Is Crime More Concentrated in Spread-out Cities?

One common argument against public transit expansions into suburbia is that better transit will enable urban criminals to visit suburbia more easily and commit more crimes; a related argument is that any non-sprawl development is crimogenic because compactness allows crime to migrate from neighborhood to neighborhood more easily.  The commonsense response to this argument is that there is no correlation between crime and transit use or density; the nation’s most dangerous cities are not the most transit-friendly or those with the best public transit.  For example, Boston (over 13,000 residents per square mile, 25 percent of commuters using public transit) had 4.7 murders per 100,000 residents and Oklahoma City (1145 people per square mile, 1 percent of workers using public transit) had 6.4. New York City had fewer murders than either city (4.3 per 100,000).

When I point these facts out, the common response I get runs something like this: “Yes, but [my spread-out Sun Belt city) only has its crime in a few bad neighborhoods, while cities like New York have crime everywhere”. Until recently, there was no way to judge the validity of this argument.

However, thanks to the creators of the American Violence (americanviolence.org) website, we do have a way to judge this argument to some extent.  The American Violence site lists homicides, fatal shootings, and nonfatal shootings for many American cities, and even has data by census tract for shootings (though not for homicides or other crimes).  So we can see whether shootings were more concentrated in a few neighborhoods for any particular type of city.

For example, Boston had 90 nonfatal shootings in 2023.* If crime was evenly spread out throughout Boston’s neighborhoods, we would find that each of these shootings was in a different tract.  On the other hand, if crime was all in one neighborhood, we would find that every single shooting was in one census tract.  As it happens, there were shootings in 42 of Boston’s census tracts (21 percent of its 207 tracts).  The worst 10 percent of these tracts (that is, the four with the most shootings) collectively had 29, or 32 percent of the city’s shootings.  The worst 20 percent (the eight with the most shootings) had 42, or 47 percent of the shootings. 

How does this compare to (for example) Oklahoma City?** Oklahoma City had 128 shootings, which occurred in 64 of the city’s tracts. The worst 10 percent (that is, the six tracts with the highest number of shootings) had 26 shootings, or 21 percent of the shootings.  The worst 20 percent (that is, the worst thirteen) had 53, or 40 percent.  In other words, shootings were actually slightly more spread out in Oklahoma City than in Boston- that is, the neighborhoods with the most shootings had a lower percentage of shootings than in Boston.

Similarly, Tulsa 100 or so miles away had 86 nonfatal shootings in 47 tracts.  The worst 10 percent of tracts (five tracts) had 22 shootings, or 26 percent- a number somewhere in between Oklahoma City and Boston.  The worst 20 percent had 38, or 44 percent of its overall shootings- also in between Oklahoma City and Boston.

Of course, Boston may be an aberration.  Is there another walkable, transit-friendly city with crime rates similar to those of Oklahoma City and Boston? San Francisco (where about 22 percent of commuters use public transit) comes to mind.  San Francisco had 79 nonfatal shootings in 45 census tracts (18 percent of its 240-odd census tracts).  However, 11 percent of those shootings were in only one census tract, and its worst four tracts (or its worst 10 percent)  had 20 shootings, or 26 percent of its shootings. In other words, San Francisco’s shootings were actually more concentrated in a few blocks than those of Oklahoma City.

What about New York City – a city with a reputation for crime worse than its actual crime rates? New York had 789 shootings in 467 census tracts (which, I note, is only about 20 percent of its tracts).  If I am counting correctly, New York had 179 shootings in its worst 46 tracts, which means its most dangerous tracts had 22 percent of its shootings, a number pretty comparable to that of Oklahoma City.  The worst 93 tracts (or worst 20 percent) had 305, or 38 percent- a number slightly below that of Oklahoma City, but not hugely below.

To sum up so far: in my three compact cities (New York, Boston, San Francisco) the worst 10 percent of tracts had between 22 and 32 percent of its nonfatal shootings.  In Tulsa and Oklahoma City, the worst 10 percent of tracts had between 21 and 26 percent of its shootings.  In other words, shootings were actually more concentrated in a few areas in the most compact cities, which sort of discredits the argument I raised above (i.e. that compact cities have crime everywhere).

So far I’ve been comparing cities that (by American standards) are relatively safe.  What about more dangerous cities?  Among the six cities that traditionally have the highest levels of density and transit use, Washington, Chicago, and Philadelphia have generally been far more crime-ridden than New York, San Francisco, or Boston.

Philadelphia had 1341 nonfatal shootings in 277 census tracts. This suggests that the greatest difference between Philadelphia and cities discussed above is that its tough areas are tougher: among census tracts with any shootings at all, the average tract in Boston, Tulsa, Oklahoma City, New York and San Francisco had between 2.2 (Boston) and 1.7 (New York) shootings.  By contrast, among census tracts with shootings in Philadelphia, the average tract had almost 5 shootings. 

Philadelphia’s worst 10 percent of tracts (that is, 28 tracts) had 453 shootings- about 34 percent of its shootings, far more than in any of the lower-crime cities discussed above.  In other words, violence in Philadelphia is actually more concentrated in a few tough areas than in any of the low-crime cities discussed above.   

Similarly, Chicago had 2513 nonfatal shootings in 523 tracts.  If I am counting correctly, the worst 10 percent (52 tracts) had about 790 shootings, or 31 percent of its shootings.  Washington had 698 nonfatal shootings in 134 census tracts. 202 of these shootings (or 29 percent) were in its thirteen most violent tracts.   

And like Philadelphia, Chicago, and Washington had about five shootings per census tract (among the subset of tracts with shootings). In sum, the three violent but compact cities had a higher percentage of its shootings in the most violent 10 percent of its tracts, compared to safer cities (whether those cities were as compact as New York or as sprawling as Oklahoma City).

How does this compare to more sprawling cities that also had lots of shootings? Is crime in those cities limited to just a few neighborhoods to a greater extent than in Washington or Chicago? Three large cities had shooting rates higher than Philadelphia and fewer than 3000 people per square mile- New Orleans, Memphis and Atlanta.  (Atlanta had 25 homicides per 100,000 residents, about the same as Philadelphia, and New Orleans and Memphis had twice that many).

In New Orleans, 118 of the 178 city’s census tracts had a shooting in 2023 and there were 450 shootings. The twelve most violent tracts had 157 shootings, or about 34 percent of the shootings.  In other words, shootings in New Orleans were about as concentrated in its worst areas as in Philadelphia.

In Memphis, there were 633 shootings in 150 census tracts, more than two-thirds of its 209 tracts. 183 of the shootings (or 29 percent) were in the most violent 10 percent of these tracts with shootings (that is, the fifteen with the most shootings)- a number comparable to that of Washington.

In Atlanta, there were 436 shootings in 113 census tracts. The most violent 10 percent of these tracts had 116 shootings, about 26 percent — slightly lower than that of Washington, but still more than in lower-crime cities. 

So to sum up: in my three high-crime sprawl cities, the most violent 10 percent of tracts had between 26 and 34 percent of the city’s shootings.  In my three high-crime compact cities, the most violent 10 percent of tracts had between 29 and 34 percent of its shootings.  So it does not appear that, among the subset of high-crime cities, shootings are any more concentrated in one group than another.

In sum, it could be argued that sprawling cities are safer because violence is more concentrated in a few bad neighborhoods than in less sprawling cities.  However, the data does not support this view.  

*I chose to focus on nonfatal shootings because they are more common than fatal shootings, which means we have more data to play with.

**I picked Oklahoma City because I saw it mentioned in a tweet comparing Boston and Oklahoma City rates.  Also, people who enjoy turning crime into a partisan issue might wish to note that Oklahoma City has had Republican mayors in recent years.