Testing for Racial Discrimination in Police Searches of Motor Vehicles

That is a new paper by Camelia Simoiu, Sam Corbett-Davies, and Sharad Goel, the abstract is familiar but depressing:

In the course of conducting traffic stops, officers have discretion to search motorists for drugs, weapons, and other contraband. There is concern that these search decisions are prone to racial bias, but it has proven difficult to rigorously assess claims of discrimination. Here we develop a new statistical method—the threshold test—to test for racial discrimination in motor vehicle searches. We use geographic variation in stop outcomes to infer the effective race-specific standards of evidence that officers apply when deciding whom to search, an approach we formalize with a hierarchical Bayesian latent variable model. This technique mitigates the problems of omitted variables and infra-marginality associated with benchmark and outcome tests for discrimination. On a dataset of 4.5 million police stops in North Carolina, we find that the standard for searching black and Hispanic drivers is considerably lower than the standard for searching white and Asian drivers, a pattern that holds consistently across the 100 largest police departments in the state.

For the pointer I thank the excellent Samir Varma.


Anecdotal but I was just stopped in Kansas while driving with the cruise control pegged on the speed limit. Cop lied and said I was speeding, presumably due to out-of-state plates and a suspiciously legal speed. Unconstitutional fishing expedition.

As it happened I also had a loaded Glock in the center console. Never came up, but I assume it would have if I were less white or less cooperative. Cop let me off with a "warning." For what I do not know.

I have a problem with the loaded Glock. Have you considered carrying a real weapon? Just kidding. My kid has a Glock .45 cal.

I get the traffic police "special treatment" because I'm over 65 years-old and white, and there's about 0.5% chance I have anything illegal in the vehicle.

Not sure what road you were on, but I've always thought that incidents like that should warrant the use of federal police to patrol Interstates and US Highways, on the basis that states don't get to "own" (in policing) the nationalized road networks.

The Interstate Highway System is not "nationalized." The states own the rights-of-way for highways within their borders, and are generally responsible for maintenance and upkeep, which is why they get to assess their own tolls as they see fit.

The federal role is primarily in the disbursement of funds, which come with enough strings attached to ensure each state plays ball with national transportation policy.

As a federal taxpayer, I'd love to see the states pay back the 90% federal largesse* they got for construction of the Interstates. Let's start with Southern states since they're in favor of a small federal government.

* https://www.fhwa.dot.gov/interstate/faq.cfm#question7
"Congress retained the Federal-State matching share of 90-10 as a reflection of the Interstate Construction Program's importance to national goals. (In the western States with large amounts of untaxed public land, the Federal share could be increased to 95 percent.)"

"Let’s start with Southern states since they’re in favor of a small federal government."

That's not likely to happen. And if Clinton wins and gets her policies enacted, you can expect billions of dollars a month in additional transfers from higher income northern states to lower income southern states.

So, if you are voting for Clinton, Observer, I'd like to say thank you. Your tax money is appreciated. ;)

I'm an observer, not a voter.

anecdotal, but I have had two friends stopped within the last 6 months, one in South Dakota, one in Tennessee, one two mph over-one two mph under, both cruise control.
The lesson: don't drive the speed limit.

It is called taxation without representation.

If I remember my history right some folks in uniform got shot over that as well.

Has anyone actually tried to read the study? The study invents a new methodology to find racial bias where other studies have failed. The other studies were based on "hit rates" (probability of finding contraband) and did not find evidence of racial bias. The obvious question is whether the new methodology is actually better than the "hit rate" approach that found little or no evidence of police bias.

Stated differently, people will try awfully hard to beat any given set of data to support their preferred conclusions. Whether they have really added anything is another matter. Here is a useful quote from the study

"In particular, drugs are discovered in the plurality of cases for all groups: 46% of searches of whites turn up drugs, compared to 49% for blacks, 34% for Hispanics, and 42% for Asians. And weapons are only infrequently recovered: they are found 6% of the time for whites, 9% for blacks, 7% for Hispanics, and 11% for Asians."

In other words, the stated conclusions of the study don't match (at all) the actual data.

Finding drugs or weapons 40-60% of the time is pretty significant. And as with most other cases of supposed discrimination if anyone has cause to complain it's definitely not black people.

Considering that some 22m adults in the USA used cannabis within the last month, stopping any random single-occupancy car should give you a 10% hit rate. If you profile by age and by neighborhood, as well as targeting cars with multiple adults on board, you can easily hit 40% without much effort.

I don't think it is surprising that the much higher rates of criminal behavior, etc. amongst the groups in question (e.g., African-Americans) would alter police perceptions & thresholds. In fact, arguably, that is critical to effective policing.

There appears to be no real acknowledgement of the trade-offs inherent in pretending that these criminal & behavioral disparities don't exist.

What is it exactly that Tyler and others would want the police to do?

For starters, if you have to lie to the motorist or the court about why the stop was made, you probably shouldn't be making the stop. See "arbitrary and capricious."

You are side-stepping the disparity question raised. If the rate of criminal behavior is much higher (which it unquestionably is) in certain groups, then it is illogical in the extreme to be so focused on why the said groups are stopped at higher rates.

Is there social benefit in discouraging the police (via incentives or otherwise) from searching / stopping high-risk and criminality-prone groups? Legal questions aside (which I admit could be troubling), the public policy case seems mindbogglingly weak.

Social benefit? Over the past couple years, have you not noticed the, ahem, social problems that racially biased policing creates?

People talk about the falling crime rate over the past decade as if it happened in spite of practices like these rather than as a result. Do we really want a replay of the 70s-90s crime spree that led to an internal refugee crisis of world historic proportions.

I would assume that racially-biased policing is actually *less* frequent now than before ... maybe it is actually helping, and we just have that much more to improve in terms of being race-neutral in our policing.

Jan, you think the average voter out there sees BLM and wants, ahem, less police protection?

Compared to what? The first year of Obama's presidency to the last year?

Its not clear to me that biased policies caused them. It could be the coincide with internet/mobile social new sharing sites. I can't help but think that discriminatory policies have been going on way before the "past couple of years".

I thin V's question deserves a non-snarky answer.

Then, for those that see bias everywhere, there is another riddle that needs a solution. If searches/stops are strictly proportional to a group's "criminality", is it still accurate or meaningful to call it bias?

Biased against a particular law-abiding individual maybe? (But presumably a search/stop against ANY law-abiding person is unfair or an error). But biased against a group?

THFMR is right to concentrate on the legality and constitutionality of the stops.

Supposing you are right and it makes sense to stop blacks more than others. That only excuses concentrating otherwise reasonable stops on them. If the cops limited themselves to that, PC activists would still wail about the stats, but it would still reduce the actual resentment among ordinary people.

Anything that makes policing less arbitrary and more lawlike will help cops get real policing done, even if it leads to and lower arrest rates for phoney crimes.

Assuming someone is criminal prone based only on skin color is what most people call racial discrimination. The trade off is between the reduction, if any, in crime due to such stops and the increase in crime because of the justified belief that the system that racially discriminates is unfair.

When all you care about is crime statistics, that's about correct, but I care about other things, including general decency, fairness, and the Constitution. Racially discriminatory policies impose additional costs if you value these as I do.

Would you object to having your tax returns audited every second or third year, if you are a white male? According to the FBI, the majority of white collar criminals are white male (https://ucr.fbi.gov/nibrs/nibrs_wcc.pdf). I'm sure there would be great social benefit if this clearly depraved group were subject to constant and rigorous oversight.

The majority of people are white. Nthis majority of people in white collar jobs even more so. White people aren't more likely to be white collar criminals. Blacks, on the other hand commit crimes at significantly higher rates than even low socioeconomic status whites.

Whether it is conscious or unconscious, if cops are incentivized to arrest people for crimes and are given any discretion at all, they are going to disproportionately to focus on blacks. From a Bayesian pov, a black guy driving a car does make it more likely that the car has drugs in it or is stolen. Try telling a cop to ignore this true information when his bonus or promotion depends on him intuiting it.

First, the data you reference is almost 20 years old. Second, the offense rate for white males is pretty close to their fraction of the population. So no, that data doesn't appear to justify special audit attention based on race. Try again.

I'll leave you to do your own research on relative offender / population rates for violent crimes.

Doesn't the IRS target people, like small business owners and the self-employed, who are most likely white and most likely to be underreporting income? Seems like a sound practice to me.

Ahh, but target people who are "most likely white" is different from targeting people explicitly because they're white. Its an important distinction.

I wonder why it isn't called discriminatory or indecent or unfair if (as surely happens) males or certain age groups are stopped/searched more frequently?

If a shooting takes place on campus, and if eyewitnesses report that the shooter was black (or white), should the cops ignore that information? Sure, it leads to hassles for blacks (or whites), but then shouldn't the people who are hassled complain to the shooter, not the cops?

If a shooting takes place on campus, and if eyewitnesses report that the shooter was black (or white), should the cops ignore that information? Sure, it leads to hassles for blacks (or whites), but then shouldn’t the people who are hassled complain to the shooter, not the cops?

You mean if the police have some concrete information about the criminal is it OK to use it? Of course it is. But that's not what we are talking about.

It is not the business of the police to randomly stop people in the hopes that they might catch a criminal that way. Not the way we do things, nor the way we should. The policed simply have no license, for good reason, to just go around making baseless stops.

What 'rates'? For example, last time I looked there were something like 16K homicides per year in the US. Blacks were the suspected killers in about half of those. That would mean about 8,000 killers (assuming we don't have multiple killings by one person). The # of black men in the US is something like 3M. So even though you may have a higher rate, you're going to have to make a huge number of unjustified stops if you think you're going to use that type of profiling to catch a homicide perp.

In NYC 'stop and frisk' had even more absurd numbers. The program's purpose was supposedly to remove illegal guns from the street. The figures proudly cited were things like 500 guns removed last year yet the # of frisks were to the tune of 500,000+. Getting a single gun off the street then required what was essentially a pseudo-strip search done in public 999 times. (Just imagine if Obama tried to get the BATF to do that to rednecks in the South BTW...but I digress)

Of course you don't just have huge numbers of stops that just serve to annoy innocent people and turn them against law enforcement. You also have the creation of what are essentially different sets of rules. In NYC pot was essentially legal for whites since they were rarely stopped and searched so you could carry it around all the time provided you avoided doing anything else blatantly illegal. For blacks it was essentially illegal since you could easily be caught with it even if you walked around doing nothing else illegal.

Remember the Ken Starr principle, if you focus enough you'll find something to convict almost anyone of.

Why focus on homicide when you could literally focus on any other violent crime

The corrollary to the Starr principle is the Bill Clinton axiom- attack the law officers enough & you can get away with anything

OK let's do 'crimes of violence'. Working off of table 42 on http://www.bjs.gov/content/pub/pdf/cvus0602.pdf which is for 2006 but should work for us:

We have about 4.4M 'crimes of violence' in the US in 2006. 1.019M or so the perceived race of the offender is black. With about 38M black people in the US and you have at best a 1 in 38 chance of catching a 'crimes of violence' offender by just pulling over black people randomly. That, of course, assumes you can figure out someone has committed a 'crime of violence' from a traffic stop and it assumes that no black person does more than one crime per year.

Let's pretend Google and Tesla invent a 'crime of violence' meter that will beep if someone did a crime in the last 12 months. If you were a cop and decided to pull blacks over randomly in order to use this magic device, you would be leaving 75% of 'crimes of violence' undetected.

How again does the 'rate of committing crimes' work?

The corrollary to the Starr principle is the Bill Clinton axiom- attack the law officers enough & you can get away with anything

We have law enforcement officers dedicated to blow jobs?

They arent pulling people over randomly, thats the point. If you narrow it down to black youths the odds of finding something goes up substantially. We expect cops to develop skills necessary to spotting criminals and apprehending them, but freak out when that means focusing on poor black youth rather than middle class white people.

Mavery's point above is solid, there are costs unrelated to crime prevention to this type of policing. But that doesnt change the fact that focusing on the most likely criminals makes sense.


In that case the metric you want is the # of stops needed to achieve an arrest. If you have more crime done by black youth than black middle aged people, for example, then yes you would get more stops of youth than older. But with unbiased enforcement you'd still get stops of both.

In that case you look at the 'cost' of the arrest as the stops that do not result in arrests. For example, we hear about the guy who gets pulled over 45 times in a 6 month period. Let's say on the 46th time he is in fact doing something wrong and that results in arrest. The metric then is 1 stop that 'scores' costs 45 stops that fail to score.

That rate should then hold cross the board. In white only stops you should see a rate of 45 'misses' per 1 'score'. Not sure if those measures are easy to come by but do you want to bet your pay they are equal cross races?

Just to clarify,

Say black youth has a rate of crime that's 2x black middle aged people. I'd expect to see something like this then:

Youth stops 92 with 2 valid arrests 1 arrest 'purchased' by 45 stops without arrest
middle age stops 46 with 1 valid arrest 1 arrest 'purchased' by 45 stops without arrest

In so far as i understand you, yes, success to pull over ratio seems like a good metric. Again, if you only care about successful law enforcement and no other social costs.

I thought that was the Lawrence Walsh principle. Didn't he say openly that since Oliver North's "real" offenses weren't crimes, he had to investigate until he found something?

Is there social benefit in discouraging the police (via incentives or otherwise) from searching / stopping high-risk and criminality-prone groups?


There is a social benefit in having the police use uniform standards that disregard race in deciding who to stop. We sometimes call it "equal protection of the laws."

Why do you have to lie? You can pull most motorists over for small small violation. No need to lie.

Assume the cop has to document what motivated the initial stop. Suppose for his stops involving speeding the average speed was over the limit by 15mph when it was white motorists but for blacks it was 5mph.

From that we could conclude either:

1. Whites don't like to speed by a little but a lot while blacks only like speeding by a little.
2. He cuts white speeders more slack than black.

you could conclude 2, but I dont think that insinuates anyones constitutionals rights have been violated.

#2 would be a violation of our constitutional rights. You can't have a high tolerance for enforcing the laws with one race and a low one for another.

>What is it exactly that Tyler and others would want the police to do?

While some of the others are seeking to end policing in black neighborhoods, the real answer is to your questions is -- nothing differently at all.

This is pure virtue signalling, straight up.

To follow suit, you'd be wise to comment sagely about injustice, and then go about your business until you are instructed to do it again.

Go read the Baltimore report the Justice department just issued and claim that there's no need for reform.

It's not that everyone is doing everything wrong. It's that in many places, there are lots of people doing many things wrong. Ignoring these by citing differentials in rates is myopic.

Another lesson to be learned from the Justice Department report on Baltimore is that democracy does not work very well when a large fraction of the population is DFASIV (= dumb, feckless, antisocial, inclined to violence).

One thing that caught my eye is that the solution to bad policing was to hire police from elsewhere. Does no one remember the practice of the Spanish fascists to never assign police to their home town so that they would faithfully do things to the people that they wouldn't do to the people they were connected to?

No one expects the Spanish fascists. Our chief weapon is surprise, fear and surprise.

Stops of persons of race or culture X when crime Y was committed should be roughly in proportion to the (statistics-based) conditional probability P (person who committed crime Y is race or culture X | crime Y was committed). The fact that this probability is not equal to P(crime Y was committed | person is race or culture X) confuses and upsets a lot of people, especially law-abiding persons of race or culture X. However, the fact that the former conditional probability is not equal for all races or cultures is the problem that should ultimately be addressed, and there is not enough discussion of that point.

You are arguing that if there are more black criminals, pulling over a black person means police are "getting closer."

Not really. As I say below, a misunderstanding of statistics.

Re. V and underlying crime statistics: perhaps it says police are not using the statistics _correctly_. Perhaps pulling over an innocent for b.s. does not actually catch more criminals.

I think the news mentioned a _never arrested_ middle aged black man who had nonetheless been pulled over 45 times in his life.

That is unfair, but also a horrible waste of police resources.

BTW, if you as an innocent had been pulled over 45 times, would you start to believe that the stops were not really about law enforcement?

45 doesn't seem like that many times. Do they even keep statistics on that? How would they know? I've probably been pulled over 10-15 times in my life (below middle-age) and I'm a white guy who lives in wealthy neighborhoods. Does he have a bunch of driving violations? You know they pull people over for that without arresting them, right?

I am about 60, pulled over twice. Spent 30 years diving fast cars too fast. Like 100 every day fast.

'course cops could see that they were expensive cars.

In fact, at times it was obvious, they'd get on my tail, then back off, let me go.

A Miata isn't an expensive car no matter how hot the pink of your paint job.

"Like 100 every day fast.

‘course cops could see that they were expensive cars.

In fact, at times it was obvious, they’d get on my tail, then back off, let me go."

LOL, I like Ray's stories better. They're more plausible.


If true, you have a uniquely anomalous experience that may be shaping your outlook. Jerky rich guys in expensive cars going 100 mph is exactly the people they love to pull over

It's pretty much a no win situation. The TSA wastes enormous amounts of everyone's time patting down every 85 year old grandmother, and every complains about how ridiculous it is. But when the cops target their searches at the demographics most likely to engage in criminal behavior, everyone complains about discrimination.

You know, an extreme feminist might make the "all men are rapists" claim. You seem to be on board with this demographic argument.

You've almost got it: not all men are rapists, but most all rapists are men. Keep trying, though, guy. You'll get there!

Amazing. So if you were pulled in once a year or so and grilled about your relationships you wouI'd be fine with that?

Nice strawman your building there, anon.

I don't see anyone arguing that minorities should be pulled over for no reason at all, just that the statistics should reflect the crime rate.

anon, as long as we're going reductio ad absurdum, if a woman is drugged and then raped, is it your position the police should expend equal efforts investigating women and men seen in the area?

Since you are not in the black demographic you are blithely saying: sure over police them, you might catch someone.

This would be like a woman who has a default of suspicion for men, because men are rapists.

Yes, TMC on this page many are.

One man's over policing is another's efficient deployment public resources. Again, Grandma isn't taking down a 747 mid-flight with a bomb stashed in her Depends, but the TSA treats her as if she's just as big a threat to do so as anyone else.


Literally no one on this entire comment thread has suggested pulling over minorities for no reason at all. No one!

Racism is at its heart a punt on statistics, so it is unsurprising to see that repeated here.

Most black people are not criminals, but it is easier to think they are.

Every argument that blacks should be pulled over falls out of that. You aren't asking police to find criminals, you are saying it is good enough if they find blacks.

Pro tip: focus on criminals

@Anon Most *people* arent criminals, black or white, and yet we still expect police to apprehend the ones who are.

Every argument that blacks should be pulled over falls out of that. You aren’t asking police to find criminals, you are saying it is good enough if they find blacks.

No one is saying that. If anyone is punting on statistics, it's you. If the police ignore the demographics of crime in this country and treat everyone as equally likely to be a criminal, they will, in fact, find fewer criminals, because they'll be looking in the wrong places. Never said I wanted anyone to be subject to random searches or not have their fourth amendment rights respected; you're just putting words in people's mouths.

If you think pulling over black people is *procedurally* I good way to find criminals, you do not understand statistics.

They are wasting time at best, fracturing a community at worst.


Not even this study finds that cops pull over black people just for being black, nor has anyone in the comments suggested that they should.

The supposed finding was that cops had a lower threshold of illegal behavior for pulling over blacks.

"The TSA wastes enormous amounts of everyone’s time patting down every 85 year old grandmother, and every complains about how ridiculous it is. But when the cops target their searches at the demographics most likely to engage in criminal behavior, everyone complains about discrimination."

This is a red herring. With regular law enforcement the purpose is not to achieve perfect enforcement of all laws but simply to put a control on law breaking. For example, there are probably hundreds of cars that zip by every day on a highway where the driver has some illegal drugs onboard. The purpose of law enforcement is to put a control on that by establishing the risk of carrying illegal drugs on the highway is greater than zero. Even so society does not want perfect enforcement.

With airline terrorism, however, we want perfect enforcement. We don't want the risk of getting caught trying to get a bomb or weapon onboard an airplane to be higher, we want it to be nearly impossible. This is compounded by Trump-like pundits who declare any terrorist attack to be a an absolute defeat for all of Western Civilization. As a result anyone and everyone is subject to random pat downs. BTW there is at least one 65+ grandmother (a Jewish one too!) who was charged a few years ago with providing aid to Islamic terrorists. The first consequence of deciding to have the TSA no longer patting down sweet grandmother types will be a race for terrorist groups to find and recruit a nice looking grandmother type...and given 5+ billion people on earth odds are they will find one sooner or later.

The analogy you might think about isn't so much TSA patdowns but customs where spot checks are more random and often more profiled (there though the grandmother might actually be a good target if your goal is to find who is trying to sneak in undeclared olive oil into the country!)

If the rate of criminality is higher (which, statistically, it is) why would we not expect the rate of indicia of criminality to be proportionally higher as well?

That is to say: if African-Americans commit crime at a higher rate, then applying equal standards should result in an equally higher rate of stops, searches, and other police activity incident to actual crime. Unless you mean to suggest that African-Americans are too crafty to be subject to the same standards as salt-of-the-earth White folk, rates should have no effect on thresholds.

The constitutional bar to a stop is first visible observation of breaking a law (i.e. speeding) or reasonable evidence that the vehicle is being used criminally (i.e. fits the description of a reported hit and run from a few minutes ago). The alternative is truly random spot checks (i.e. stopping every 5th driver at a DUI checkpoint on New Year's Eve).

'Rates of criminality' do not clearly factor in here. Men beat their wives more than wives beat their husbands. That is a fact too, would that justify a cop stopping male car drivers, checking their knuckles for cuts or bruises and then calling in for another officer to go to their homes to see if there's a battered wife or gf there?

If a cop did this what would be the objection? Certainly many innocent men would get pulled over. Certainly many cases of men with bruised fists would not be cases of domestic abuse, but some no doubt would.

But everyone is breaking the law all the time, so legally a stop is always justified. That is why it breaks down to pulling over people who "look suspicious"

Everyone is not breaking traffic laws all the time. Pulling people over on only the grounds they “look suspicious” is problematic.

90% of people are speeding at all times

In NJ it's a challenge to break 50mph. When a cop is roaming most cars stick to the speed limit.

I guess it went over your head that they get searched more often. Which implies that any estimate of higher criminality among that sub-population is going to overestimate the difference (I assume, for various reasons, that the underlying rate of criminal acts is actually higher, but this is at least one reason that it would be an exaggerated statistic).

So, the stats kind of become self fulfilling. It might be more of a chicken and egg issue than you think. At the same time, there are legitimately higher needs for policing in a lot of those very same communities, at least in the short and medium term.

Sure, for drug crimes that's plausible. For homicide? Rape?

Umm. A culture of violence more related to gang culture related to turf, etc., in distribution of drugs? And then general insecurity differences causing additional criminality due to lack of police resources allocated in that way?

I dunno, it certainly wouldn't stun me if, some time down the road, when we know lots more, to find out that there was some, say, very small genetic differences in distributions of genes among identifiable sub-populations in a particular direction which would lead one to expect a difference in the types of crimes you refer to. However, considering the plausibility of numerous other social explanations, I do not see it generally fit to assume non-social and non-cultural explanations for these differences, also accounting for history and development of families and communities over generations, including their interactions with "the system" and related effects on family, community, etc.

Umm. A culture of violence more related to gang culture related to turf, etc., in distribution of drugs?

If it helps you feel better, go with that.

People have been fighting over turf for a very long time. But we can pretend that this stuff would never usually happen when stuff that is demanded gets banned ...

I don't think it matters whether the explanation is genetic or not. Why even bring that up? Isn't that a distraction?

OK Cliff. I promise not to pre-oppose arguments. I'll just wait for you to bring it up. Or maybe not ...

I guess it went over your head that they get searched more often. Which implies that any estimate of higher criminality among that sub-population is going to overestimate the difference (I assume, for various reasons, that the underlying rate of criminal acts is actually higher, but this is at least one reason that it would be an exaggerated statistic).

Survey research on victimization is congruent with arrest and conviction data re the distribution of perpetrators among broad racial categories.

There are probably some kind of half excuses relating to possible stats/data bias issues. But I would like to point out that if this leads to a higher false positive rate then naturally you should expect a larger body of discontent towards the police (and/or populations responsible for tolerating differentiated privacy invasions of innocent people).

An interesting (vital??) supplement to the study might be what are the statistical comparative results of the searches (& any seizures) on that same "threshold" basis.

Do police get a higher proportional number of "hits" in one set of "racial" stops than another?

Betcha !!

Mort Dubois, how do you know white males aren't over sampled on audits?

Audits are rarely (ever?) triggered on the basis of name or colour. Much of the time, the person who might trigger an audit would not have a clue of the colour of the person/people involved.

True. Thanks to Obama and Lerner, audit risk is now based on political party affiliation.

The alternative explanation being that, among the list of numerous left, right and other wing associations audited in relation to political activity, that there were actually more offenders in one camp than the other.

If there was such a higher percentage among the right wing groups with charitable status despite a lot of political activity, then perhaps the correct explanation is that they underinvestigated in that direction and the high rate of lost charitable status was evidence in the opposite direction?

I dunno, as compared to the fairly systematic auditing of environmental and social justice NGOs in Canada despite basically no mention of significant auditing among NGOs, etc., more aligned with the government's agenda, it seems that the situation relating to audits of charitable status and political activity in the USA covered quite a lot of angles across the ideological spectrum and in many other directions as well. There were somewhat more obviously right wing charities audited than obviously left wing charities, but some left wing ones also lost charitable status due to the fact of their primarily political activity.

Right, but they might be triggered by income, itemized deductions, the presence of a Schedule C, etc., all of which might also correlate positively with being white.

You are obviously not aware of the attempts at use of "surname association" by the CFPB to "find" racial "disproportionate impact."

Well intended, of course.

I don't. There was almost no information available (in a quick search) on rates of white collar crime by race - the best I could do was the ancient study I linked to. Apparently the racial demographics of the perps was collected, but not reported other than a sentence which said that most white collar criminals are white. So the proof behind my question is weak. And maybe that's a sign of the racial fears prevalent in our society - when it comes to crimes disproportionately committed by minorities, we have a ton of current data, when it comes to the other kind, not much to work with.

Is there any other kind?

A further control to the study should have been crime rates in the respective areas measured. Policing in higher-crime areas naturally leads to a lower standard for stops; comparison should be made between the higher-crime areas populated by whites and blacks, not between high and low crime areas.

And writers should promote their own work, especially when it explores counterintuitive findings. http://bloom.bg/2aU7Ft2

Were olfactory data among the "omitted variables" of this statistical accounting?

When the actual physical circumstance of a police encounter with the driver or occupants of a motor vehicle entails unambiguous detection of illicit combustion, should police pretend that their olfactory equipment is malfunctioning? Relevant olfactory data seem routinely never to be represented in statistical accounts of police encounters and circumstances of arrest.

Are the circumstances of physical encounters adequately represented in mere statistical accounts?

First, you would have to demonstrate that black people smoke marijuana while driving more often than white people. Which I somewhat doubt, considering that they aren't stupid and know they are far more likely to get pulled over and/or done for in such a situation. But ... maybe? If I'm wrong and you're right, then I think it would explain only a small part of the observed difference.

Tyler- it's a little irresponsible to post a paper like this without reading it or indicating whether you endorse its methods, conclusions, and significance.

Using esoteric statistical techniques to attempt to overcome a lot of known confounding from unrecorded variables (e.g. demeanor during traffic stop, scent of drugs, condition of vehicle, attire of occupants), particularly where there are known disparities in criminal behavior (e.g. murder rates, which are pretty rigorously gathered stats) is a tricky business. It's hard to parse the extent to which this analysis shows anything new or interesting.

Statistical disparities in "thresholds" for stops between races with far lower rates of criminality and those with higher could be justified based on any number of factors not captured in the threshold, but which legitimately do correlate with (and drive reasonable suspicion of) criminal activity.


If the contraband find rates are similar, then isn't this evidence that the higher search rates are wrong?

Also, if there is a higher search rate, I would assume that a higher effort rate per search would also apply. Like, do they check your pockets, glove compartment and trunk, or do they pull the entire car into pieces? 'Cause y'know if you talk like you're from the wrong part of town, maybe they're just primed to give you shit.

Maybe one guy smells a bit of weed, checks your glove box and pockets, and believes you that it's gone and there's no more. But I betch your ass if he was black, the odds of going full anal cavity before realizing what a bunch of twats they are is much higher.

(Incidentally, a few years back this seemed to be a tit for tat strategy between Turkey and Israel - something about ridiculously unnecessary anal cavity searches for certain citizens travelling between the countries. Y'know, this kind of stuff might just not impress people very much. ... - I forget who started that particular round of nonsense, but I think it loosely stemmed from the IDF attack on a humanitarian boat a few years back)

If one group gets searched more and you get the same hit rate then the higher search rate seems completely justified. As you do more marginal searches, hit rate should drop off.

Imagine that you give out detentions for not doing your homework.

You observe that white children are skipping homework half as often as black children. So you decide to check the homework of black children twice as often.

In this simple scenario, we KNOW that black children are only skipping homework twice as often. But, having done our maths homework as children, it is not difficult to see that four times as many black children will end up in detention, despite only twice as many not doing their homework.

According to what you're defending, this is the correct way of things.

You think an 8-year old child cannot see this stuff plain as day right in front of them? I mean, at least when this stuff happened in my school it was "mandatory after school help sessions", and not "punishment".

You see, the way you would have it, a twice as high violation rate would lead to a four times higher punishment rate. In order for twice as much violation to lead to twice as much punishment, you have to search each group at equal rates, not different groups.

'cause when it carries on from kindergarten through all of life, in a group twice as likely to offend but four times as likely to be punished (twice as much punishment per offense), you gotta not be surprised to hear people going around saying things like "you gotta FUCK UP a system like that, yo!" And obviously, cops are going to bear the brunt of it, because most pro-discrimination people don't exactly go around with bumper stickers promoting their perspective.

You do understand the basic math that doubling the search rate on a group with twice the offense rate implies four times as many punished, not the proportional twice as many punished, right?

Which would mean that "punishments per violation" rate would be twice as high for black people.

If I were twice as likely to get punished for the same thing, I would not recognize any legitimacy of a system that would uphold such an injustice. Because the US is built on individual freedoms, and collective punishments are universally despised.

A topic like race and policing always does a good job of reminding me that under most libertarians is a in reality a big-government conservative waiting to run free.

There seems to be no logic behind this statement other than your own biases. You've failed to making any connection with specific Libertarians adopting a "big-government" position. Let along it being a conservative big-government position.

I'm not quite sure what 'Observer' is saying, but I think it is fair to say that academic libertarians nowadays (the Mercatus crew in particular) have sensibilites and biases very similar to ordinary rank-and-file faculty or are at pains not to take stances which lower their status on the faculty. Ergo the attacks on the police, the open borders discourse, the complete indifference to freedom of contract and freedom of association, and various and sundry other curios (by way of example, the most quoted 'Republican' 'economist' on this site is...Bruce Bartlett, an opinion journalist / lapsed congressional aide / lapsed history teacher with almost no academic training in economics but with a 12 year history of savaging Republican politicians).

"I’m not quite sure what ‘Observer’ is saying,"

When I see comments like his, I wonder if I'm not guilty of making the same kind of comments, that seem perfectly 'reasonable' to me, but make no sense at all to someone else. I know I've been guilty of such comments in the past. Observer might have had a good point, but he didn't make any logical connection so it's impossible to say.

Maybe I see a thread of logic. If by big government Observer means aggressive policing, some are supporting that. Same as the people who normally want to reduce the coercive power of the state? I don't know.

Even if it is the same group of people, it's not contradictory to want both smaller Federal government and the current level of local policing. They are not mutually exclusive.

Only two solutions that I see, end the war on drugs and/or much more separation between blacks and whites.

The only real choice is to end the war on drugs. Medicalize it. Treat drug abuse as a medical problem, and not a crime against society. Make nonviolent dealing illegal, but a nonfelony.

Drug abuse is more a social problem than a medical problem.

Use or genuine abuse?

I thought the whole story of seduction and gateway drugs was being undermined, that it was being found more that people with problems self medicate.

Careful though. Depending on how things are defined, that could involve people in white coats with the power, at a stroke of a pen, to effectively incarcerate people against their will. The one most likely to come back on this comment does not seem to be the type to abuse this power, but if there is a power that can easily be abused, it would be dumb to think it wouldn't happen at least sometimes.

The debate about whether it is a medical or social issue is complicated for a few reasons. Not the least of which is that the people with certifications relating to drugs and health are basically the least scientific and most socially-driven of "medical" practitioners.

I think that there can be much value in things which are imperfect science. But not if the people using those tools have deluded themselves into believing that their pseudo-science is in fact science.

Some pathways to medicalizing drug abuse could lead to extremely arbitrary powers on the part of a profession which within the nether regions of the state is often linked to some pretty sick and dark stuff, as compared to other views which suggest that general practitioners should have more liberty to prescribe, discuss and treat about drug-related issues.

Personally, i think the best approach is to have sort of "centralized" outlets in the sense that there are places where you get drugs, and along with the transaction, the buyer always receives materials to promote the ease of reducing usage if this is what the buyer wants.

Not knowing that something is addictive is very dangerous. But telling people "this is sooooo addictive. You'll probably never be able to quit", well, maybe they decide to try, and are already primed to believe that there is only one path forward - MORE!

Like selling untaxed cigarettes? Ask Erik Garner how that worked out.

He had a heart attack, right?

If the war on drugs ends, the black crime rate will soar. Not the reported crime rate necessarily, because it is harder to catch someone for an assault or murder than it is to catch them with crack in their pocket.

Would we also reduce crime catching more white pot smokers?

Depends on the crime rate in the area/demographic. If an area is low crime, a turnstyle jumper is a turnstyle jumper. In high crime areas, a turnstyle jumper often has outstanding warrants.

That sounds like a story, in the sense of Dan Ariely and Behaving Irrationally.

Much bias is based on lazy statistcs, mental shortcuts.

That sounds like projection.

Gosh, you said a magic word, you win the internets.

Neither of you want to prove your points with anything like facts, so you resort to name calling. Well done.

I am pretty done with this page, but seriously MOFO? One guy makes up a story about turnstile jumpers, and the other guy should have ready data in turnstile jumpers?

In the never ending effort to be offended and outraged this is just another stupid example. Since hispanics commit crime at 2-4 times the rate of whites and blacks commit crimes at 8-10 times the rate of whites is it any surprise that blacks and hispanics get stopped more??? I can remember passing a cop car 10 mph over the speed limit in my older station wagon with two kids and my wife in the car. The cop rightly concluded I wasn't worth the effort. On the other hand when I was single and had a red Camaro I would get stopped for nothing especially on Friday and Saturday nights. Cops aren't stupid, they know who to stop and if a reason is needed it may be a trivial reason.

Blacks are perpetrators of certain categories of violent crime at rates around 7-8x that of non-blacks. That doesn't apply to property and drug crimes, for which the ratio is (IIRC) more along the lines of 3x.

If you watch one group of people twice as much as you watch the second group, the first group will "commit more crimes", no?

Compare two situations. 1) You're a drug dealer in a white upper middle class area that is generally safe. You are white. 2) You are a drug dealer in a largely low-income neighbourhood that is generally safe. You are black.

Same crime. Who's more likely to be a) questioned, and if so b) investigated, and if so c) arrested, and if so d) charged, and if so e) brought to trial and if so f) found guilty. Assuming that all other details of the situation are identical?

Don't bullshit me. You know that there is existing racism within the system from ABCDE right through to F. Which doesn't exactly breed respect for the laws that have been established, considering that in the macro sense it is largely a product of the same system and people.

Is there enough racism from A to F to explain the multiples differences you mention? I doubt it. A middle finger in the general direction of the system probably explain a good deal of the rest of it though, although this will become manifest in quite a large varieties of ways and would rarely be the motive in the act of specifically committing some particular criminal act.

The evidence isn't there for that conclusion Nathan. There are many studies on the issue

Considering that we're looking the evidence in the face that more black people are getting pulled over, by much higher margins than the difference in contraband findings would warrant, by definition the group is already larger as a result of discrimination by the time you've gotten through steps A and B.

I dunno. Maybe you have a way to spin that away? Or am I missing something real obvious here?

What cops know is who they can stop and meet their ticket and arrest quotas without there being any blow back. Ironically its the same set of people that they are deathly afraid of so we end up in a more dangerous situation for both parties.

Indeed cops are not stupid, and they are doing exactly what society has asked them to do.

The Manhattan Institute conducted a study of traffic stops in New Jersey some years ago which demonstrated that the racial composition of those stopped was almost precisely equal to the racial composition of the pool exceeding the speed limit by 15 mph.

One does get the impression that studies like this are battlespace preparation in a campaign to destroy one of the few effective policy initiatives of the last 50 years: the partial re-establishment of a measure of order in urban areas, and that the motivation to wreck it is that the re-establishment of order was achieved by methods (hiring police and prison guards and reducing the discretion of judges) which are simply unacceptable to people who verbalize for a living.

I never would have imagined that people would think the "getting pulled over while black" argument could be resolved by a dataset of people pulled over at 15 mph over the limit.

"Did you see the colour of the guy who just blasted by at 15 mph over the limit?"
" No."
"Me neither".
"Well, who cares! He's going 15 mph over the limit. This one will be good for today's quota."

I never would have imagined that people would think the “getting pulled over while black” argument could be resolved by a dataset of people pulled over at 15 mph over the limit.

Well, you don't think very clearly, which is not my problem

You know the race of people who pass you by 15 mph in a vehicle?

Here's an idea. To test for bias, let's create a study condition where obviously bias will be minimal or zero. And then use it to demonstrate that there is no bias. I sure hope the original study had at least a little more discussion relating to non-15mph-over circumstances.

However, it might be worth noting that an absence of a bias where there bias should have also been minimal or zero is actually a pretty good demonstration that the underlying rate of deviant behaviour was similar among groups. Right? It's a pretty big whole in the theory kind of observation with respect to black criminality, but certainly I wouldn't assume that just because they speed at similar rates that therefore the underlying levels of other violations are similarly equal.

You know the race of people who pass you by 15 mph in a vehicle?

The Manhattan Institute has published a summary of the study and it's aims and methodology if you'd care to learn something in between bouts of pontificating.

Interesting stat. For the opposite reason to I think you might have introduced it for though.

You're confused. They were looking at the composition of ALL stops, not just those of people going 15 mph over.

Art referred to the ones going 15 over. So I talked about the ones going 15 over. It's really not very confusing.

"Black drivers in Raleigh are searched at a much higher rate that whites (3.9% vs. 1.9%), but when searched, blacks are also found to have contraband at a higher rate (16% vs. 13%)This suggests that officers can more easily determine which black drivers are carrying contraband, which causes their searches of blacks to be more successful than their searches of whites..... In spite of the higher hit rate for black drivers, we find that blacks still face a lower search threshold (6.4%) than whites (8.8%), suggesting discrimination against blacks."

I don't know if this is discrimination against blacks, or just willful ignorance toward white offenders.

What??? By your stats more blacks than whites where found in violation therefore justifying the stop, how is that evidence of police discrimination? if anything the stats show the instincts of the police officers was fairly accurate. Discrimination would be demonstrated if the rate of contraband found on black detainees was LOWER than whites. The lower threshold for blacks is supported by the higher rate of contraband found on them. Most likely the police recognize factors indicating illegal behavior not captured by the researchers performing the study- a sign of experienced police personnel.

Sounds pretty dumb to me. They search at twice the rate (100% higher), but the difference in what they ultimately found was about 20%. Based on the 16% and 13% accuracy rates, you just cannot possibly tell me that this additional 2% is comprised of those 3% who actually had contraband by virtue of a 70% correct rate for the officer.

The evidence actually points to the contrary. That they are making a much higher rate of mistakes among blacks than whites. Among whites, you have 1.9% searched and 13% correct. Among blacks you have 3.9% searched and 16% correct.

1000 white guys. 19 searched. 2.47 found to have contraband. That's 16.5 guys pissed off at being searched per 1000.
1000 black guys. 39 searched. 6.47 found to have contraband. That's 32.3 guys pissed off at being searched per 1000.

Is it worth pissing off an additional 16 people per 1000 with unjustified searches, in order to find four joints? Police always tell us how the trust of the community (e.g., in this case different racial communities in the US) is important for them to do their jobs. But some people sure seem to get pretty uppity when someone suggests they might try to actually earn that trust.

the correct metric is stop rate X contraband rate ("yield rate")- and for blacks the number was several fold higher- i.e. no prejudice against black drivers. Now you change the argument & worry about "pissing off" people- public relations not prejudice- you may be correct that there will be bad PR but there is still no evidence of racial bias UNLESS the yield rate for blacks is less than whites. Police look to prevent crime therefore they will error in favor of making a few mistakes that inconvenience people rather than miss crimes. The NYPD earned that trust by preventing the deaths of thousands of minority males over the last 25 years- hard to overlook saving lives isn't it? Sure there are mistakes & bad cops- but thousands of living young men are nothing to sneeze at.

If you're stopping twice as many of one group, but the yield rates are not very different, then that is discrimination.

The metric you suggest, I could search 100% of blacks and 1% of whites, and so long as the yield was marginally higher among blacks. Taking it to the logical extreme should suffice to demonstrate that your reasoning doesn't work that well. I don't think I've hit upon the best figure quite to focus on, but I do not believe your suggestion would be an improvement, because, just look, it is consistent with a police state life for an entire group and no troubles for another, on the basis of a marginal underlying difference. But I don't imagine you were quite thinking of it that way.

Several times higher is very different

Twice as high is quite a lot.

Imagine that you belonged to that group where twice as much harassment applied. People being what they are, do not generally respond positively to such a situation.

twice is a lot but not as much as the difference in rates of crime by race- which is a lot more than twice

In pretty much any case where there is a racial disparity in some law enforcement or judicial statistic, there is also a gender disparity that is an order of magnitude larger than the racial disparity. Everyone understands the reasons for gender disparity in things like police shootings or incarceration, and it doesn't trouble anyone.

I'm not clear it's there. I provided a hypothetical above. A cop decides to start pulling over male drivers and visually inspects their knuckles. If they appear bruised or cut she will call in for a cop to visit the driver's home and see if his wife or gf was beaten. What's wrong with this?

In terms of domestic violence, men more often beat their wives than wives beat their husbands. In terms of violence the male rate of inflicting violence is much higher than the female rate.

On the flip side a cop will stop a lot of innocent men and even those with bruised knuckles will have them for reasons other than recently attacking someone. Yet I'm sure if any police department tried this there's be a huge uproar. Yet the same logic happens with a strategy of trying to pull over black motorists on the theory that they have a higher rate of committing crime. They do but that doesn't change the fact for every one black criminal you might catch you'd still be jamming up a lot of innocent black drivers.

If police are trying to detect and stop violence, then it makes sense to focus on men more than women as men are more prone to violence. You are constructing needlessly convoluted hypothetical to confuse the obvious fact that police will focus on who they think are the most likely to be guilty of a crime.

"most likely to be guilty of a crime" means you're using crime rates which means randomly screening all men for evidence they were recently involved in violence makes just as much sense as pulling over black motorists "just because".

If 10 minutes ago a woman got mugged a block away then yes stopping every male or every young male may make sense given odds. Stopping males in general, however, doesn't make sense even though they commit crimes at a rate far higher than blacks do.

You are assuming traffic stops are supposed to be a type of screening investigation when in fact they aren't. They are supposed to simply be traffic stops for traffic offenses and discovery of other types of crime are supposed to be incidental to that.

No one is suggesting pulling over black motorists "just because"

I said this in the comments no more than 2 or 3 days ago to Mulp.

A lot of what is obvious has been said already (regarding different crime rates for different groups being intrinsically linked to higher stop rates). What I'd be really curious to know is how far the so-called "equality warriors" would be willing to take this kind of rationale. Should we say it is discriminatory to have more police in more violent areas? How about number of police per capita? Spending on police per capita?

I might be exaggerating here but it seems to me that *a lot* of the background in this discussion here is based on the idea that higher crime rates are caused by additional policing. That is, some people believe that whites and blacks are committing as much crime as the other, but police is only catching the black criminals... Sounds completely absurd to me (you can look at the murder rate for instance) but these folks really seem to believe that!

What would take these people to see this is absurd? Maybe there is no way. See what happened in Milwaukee. Guy killed had a loaded gun, and riots are still happening. People want to fight reality. Really tough to imagine what we can do about this. Get rid of tenure?

As I ask above, men do far more violent crime than woman. Suppose a cop decided to start randomly pulling over men and searching their bodies for evidence they recently committed some crime of violence. I'm sure such a program would be stopped rather quickly.

Your 'rate of crime' justification IMO only works if you are talking about actual valid reasons to pull someone over (i.e. speeding, making an illegal turn, etc.). If you're talking about some generic 'crime rate' then what does that have to do with pulling people over? Should a cop randomly pull over upper class white men because they have a 'higher rate of insider trading' or 'tax fraud'?

In general I think the police already does gender profiling, and I think it is totally appropriated. Just like age profiling is also appropriated. We must have equality in regards to what is considered a crime, but policing cannot be equal. Policing is reactive by nature, it must follow society trends and characteristics. To try to fight that based on "equality" is not only ineffective, it is ultimately unfair to the law abiding citizens who live in high crime areas.

If pulling over random upper class whites actually lead to a meaningfully higher conviction rate for insider trading or tax fraud, then sure, why not. Counter question, do you think the FTC should investigate poor blacks for insider trading in proportion to their race or do you think its reasonable to focus on people who are likely to be engaged in insider trading, irrespective of their race?

I am 100% certain that the FTC does profile on their investigations. Just like the IRS does! I have not received my refund yet, and I bet this is directly related to the amount of such refund, previous returns, income level, etc. The fact that bothers people is that profiling by traffic stops only helps catching certain crime (drugs, outstanding warrants, etc.). No one really is against profiling, so much so that we don't even call it that in many cases.

I suspect few if any insider trading convictions are generated by profiling. I think most or all of them are generated by evidence that something is suspect or by those who are already caught giving up others they know who are involved.

I don't generally assume that driving is an activity that I highly associated with dealing drugs. So, using your logic, it is a piss poor way to catch people dealing drugs.

I suggest paying attention to the guys on the street calling out "weed, weed" to passers by and then trying to figure out whether their supply is a major criminal network or just some homegrown nonsense down the road. Simple, stupid. And anyways, what will the cops always tell you? Those aren't even the guys they want - they want important middle men, king pins, the hit men, etc. Which you will never ever ever in a million years get by traffic stop discrimination.

Yes, indeed. Let's focus investigative efforts where the crimes are committed. Dragnet approaches have always been dumb and a threat to liberty, and always will be dumb and a threat to liberty.

(Cute example by the way. But audits are generally triggered for very specific reasons, as compared to the often arbitrary application of police powers.)

I think you are indeed, exaggerating. There is a viewpoint that the policing itself drives up crime rates and, to be honest, I haven't fully considered the merits of that argument. That said, I don't think most people are as entrenched in this view as you suggest. What I find, anecdotally, is that progressives see the root cause of the crime rate gap as being uneven school quality, discrimination in employment, lending, etc. and general social stigmatization of blacks. Basically, that blacks (and hispanics) are somewhat of an underclass because the dominant White/Asian culture seeks to maintain a dominant position in the social hierarchy.

It's not so much fighting reality as just viewing it through a very different frame. Some see systemic racism very vividly and others just don't. Most acknowledge that there is no smoking gun. No overtly racist official policy. Rather, we have a cascading of small biases and preferences that add up, or don't. It's hard for people to avoid talking past each other on these points when they both see the other as denying the plain reality.

The most extreme version of the progressive case comes from the far left / Marxist camp. Basically, they see material inequality as the fundamental social ill. Furthermore, wealth is acquired primarily via inheritance and/or exploitation/cheating. The job creators of right wing lore are exceptional. Racial disparities are a consequence of this basic material inequality. The only solution, in their mind, is radical wealth redistribution and collectivization of production. There are indeed softer versions of this stance (social democracy) as well. It's an increasingly influential framing nowadays. Again, when you are mentally trying to support a very grand theory of everything, that tends to influence how little bits of evidence are interpreted on key points. A lot of people see some key parallels in gender, race, and wealth inequality and tend to overreach. Self-righteousness is a factor here.

I'm being a bit tough on this point of view here. I think the conservative framing is equally problematic, but in very different ways. I think there is some merit in each mode of thinking but not enough listening by either side.

If you put twice as many cops on patrol tonight in a given city, I would bet good money that the number of phone calls they receive to report crimes or problems will be lower. (The stat on how many arrests will be misleading of course, because, no shit, twice as many cops on patrol gets you more arrests.)

That night, crime rates are not driven up. In fact, you might even be fine if you double all patrols evenly throughout the city.

The problem is not the quantity of patrols or interactions. It's that time, and again, and again, when you're out shooting the shit with some people, and recounting that story about how you've gotten carded for the 10th time, pulled over for no reason for the fifth time, and all the black guys are nodding like the know exactly what you're talking about, and all the white guy can do is shrug his shoulders and say "Sorry guys. I didn't do it, but I never get that. Ever."

When you know that as a result of discrimination, you are more likely to be questioned, investigated, arrested, etc., even when innocent,

A lot of people here seem to be optimizing from the perspective of having a maximum number of convictions for a minimum amount of police effort. Mathemtically, the argument is sound. Until math meets people, with the result of proliferating views that it is acceptable to disrespect or even kill police.

I used to have a neighbour who was a cop. He was a light blue button t-shirt with a collar, dark blue pants with a single stripe down each leg, black shoes which looked nice but are good for walking/running in, and sometimes sunglasses. That was 20 years ago. There are not more guns on the street, but today, cops rarely exit their vehicles, and when they do they are dressed for war and treat people like subjects, not citizens.

Cowen should translate this into laymen language. And explain how this addresses the obvious counter points you would expect fromt his crowd. How does this study show that more police searches by race, are not cause by different behaviors by race?

Not sure about the language here but I'd argue you could look at it by looking at the rates where searches come up with something.

Say blacks drive around with drugs 2x as often as whites. Let's say the rate for whites is 5% and blacks is 10%. OK so if police using 'hunches' that are not racially biased search black cars 2x as much as white cars, they should achieve a 'capture rate' that equals whites then the ratio of 'hits per search' should be equal.

Say a cop hearing the meme 'blacks are more likely to carry drugs' decides he will simply search every black car he encounters and no white cars. He will score a lot of black arrests for drugs but he will log a huge number of searches per actual arrest. Say a different cop is much more savy but unbiased in his hunches. He will search black cars more often than whites but the # of searches in order to score an arrest for him should be the same.

Think of a search as a cost. If you stop someone 45 times turning up nothing then you're wasting a lot of resources. An increased rate of doing something illegal only justifies doing more searches based on race to a very small degree. After that it becomes wasteful.

And a search that results in nothing should be considered a big cost. Not only do we have to pay the cop and his support system to do a search but it violates the freedom of an innocent person, likely makes him angry and annoyed and increases disrespect for law enforcement. A police style that results in hundreds of stops that result in either no valid crimes or only trivial ones should be viewed as worse than any onerous tax rate!

I think you are assuming a skill that is simply not existent. In most cases, cops will find drugs randomly. So using your example above, it is logical and effective for a cop to stop more blacks than whites, since the percentage of drugs captured will be higher on average.

1. The purpose of traffic stops is traffic law, not to find illegal drugs. Should a cop audit someone's tax returns to find tax fraud when he pulls them over for speeding too? I understand if the drugs are obvious when he pulls someone over (i.e. crack pipe on passenger seat) or if behavior leads him to a reasonable search (i.e. guy keeps nerviously looking at glove compartment).

2. The 'random' aspect only makes sense to a degree. A cop wants to avoid doing a search that doesn't turn up anything (this annoys the person and wastes everyone's times) and only do searches that turn up something. So the metric then would be to try to get the ratio of misses to hits as low as possible. If blacks carry drugs, say, more often than whites that will lead to more stops 'at random' but only to the degree they carry it more. Going beyond that is not justified randomly.

In other words if blacks do something 2x as often than whites then maybe it would make sense to have the odds of a search be 2x as great all else being equal. If however a rookie cop gets it his head "twice as often means I should search all blacks I find while ignoring whites" will result in more bad searches per positive result.

This is using semantics to either prove your point or to enforce your ideology. The goal is policing against crime. If a cop stops a car and suspects another crime, obviously he has the right to do so. It is only logical that the cop needs to keep law enforcement in general in his mind while patrolling. Just because you are stopping cars it does not mean you are only concerned about car related crime! If tax audits were somehow identifiable when you stop someone you would definitely use that as part of the equation (see just how illegal immigration status has come into play recently).
Of course you want to have the hit ratio high, the problem is the assumption that you have this fine tune level of control you assume. Stopping a law abiding black guy is a lot worse than stopping a white criminal, no one in their right mind is arguing about that. However, this is a not a videogame. You cannot account on the fly how many whites and how many blacks you stopped. You have neighborhoods who are more / less diverse. There's also the fact that, if a neighborhood that is predominantly back is a lot more dangerous than a white one, that one will (and should!!) get more policing. So the math here is not directly related to profiling by itself, but to crime and police resources.

If you search 2 times as many because the underlying rates is twice as high, then you will catch four times as many, not the proportional twice as many.

The only way to have punishment in the same proportion to incidence is for the search rate to be the same.

Usually, notions of justice revolve around "similar punishments for similar crimes", and at the collective level with your proposed strategy, you get twice as much punishment per unit of crime. Some people would consider that an outrageous violation of justice - one group getting punished twice as much per unit of offense.

An increased rate of doing something illegal only justifies doing more searches based on race to a very small degree.

Race is not the only factor. Youth, poverty, gender, observed behavior, etc.

Yes, and we should profile based in all of those.

And then scratch our heads, perplexed by the very confusing situation that those same people who were discirminated against and searched more often had higher rates, and in time lost all respect for the law and society as a whole.

I don't think you guys understand how much the "fuck the system, fuck everybody, especially anyone who has anything to do with the system" kind of mindset dominates, when the system Fs you up for no GD reason. To those who don't know personally, just admit it, you don't have a bloody clue.

There's a better way, and let me say that a million times over, but when I hear about cops getting shot, I think "well, glad that someone is standing up for themselves, but it would be better if it weren't done in a way that will further legitimize police abuses". I do not support it. But if you understand how it can come to that, you understand it, and if you do not, then you should be listening, not speaking.

There's a reason that "FUCK THE POLICE!" is among the most commonly used daily expressions once getting beyond those in the range of "y'alls", "yo", "bro", "ni**er" and the like. And in America, the special dynamics of this goes back generations.

"In the course of conducting traffic stops, officers have discretion to search motorists for drugs, weapons, and other contraband. "

Wait, what?

No they don't. I have to exhibit probable cause first. In my experience, this is being followed.

I believe the higher rate of search is due to the higher incidence of displaying probable cause.

Good point. Many people consent though

A lot of people here are saying 'higher incidence' justifies higher search rates therefore everything is find. No it doesn't.

Say whites carry contraband in their cars 5% of the time and blacks do 10% of the time. If those were the only two facts then it wouldn't make sense for a cop to search all blacks. Lacking all other information a rational cop would toss a 20 sided die and search a white car if it lands on 1 (5% of the time randomly) and a black car if it lands on 1 or 2 (10% of the time randomly).

What would happen if a rookie cop heard the metric about blacks being twice as likely to carry contraband and decided to search black cars 100% of the time but white cars 0%? How would you tell there is something off in his method?

You would look at his failure rate. The cop who did that would have a higher failure rate than a cop who followed the first method. Even if you didn't know for sure how many white cars and how many black cars carry contraband, you would still be able to compare the first cop to the second and see by failure rates that the second cop's method is less efficient than the first. What counts here is viewing the # of failures as a cost. A cop who racks up 100 failed searches but gets 5 goods ones is nowhere near as good as the cop who does 30 failed searches and 3 good ones. While first cop can claim he is breaking a sweat doing 70 more searches than the other cop and managed to get 2 more criminals than the other cop, he has effectively imposed on society 100 bad searches which means 100 innocent people jammed up. What the second cop did was save society 70 bad searches and only inflicted 30 bad ones on us.

View the bad searches as a cost and you can view the first cop is 'selling' captured criminals at a price of 20 bad searches while the second is selling at a price of only 10.

View the metrics right and they'll tell you everything. I suspect a lot of policing in black communities is done with the premise that failed stops/searches are free therefore the first method is better.

Even if you have higher incidence in black populations, that should only slightly increase the chance of a search for any particular stop (as well as the chance of a stop to begin with absent anything clearly observable). Anecdotal stories about individuals with not obvious criminal background getting stopped dozens of times therefore could not be justified by any increased rate of incidence observed objectively in any world we exist on.

You seem to be going hard after a complete strawman that cops only search blacks and never whites which is not at all the case

It's not a strawman, I'm comparing two extreme possibilities. The perfectly balanced cop who only does random searches in exact proportions as different types of people offends versus a totally incorrect cop (who perhaps has read too many comments here that imply things like 'if blacks commit crimes 2x as often as white that justifies any amount of additional search rates). Few cops would be perfectly good or perfectly bad but by establishing the extremes we could begin to develop the metric needed to measure where they are and establish what moves improve things versus making them worse.

Does this paper not exist to argue that the lack of substantially lower hit rates does not diaprove unfair bias? Your hypothetical apparently doesn't exist, so if you were making an honest argument, apparently the status quo of much higher search rates but close hit rates is acceptable to you?

As interesting as the finding is, in terms of the general conclusions some might be tempted to draw from it, the way you put it is quite clear in that actually we cannot conclude anything particularly strong here. What does it mean? Not obvious. If you assume that cops are perfectly unbiased, then the results will tell you that blacks get what they deserve. If you assume that some are at least somewhat biased, then the results simply give you a hint at what this means on the street.

Here's the unadjusted data from page 16:

Race Search Rate Hit Rate
White 3.1% 32%
Black 5.4% 29%
Hispanic 4.1% 19%
Asian 1.7% 26%

This doesn't look very dramatic to me. In the case of severe bias, I would expect radically and consistently lower hit rates. The search rate for blacks is elevated, but the hit rate is only slightly lower than the white rate.

Maybe you are not understanding the "infra-marginality" issue the paper is designed to address?

As I understood it, they presented a hypothetical example showing that you could have different (i.e., discriminatory) thresholds for search and yet still have the same hit rate (if you have sufficiently different "signal distributions"). Okay. But I always anchor in the straight numbers. The white search rate is 3.1% and the black rate is 5.4%. I don't know what the black rate *should* be (whatever that even means), but it's not out of the question that it might legitimately be higher.

Here is what they say:

"In this hypothetical world, the search rate of black drivers would drop to 3.5% (compared to the existing search rate of 5.4%), in line with the 3.1% search rate of whites. We estimate that the search rate of Hispanic drivers would drop even further, to 1.9%, below that of white drivers. It bears emphasis that this exercise is inherently speculative, and while useful for understanding the magnitude of the threshold differences we find, it is difficult to accurately predict what would have happened under such a counterfactual scenario"

So according to their "inherently speculative" exercise the black search rate should only be 3.5%, practically the same as the white rate. Hmm. Which would almost certainly mean a way higher hit rate for blacks (which they offset with the signal functions). I can't critique all the technical details, but I don't really believe the 3.5% and even using that as a worst case we're talking about 1.9%. I was expecting something a lot more dramatic to drive Tyler into a depression.

Say 2.5% of whites carry contraband and 5% of blacks do.

Let's say there's 1,000,000 black drivers so that will mean there's 50,000 carrying contraband. OK so let's say a cop randomly pulls over 1000 in a year. What would happen if he searched all of the cars? Well he would find about 25 hits if those are white drivers or 50 hits if they are black, most likely it will be between 25-50 since a cop will have a mix of black and white drivers assuming he is going to pull over drivers as he finds them and not do something stupid like pull over any and all blacks he sees while ignoring all whites no matter what.

But searching 1000 cars takes a lot of time. If the cop does fewer searches he has time to pull over more cars. Assume an analogy here with fishing, the cop only cares about scoring catches and nothing else.

Some ponds have more fish than others however it doesn't follow that a rational fisherman only fishes in the pond with the most fish. For one thing a pond that's very dense in fish but has a lot of other fishermen fishing it may end up being harder to catch fish than a pond that's less dense but has fewer people fishing it.

So my thinking is this, if you are 'overfishing' a profile type then what you are going to see is a higher search rate but a lower 'hit' rate. If blacks do carry contraband more often than whites then the NC data doesn't make sense unless you cops are 'overfishing'. Experience and word of mouth has told them blacks have a higher rate of criminal behavior but they overestimated how much that will result in 'hits' if they search blacks more often. As a result blacks get searched more often and more often the search turns up nothing.

Game theory also says blacks that are involved in moving around contraband should respond to these incentives. They would almost certainly respond by taking steps to minimize 'riding dirty' (taking fewer trips with contraband, driving more carefully when doing so, recruiting whites and other non-profiled groups where possible).

The problem with defenders of profiling is that they are often so proud of themselves for being level headed rationalists who reject political correctness in favor of hard reality that they miss the reality of their own logic.

I'm thinking the data on page 16 does show the problem.

First blacks make up about 40% of the stops. NC blacks make up about 20% of the population so stops are represented twice as much as you might expect. I'm not sure what portion of NC drivers' licenses are issued to blacks so it might be even more than twice.

Second, we've been told that blacks commit crimes at twice the rate of whites. But if that's the case wouldn't the 'hit rate' on searches be higher for blacks? Instead they are lower.

This would be consistent with cops overstopping and oversearching blacks. Perhaps a higher rate of drug use or drug carrying initially justifies a higher stop/search rate but if you take that too far you'll actually end up with a hit rate that is lower. If you take the idea 'blacks commit more crime' too seriously you'll end up stopping many blacks and you'll end up with a low hit rate.

Blacks are suffering about 70K failed searches per year in NC (leave aside the issue of 'failed stops' where the stop wasn't really necessary to begin with but that's harder to objectively measure). Whites are suffering about 49K failed searches despite being much more of the population and more of the initial stops.

So look at it this way. We have combined about 120K failed searches. That's wasted time on the cops behalf but that's also a type of tax. It's probably quite nerve racking to have an encounter with a cop that goes bad and results in a full search of your car when you have no contraband on you. It's also likely to lower your respect for police and increase the chances of assuming they are untrustworthy.

Whites make up 75% of the population yet blacks are paying about 60% of that 'tax'.

So here's the problem, you have concerns that blacks have a higher crime rate yet you are supporting a policy that decreases the police's ability to be effective in black communities.

How about zero-yield searches per 10,000?

You search 200 white people and 400 black people. So you get 200*0.13=26 busted white guys and 400*0.16=64 busted black guys.

Mind you, the actual difference is 13% and 16%, but now we're comparing 16/13 = a 23% higher rate of contraband among blacks to a 64/26 = 146% higher rate of potential convictions (assuming there are no further disproportions).

Not only does this practice lead to 146% higher conviction rates (like those stats you hear .. "black commit twice as many of such and such .." despite only 23% higher underlying deviance - per 10,000, now you've got the remaining 336 pissed off black guys and the remaining 174 pissed off white guys.

With a bust rate differential of five times higher than the underlying differential in violations, and about double the rate of blacks being innocent and violated, well what, do you want them to bend over and hand you something long and rounded at the end? Or maybe there's a reason they aren't too cool with that situation.

Comments for this post are closed