometimes important things are not easy to do.
The problem with your (long winded) proposals is that they aren’t likely to move the needle one iota.
So, let’s go ahead and do national de-certification and asset forfeiture reform, etc [insert other silly libertarian ideas here]. Will life be materially different 5 years from now vs. today. If so, why? What will change?
BTW – the easiest and most efficient path out of this mess is for citizens to STOP RESISTING ARREST and let the civil and criminal system adjudicate the matter. That would solve like 99% of all of the high profile cases.
STOP RESISTING ARREST
Think of this as the Ronald Coase solution to all of the high profile cases.
> The problem with your (long winded) proposals is that they aren’t likely to move the needle one iota.
Wow, very convincing.
> So, let’s go ahead and do national de-certification and asset forfeiture reform, etc [insert other silly libertarian ideas here]. Will life be materially different 5 years from now vs. today. If so, why? What will change?
On decertification, Florida has > 1000 previously fired officers on active duty in other districts, so something in the tens of thousands of the worst police officers would be removed from duty.
On removing police unions, at least another 500 or so officers who appealed through arbitration would be re-hired. A much larger number of officers who get saved by union red tape would also be removed.
Remember, the majority of actual misconduct cases come from a very small percent of officers. So these two together could prevent a very large number of the incidents that erode community trust.
> BTW – the easiest and most efficient path out of this mess is for citizens to STOP RESISTING ARREST and let the civil and criminal system adjudicate the matter. That would solve like 99% of all of the high profile cases.
So long as we’re commanding hypothetical citizens by fiat, the easiest way to fix the problem is for nobody to do any crimes, and then we don’t need police and we fix the crime problem. But I guess people do crimes, so we need police. Maybe if police officers always did the right thing, we wouldn’t need to have any oversight or rules for police officers. But they don’t always do the right thing, so we do need oversight and rules for police officers.
“But they don’t always do the right thing, so we do need oversight and rules for police officers.”
We have oversight and rules for police. They work most of the time. We also have departments of ever higher levels of authority that can come in and provide oversight and rules for entire police districts as well as individuals. They work most of the time.
The bottom line is that the current system provides “good enough” oversight of police. It’s never going to be perfect. Attempts to achieve perfection require a level of “oversight” that appears to make it impossible for police to do their jobs, hence why current efforts have resulted in staggering increases in crime and murder rates.
Things like police unions and qualified immunity came about because police needed protections to do their jobs correctely. If they can’t we all suffer. Lots of people who want to provide “oversight” really want to prosecute police officers who are doing their jobs correctly from a public welfare point of view but incorrectly from a political point of view.
Bob is right that 99% of the problem is resisting arrest, specifically blacks resisting arrest. This is due in part to a culture telling them they are justified in resisting arrest. The single best method we have for reducing these high profile incidents is an uncompromising cultural stance against resisting arrest in all circumstances with no excuses.
This assumption is, generally speaking, not true. See Fig 7 of this paper, for example:
https://arxiv.org/pdf/1610.02413.pdf
At a fixed FICO cutoff, blacks are more likely to default than average and Asians less likely. The disparity is not even small. (You can find similar results in educational settings.)
A case where this is true is in recidivism prediction, c.f. COMPAS. This leads you to impossibility theorems which say that calibration and no-disparate-impact are mutually incompatible – you need to pick one or the other. Example: https://arxiv.org/pdf/1609.05807v1.pdf
https://arxiv.org/pdf/1609.05807v1.pdf
No metric can be perfect. Credit Scores can’t capture all relevant data or properly predict how that data will effect the desire metric (repayment).
It makes sense that blacks would underperform even relative to objective measures built for the generals society as not all of their inferiorities can be captured by such metrics.
Minor correction that doesn’t really impact the larger philosophical point on equity
> I believe that credit scoring is non-discriminatory with disparate impact. That is, a black borrower and a white borrower who each have a credit score of 650 will have the same probability of defaulting on the loan. But if you set 650 as a cutoff for approving loans, then the proportion of loan approvals that are for blacks will be less than their share in the population.
is empirically just not true in the status quo. A black borrower with a credit score of 650 is more likely to be denied for a loan from a face-to-face lender and still more likely to be charged a higher rate even with algorithmic online lenders. The latter is in large part driven by differential fixed costs of lending and competition across regions that are geographically segregated, and I don’t think there’s any question that geographic segregation is the lingering result of overt racism up through the late 20th century.
http://faculty.haas.berkeley.edu/morse/research/papers/discrim.pdf
I think you might appreciate a more sophisticated treatment of equity in mechanism design. For this, I’ve found some of the literature on fair machine learning to be helpful
https://5harad.com/papers/fair-ml.pdf
Loans are made or rejected on many criteria, and not just credit score. I don’t believe that the studies you cite prove systemic racism. Try the following thought experiment: Imagine a completely racist loan officer who hates minorities with a passion. When a person of color requests a loan, he looks for any excuse, no matter how inconsequential, to deny the loan. What prediction could we reasonably make about the few loans he does grant to persons of color? That their default rate would be near zero, given that the borrowers’ qualifications had to have been stellar for them to get the loan. So, is the default rate for all Black borrowers significantly lower than that for all white borrowers? No, according to economist Thomas Sowell, it’s virtually the same.
> Loans are made or rejected on many criteria, and not just credit score.
That’s my point. Given the topic of this blog post being “should institutions be reformed for equal outcomes,” my point is that “a fixed FICO score threshold” clearly does not represent how decisions are made in the status quo. As Chris Stuccio pointed out, even if FICO score were the only determinant, trying to draw a policy based on that which is “fair” even by the standard the blog author wants is a fair bit more nuanced than a fixed cutoff.
Maybe it’s meant to be a hypothetical thought experiment, but I don’t think it’s one that really serves the point of the argument.
To be clear, if “fair” meant “a classifier that is perfectly calibrated on each subgroup”, then we would actively discriminate against blacks (as in, treat “black with equal credit history” as -100 or so FICO pts) and in favor of asians (+20 or so FICO pts).
Most credit datasets tend to show that lending to blacks is less profitable than lending to others due to high default rates not paid for via higher interest. So while you can certainly find racial disparities, in aggregate they tend to transfer wealth towards blacks.
> To be clear, if “fair” meant “a classifier that is perfectly calibrated on each subgroup”
I don’t particularly like that definition of fairness as it’s possible to engage in targeted malicious discrimination, even at the expense of one’s own profit, while remaining perfectly subgroup calibrated, as pointed out in section 3.3 of
https://5harad.com/papers/fair-ml.pdf
and borne out in actual recent history through redlining practices.
In the end, I think doing anything other than trying to be as accurate as possible as a modeler is incorrect. There are times, particularly in the regime of “wide” rather than “tall” data, when decisions made by modellers meaningfully impact the residual risk distributions, and in those cases the goal of subgroup equity may coincide with the goal of improving overall accuracy. But when it comes to making decisions based on calibration parity or misclassifications parity, I think that’s a question for policy design (i.e., should we do this at all if it can’t be equitable) rather than mucking about with propensity thresholds.
“… it’s possible to engage in targeted malicious discrimination, even at the expense of one’s own profit…”
It’s possible, but it’s extremely unlikely to last long unless government mandates it across the board. If I lose money to satisfy my prejudices but my competitors don’t, I’m likely going to be out of business very quickly.
@somebody, the example described in that paper is more or less the exact opposite of what happens in real life.
Real life:
Data geek: “Boss I came up with a way to improve the accuracy of our underwriting! I use feature_X and ….math nerd stuff here…”
Data geek: “Boss I tried deep learning on our data set and roc_auc went up! This should justify a big bonus for me…”
Boss: “Lets run it by legal.”
Legal: “No, this is probably racially discriminatory. Can’t risk it.”
I mean, I’m sure that’s true now, but the example is exactly what happened throughout most of the 20th century.
Regardless, the point is to quibble with the definition of fairness, since I think definitions should be robust even against hypotheticals.
Note: See Thomas Sowell’s book, “Economic Facts and Fallacies,” pages 178 – 183.
The “white privilege” meme is based on a truth – that there are benefits to being in the majority. This truth applies not only to people but to most other biological organisms as well. The meme, however, racializes this truth and makes it toxic by implying that majoritarian benefits accrue not just from greater numbers but from malign intent.
Focusing on this truth blocks out other truths. There is, for example, the truth that the black poverty rate in the United States fell to 47% in 1960 from 87% in 1940. This before the Civil Rights Act, before Affirmative Action, and during a time when anti-black discrimination was the law in much of the South. Blacks pulled themselves out of poverty at an astonishing rate in a time of naked white hostility and aggression.
This truth, which should be celebrated as a triumph of the human spirit, has been buried under a great lie – that blacks can only advance with the help of white people. That if some white people discriminate against them, they are helpless. Black people have agency – they can and do improve their lives and those of their families by their own actions. Kendi places Black agency squarely in the hands of white people.
Privilege *is* real. Life *is* unfair. That unfairness stings: it is a wound that it hard and costly to close and easy and profitable to reopen and exploit. This is a problem for which secular liberal meritocracy has no satisfactory answer. How do you convince losers to suck it up and simply accept their losing lot in life? Why should they?
Even with zero culpability on anyone’s part, and even in highly homogeneous social circumstances, some people have better luck and an easier time than others getting desirable things.
Now it is possible to get people to accept that there is always good and bad luck and deviations from what people should get in some idea social conception, but at some point the deviation from the culturally prevailing intutions of ‘fairness’ become too large and the legitimacy of the whole order comes into question.
But it turns out to be far too easy to tell big lies and convince lower-status people of completely exaggerated narratives of the current order’s illegitimacy and the degree of culpability of higher-status people in intentionally keeping the status of members of other groups down.
“How do you convince losers to suck it up and simply accept their losing lot in life? Why should they?”
No one has to “accept their losing lot in life.” What people should accept is that they are responsible for their actions and that unwise actions lead to loss. The problem is that we’ve cut the link between actions and consequences. Kids get trophies whether they win or lose. They pass on to the next grade even if they haven’t learned anything in their current classes. People who choose not to become educated, who choose to have children out of wedlock, and who choose to stay unemployed can get along well enough on the welfare system.
Imagine how dangerous the world would be for a person without the ability to feel pain – as happens with certain forms of leprosy. Such a person could hurt himself terribly by continuing to walk on a badly sprained ankle or putting his hand on a hot stove without knowing it.
Governments and doting parents can create a sort of moral leprosy by not allowing people and children to face the consequences of their actions, weakening or even destroying the feedback loops linking cause and effect.
As the consequences of self-destructive actions (such as dropping out of school, having children out of wedlock, and drug and alcohol abuse) are increasingly borne by others, the incidence of such behavior will rise. At the same time, as the rewards for hard work, perseverance, and integrity fall, such virtues can be expected to fade.
There are indefinitely many racial inequities; and, because of the way they are interconnected, it is impossible to eliminate all of them. For example: If Harvard admits new undergraduates on the basis of test scores, proportionately fewer blacks than non-blacks will be admitted—a racial inequity. But if instead it admits blacks in proportion to their numbers in the general population most of these admittees, being ill-matched with the rest of the student body, will have comparatively unsatisfactory experiences: they will benefit less from going through Harvard than will the non-black students. This is a different racial inequity, caused by “rectifying” the first racial inequity.
No reasonable course will eliminate all racial inequities. If racial inequities occasion complaints, there will be no end to the complaining.
Let’s be real here. There has been no serious attempt by so-called anti-racists to remove disparities. As with feminism the goal is to remove invidious inequality where they are on the losing side but never to change things (e.g. sports or nursing) where they are the majority. And of course, anti-Asian AA in universities shows that it’s permitted to hurt a group that suffered from oppression and formal discrimination (anti-Asian laws lasted longer in many states than anti-Black laws) if that group overcomes its disadvantages and does well.
“Rather than turn our focus away from invidious discrimination, we must confront it directly.”
How?
Generally, it seems, anti- discrimination measures act to increase the cost of labor of a protected class. Anti-discrimination measures must be understood as a component of compensation. Hiring members of protected classes decreases the potential penalties an employer may have to pay for not hiring members of the class, but that must be balanced by the additional penalties the employer may pay to resolve complaints by protected class employees and any loss of productivity the employer might suffer if the protected class employee was hired in lieu of a more productive non-protected class applicant.
The added cost of employee compensation due to anti discrimination primarily goes to the legal guild/anti-discrimination industry. This skim may reduce employee compensation from what it might have been, or it can lead the employer to forego additional employees, substitute contractor/gig employees, automate, or offshore. It would be useful to have additional data available on these dynamics and cash flows.
But at any rate, the anti-discrimination industry is not going away. And it will continue to expand its share of the economy through rent-seeking legislation, novel court decisions, etc. We can see this clearly in California where the AB 5 anti-gig provisions enrich the powerful legal guild and where, ironically, the Proposition 209 anti-discrimination law is on the ballot to be repealed opening a gold mine for the legal guild and anti-discrimination industry.
Of course there are other industries devoted to taking a cut of the action between employers and employees. And countless other lucrative anti-discrimination angles. But the cost of these activities is not an economic multiplier.
Since, at this late stage in the game, there are no reforms imaginable, people will find ways to adapt. These adaptations may include reduced labor force participation, increased participation in the informal economy, more individual moves to achieving personal autarky, and increased efforts to conceal activities from the surveillance state. Things will get much worse and there is no guarantee they will ever get better.
And I suppose it should go without saying that these dynamics are unique to two-party systems of governance. In authentic democracy, multiple parties make it more difficult for such industries to rent-seek with such success.
I have seen no evidence that rent seeking is less common where there are more than two parties.
In two party systems, coalitions are created within parties. In multi-party systems, they are created between parties.
It is true that political scientists offer competing analyses as to whether various types of government/electoral system are more or less prone to rent-seeking, I am unaware of any widely accepted metric of rent-seeking.
So let’s go back to basics, Buchanan and Tullock. (Between the two there is too much material for me to sort through this morning to find out whether they specifically addressed this question but it’s probably in there somewhere.) In particular the Tullock paradox and the basic public choice model in Calculus of Consent. Gordon Turlock asked why is there so little money in politics given the rich potential returns of rent-seeking. To my uninformed way of thinking two party systems are somewhere to the left of center of the U shape derived on the expected expected costs of collective decisions /number of individuals required to take collective action graph because there are fewer individuals needed to grant a rent and the those who can grant a rent are competing with each other to get the payoff. In addition, the parties themselves offer structure to minimize the transaction costs of rent-seeking. In contrast, simply because more individuals are necessary to act to grant a rent in a multiparty system, log rolling is costlier and the multi-party systems will always be to the right of two party system and in most cases closest to the bottom of the U.
I believe Arend Lijphart’s findings in Patterns of Democracy bears this out. He basically found that consensus (multi-party) forms of democracy out perform majoritarian (two-party/FPTP) systems on dozens of wide-ranging metrics.
More specifically to the cultural division question, I would cite Appendix 2 “Political Accommodation in Culturally or Ethnically Divided Countries” of the second edition of Robert A Dahl’s On Democracy. Belgium, Netherlands, and Switzerland achieve successful consociational democracy despite ethnic divisions due to the pluralism offered in multi-party systems. Dahl references Lijphart specifically but this also goes back to Buchanan whom I recall saying something to the effect that democracy is a conversation. In a two party system you are more likely to get one-way conversations in which selling the rents is monopolized and brazen. For example, see Obamacare.
Finally, looking at proxy indices, we find that all of the 21 countries that out perform the USA on the Transparency International perceptions of corruption index are multi-party states. As are 5 of 6 that out perform the USA on The Economist’s crony capitalism index of 23 countries. China was the exception.
I think the charitable interpretation of the “anti-racist” model is that it’s proponents think that ‘given our country’s history,’ we should have an overwhelming prior in favor racism being the explanation for disparities. They might say if you see someone limping who also has a bullet in his leg, it’s safe to assume they are limping because they were shot. I don’t think this prior is warranted because there are multiple plausible explanations for the proverbial limp (racial disparities), some of which would even also likely cause the observed racial discrimination, and we also know that for many other groups intense racial discrimination isn’t sufficient to incur poorer performance, while for some others we know it isn’t necessary to incur poor performance.
The less charitable interpretation – which seems to describe Kendi – is that disparities are injustices by definition, and the model openly adheres to racial collectivism, and individuals don’t have rights irrespective of their race, but rather races, as groups, have collective rights.
The problem with making racism the default explanation for all disparities is that it takes agency out of the hands of people of color and places in the hands of white people. In effect, we’re telling minorities that they’re victims of oppression and, because of mean white people, any attempts on their part to improve their lives and those of their loved ones is futile. Instead, they must wait until white people perfect themselves and their racist institutions.
While it’s true that there’s racism in America, there are other truths. It’s also true that somewhere between two-thirds and three-quarters of Black Americans are middle class or higher. Black Americans dominate American sports and and extremely influential culturally and in the entertainment industry. And that happened because black people took action.
The truths we on which we choose to focus can make all the difference in our lives.
The problem is that it’s unfair to white people. That it may also infantilize brown people is also unfortunate, but the PRIMARY injustice is done to the ones being demonized and discriminate against, whites.
I dislike the emphasis on how terrible affirmative action and all the rest is for brown people because it implies that if it didn’t harm brown people the damaging effects on whites would be unimportant. Whites are not slaves who exist for the benefit of brown people.
Re: “I am inclined to get rid of the college admissions process and replace it with open admissions rationed by a lottery system.”
I would highlight three aspects of educational productivity at residential colleges: a) Interactive pedagogy is more effective when students have roughly similar aptitude. b) Students learn more from one another than from Faculty. c) Interactive diversity helps prepare students for the workplace and citizenship.
Therefore, in order to maximize educational productivity, a residential college should balance a target level of academic selectivity and a pragmatic level of interactive diversity.
More often than not, selective residential colleges don’t achieve interactive diversity. Instead, the outcome is statistical diversity in admissions, followed by racial self-sorting across academic fields and in social life. Case studies by Peter Arcidiacono (Duke U.) find that academic mismatch (i.e., overcorrection for disparate impact of standardized test scores) exacerbates self-sorting. For example, Black students at the selective universities have fewer interracial friendships than they had in high school, even though the selective universities have greater statistical diversity than did the high schools, which the Black students had attended. It seems that only moderate affirmative action achieves educational productivity and integration. Real-existing affirmative action partly backfires.
A pure lottery system in admissions would be inefficient. It would require very high attrition, dilution of pedagogy, or both. One might balance efficiency and fairness by imposing an aptitude filter (for example, a threshold standardized test score), followed by random selection among the applicants above the threshold. Presumably, MIT would set a higher aptitude threshold for inclusion in the admissions lottery than would Brown University, and so on.
I hasten to add that an admissions lottery (with or without an aptitude filter) would vastly increase decision-making efficiency in admissions. The large admissions bureaucracy would either shrink or switch from very elaborate gatekeeping to recruiting students.
Please define “interactive diversity.” Could you build this out a bit?
Just curious what you mean and not trying to be provocative.
Also, note that at least 70% of the population will never graduate from college. How do you get interactive diversity with these folks?
Sorry I didn’t catch your comment sooner.
Re: interactive diversity. It’s a phrase I contrived to highlight the fact that diversity in college admissions doesn’t improve education unless it is followed by real integration—in courses, Majors, group projects, research, friendships, extra-curricular activities, leisure. I should have said diversity with integration.
Integration would prepare students better for teamwork at the workplace. Integration might also indirectly enhance academic education in the social sciences, if integration fosters understanding of norms and culture.
Alas, residential colleges typically pursue policies that undermine integration. Colleges have separate orientation programs for minority students, social houses for minority students, periodic off-campus retreats for minority students. The stated rationale of these policies is to support minority students. However, the policies promote internal bonding at the expense of integration.
You’re correct about intrinsic social limits of any real integration in higher education. Bryan Caplan (“The Case Against Education”) and Charles Murray (“Coming Apart”) provide systematic evidence that educational attainment constitutes a great social divide, and that fewer students should attempt higher education.
I’m not familiar with whatever academic studies might say about this, but it was most definitely not true for me.