Location data show that race affects citations and fines for speeding
73 points
21 hours ago
| 10 comments
| science.org
| HN
djoldman
20 hours ago
[-]
> ...rideshare data from Lyft in the state of Florida to compare minority drivers with their white counterparts. Lyft objectively measured drivers’ locations, driving speed, and location speed limits .... White and minority drivers showed no discernible differences in speeding behaviors or traffic violations. However, when both drove at identical speeds, police were still 33% more likely to issue speeding citations to minority drivers and charged 34% more expensive fines, unequivocally revealing bias.

Amazing to have access to that data. So many questions...

What's the variation in ticketing with respect to different police departments? How about geographic location?

reply
bilbo0s
19 hours ago
[-]
What's the variation in ticketing with respect to different police departments? How about geographic location?

Teasing out that data is actually a more fruitful exercise. I don't mean to sound callous at all when I say this, but having a study that says "In the US blacks are discriminated against" is a bit lacking in new data or understanding. Any intellectually honest observer already knows they face discrimination.

reply
stevenAthompson
19 hours ago
[-]
> Any intellectually honest observer already knows they face discrimination.

Yep, but a significant percentage of the population are not honest observers. This is pretty hard to refute even for those that would normally attempt to hand wave the data away. This time those folks will have to get more creative.

reply
kube-system
20 hours ago
[-]
Not surprising, however, something I'm curious about is how much of the effect is due to first order bias directly due to racial prejudice -- and how much of it is second order biases and miscommunications due to the inherent social tensions of a minority traffic stop in the US.

For sure, I am immediately privileged by my color when I am stopped by the police, but the social ease by which we communicate and interact during the stop also intuitively plays a role in whether I'll get off with a warning or not.

Although I guess it's probably a distinction without much of a practical difference.

reply
ceejayoz
19 hours ago
[-]
https://engineering.stanford.edu/news/black-drivers-are-less...

> The largest-ever study of alleged racial profiling during traffic stops has found that blacks, who are pulled over more frequently than whites by day, are much less likely to be stopped after sunset, when “a veil of darkness” masks their race.

Very clever methodology on this one: they used Daylight Savings to rule out other variables.

> Next, they took advantage of the fact that, in the months before and after daylight saving time each year, the sky gets a little darker or lighter, day by day. Because they had such a massive database, the researchers were able to find 113,000 traffic stops, from all of the locations in their database, that occurred on those days, before or after clocks sprang forward or fell back, when the sky was growing darker or lighter at around 7 p.m. local time.

> This dataset provided a statistically valid sample with two important variables — the race of the driver being stopped, and the darkness of the sky at around 7 p.m. The analysis left no doubt that the darker it got, the less likely it became that a black driver would be stopped. The reverse was true when the sky was lighter.

reply
kube-system
18 hours ago
[-]
Well yes, clearly the difference in the reason for initiating two otherwise equal stops can only be due to appearance.

I was questioning whether there are differences in interactions during the stop, because police have a lot of discretion in whether or not to issue a ticket.

reply
jrm4
19 hours ago
[-]
As a Black person, I don't much see a large difference between the two things you mentioned?

Which is to say, it's certainly convenient to think that racism works like "there are racist cops on one hand and innocent well-meaning cops on the other," but that strikes me as an unhelpful oversimplification.

More specifically, I generally believe that most negative effects of racism come from folks who don't much realize how racist they are.

reply
kube-system
17 hours ago
[-]
Yeah I’d agree that definitely sounds like an oversimplification.

I’m more wondering if there is a sort of social bias that isn’t directly race related. More of a socioeconomic or cultural thing.

I could certainly imagine a cop treating someone in a pickup truck wearing a cowboy hat differently than the same person in a Nissan Altima with chrome wheels playing rap music. Or treating people differently based on how they speak during a traffic stop.

Ultimately I understand there are deep links between culture, socioeconomics, and race, but I’d be interested to know more specifically which of those is more closely linked to this bias.

reply
balderdash
20 hours ago
[-]
Putting race aside for a moment - I completely agree, my experience (though not a huge sample set) is that most officers get to your car expecting confrontation, and when they don’t get it end up just issuing a warning / writing a less severe ticket. Given that it’s easy to get out of a ticket afterwards, why argue it on the side of the road?
reply
aerostable_slug
19 hours ago
[-]
It would also be interesting to see reporting on officer's perception of the offenders. Are we seeing the results of "passing the attitude test"?

Then it would be interesting to study if identical actions of racially diverse drivers are perceived differently (i.e. if the same actions are perceived as polite with one race, confrontational with another).

reply
javman
19 hours ago
[-]
Can you clarify how it is easy to get out of a ticket afterwards?

In my experience from sitting in a traffic court a couple times (in Washington state), the only people who get out of a ticket are those who have squeaky clean driving records for 10+ years, and those who have lawyers. But most people can't afford lawyers for traffic violations.

reply
balderdash
19 hours ago
[-]
I’ve usually just hired a lawyer, but went once myself, I’ve never paid more for the lawyer than I would have paid for the ticket so it’s usually been a 0%-50% discount to the ticket. But the real cost a speeding ticket is the 5%-30% increase in your insurance which could be hundreds of dollars a year ANNUALLY so it always just seemed cheaper short and long term to pay someone to deal with it (mind you I’m talking a grand total of four speeding tickets)
reply
buildsjets
18 hours ago
[-]
Pay a lawyer, which genrally is around the same cost as paying the fine, but you don't get it on your driver's record. If you are in Washington State, you have probably seen Jeannie Mucklestone's billboards on the side of the road attesting to her success rate in beating traffic tickets. My experience is that her claims are accurate.
reply
dullcrisp
19 hours ago
[-]
It’s possible that you answered your own question.
reply
hydrogen7800
19 hours ago
[-]
>how much of the effect is due to first order bias directly due to racial prejudice -- and how much of it is second order biases and miscommunications due to the inherent social tensions of a minority traffic stop in the US.

To me, this sounds like the difference between individual (your "first order") and systemic racism (your "second order'), which is pretty well studied. Some people only choose to recognize the existence of the former.

reply
sunflowerfly
18 hours ago
[-]
A black friend of mine gets stopped constantly. He is highly educated (well beyond a bachelors degree) in a stem field. He dresses professionally. He drives cautiously and does not speed. He once joked he does not feel as bad getting stopped at night because it is less likely racial profiling.
reply
xnx
20 hours ago
[-]
> our findings suggest that compared with enforcement by police officers, appropriately located automated technologies, such as speeding cameras, could help reduce selective enforcement of traffic violations.

Every stoplight should be a speeding, red-light running, and missing tags enforcement camera.

reply
haliskerbas
20 hours ago
[-]
Shall we chuck in facial recognition too, just in case you half-inched a pack of gum at the Tesco self-checkout?
reply
kypro
19 hours ago
[-]
Tesco self-checkout already has facial recognition connected to other local stores (including those of other major supermarkets), so this probably isn't needed.

If you were stupid enough to steal a pack of gum (some of my family are) you'll quickly find iyou're unable to buy groceries from any store in your area [1].

[1] https://bigbrotherwatch.org.uk/blog/the-spectator-has-your-l...

reply
dullcrisp
19 hours ago
[-]
I don’t like the authoritarian-sounding phrasing here, but upon consideration, would this method of enforcing traffic laws really be much worse than having people with guns chasing people down at high speeds?
reply
lcnPylGDnU4H9OF
19 hours ago
[-]
> would this method of enforcing traffic laws really be much worse than having people with guns chasing people down at high speeds?

Yes. People don't scale the way computers do.

reply
xnx
18 hours ago
[-]
A traffic camera never planted drugs in my car.
reply
lcnPylGDnU4H9OF
10 hours ago
[-]
This is still possible even with cameras running recognition software at every intersection.
reply
dullcrisp
19 hours ago
[-]
Hmm there might be a third rail that I don’t want to touch here. Pardon the phrasing.
reply
kube-system
20 hours ago
[-]
That is cost prohibitive. I'd rather replace red lights with traffic circles.
reply
djoldman
15 hours ago
[-]
Yes, but a couple problems:

1. Where to put them? In Chicago these cameras unfortunately end up being a regressive tax.

2. What to do with the money? Anytime we give the government money from things like this, we create weird incentives. I've been wondering lately what it would be like if we required all the fees to be returned to the federal treasury / burned... no incentives!

reply
xnx
15 hours ago
[-]
> 1. Where to put them? In Chicago these cameras unfortunately end up being a regressive tax.

Everywhere. Starting with highest volume intersections.

> 2. What to do with the money? Anytime we give the government money from things like this, we create weird incentives.

Good concerns to have. I'm less concerned with bad incentives (e.g. vs. civil asset "forfeiture") because the enforcement is automated. There can still be bias in roll-out of the cameras, but this wouldn't be the same degree.

> I've been wondering lately what it would be like if we required all the fees to be returned to the federal treasury / burned... no incentives!

I like that thinking. This makes everyone's money worth ever so slightly more. (which is also minusculey regressive)

reply
mjamesaustin
20 hours ago
[-]
Red light cameras increase the rate of rear ends, sometimes dramatically, as drivers unsafely brake to avoid the fine.
reply
asoneth
19 hours ago
[-]
My understanding is that this sort of rear-end crash generally tapers off as a sufficient number of local drivers become accustomed to driving safely instead of racing to make it through a yellow light.

But even if it doesn't, it might still be a reasonable tradeoff. A higher number of minor rear-end crashes (typically between unsafe drivers) might be acceptable if it helps to reduce more t-bone crashes that are more likely to be fatal.

reply
Tostino
4 hours ago
[-]
In Pasco county FL it didn't taper off. We just had a huge increase in collisions until they took the cameras down.

https://www.tampabay.com/opinion/editorials/red-light-camera...

reply
asoneth
3 hours ago
[-]
Fatalities seems like a more relevant safety metric than collisions.

"IIHS research examining large U.S. cities with red light camera programs found significant reductions in citywide fatal red light crash rates (and fatal crash rates of all types) at intersections with traffic signals, including intersections that didn’t have a camera. ... Cities that discontinue their programs tend to feel the sting. The fatal crash rate was 30% higher in cities that shuttered their red light camera programs." https://www.amica.com/en/resources/auto/safety/ways-red-ligh...

But the data do indicate that the benefits of automated traffic enforcement vary widely by community. I suspect any of poor road design, reckless driving culture, or local corruption that prioritizes income over safety can lower their net utility, and some combinations of those might even make their utility net negative.

reply
dragonwriter
18 hours ago
[-]
No, if it increases the rate of rear end collisions, it means that drivers are unsafely following too close to react to the driver in front of them, behind the drivers that are planning to unsafely and illegally run the red light and then changing their mind late in reaction to the camera.
reply
stevenAthompson
19 hours ago
[-]
Surgery increases your odds of dying during surgery. It also saves far more lives than it costs.

Unless you've seen data that says otherwise, I think this is the same.

reply
mitthrowaway2
19 hours ago
[-]
For that to be true, how fast must they have been going through the yellow?
reply
dragonwriter
18 hours ago
[-]
It just requires that the driver behind them was unsafely and illegally failing to maintain a safe following distance (and/or failing to pay attention to the car in front of them.)
reply
mitthrowaway2
16 hours ago
[-]
That's true of all rear-ends, but for the ones that are influenced by the presence of a red light camera, it must also mean that the crash wouldn't have happened without a camera-induced behaviour shift in the leading car. So it requires not only an unsafe following car, but also the lead car going too fast towards a yellow and then braking suddenly, upon remembering that they'll lose money for going through a red light. Still the fault of the car in the rear of course, but if the car in front is driving reasonably in the first place, then the red light camera won't have any effect. The ones who get rear-ended are the ones who were going to anyway, camera or not.
reply
chrislongss
20 hours ago
[-]
You forgot tracking. Tracking and recording every trip you take.
reply
xnx
19 hours ago
[-]
Everyone's phones already collect and sell this information. If it's useful, and the citizens vote for it, why shouldn't cities collect this data? It already happens on toll roads
reply
chrislongss
19 hours ago
[-]
Maybe we should strive to make it illegal for companies to collect and sell phone data instead of allowing more mass data collection?
reply
lcnPylGDnU4H9OF
19 hours ago
[-]
> companies

Different from the government (in theory). In the US in particular, any data that is generated from a tax-dollar-paid system is by default public information that can be requested either by FOIA or a state-equivalent law. I think that is a qualitative difference that makes this genuinely worth considering. Imagine if Google et. al. were legally required to make their location data publicly available. I mean, assume that were the case from the inception of the idea to gather that data because that's how it is with governments in the US.

(Granted, there are further considerations around the potential for abuse, especially when considering victims of domestic violence/abuse. But the idea is far more palatable than what we currently have with large corporations.)

reply
Carrok
18 hours ago
[-]
You consider it _more_ palatable that every trip you take could potentially be publicly accessible information?
reply
lcnPylGDnU4H9OF
10 hours ago
[-]
> than what we currently have with large corporations

Let’s be clear: the current situation is that every trip I take is privately accessible information that is sold for pennies but otherwise difficult to access; critically, there is no obligation for the information to be shared. Yes, this is a far less palatable situation than if such data is simply and plainly public: the gatherers, in this case governments, would have an obligation to share it with the subjects of the data.

I’d personally prefer no data collection, but that’s just where my vote goes. Unfortunately, in this hypothetical my vote was the less popular one:

> If it's useful, and the citizens vote for it, why shouldn't cities collect this data?

reply
slt2021
19 hours ago
[-]
you described a life under techno-fascism
reply
stevenAthompson
19 hours ago
[-]
Enforcing the laws that already exist via technological means instead of with personal prejudice, guns, and high speed pursuits is not fascism.

Could it eventually lead to fascism? Maybe, but so could enforcing the law with fast cars and guns if we aren't careful.

reply
slt2021
11 hours ago
[-]
the government will always have a way have non-uniform enforcement. Elites and important people won't be enforced, while everyone else will get mass surveillance, loss of privacy, and techno-totalitarianism.

Everything will be according to the laws, but you forgot who is writing and passing the laws and in whose interest. It is rarely the interest of the common folk.

reply
mitthrowaway2
19 hours ago
[-]
Is that the term for a uniform, unbiased, automated enforcement of law?

Do we prefer that speeding should only be punished when the wrong kind of people do it? Perhaps encourage everyone to make a habit of breaking the law, so that we can use it against them when they say something we don't like?

reply
slt2021
15 hours ago
[-]
>uniform, unbiased, automated enforcement of law?

the only uniform thing would be data collection, loss of privacy, and being a victim of gov-t oppression.

"unbiased" - depends on the bias inherent to the training dataset and enforcement algorithm. Areas with more sensors will get more enforcement.

what if you a victim of false positive? you wont be able to appeal to "unbiased automated enforcement"

just look at recent events, when Trump just stripped some students of their greencards and deported them with 0 due process.

or took a guy with some random tatoo and sent one way to El Salvador...

reply
mitthrowaway2
11 hours ago
[-]
How is getting ticketed for speeding "being a victim of government oppression"?

By "bias inherent to the training dataset and enforcement algorithm", are you talking about anything other than "areas with more sensors"? These aren't exactly LLMs; it's kind of like saying vending machines might be biased because there are more of them in certain areas.

The areas where there are more of them will indeed be the areas where speeding rules get more strongly enforced (eg. school zones), so if you don't want cars speeding through your neighbourhood, you'd want to have better coverage. The areas with no enforcement would be the neighbourhoods being discriminated against, so is your view that it's better to have speed cameras nowhere?

As for due process, that seems entirely unrelated but if anything this puts you in a better spot. If you are a victim of false positive, the recording is right there, you download it from the web portal, review whether it was really a false positive or not, and if so, take it to court to dispute the charge. You don't have to he-said-she-said with a police officer who might, if they were feeling especially vindictive, plant some evidence on you during the traffic stop.

reply
slt2021
44 minutes ago
[-]
tickets is only a pretext, surveillance and loss of privacy will affect anyone regardless of whether they drive the speed limit or exceed it.

this is what fundamentally people don't understand. They look at one government failure (inability to enforce order on roadways), and cheer to spend taxpayer money to erect a techno-fascist surveillance state.

Sure, if you speed you will get ticket, but also if you post anything against the US government or against Israel, you will be tracked and thugs will come knocking your door.

It is fundamentally wrong to allow government to massively track and collect data on citizens, as it makes it trivial to track and harass people since all your data is in one place, one SQL query away and the state will use it against you

reply
woah
19 hours ago
[-]
Every subway gate should be a ticket checker
reply
dullcrisp
19 hours ago
[-]
Fascist!
reply
clown_strike
14 hours ago
[-]
You see techno-fascism, others here see a business opportunity.
reply
seryoiupfurds
18 hours ago
[-]
People are free to choose not to run red lights.
reply
kypro
19 hours ago
[-]
Not really. It's like this in many cities in the UK.
reply
g8oz
19 hours ago
[-]
This seems like the kind of research that would invite retaliation from the current American administration. I hope none of the authors are visa holders.
reply
gtsop
19 hours ago
[-]
I wonder why the downvotes. I assume people believe this comment is overly sarcastic... but is it? The absurdity of what the administration is doing automatically makes this comment a legitimate one, and not a troll.
reply
dragonwriter
19 hours ago
[-]
> our findings suggest that compared with enforcement by police officers, appropriately located automated technologies, such as speeding cameras, could help reduce selective enforcement of traffic violations

This seems to presume that automated enforcement replaces, rather than enabling departments to even more precisely focus, the selective human enforcement of traffic laws. Given the role of selective traffic enforcement as a conscious tool in generating contacts to look for non-traffic issues, rather than a product of mere implicit bias, I find it extremely unlikely that this would happen with any real-world police department.

reply
josefritzishere
20 hours ago
[-]
That is positively shocking. Police who are already widely known to target minorities for criminal offenses also target them for minor traffic violations.
reply
vharuck
19 hours ago
[-]
This analysis doesn't say whether minority drivers were more likely to be pulled over for traffic violations, only that they were more likely to be issued a citation and that those citations came with higher fines. I would guess a cop rarely knows the race of a speeding driver before pulling them over. The discrimination would happen either when deciding whether to issue a citation, or after looking up the driver's info on the police car.

When the police look up your info, do they see past citations? If so, it's a self-reinforcing cycle: people more likely to receive citations are more likely to have citations listed, and a cop who sees past citations might be less willing to let the driver off with a warning.

reply
readthenotes1
20 hours ago
[-]
Given the different rate of death-by-cop, I'd be surprised if the "Asian" contribution to the citation effect wasn't actually negative.

https://www.statista.com/statistics/585152/people-shot-to-de...

reply
ReptileMan
19 hours ago
[-]
> Our main analysis assigns drivers to one of two racial and ethnic categories: white or minority, where minority includes Asian and Pacific Islander, Black, or Hispanic (31).

Whenever you see this kind of aggregation you can be certain that you there will lies, damned lies and statistics. Let me guess - if the authors did by 5 groups the results they got were not saying something they liked. Or at least they should have put any kind of proof that the rate of ticketing for all of those aggregated groups were quite similar.

reply
dragonwriter
19 hours ago
[-]
> Let me guess - if the authors did by 5 groups the results they got were not saying something they liked.

The main reason likely is that US race (treating Hispanic ethnicity as a race trumping nominal racial categories, so the other categories all are implicitly prefixed by “non-Hispanic”) is something like (in 5 categories):

59% White, 19% Hispanic, 13% Black, 6% Asian, 3% Other/Mixed.

Or, in two categories:

59% White alone, 41% Other

The two-group breakout takes far less data to get equal statistical utility for its smallest group when compared to the 5 group breakout (that's true even if you don't really care about the Other/Mixed catch-all group in the 5-group breakdown, but only the other four.)

reply
pclowes
20 hours ago
[-]
I very well believe this could be true. However, after hearing of many studies that don’t get published unless they confirm the suspicion of racial prejudice I don’t even know what to believe anymore. Side effect of a post-truth society locked in endless red-blue proxy fights I guess.

I don’t know why we need police doing traffic enforcement. The cars know how fast they are going, the cars know where they are, the cars know what the speed limit is, why are they able to speed? Enforcement should be fully automated and violations should be forced to provide evidence of a special situation id they do speed eg: in labor etc. or a ticket is automatically issued.

Also in favor of governors on max acceleration and relative force generation for large heavy vehicles. Eg: 9k lb electric truck can do 0-60 in a minimum of 8seconds.

reply
viraptor
20 hours ago
[-]
> I don’t even know what to believe anymore

Good research? Even if studies which did not find this problem were not published, if the methodology is correct this study is correct regardless of external pressures.

> The cars know how fast they are going

They don't. My car's displayed speed is 5% higher than actual speed. It will change a little bit again after the next tire replacement.

> the cars know where they are

They don't. Not only are there no up to date maps for permanent roads, roadworks will add changes and in downtown gps will happily put you on the road on the other side of a high-rise.

> the cars know what the speed limit is

Again, up to date global, correct, precise maps do not exist and will not exist. Vision can't be relied on either, because bad luck can make a speed limit invisible.

reply
pclowes
19 hours ago
[-]
You don't believe there has been widespread failures if not malpractice and outright fraud in many journals and published studies resulting in replication crises and performative/self-serving peer review over the last several years?

Also cars can read speed limit signs pretty accurately. GPS is pretty accurate. It should not be possible to go highway speeds in areas that are purely residential and miles away from the highway.

Also we could make a more accurate map enforcement system. Not trivial but straightforward.

reply
viraptor
18 hours ago
[-]
> ... over the last several years?

I'm taking about this specific research here.

> cars can read speed limit signs pretty accurately.

Only when they're not covered by trees, snow, other cars, are not damaged, workers didn't forget to uncover them, etc. And from time to time you will sync with a car in another lane just the right way to miss the sign.

> GPS is pretty accurate

Often. Yet, I will regularly get placed on the wrong road in larger cities. It doesn't have to be a highway speed problem - since the idea is automated enforcement, going 10k/h over the limit of the parallel street is already an issue.

> we could make a more accurate map enforcement system

Maybe in a distant future. Right now, we're at step 0: there's no publicly available accurate digital map of all streets. There are various segments in different agencies, there are global commercial maps, and there are community efforts. Until that's solved, the issue is moot.

reply
pclowes
18 hours ago
[-]
I just don’t think it is that hard of a problem. The legislative hurdle is much higher than the technological difficulty.

You can always fail safe. Eg (extreme case) unless your car can prove to itself it is on a highway or a high-speed street it can only go 25 mph.

reply
viraptor
17 hours ago
[-]
> prove to itself it is on a highway or a high-speed street it can only go 25 mph.

GPS attacks are real and even common in some areas. Do you know how much havoc could an adversary cause on traffic / cargo transport if this was implemented? Imagine a large American highway near a big city being blocked that way for days.

reply
pclowes
17 hours ago
[-]
You could override it and i am sure “gps attack” would be an allowed exception.

People can hack cars already.

We didnt always have to make everything to serve cars and we can stop anytime we want.

reply
AlexandrB
20 hours ago
[-]
It's all fun and games until these same concepts are applied to computers and Windows prevents you from typing anything potentially "hateful" while pinging the government so you can get fined appropriately.

We're nearly there actually with technologies like Gatekeeper and OSs requiring account creation to use, so I expect to see it in my lifetime. The missing piece was the ability to classify content automatically in real time, but luckily AI came along.

reply
pclowes
19 hours ago
[-]
A large heavy car going from 0-60 in 4 seconds through a mass of innocent people on Bourbon St is different than words online.
reply
dmicah
19 hours ago
[-]
Also consider that traffic violations are a source of funding for a municipality. Using technology to eliminate violations completely may not be desirable.
reply