Join 3,520 readers in helping fund MetaFilter (Hide)


Mapping Data
February 12, 2013 12:41 AM   Subscribe

In December, the Philadelphia police department released a csv database of major crimes (murder, rape, burglary, etc) since 2006. Since then, community software developers have been mapping the data. The community involvement is hoped to spur the future release of large city data sets.

The mapping projects:
PHL Crime Mapper
Philly Crime Map
Philly Rap Sheet (Using court data)
AXIS Philly: Changes in Crime by Neighborhood
Philly Homicides Animation
posted by kaibutsu (16 comments total) 23 users marked this as a favorite

 
Philly is also making good use of this stuff internally. A friend of mine works on GIS projects for the city and has built out the system that supports GunStat, which helps to focus police enforcement on areas of gun violence.
posted by bfranklin at 12:54 AM on February 12, 2013


And if they're wrong?
posted by unliteral at 2:16 AM on February 12, 2013


one of the problems with these projects is that they tend to amplify the idea you get from the evening news that philly (or take your favorite big city) is full of crime, because they don't go outside of philly. you can see that pretty dramatically with the homicide animation where the murders stop as soon as you hit the boundary of philly. which gives you a pretty accurate picture of what you would think watching the evening news since every murder which occurs in philly is labeled by philly, no matter the neighborhood, whereas outside of philly it will be announced with whatever smaller suburban area it is located in.

but the other thing is that, if you, say, focus in on homicide. much of the excess of homicide in philly is likely tied to the drug trade. ( it's not clear how to clean the data so that you can only see murders which occur on the street, which are going to especially correlate with the drug economy.) the drug trade is a retail business and most murders are going to occur on the retail side reflecting sales competition. what you should see is that the murders cluster around retail corridors and like other retail businesses these corridors are tied to traffic patterns linking customers to retailers. now, many of the customers in the drug trade are coming into the city from the suburbs to buy. so to understand the pattern of murders in philly you very much have to see how car traffic in philly is linked to the suburbs. so again, you don't see the real picture of the data unless you look at the connections between the city and the suburbs.
posted by ennui.bz at 2:34 AM on February 12, 2013 [8 favorites]


Sure; the mandate is then for more access to data in order to draw a more complete picture...

I was actually liking the AXIS change in crime visualization, which does a decent job of showing that crime is overall on the decline. In the places where we see red, it seems like there's often just one or two major crimes in a mostly quiet area; some random noise meaning that the crime rate has gone up 400% or more. Ideally, there would be something in the visualization to indicate 'not statistically significant' for some fluctuations.

Personally, I'm mainly interested in seeing interesting community uses of the open government data sets.
posted by kaibutsu at 2:56 AM on February 12, 2013 [2 favorites]


Sure; the mandate is then for more access to data in order to draw a more complete picture...

since so many crimes in places like philly are tied to the drug trade the data is hugely biased while at the same it doesn't actually map the drug trade very well.

depending on what questions you are asking about this data, as an experimental assay, city reported crime figures can be very bad data.
posted by ennui.bz at 3:40 AM on February 12, 2013


(i don't want to be down on this stuff... maps are neat! however, you get this sort of implied objectivity when you start looking at datasets, but if the assumptions behind your data collection are flawed, all of those numbers can be somewhere between meaningless to actively misleading. science is hard and social science is even harder. there is this idea that if we just had all the data, then efficient technocratic solutions would be possible and I think it's really false when you drill down into the social science in an honest way. the whole history of the "broken windows" paradigm in NYC illustrates this, I think.)
posted by ennui.bz at 3:46 AM on February 12, 2013 [2 favorites]


The news actually focuses a great deal on crimes committed out in the grittier suburbs, because that's "news." The week we moved into my Philly neighborhood a body was found in an abandoned house around the corner and someone had attempted to set the house on fire. It didn't make the news.
posted by Peach at 4:02 AM on February 12, 2013


Because I have completely juvenile sensibilities, I just had to check out the Vaguely rude place names around the world map at the second link.

A bit of travel advice: Meat Cove, Nova Scotia is incredibly popular at this time of year, but stay away from Tampon on Reunion Island in the Indian Ocean. Also, the city of Shit in Mazandaran province in Iran is to be avoided at all costs.

Apologies for the derail.
posted by psmealey at 4:14 AM on February 12, 2013


I have it on good authority that the gentrification that goes along with the development of University City, a lot of the dealers are heading to my hood, which is just outside of the city limits and therefore dark on the animation map. Sort of underlining ennui.bz's point. Great post.
posted by angrycat at 4:50 AM on February 12, 2013


These are a lot prettier than the (still useful) mapping programs we have in Milwaukee. It's cool to see people work with what's been provided and do third-party maps. MapMilwaukee provides more general information and COMPASS provides more detailed information. Both of them are pretty ancient, but I do think they also release the information the maps are based on. I've never seen any sort of third-party web app use it, though, not really sure why.
posted by nTeleKy at 7:27 AM on February 12, 2013


For about two years now, I have been working in a strictly amateur capacity with City of St. Louis crime data, for bicycle thefts on up to homicide, with crimes dating from 2001 through 2011.

One of the things I like to do for a given neighborhood (and there are many different definitions of "neighborhood" here) is to, for a given neighborhood, map total crimes, total crime per capita, and total crime per acre, then place these side by side. I ask the question, are you likely to be involved in a crime because you are in the "wrong neighborhood" (total), you run into the wrong person (total per capita), or because you're in a spot where crime is likely to occur (total per acre)?

Note: There are some spots on the map where the crime is high on all three metrics and that is not comforting.
posted by adipocere at 7:35 AM on February 12, 2013 [2 favorites]


It's also very important to map crime with population density. Where I lived in Birmingham there were these interesting crime hotspots where there were significantly more police calls than other places. They were called 'Apartments'. Once you normalized for population density there was no more crime than anywhere else (in fact it worked out to be less crime)
posted by srboisvert at 8:03 AM on February 12, 2013


This is interesting. I've been playing with the same Uniform Crime Reporting data for Indianapolis and it is very easy to create a false or misleading narrative because the data has a lot of weaknesses - under reporting is just one problem.

I think it would be a lot more interesting and useful to get all incidents, not just Part I (rape [which for UCR purposes only includes women], murder, etc) offenses. I'm currently struggling with our police department over access to that data - they have their own ArcGIS web app that displays the past six months of data but there is no way to get the data out of that system for further analysis and so far they refuse to release the raw data directly. If that data were available I feel like there could be a lot more interesting analysis done, things like are homicides preceded by an increase in reports of shots fired? How have quality of life offenses moved around the city? What has been the effect of the switch from beat policing to zone policing?

Cities over a certain size (100k?) are also required by the FBI to collect data on the age, sex and race of people they arrest. I just received that data for Indianapolis from the FBI and while I haven't had a chance to analyze it yet I think it will be interesting to compare the arrest rate with the crime rate.

I have a small blog where I talk about open data in Indianapolis here and I'm interested in connecting more with other people doing similar work.
posted by ChrisHartley at 8:08 AM on February 12, 2013


Personally, I'm mainly interested in seeing interesting community uses of the open government data sets

The problem is companies like Zillow take this information then apply economic alogorithms to it. Then some people make financial decisions on it.

As for the gentrification of U City - Penn and their police department might have something to do with it... ditto with angrycat, great post.

I used the CML and the GIS info to find my way around the city for the years I spent there - invaluable!
posted by vonstadler at 9:25 AM on February 12, 2013


In December, the Philadelphia police department released a csv database of major crimes (murder, rape, burglary, etc) since 2006. Since then, community software developers have been mapping the data.

These are their stories.

CHUNG CHUNG

Law and Order: CSV
posted by condour75 at 11:51 AM on February 12, 2013 [4 favorites]


It seems like the trivial use case of this sort of data is to make it into a map, but that's somewhat misleading for all sorts of reasons that people have discussed above--the amount of crime and where it happens has to do with the underlying dynamics of the community & how people interact with the built environment throughout the day.

I'm not sure how feasible it is with the data set that the PPD released, but it would be much more interesting and useful to build a statistical model that reflects the underlying processes so that you could then rigorously test various interventions (police or otherwise) and see what works and what doesn't. Even without a super sophisticated model that tries to model the underlying processes, a spatio-temporal model could be useful in terms of figuring out when and where to place extra resources for the police.
posted by scalespace at 1:45 PM on February 12, 2013


« Older Neal Goldman and his company Relationship Science...  |  Around midnight local time (UT... Newer »


This thread has been archived and is closed to new comments