A statistical analysis on Driving While Black

[Jarhyn]

https://www.economist.com/graphic-d...-in-america-face-discrimination-by-the-police

To assess racial disparities in police interactions with the public, we compiled and analyzed a dataset de- tailing nearly 100 million municipal and state pa- trol traffic stops conducted in dozens of jurisdictions across the country—the largest such effort to date. We analyze these records in three steps. First, we measure potential bias in stop decisions by examin- ing whether black drivers are less likely to be stopped after sunset, when a “veil of darkness” masks one’s race. After adjusting for time of day—and leverag- ing variation in sunset times across the year—we find evidence of bias against black drivers both in highway patrol and in municipal police stops. Sec- ond, we investigate potential bias in decisions to search stopped drivers. Examining both the rate at which drivers are searched and the likelihood that searches turn up contraband, we find evidence that the bar for searching black and Hispanic drivers is lower than for searching whites. Finally, we examine the effects of legalizing recreational marijuana on policing in Colorado and Washington state. We find evidence that legalization reduced the total number of searches conducted for both white and minority drivers, but we also find that the bar for searching minority drivers is still lower than for whites post- legalization. We conclude by offering recommenda- tions for improving data collection, analysis, and re- porting by law enforcement agencies.

Saw this on Reddit, figured there would be a few posters here who would like to chew on this for a while.

 

[Loren Pechtel]

The actual study: https://5harad.com/papers/100M-stops.pdf

I note a couple of problems:

1) As always, no attempt to address socioeconomic bias. Cops tend to go where the hunting is good (I have yet to see a cop staking out a location where there isn't some problem that causes an increased rate of traffic violations. Also, if the problem gets fixed they cease to stake out that location)--and the hunting is on average better in poorer areas. Poorer areas are disproportionately minority, also--hence a bias is to be expected.

2) Minorities are more likely to be stopped because of warrants. There's an automatic search as part of an arrest even if there is no reason to suspect contraband. This will cause searches of minorities to on average have lower hit rates. If you want to look at how accurate cops are at choosing to search you have to restrict your data to cases where the cops actually made the search decision. The study does not discuss this factor at all.

3) The veil-of-darkness data isn't so obviously flawed but there's a big swing in rates as the time of darkness changes. This is as big as the effect they supposedly found and is completely unexplained. I have a hard time attaching much importance to it.

 

[Testy Calibrate]

The actual study: https://5harad.com/papers/100M-stops.pdf

I note a couple of problems:

1) As always, no attempt to address socioeconomic bias. Cops tend to go where the hunting is good (I have yet to see a cop staking out a location where there isn't some problem that causes an increased rate of traffic violations. Also, if the problem gets fixed they cease to stake out that location)--and the hunting is on average better in poorer areas. Poorer areas are disproportionately minority, also--hence a bias is to be expected.

2) Minorities are more likely to be stopped because of warrants. There's an automatic search as part of an arrest even if there is no reason to suspect contraband. This will cause searches of minorities to on average have lower hit rates. If you want to look at how accurate cops are at choosing to search you have to restrict your data to cases where the cops actually made the search decision. The study does not discuss this factor at all.

3) The veil-of-darkness data isn't so obviously flawed but there's a big swing in rates as the time of darkness changes. This is as big as the effect they supposedly found and is completely unexplained. I have a hard time attaching much importance to it.

One of the two of us missed the point of the veil of darkness issue. Can you explain what you mean?

 

[ronburgundy]

The actual study: https://5harad.com/papers/100M-stops.pdf

I note a couple of problems:

1) As always, no attempt to address socioeconomic bias. Cops tend to go where the hunting is good (I have yet to see a cop staking out a location where there isn't some problem that causes an increased rate of traffic violations. Also, if the problem gets fixed they cease to stake out that location)--and the hunting is on average better in poorer areas. Poorer areas are disproportionately minority, also--hence a bias is to be expected.

2) Minorities are more likely to be stopped because of warrants. There's an automatic search as part of an arrest even if there is no reason to suspect contraband. This will cause searches of minorities to on average have lower hit rates. If you want to look at how accurate cops are at choosing to search you have to restrict your data to cases where the cops actually made the search decision. The study does not discuss this factor at all.

3) The veil-of-darkness data isn't so obviously flawed but there's a big swing in rates as the time of darkness changes. This is as big as the effect they supposedly found and is completely unexplained. I have a hard time attaching much importance to it.

One of the two of us missed the point of the veil of darkness issue. Can you explain what you mean?

I think he's referring to the fact that the % of stops with black motorists goes up as the time of sunset gets later. This could be related to the time of year (later sunsets = Summer) and weather conditions that impact traffic stops. In unpleasant weather, either too cold or too hot, cops would be less motivated to leave the comfort of their car, and thus would be more likely to let minor infractions go. If racial groups differ in the types/severity of infractions they tend to commit, then the proportion pf blacks stopped would vary with factors (such as season) that impact what cops are willing to pull someone over for.

This same explanation might also relate to the day/night differences. Serious violent crimes occur far more often at night, as does gang activity and the probability of a vehicle with multiple young males, regardless of race, being up to no good. Thus, traffic stops are more likely to turn into something more serious (and thus worth a cops trouble) at night. This would make cops lower their threshold for traffic stops at night, which would increase the % of stops involving minor infractions that cops sometimes ignore.
If racial groups differ in the types/severity of their infractions, then they would differ in their relative proportions of stops in day night.

All that said, the burden is on these alternative explanations to show that the types/severity of infractions varies with race and time of day in a pattern that could account for the observed day/night difference. Otherwise, visibility of race to the cops is a rather parsimonious explanation.

 

[Jarhyn]

The actual study: https://5harad.com/papers/100M-stops.pdf

I note a couple of problems:

1) As always, no attempt to address socioeconomic bias. Cops tend to go where the hunting is good (I have yet to see a cop staking out a location where there isn't some problem that causes an increased rate of traffic violations. Also, if the problem gets fixed they cease to stake out that location)--and the hunting is on average better in poorer areas. Poorer areas are disproportionately minority, also--hence a bias is to be expected.

2) Minorities are more likely to be stopped because of warrants. There's an automatic search as part of an arrest even if there is no reason to suspect contraband. This will cause searches of minorities to on average have lower hit rates. If you want to look at how accurate cops are at choosing to search you have to restrict your data to cases where the cops actually made the search decision. The study does not discuss this factor at all.

3) The veil-of-darkness data isn't so obviously flawed but there's a big swing in rates as the time of darkness changes. This is as big as the effect they supposedly found and is completely unexplained. I have a hard time attaching much importance to it.

One of the two of us missed the point of the veil of darkness issue. Can you explain what you mean?

I think he's referring to the fact that the % of stops with black motorists goes up as the time of sunset gets later. This could be related to the time of year (later sunsets = Summer) and weather conditions that impact traffic stops. In unpleasant weather, either too cold or too hot, cops would be less motivated to leave the comfort of their car, and thus would be more likely to let minor infractions go. If racial groups differ in the types/severity of infractions they tend to commit, then the proportion pf blacks stopped would vary with factors (such as season) that impact what cops are willing to pull someone over for.

This same explanation might also relate to the day/night differences. Serious violent crimes occur far more often at night, as does gang activity and the probability of a vehicle with multiple young males, regardless of race, being up to no good. Thus, traffic stops are more likely to turn into something more serious (and thus worth a cops trouble) at night. This would make cops lower their threshold for traffic stops at night, which would increase the % of stops involving minor infractions that cops sometimes ignore.
If racial groups differ in the types/severity of their infractions, then they would differ in their relative proportions of stops in day night.

All that said, the burden is on these alternative explanations to show that the types/severity of infractions varies with race and time of day in a pattern that could account for the observed day/night difference. Otherwise, visibility of race to the cops is a rather parsimonious explanation.

Actually, the veil of darkness data is about a phenomenal wherein the lighting conditions veil the race of the person driving the car.

This allows the researchers to track the difference between police behavior where they can reasonably be expected to have known the race of the driver vs when they are (literally) colorblind.

The study shows that during these particular time periods, the racial disparities more closely echo population data, indicating that cops are engaging in racial profiling.

As to LP's gripe about how the data swings a lot in this particular measure over the course of time, I might remind the forum that there are, in fact, times of the year where the lighting conditions that lead to "the veil of darkness" are more or less obfuscated by cloud cover.

This particular part of the study is the most interesting, I think, and it heavily implies that the data indicates much more than simple "socioeconomic" factors.

 

[Loren Pechtel]

The actual study: https://5harad.com/papers/100M-stops.pdf

I note a couple of problems:

1) As always, no attempt to address socioeconomic bias. Cops tend to go where the hunting is good (I have yet to see a cop staking out a location where there isn't some problem that causes an increased rate of traffic violations. Also, if the problem gets fixed they cease to stake out that location)--and the hunting is on average better in poorer areas. Poorer areas are disproportionately minority, also--hence a bias is to be expected.

2) Minorities are more likely to be stopped because of warrants. There's an automatic search as part of an arrest even if there is no reason to suspect contraband. This will cause searches of minorities to on average have lower hit rates. If you want to look at how accurate cops are at choosing to search you have to restrict your data to cases where the cops actually made the search decision. The study does not discuss this factor at all.

3) The veil-of-darkness data isn't so obviously flawed but there's a big swing in rates as the time of darkness changes. This is as big as the effect they supposedly found and is completely unexplained. I have a hard time attaching much importance to it.

One of the two of us missed the point of the veil of darkness issue. Can you explain what you mean?

The veil of darkness is supposed to hide the race of the person being stopped and thus prevent a racial bias in who to stop. On the surface it seems quite reasonable. The problem is they have three graphs showing the effect at different times--and getting very different results for the three times.

 

[Loren Pechtel]

The study shows that during these particular time periods, the racial disparities more closely echo population data, indicating that cops are engaging in racial profiling.

As to LP's gripe about how the data swings a lot in this particular measure over the course of time, I might remind the forum that there are, in fact, times of the year where the lighting conditions that lead to "the veil of darkness" are more or less obfuscated by cloud cover.

This particular part of the study is the most interesting, I think, and it heavily implies that the data indicates much more than simple "socioeconomic" factors.

Your comment about lighting makes no sense. This is data over several years, weather will average out. What they did there was look at stops before and after sunset--the same lighting conditions in each chart. By looking at it as the time of sunset changes that removes most time-based effects from the data.

It certainly implies there's something going on. The problem is we have a completely unidentified factor that's changing the ratio considerably and isn't explained by the cops being racist. I have a problem with trying to draw conclusions from the data given this big unknown sitting there, though.

(And note, also, that the veil of darkness hides other things than just the race of the people in the car. Thus seeing a change from hiding the occupants doesn't prove racism, it only suggests it.)

 

[ronburgundy]

I think he's referring to the fact that the % of stops with black motorists goes up as the time of sunset gets later. This could be related to the time of year (later sunsets = Summer) and weather conditions that impact traffic stops. In unpleasant weather, either too cold or too hot, cops would be less motivated to leave the comfort of their car, and thus would be more likely to let minor infractions go. If racial groups differ in the types/severity of infractions they tend to commit, then the proportion pf blacks stopped would vary with factors (such as season) that impact what cops are willing to pull someone over for.

This same explanation might also relate to the day/night differences. Serious violent crimes occur far more often at night, as does gang activity and the probability of a vehicle with multiple young males, regardless of race, being up to no good. Thus, traffic stops are more likely to turn into something more serious (and thus worth a cops trouble) at night. This would make cops lower their threshold for traffic stops at night, which would increase the % of stops involving minor infractions that cops sometimes ignore.
If racial groups differ in the types/severity of their infractions, then they would differ in their relative proportions of stops in day night.

All that said, the burden is on these alternative explanations to show that the types/severity of infractions varies with race and time of day in a pattern that could account for the observed day/night difference. Otherwise, visibility of race to the cops is a rather parsimonious explanation.

Actually, the veil of darkness data is about a phenomenal wherein the lighting conditions veil the race of the person driving the car.

This allows the researchers to track the difference between police behavior where they can reasonably be expected to have known the race of the driver vs when they are (literally) colorblind.

The study shows that during these particular time periods, the racial disparities more closely echo population data, indicating that cops are engaging in racial profiling.

I realize that is one possible interpretation of the data, which I acknowledged is a parsimonious one. It is some of the more compelling evidence of skin color itself being a factor in stops, but still not without it's flaws and alternative interpretations, one of which I presented. Namely, the probability of finding drugs/guns/warrants/etc. goes up among nighttime drivers, so cop's are more willing to pull over drivers for minor infractions that they often let go in t the daytime. If the black/white ratio for serious infractions is higher than minor infractions, then probability of a person who gets pulled over being black will be lower at night than in the day.

As to LP's gripe about how the data swings a lot in this particular measure over the course of time, I might remind the forum that there are, in fact, times of the year where the lighting conditions that lead to "the veil of darkness" are more or less obfuscated by cloud cover.

I don't see how that addresses the issue. Blacks are more likely to get pulled over in the Summer than Spring, regardless of weather it is day or night. Cloud cover doesn't matter at night, so it's not likely relevant. Also, they exclude dusk times, and cloud cover 90 minutes before dusk does not conceal the race of drivers.

This particular part of the study is the most interesting, I think, and it heavily implies that the data indicates much more than simple "socioeconomic" factors.

Possibly, but the data don't directly implicate race as the causal factor, only a correlate. It is purely an assumption, albeit a plausible one, that visibility of the driver's race plays a role. But it's also plausible that driver behavior (which may be correlated with race) interacts with time of day factors related to varying thresholds cops have for making stops. Plus, the type of car, it's condition, it's "flare", it's number and age of passengers, etc. are all also less visible at night and are things that impact traffic stops. If any are correlated with race due to SES factors or gang membership, etc., then that could account for the "veil of dark" results as well.

I don't doubt the plausibility of your interpretation, but the data presented are not the smoking gun you're painting them as.

 

[Jolly_Penguin]

Pulling people over based on race seems to me another case of judging individuals by the groups they belong to and charicatures they associate with the group.

It is the same thought process as in the racial discrimination in college admissions. Black people are presumed to be this or that because they are black.

 

[repoman]

I bet the cops are also thinking black and hispanic drivers have less of chance of spending a lot of money on good lawyer for legal proceedings.

Imagine a cop going to an extremely rich area and pulling over lots of people pulling out of their driveways. That is a career deathwish.

 

[Jarhyn]

Pulling people over based on race seems to me another case of judging individuals by the groups they belong to and charicatures they associate with the group.

It is the same thought process as in the racial discrimination in college admissions. Black people are presumed to be this or that because they are black.

Congratulations, you just described the dictionary definition of unethical prejudice/stereotyping/RACISM

 

[Jolly_Penguin]

Pulling people over based on race seems to me another case of judging individuals by the groups they belong to and charicatures they associate with the group.

It is the same thought process as in the racial discrimination in college admissions. Black people are presumed to be this or that because they are black.

Congratulations, you just described the dictionary definition of unethical prejudice/stereotyping/RACISM

Aye, but sadly many have redefined it so its ok so long as their chosen group is on one side of it.