Much of unsafe driving can be caught automatically, and it is much more automated in many countries. Red light cameras, speeding detection are very mature technology. I'm guessing detection of other violations can be automated with modern technology. This would not only free up police but also be much more reliable.
The Stanford cite actually calls out the difficulty finding good data, and they only have data from 21 state and 29 municipal departments. I couldn't find how they defined a "stop," either - the actual New York State site has ~10M actual *citations* over 4 years (and there are necessarily significantly more stops than citations given people are often "let off with a warning"), or 2.5M per year, yet the Stanford data only has 920k "stops_per_year," for the state of New York, so I really don't know what to think. It certainly seems like the Stanford source probably materially underestimates the total stops.
The math checks out, and the sources seem legit on their face. Not convinced that the inputs represent reality in 2024.
Two hours for a traffic stop seems extremely high. Counting time actually interacting with the public, it looks like 5-10 minutes. Hard to imagine that every single stop involves staking out a speed trap for an hour or more with no other activity.
[2] Are we sure all 10M citations were issued by humans, during traffic stops? vs. traffic cameras, parking attendants? 400M seems incredibly high, with ~230M drivers in the US, that would mean the average driver is getting pulled over by police more than once per year.
OTOH taking a conservative estimate of total traffic stops per year (60M) and hours per traffic stop (0.5) doesn't actually change the headline that much. In the OP, police are spending about 18% of their time on traffic stops. The 60M/0.5 version would lower that to 2%, but that still leaves only 30% left, since the post already assumes they're spending 65 to 70% of man-hours on admin, paperwork, driving, etc.
This is also sort of orthogonal to the question from the ACX thread, which is less about how police hours are currently distributed, and more where the marginal additional hour will end up. Anyway that's all I got.
Thanks for the triangulation, I truly appreciate it.
On traffic stops, it's really hard to find good attribution, which is why I went with such an old one for the initial estimation.
If you search online for things like "how many speeding tickets per year," you'll find many websites citing 41M, or sometimes numbers between 112k per day and 125k per day, yet they're all unattributed insurance company or lawyer sites that never give a source for the data.
And then of course, speeding tickets aren't all traffic stops, but a subset of citations, and citations aren't even all traffic stops, as some (probably fairly large, given many people's recurrent experience of "being let off with a warning") number of traffic stops end with no citation.
The Stanford cite actually calls out the difficulty finding good data, and they only have data from 21 state and 29 municipal departments. I couldn't find how they defined a "stop," either - the actual New York State site has ~10M actual *citations* over 4 years, or 2.5M per year, yet the Stanford data only has 920k "stops_per_year," for the state of New York, so I really don't know what to think. It certainly seems like the Stanford source probably materially underestimates the total stops.
In terms of the New York State data source, about 8M come from TSLED and 2M from TVB, and that translates to Traffic Safety Law Enforcement and Disposition, and Traffic Violations Bureau. But both have similar tickets - both dispense speeding tickets, and registration tickets, and insurance tickets. They both have police departments associated with each ticket, so I couldn't really see a material difference between the two sources, and the handbook didn't have any further info. I've submitted a question asking about any such difference.
Still, as you say, even if we use significantly smaller numbers, we can see that a very significant amount of discretionary police-hours are used for traffic stops, while many rather serious crimes are going unsolved.
Additionally, of the "free" time remaining for actually solving crimes, as I said in the analysis, I wasn't considering any of the things like "parking enforcement, the time they spend harassing minorities and teenage skateboarders, any time spent racking up citations and arrests for petty victimless crimes like drugs, trips to the donut shop, general time wasting and kibitzing, etc. All the non-traffic-but-not-useful stuff we know they do, ALL of which would cut into that 12-20% time they could be spending solving actual crimes."
So I also think the overall conclusion still more or less stands. I certainly do appreciate the triangulation and the additional sourcing, kudos for doing it.
Yet another reason to get police out of the traffic and parking enforcement business. Cameras and civilian metermaids can do it just fine.
Much of unsafe driving can be caught automatically, and it is much more automated in many countries. Red light cameras, speeding detection are very mature technology. I'm guessing detection of other violations can be automated with modern technology. This would not only free up police but also be much more reliable.
Thanks for the post.
Why does this source say that there are 20 million traffic stops per year rather than 100 million?
https://openpolicing.stanford.edu/findings/
Yeah, just looked into that.
The Stanford cite actually calls out the difficulty finding good data, and they only have data from 21 state and 29 municipal departments. I couldn't find how they defined a "stop," either - the actual New York State site has ~10M actual *citations* over 4 years (and there are necessarily significantly more stops than citations given people are often "let off with a warning"), or 2.5M per year, yet the Stanford data only has 920k "stops_per_year," for the state of New York, so I really don't know what to think. It certainly seems like the Stanford source probably materially underestimates the total stops.
Sorry, I realised that zahmahkibo already noticed this possible discrepancy - it's better to respond to their comment below
via https://www.astralcodexten.com/p/highlights-from-the-comments-on-prison
Intrigued by the top-level claim. Gonna try to replicate the work and report back. Consider this my study pre-registration
The math checks out, and the sources seem legit on their face. Not convinced that the inputs represent reality in 2024.
Two hours for a traffic stop seems extremely high. Counting time actually interacting with the public, it looks like 5-10 minutes. Hard to imagine that every single stop involves staking out a speed trap for an hour or more with no other activity.
On footnotes:
[1] Lichtenberg & Smith (assume it's this paper: https://www.researchgate.net/publication/237692372_How_dangerous_are_routine_police-citizen_traffic_stops_A_research_note) sources their 120M and 180M estimates from 1994 and 1981, respectively. The 60M estimate comes from the "National Center of State Courts" which records "traffic filings", which afaict includes parking violations, collisions, and other stuff that wouldn't require a stop.
* https://openpolicing.stanford.edu/findings/ claims "more than 20 million" per year.
* https://idot.illinois.gov/content/dam/soi/en/web/idot/documents/transportation-system/reports/safety/traffic-stop-studies/2023%20Illinois%20Traffic%20Stop%20Study%20-%20Part%20I.pdf says 2.26 million traffic stops were reported in Illinois in 2023. If we un-rigorously extrapolate to the entire US (~27 times the size) we get about 61 million.
* https://www.impactforequity.org/wp-content/uploads/2024/04/2023-Traffic-Stops-Data-Report.pdf reports 537k for the City of Chicago, pop 2.67M. The national rate (~127x) would be about 68M.
[2] Are we sure all 10M citations were issued by humans, during traffic stops? vs. traffic cameras, parking attendants? 400M seems incredibly high, with ~230M drivers in the US, that would mean the average driver is getting pulled over by police more than once per year.
OTOH taking a conservative estimate of total traffic stops per year (60M) and hours per traffic stop (0.5) doesn't actually change the headline that much. In the OP, police are spending about 18% of their time on traffic stops. The 60M/0.5 version would lower that to 2%, but that still leaves only 30% left, since the post already assumes they're spending 65 to 70% of man-hours on admin, paperwork, driving, etc.
This is also sort of orthogonal to the question from the ACX thread, which is less about how police hours are currently distributed, and more where the marginal additional hour will end up. Anyway that's all I got.
Thanks for the triangulation, I truly appreciate it.
On traffic stops, it's really hard to find good attribution, which is why I went with such an old one for the initial estimation.
If you search online for things like "how many speeding tickets per year," you'll find many websites citing 41M, or sometimes numbers between 112k per day and 125k per day, yet they're all unattributed insurance company or lawyer sites that never give a source for the data.
(see an example of 112k speeding tickets daily here: https://www.michigan.gov/msp/-/media/Project/Websites/msp/ohsp/1_June-2023/June-2023---Speeding-FACT-SHEET.pdf?rev=24637edc6c814fd7a6d961f88c63a38d&hash=FC91AF44AD13BE6A561E6A0CF4861917)
And then of course, speeding tickets aren't all traffic stops, but a subset of citations, and citations aren't even all traffic stops, as some (probably fairly large, given many people's recurrent experience of "being let off with a warning") number of traffic stops end with no citation.
The Stanford cite actually calls out the difficulty finding good data, and they only have data from 21 state and 29 municipal departments. I couldn't find how they defined a "stop," either - the actual New York State site has ~10M actual *citations* over 4 years, or 2.5M per year, yet the Stanford data only has 920k "stops_per_year," for the state of New York, so I really don't know what to think. It certainly seems like the Stanford source probably materially underestimates the total stops.
In terms of the New York State data source, about 8M come from TSLED and 2M from TVB, and that translates to Traffic Safety Law Enforcement and Disposition, and Traffic Violations Bureau. But both have similar tickets - both dispense speeding tickets, and registration tickets, and insurance tickets. They both have police departments associated with each ticket, so I couldn't really see a material difference between the two sources, and the handbook didn't have any further info. I've submitted a question asking about any such difference.
Still, as you say, even if we use significantly smaller numbers, we can see that a very significant amount of discretionary police-hours are used for traffic stops, while many rather serious crimes are going unsolved.
Additionally, of the "free" time remaining for actually solving crimes, as I said in the analysis, I wasn't considering any of the things like "parking enforcement, the time they spend harassing minorities and teenage skateboarders, any time spent racking up citations and arrests for petty victimless crimes like drugs, trips to the donut shop, general time wasting and kibitzing, etc. All the non-traffic-but-not-useful stuff we know they do, ALL of which would cut into that 12-20% time they could be spending solving actual crimes."
So I also think the overall conclusion still more or less stands. I certainly do appreciate the triangulation and the additional sourcing, kudos for doing it.