NYPD: Public Too Stupid to Understand a Citywide Crash Map

This morning’s City Council transportation committee hearing covered a number of bills, including one that would require NYPD to release data to the Department of Information Technology and Telecommunications for a public map of crash locations and traffic fatalities, to be updated monthly. NYPD testified in opposition to the bill, claiming that it was already doing enough to release information to the public. A panel of technology and street safety experts testifying later disagreed, and were joined in their skepticism by some council members, including committee chair James Vacca.

NYPD thinks its data, shown above on a third-party map, is too confusing for the public. Image: ##http://nyc.crashmapper.com/11/8/13/8/standard/collisions/2/15/40.715/-73.990##NYC Crashmapper##

One of NYPD’s main objections to the crash map bill is that crash reports, which must use forms from the state Department of Motor Vehicles, map incidents to the closest intersection, not geographic coordinates or addresses. Without having a precise location, NYPD says the public might be confused about where crashes actually occur.

“Putting it on a map is inherently somewhat misleading,” NYPD assistant commissioner of intergovernmental affairs Susan Petito said.

Council Member Dan Garodnick asked if the agency would be interested in joining the council to advocate for a change to the state form to allow for addresses or geographic coordinates. “I don’t think so,” Petito replied. “The utility of a street address, I can’t sit here and tell you that would add anything.”

While arguing against sharing more detailed information with the public, Petito said that the police department’s own access to high-quality information about where and why crashes occur give the department a better perspective on traffic safety than the public has. “We look at it a little differently from the way a member of the public would,” she said. “We have access to so much more information, including everything on the police accident report.”

“I’m not worried about confusing the public,” Vacca said after the hearing. “I think people understand what’s released more than the police department would give them credit for, and I think we should have the information.”

While having the city create its own crash map would be a step forward, transportation and technology advocates testifying today said it’s more important for city agencies to release quality data to the public, which would be easy to access by coders and interested communities. At present, crash and summons information is released in PDF and Excel formats that the police department must compile each time it releases data, and developers must reconstruct to create complete, geographically-tagged data sets.

“The data isn’t truly open,” Transportation Alternatives general counsel Juan Martinez said. Though Local Law 12 of 2011 was a “landmark bill” that helped open up city data to the public, Martinez said, compliance by city agencies, including NYPD, has been less than comprehensive. “We are strongly recommending that in addition to making a map, which is one way to present the data, you also make the data available,” Martinez said.

AAA New York, an infrequent ally of street safety advocates, also supported the crash map bill, saying that it would use publicly-accessible data to assess the effectiveness of red light cameras. AAA has said it supports automated enforcement “in concept” but has regularly attacked NYC’s traffic camera programs.

One of the reasons agencies are reluctant to release data is that requests often come from those who are battling an agency over a specific issue, said Noel Hidalgo of Code for America. “They are scared to be transparent,” he said. “We need advocacy from the council to demand that data is released in a non-judgmental way.”

Freelance web developer John Krauss created a website that saves and archives NYPD crash and summons data, called the NYPD Crash Data Band-Aid, and another site that maps crash data. “I call it the crash data band-aid for a reason,” Krauss said. “We have to establish a high-quality original source.”

“This all comes down to departments releasing data in a timely manner in a way that programmers can use,” programmer Nathan Storey said, in response to questions from Council Member Vincent Ignizio, who himself was clearly intrigued by the notion of improving the city’s open data practices.

“Let these guys have a hack at it,” Martinez said.

  • Anonymous

    Yes, knowing that a particular intersection is dangerous wouldn’t help anyone. Please keep that confusion knowledge away from the public who might actually use such data.

  • My Taxes Go To NYPD

    Where did PD find this embarrassing dimwit? Of course it would help to have specific crash locations. There are thousands of crashes in places a long way from the nearest intersection — half-mile long viaducts, roads along long frontages like zoos or art museums or warehouses, to name a few. You can’t fix the problem if you don’t really no where it happened. Of course, in Ray Kelly’s “shit happens” world of traffic safety, none of it really matters anyway.

  • Janus

    “Without having a precise location, NYPD says the public might be confused about where crashes actually occur.”

    [Five minutes later, when NYPD is asked if they’d like accident report forms to include precise locations.]

    “I don’t think so,” Petito replied. “The utility of a street address, I can’t sit here and tell you that would add anything.”


  • Noel Hidalgo

    If you are interested in hacking on what crash data we do have, join us at BetaNYC. We have weekly hacknight discussing these issues .

  • Community Member

    When it comes to traffic safety there’s the DOT, which attends countless community board meetings, listens to feedback, holds public forums, and posts its plans online, and develops projects that are requested by local neighborhoods.

    Then there’s the NYPD, which basically says, “Trust us, guys.”

    So tell me: which one is the city agency that the tabloids are always saying doesn’t listen tot he community?

  • Joe R.

    What do you expect from an agency whose favorite phrase seems to be “no criminality suspected”? The NYPD has always had contempt for the general public it’s supposed to serve. This just confirms it.

  • Bolwerk

    I’m rather tired of everyone thinking the solution to traffic violence is policing. As if having police does a whole lot of good for you when it’s time to scrape you off the road, “criminality suspected” or not.

    Policing may play a necessary role somewhere, but the vast majority of the solution has always been about design. Design for high traffic, you get high traffic, and with high traffic you get casualties.

  • Joe R.

    No argument here that infrastructure, and especially reducing traffic volumes, are the best solutions. I’ve written many posts advocating both approaches. I’m merely saying here that the police (and legislatures) are out to lunch when it comes to doing the one important thing they can do to make the streets safer-namely getting drivers who demonstrate they can’t drive off the roads, preferably permanently. Unlike others here, I don’t even care that they go to jail. Just keep anyone who kills or seriously injures someone while driving from ever doing so again.

  • Not Paul Browne

    Suffice it to say that Ray Kelly is not dedicating his agency’s top personnel to the problem of reducing motor vehicle violence on NYC streets.

  • JimthePE

    I think what Susan Petito was trying to say is the number of crashes per year at a certain location is only half the story. The big intersections will swamp out all the other locations, just because of the sheer numbers of people and vehicles going through them. You can’t directly compare Times Square to an intersection of two minor back streets in Queens. You are less likely to be in a crash in a busy intersection with 100 crashes per year than an intersection with a quarter the traffic and 75 crashes per year.

    A good GIS person could pretty easily present the data in a way that makes sense to the average person. Crash risk per 1,000,000 vehicles is a standard way to measure it, but a percentile rank compared to similar locations would be as good, if not better.

    She’s also wrong about the NYS DMV forms. They have fields for precise latitiude and longitude on the crash report form. Most modern police departments have computers and GPS in each squad car that automatically fills these fields in.

  • Petito sounds ridiculous. Is she against having accurate information?

    The State of Illinois Division of Traffic Safety has a great crash dataset, which I’ve put into an online map that finds the number of crashes near any specific point. The crash locations are very precise and are set by GIS technicians who review the police departments’ crash reports.


  • A. Scott Falk

    No, no. I was there and I assure you, she was saying that the intersection info is too inaccurate. Then, of course, she told Dan Garodnick that she didn’t think they could partner with the City Council to lobby Albany to add accurate location info on these crash report forms.

    She was just slimy as hell.

  • JimthePE

    IT’S BEEN ON THE FORM FOR 10 YEARS! NYPD was the first dept in the state to get the hardware!

    Do you think she was BSing to avoid looking uninformed? Or did she have an agenda? Either way, someone should get a blank form for the next hearing, and call her on it.

  • Anonymous

    Will second this, as I was there and wrote the piece. Petito was not making an argument about exposure risk; her argument was that the public would be confused because crashes are mapped to the nearest intersection.

  • Michael Klatsky

    The state dmv office is understaffed with old computers running Map Info. (Not arcinfo.)

    The points, are assigned a nearby node for association – not for representation – state dmv, when developing the system of assigning points many years ago, never conceived of mapping them – so it wasn’t a consideration.

    Therefore, unless pressure is directed at NYS DMV to update their methods & systems to something modern like the NYCDOT Traffic Information Management System (TIMS), the data will be unavailable, because it is a misleading and bad (geo)dataset.

  • Michael Klatsky


    I agree. Designing local streets with an 85′ roadbed, 4 travel lane + 2 parking and no median is ridiculous.

    I don’t care about how much LOS is going to decrease – it isn’t a usable metric for an urban grid, because of the nearly unlimited alternative routes.

  • Michael Klatsky

    Before you begin hacking the data – make sure you understand the data source. The raw data is bad geodata.

  • Michael Klatsky

    The precisely is the point. The data would indicate that a problem intersection exists, when in reality, a midblock high volume driveway can be the source of the crashes. Perhaps there is a destination midblock, with a poorly maintained sidewalk on that side, with people walking halfway on the other side and then jaywalking across.

    Those are mapped to the nearest node and are therefore bad data for map representation. They nodes are for association for easy lookup.

  • Jonathan R

    I am not too convinced of the utility of crashstat-type maps. It seems to me
    that crashes are highly dependent on the number of vehicles passing by and on the number of pedestrians and bicycles passing by. I expect that with the use of
    factor analysis, one can accurately forecast the likelihood of crashes at any
    location if provided the number of vehicles and the number of pedestrians in
    that location. If so, why spend time and energy on obtaining data that has very
    little marginal predictive value?

  • david

    “It seems to me that crashes are highly dependent on the number of vehicles”

    What is this nonsense? Factor analysis? Care to back up that massive assertion with at least one reference? Are you assuming that all streets behave equally, considering visibility, construction work, sidewalk width etc.

  • Noel Hidalgo

    Yes. We know that and we testified as such. Here’s our point of view of yesterday’s hearing… http://blog.betanyc.org/post/63680541609/three-wins-a-recap-on-todays-safestreetsmap-hearing

  • Noel Hidalgo

    we’ve looked at the data NYPD provides. it isn’t good enough. we can get the raw data but we must FOIL it and that is a slog. we want this data to be open, good, and frequently updated. If you are as good as you say you are, join us at a betanyc.org hacknight. we could use your skills

  • Noel Hidalgo

    that’s cool. that’s your opinion, but developers and advocates say otherwise.

  • Anonymous

    While a measure of the crash rate per vehicle or per person going through an intersection makes for interesting comparisons, I would argue that the absolute number of crashes is much more relevant for safety advocates.

    To give an extreme example, which intersection would you fix first?

    a) an intersection that sees 100 crashes per year, with a traffic of 100 million cars per year.

    b) an intersection that sees 1 crash per century, with a traffic of 1 car per century. Hey, the crazy person who drives there once a century always crashes!

    Even though b) has per-vehicle rate that is one million times higher than a), it is not really worth the effort compared to a).

  • BornAgainBicyclist

    Seems more deliberate than dimwit.

  • Jonathan R

    I definitely respect your approach, and we are both advocates. It’s plain as day obvious to me as a bicycle and pedestrian advocate and as a parent of a two-year-old that many NYC streets are dangerous as laid out today.

    Using crash data to decide which dangerous streets to fix first to me seems ghoulish. I would like to fix the dangerous streets without sacrificing more children.

    If you have another more pleasant interpretation of how to use crash data please share it with me.

  • Jonathan R

    Look up “factor analysis” on wikipedia. It’s a technique to reduce the number of variables in the dataset.

    It costs more to add more variables to the dataset. Mr. Hidalgo and Mr. Krauss had to take the time out to testify at this hearing for instance, instead of doing paid work in other contexts. Glad they could do it, but there’s a cost involved.

    Therefore it makes sense to me to use statistical techniques to reduce the cost of data collection without compromising the validity of the conclusions we can draw.

  • david

    I was not asking what factor analysis is – I am extremely familiar with it. I was more astounded that having an intuitive feeling about a statistical procedure is a justification to ignore actual data about how people are killed on the street.

    I’m not suggesting we don’t show other variables too, such as total number of cars/people (I think that is a good idea) but your framing of the problem suggests that that actual fatality/accident data does not matter.

  • Jonathan R

    Sorry, but to me as a parent, the most straightforward way to parse reliance on “actual fatality/accident data” is “children must be sacrificed to make streets safer.” If we can figure out which streets are dangerous without killing more kids, why don’t we?

  • david

    Yes, but how does not releasing the data in a form that facilitates analysis achieve this? You started off by suggesting that these types of maps are not useful as the data can be explained through statistical analysis. I’m arguing that 1) the data is useful and 2) analyzing it in a meaningful way is also useful.

    To perform analysis you actually do need some underlying data, not speculation.

  • Bolwerk

    I wasn’t disagreeing to you, just adding to what you said.

    They really aren’t bothering, despite people demanding ever more police and ever larger police budgets.

  • It’s interesting how different states are set up. The DMV here, run by the Secretary of State’s office, isn’t involved in crash data.

  • Jonathan R

    Sir, if you can explain how to use the crash data in a
    sensitive way go right ahead. I would choose to avoid basing advocacy decisions on crash data because of the sacrificing-children problem as outlined elsewhere
    in this thread.

  • david

    I’m not saying that crash data has all the answers – it clearly doesn’t; there’s much more context needed . But not releasing it so that other people can’t analyze is illogical.

    I strongly disagree that releasing data of this nature “has very little marginal predictive value?”

  • Jonathan R

    OK, to the marginal predictive value point.

    Imagine a five-way intersection where a major two way arterial street (with bike lanes), meets a cross street on its curved way down (and up) a steep hill. There is a slip lane between the arterial and the cross street. Motor traffic regularly runs the red lights and there is a streetside flat-fix shop at the corner. Does this sound like a dangerous place for pedestrians and bicycles? Sounds like one to me.

    So when I open the crashstat browser to Edward L Grant Highway & West 170th Street, Bronx, NY 10452, and see a large blue dot, I am not too surprised. That’s what I mean by crash statistics having “little marginal predictive value.” Anyone standing at this corner can figure out that it’s dangerous.

  • Anonymous

    Sure, let’s make all the proactive changes we can, and sure, let’s try to predict stuff. But predictions from a theoretical model will be far from perfect, and crashes still happen, even in Sweden. It is absurd not to use that information. Using actual crash information after a tragedy is not “sacrificing children”; it’s like looking at the black box after a plane crashed despite the best effort of the engineers who designed the plane.

  • Anonymous

    Sure, let’s make all the proactive changes we can, and sure, let’s try to predict stuff. But predictions from a theoretical model will be far from perfect, and crashes still happen, even in Sweden. It is absurd not to use that information. Using actual crash information after a tragedy is not “sacrificing children”; it’s like looking at the black box after a plane crashed despite the best effort of the engineers who designed the plane.

  • Anonymous

    Sure, let’s make all the proactive changes we can, and sure, let’s try to predict stuff. But predictions from a theoretical model will be far from perfect, and crashes still happen, even in Sweden. It is absurd not to use that information. Using actual crash information after a tragedy is not “sacrificing children”; it’s like looking at the black box after a plane crashed despite the best effort of the engineers who designed the plane.

  • Noel Hidalgo

    What is your baseline for a safe street? Are you already prioritizing your street over mine?

    That’s not what I’m talking about. I am not JUST talking about deaths or injuries. I’m talking about violations too! As not a parent, but someone who likes all beasts – two and four legged – of all variety, I don’t understand why you don’t want your tax dollars to work for you?

    I want data. This is an additionally step to what you are already working on and advocating for. It would enable me to help you. It brings us together and enables us to see deeper into the actions of our government AND how to improve conditions. How are we divergent?

  • Noel Hidalgo

    Actually, I get paid to do this… Join one of our meetups and see how factoring together….

  • Mark L

    Obviously, raw numbers have to do with volume. But, it is very helpful to look at routes you typically walk, bike or drive and be aware of the danger points.

    I never realized, for example, that Sterling and 5th Ave is a bike hotspot. I’ll be more careful there.


Advocates Urge Lander to Upgrade NYPD Crash Data Bill

A bill that would have pushed Ray Kelly’s police department one step closer to opening up crash data has been reintroduced by Council Member Brad Lander. But with new leadership, NYPD is dropping hints that it will release better public data soon. Advocates say Lander’s bill could use some upgrades to help the public get […]

NYPD Crash Data Now Easier to Use and Updated Daily

The city went live with a major upgrade to NYPD’s crash data today. Information about traffic crashes was previously released via difficult-to-use monthly updates posted on the police department’s website. Now it’s available through a standardized feed updated daily on the city’s open data portal, allowing the public to sort crashes by time of day, […]

City Council Bills to Release Traffic Data Pass Committee Unanimously

Three bills to open up information about traffic and street safety to the public cleared the City Council’s transportation committee unanimously today. According to committee chair James Vacca, the bills are scheduled for a floor vote this Wednesday. Two of the bills, in particular, should provide New Yorkers with a much clearer picture of what’s […]

After NYPD Kills Bill, Council Pushes for Traffic Safety Data From DOT

The City Council Transportation Committee held a hearing yesterday on four bills that would release new information about traffic crashes and how the Department of Transportation decides whether to install traffic calming measures and traffic control devices like stop lights and stop signs. All together, the bills would cover a wide spectrum of information, but […]