I hold these truths to be self-evident:
- Any law whatsoever, no matter how desirable on balance, will impose some costs on someone somewhere.
- In any society with more than about 12 people, it is virtually certain that those costs will be borne unequally.
- If the costs are borne unequally, then the costs borne by various individuals are virtually certain to be non-trivially correlated with at least one observable characteristic.
For example: A law that says you have to pick up after your dog will be costlier for dog owners than for non-dog-owners. Dog owners, depending on the community, will be either disproportionately old or disproportionately young or disproportionately rich or disproportionately poor or disproportionately ill or disproportionately healthy.
Therefore, unless you are willing to conclude that all laws are bad essentially without exception, you cannot argue that a law is bad just because it imposes individual costs in a way that is correlated with observable characteristics.
Therefore when Michael Bloomberg is criticized for supporting a stop-and-frisk policy because it caused disproportionate pain to young people with dark skin, his critics are being either disingenuous or unthoughtful.
Bloomberg’s recent public apology incorporates some of his critics’ language, which might lead a casual reader to believe that he either is, or is pretending be, just as thoughtless as those critics. But that strikes me as a bum rap.
Bloomberg’s exact words were “I didn’t understand then the unintended pain it was causing to young black and brown families and their kids”. That seems fine. After all, if one recognizes after the fact that one’s policies were responsible for higher costs than anticipated, one might well feel the urge to apologize. And in the course of apologizing, one might legitimately mention in passing some observable characteristics of the people who bore those costs. It looks to me like that’s what Bloomberg did, and it leaves him on solid ground.
We can of course still debate whether the costs of stop-and-frisk were or were not unacceptably high — in other words, we can still debate whether Bloomberg’s apology is appropriate, but it seems to me that it’s at least clearly within reason. Which is more than one can say for the rhetoric of those who demanded this apology on the ludicrous basis that stop-and-frisk, like every other policy that has ever been implemented by any government anywhere, affected some people more than others.
How about criticizing the policy not because it imposes some costs on someone, but because it is a clear violation of the 4th amendment?
Jens: That is, of course, an entirely separate issue. But if we’re going to start disallowing policies just because they’re clearly unconstitutional, then Bloomberg has a whole lot more to answer for, as do the rest of them.
“like every other policy that has ever been implemented by any government anywhere, affected some people more than others.”
It is not just that, though, is it?
“you cannot argue that a law is bad just because it imposes individual costs in a way that is correlated with observable characteristics.”
But if you observe a pattern of laws that all adversely affect one group, there might be cause for concern.
It is not just laws either, but the implementation of laws. If we have evidence that laws will be applied unevenly we may wish to re-consider that law.
There is also the effectiveness. If the law has a downside but no or little up-side, we might wish to reconsider.
Looking back at USA history, I notice a few laws that do seem to do this. Slavery, Jim Crow for example. There is evidence of uneven application of the laws, laws and policies being targeted at specific groups both deliberately and as an unintended consequence. The value of the stop and frisk is also questionable, whilst there are certainly downsides.
Apart from the obvious, a very quick look shows that stop and frisk may lower house prices, for example, so is seen locally as a dis-amenity.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2695584
On the face of it, there is a justified concern here. There is something worth looking into.
#2 It is not so much clearly unconstitutional, but has been ruled so by a court.
To be clear, stop and frisk was not ruled unconstitutional per se, but the way NY did it was. The judge did not consider whether it was effective or not. It is not only, or even mainly about the law itself.
I don’t know if you are saying that you think lots of other laws are unconstitutional, or whether you support laws that have been ruled to be unconstitutional.
I think what people may be objecting to is not that the law against carrying illegal goods affect some people more than others but that there was unfairness in the way the law was applied.
Lets say that 70% of people carrying illegal goods are from a group that make up roughly 50% of the population but 90% of people chosen to be stopped and searched for possession of these illegal goods are from this group. While this policy would probably lead to more violators being identified it would also mean that members of this group would have a greater chance of getting caught for breaking this law than non-group members. This sounds like reasonable grounds for objecting to the way this law is enforced as it unfairly affects some people more than others.
I assume part of his critics’ reasoning is that it’s worse for laws to disproportionately harm people who are already disadvantaged, and that black New Yorkers were already on average poorer and more frequent targets of racism. (Bloomberg’s critics probably wouldn’t have attacked him for supporting affirmative action policies that “victimize” white people.)
Whether that is logically defensible or not, it does seem to square with our moral intuitions about things like “kicking someone while they’re down”.
Steve,
I think that, as with virtually all your posts, there’s a lot of clear thinking here.
The thing that didn’t ring true for me was Bloomberg’s statement, “I didn’t understand then the unintended pain it was causing to young black and brown families and their kids.” Whatever else we think about Bloomberg, he’s not a dumb guy and he seems to be someone who can think through consequences. So during years of this policy he didn’t understand the pain it was causing? I just don’t buy it. Now if he had said, “It’s causing pain, but it was worth it for the good consequences,” that would be different.
Academic experts on the ethics of AI are hopelessly stuck on this issue. If, say, police are guided by some AI program to tell them where to look for criminals, they will look in high-crime areas, and some people will complain that he AI is discriminatory.
Does it not make a difference whether the characteristic that determines who bears the costs is due to a choice made?
I can choose whether or not to own a dog. I can choose whether or not to own a gun. I can’t choose whether or not to be black. A teenager in high school can’t choose whether they live in a poor or rich area etc.
A young, law-abiding, black man should have just as much right to walk around without being molested by the police as an old, law-abiding, white woman.
” and some people will complain that he AI is discriminatory.”
Famously the Amazon recruitment AI was biased.
The issue of whether things like stop and frisk are discriminatory is a difficult one to determine. One usual measure is outcome test. If loans are repaid by black people at a higher rate than white people, that suggests that only exceptional black people are accepted and indicates discrimination against them. The other is benchmarking, which looks just at the rates of hits for the at-risk population – so offers of fewer loans to black people would indicate discrimination by this measure. Outcome tests are supposed to correct for the problems with benchmarking, for example if black people are actually less credit worthy than white people offering them fewer loans would not necessarily be discriminatory.
These suffer from a problem called infra-marginality. This is effectively that the distributions of each population are not the same. Simoiu et al [2017] offer a different test – the threshold test.
https://arxiv.org/pdf/1607.05376.pdf
Basically, this looks at what threshold each population must exceed in order for the action to take place, for example the probability in the mind of the assessor that the loan would be repaid. For the loans, it would be discriminatory if black people needed to pass a 70% threshold for assessed pay-back probability, whereas white people only needed to exceed a 65% threshold. It is easy to construct realistic scenarios where the threshold is not the same as the actual payback rate, so discrimination can occur when outcome tests do not show it. The authors claim “Addressing this challenge, we have developed a new statistical approach to detecting discrimination that builds on the strengths of the benchmark and outcome tests and that mitigates the shortcomings of both.”
The authors studied traffic stops and the rate at which a search resulted, using a database of 4.5 million stops. They found the search thresholds as follows:
White – 15%
Black – 7%
Hispanic- 6%
Asian – 13%
This shows discrimination against blacks and Hispanics compared to whites and Asians.
Whilst I am sure this is not the final word, I think it is a productive way to view the data.
Another point, from the same study above:
“Table 1 shows that the search rate for black drivers (5.4%) and Hispanic drivers (4.1%) is higher than for whites drivers (3.1%). Moreover, when searched, the rate of recovering contraband on blacks (29%)and Hispanics (19%) is lower than when searching whites (32%). Thus both the benchmark and outcome tests point to discrimination in search decisions against blacks and Hispanics.
So overall, the results are the same for all tests, benchmark, outcome and threshold. There was discrimination against blacks and Hispanics. The threshold test did show a more consistent pattern when looking at district data.
David R. Henderson: Yes, your point is very well taken.
Eric Kehr: Your argument strikes me as backwards. A law that is costly for dog owners discourages people from buying dogs, which is an unfortunate side effect. A law that is costly for blacks does not discourage people from becoming black, because that’s (for all practical purposes) not an option in the first place. So the unfortunate side effect is avoided. If a law is going to impose costs on a specific group of people, I usually want that group to be that you can’t opt in or out of, lest we set up all sorts of perverse incentives.
Also—and this is a separate point—a law that is costly for dog owners is ipso facto costly for people who want to be dog owners, and wanting to be a dog owner is not a choice.
Harold:
But if you observe a pattern of laws that all adversely affect one group, there might be cause for concern.
Yes, point well taken.
All policies are discriminatory, in the sense that common outcome measures will show different effects for different groups. Steve’s post explains why. Studies claiming to show discrimination are just a big waste of time.
Steve, isn’t the whole purpose of laws to affect people’s choices? And if it is a good law, that’s often because those choices negatively effect other people.
A law that dog owners need to pick up after their dog does not impose costs on dog-owners per se (or people who want to be dog-owners); it imposes costs on “dog-owners who wouldn’t otherwise pick up after their dogs”.
A dog owner who would pick up after their dog is not at all negatively affected by this law. Indeed they are positively affected as they get to walk their dogs on cleaner streets.
#16 Oh no, externalitites!
In reply to this specific point, a law prohibiting murder is costlier to murderers than non-murderers, but nobody suggests we should not have one. Even if murderers are over-represented by some identifiable group.
I don’t think the issue of dog waste was the main point in this discussion, but it does raise questions.
The economic way to go about this would be not to ban leaving poo, but to either incentivise picking it up or discourage it with a financial penalty proportionate to the external harm.
With dog poo, we could probably incentivise collection with a small fee paid per pound of poo. This would allow dog owners to either leave their own (actually their dog’s) poo for collection by someone else or collect it themselves. The problem is that collecting and transporting the poo to a central collection area is rather onerous – probably more so than the initial harm. This would probably leave it to almost destitute people who would gather it in a similar manner to collecting cans for recycling.
We cannot practically impose an appropriate tax on poo as enforcement is difficult and would be costlier than the harm.
Banning it seems a reasonable way to go in this case to reduce the external harm with the smallest cost. The threat is of a punitive fine much larger than the individual harm caused by the waste.
What actually makes the difference is a change in attitude. In the UK, when I was a kid nobody picked up dog poo – the idea would have seemed very odd indeed. Now nearly everyone does and anyone not doing so is likely to receive negative attitude from any witnesses. It is no longer seen as socially acceptable to leave it on the street. A survey in 2010 showed “There was strong agreement with the statements that dog walkers should always clean up after their dogs on street pavements (98%) and in public parks and playing fields (97%) however only 56% strongly agreed that dog walkers should clean up after their dog(s) in all locations.” Respondents considered that “It was the right thing to do” to be the most important factor. The provision of waste bins was also important, which imposes a cost on the local authority, although not nearly as large a cost as cleaning the streets. In 1996 the UK law changed to allow “dog control orders” on specific areas of land, allowing fines for fouling. There have also been public awareness campaigns. These seem to have been successful in changing attitudes towards responsible dog ownership.
For this problem, the usual economic solutions do not seem to be available, although I am open to suggestions. The more important factor is public awareness and social responsibility. This has parallels with the Haifa childcare experiment by Uri Gneezy and discussed in Freakonomics.
There has been a similar change in attitude to drunk driving, which is generally seen as socially unacceptable (although less so in rural areas).
How is it in the USA, particularly in street environment?
Bloomberg may need to argue further (and soon) that the benefits of the policy not only exceeded the costs (assuming they did), but also went disproportionately to people bearing the same characteristics.
The criticism tends to make it seem like this was done on a whim, but there was a problem ex ante…disproportionately distributed.
Concerning differential incentives/costs from legislation, one has to bear in mind both the incentive effects and the distribution effects. We WANT people to internalize the costs of dog poo externality, that it falls deferentially on different people might be a reason not to impose the “tax.” Ditto stop and frisk. Incentivizing people not to carry guns is fine but how intense should the disincentive be if it falls disproportionately on people already pretty low on the income/social distribution? I hear Bloomberg admitting his PD did not get the intensity right.