Traffic deaths and safety: who's really the safest?

Michael Lewyn's picture

William Lucy of the University of Virginia has written extensively on the question of whether outer suburbs are safer than cities or inner suburbs; he argues, based on traffic fatality data, that outer suburbs are certainly less safe than inner suburbs, and maybe even less safe than cities. (1) 

However, Lucy's analysis is not particularly fine-grained: it analyzes data county-by-county, rather than town-by-town. What's wrong with this?  Often, suburban cities within a county are quite diverse: some share the characteristics of inner suburbs (e.g. some public transit) while others look more like exurbs.  So I wondered whether there is any significant 'safety gap" between inner and outer suburbs. 

Thanks to the City Data website ( this problem is at least partially soluble. The City Data website has traffic fatality data for some individual communities.

So I decided to give the data a try for one metro area I am somewhat familiar with. Rather than trying to analyze every single suburb of a region, I decided to make a case study of one or two metro areas that I am at least somewhat familiar with, and to focus on suburbs that compete with each other for residents of roughly the same social class. 

First, I examined St. Louis and some of its suburbs.  In 2009, the city of St. Louis had 10.9 car crash fatalities per 100,000 people, and 40.3 murders,(2) for a total violent death rate of 51.2 per 100,000.  How do its suburbs compare? 

First I looked at two central-western inner suburbs popular with upper-middle-class professionals: University City and Clayton.   Between 2005 and 2009, University City averaged 3.8 murders and 1.1 car crash fatalities per 100,000, for a total violent death rate of 4.9/100,000. During the same period, Clayton averaged 2.4 murders and 1.2 car deaths per 100,000, for a total violent death rate of 3.6 per 100,000.  Obviously, these suburbs are far safer than the city of St. Louis, whether measured by car crashes or murders.  

How do outer suburbs compare?  Rather than focusing on the outermost exurbs, I wanted to look at places that compete with University City and Clayton for residents: central-western outer suburbs where bus routes begin to thin out, but are still developed upper-middle-class suburbs rather than ruralizing exurbs.  In particular, I focused on Chesterfield, a booming, job-rich suburb far beyond the region's outer beltway (I-270).  During 2005-09, this suburb had 2.6 car crash deaths per 100,000 and 1.3 murders, for a total of 3.9 violent deaths per 100,000- a little safer than University City, a little more dangerous than Clayton.  (I note that some nearby suburbs, such as Manchester, had even fewer fatalities). 

Then I jumped the St. Louis county Line into St. Charles County, and looked at St. Peters, 25 miles from the city of St. Louis.  St. Peters had no murders, but had 6.2 traffic fatalities per 100,000 residents during 2005-09- definitely more than any of the other suburbs listed above.  Nearby St. Charles had 1.3 murders and 5 traffic fatalities per 100,000 people per year during this period, for a violent death rate of 6.3 per 100,000.

St. Charles County as a whole (which is less dense and thus less developed than St. Peters) looks much more dangerous than St. Peters. Although no statistics were available for 2007-09, 2005-06 traffic deaths averaged 11.2 per 100,000- a result consistent with Lucy's analysis.

Then I looked at another metro area, Buffalo,  Here I varied the analysis by focusing on the middle-class eastern suburbs, as opposed to the most affluent suburbs (which, in Buffalo, are northeast of the city).  The city of Buffalo averaged about 20.4 murders and 10 traffic deaths per 100,000/year between 2005 and 2009.  Cheektowaga, a large inner suburb, averaged 0.5 murders and 4 traffic deaths per 100,000, for a total of 4.5 violent deaths per 100,000.  Because the region's transit system mostly ends at Cheektowaga's eastern boundary (Transit Road) I would guess (if my St. Louis findings were typical) that suburbs further east had higher traffic death rates.  In fact this was the case.  Lancaster, the suburb immediately to the east of Cheektowaga, averaged 21.2 traffic fatalities per 100,000/year from 2005-09. and Elma (just south of Lancaster) averaged 14.3 traffic deaths.  Again, inner suburbs were safer than outer suburbs- though here the gap was much greater than in St. Louis.  (But assuming arguendo that Elma and Lancaster had zero murders, they were still safer than Buffalo).

A look at the more affluent northern suburbs showed a similar growth of traffic deaths as one moves further out.  Kenmore, a tiny, walkable mini-suburb, averaged 1.3 violent deaths per 100,000 (all traffic) from 2005 to 2009.  Amherst, which contains some totally car-dependent neighborhoods and a few with adequate bus service, averaged 4.2 violent deaths per 100,000 (1.1 murders, 3.1 traffic). Tonawanda, a similar northern suburb, also had 4 traffic deaths per 100,000 (and zero murders, giving it a roughly identical violent death rate to Amherst). But exurb-like Clarence averaged 9 traffic deaths per 100,000.  

So if St. Louis and Buffalo are any guide, it appears that the impact of traffic deaths (and the overall level of violent death as well) drops dramatically between the city limits and the innermost suburbs, doesn't change much between the first couple of rings of suburbs, and then rises dramatically in the least developed, least transit-friendly, most exurb-like suburbs. 


(2) Lucy emphasizes the number of homicides caused by strangers.  However, these statistics are far less easy to access and so I have not used them.   Moreover, some homicides are of uncertain origin.


Michael Lewyn is an assistant professor at Touro Law Center in Long Island.



Great analysis, some questions

Great, detailed analysis, as usual, Michael. I charted your numbers for future use.
I had trouble understanding the particular mix: deaths by assault and deaths by crash. One has socio-cultural foundations and the other technological and planning bases; totally unrelated etiological grounds. Location can hardly serve to explain both.
When it came to understanding crashes the inference is:
"........and then rises dramatically in the least developed, least transit-friendly, most exurb-like suburbs. " ......... that transit unfriendliness could be a contributing factor. That potential attribution, if it is truly intended, leaves the high number of city crashes to be explained by other factors. Which would they be?

Urban Pattern Associates

Missing the Point

Michael, this is all very interesting, but you have politely discredited Lucy's research unfairly. Your second footnote may as well be the headline. The reason that your data is so different from Lucy's, I believe, is that you have not limited murder to murder by strangers. My understanding of the careful study that you undermine so hastily is that it depends utterly on this qualification. Disagree with the premise if you like, but don't compare apples to oranges and only mention the fruit distinction in your footnotes. Bad science.

Jeff Speck AICP

Michael Lewyn's picture

good point

I agree that I probably should have discussed Lucy's "stranger homicide" point in more detail. Having said that, part of the reason I didn't is that I wasn't sure the difference mattered as much as one might think, at least for more dangerous cities.

Nationally, about 12 percent of homicides involved strangers, and about 44 percent involved unknown relationships ( ).

So let's apply that to St. Louis. Even if you use the 12 percent figure for St. Louis, St. Louis has about 5 stranger homicides and 11 car deaths per 100,000 for a total of 16- far more than any inner suburb, and even a bit more than St. Charles County.

For safer cities, the results are somewhat different. If you use the 12 percent figure for Buffalo, it has about 2.5 stranger homicides and 10 traffic deaths per 100,000, for a total violent death rate of 12.5- less than the outer suburbs but still far more than Cheektowaga or other inner ring suburbs. If you assume most of the "unknown relationship" homicides are stranger homicides, Buffalo's violent death rate rises to outer suburb levels.

Either way, first-and second-ring suburbs are far safer than either cities or the outermost suburbs.

Having said that, there is another respect in which comparing central cities with smaller jurisdictions is off base. People don't choose cities as a whole but neighborhoods within cities. It seems to me logical that crime rates differ far more in a city of 400,000 than in a suburb of 50,000. In a future blog post maybe I'll find a way of addressing THAT issue.

Some relevant points, perhaps?

Michael: your data-crunching exercises are always among the most interesting things on Planetizen. I then always like trying to come up with explanatory factors for you.

I suggest that the lowest socio-economic groups are "priced out" of the innermost suburbs, and as time goes on, tend to "cluster" on the least unaffordable land, which often may be exurban.

The presence of lower socio economic groups in inner cities is an anachronism where it still exists. Land in inner cities is meant to be too valuable and too capable of exploitation for agglomeration efficiencies, to remain "blighted" indefinitely.

The way urban economies and real estate markets evolve, the locations most affordable to lower socio economic groups will always be trending "further away" from the urban centre and from nodes of amenity value such as frequent transit services.

Entrenched "blight" is actually the only way lower socio economic groups will remain located in what should be high value locations. In cases still extant, the tragedy is that the relevant sources of employment for these people have "suburbanised" and their location close to CBD jobs is of little benefit to them. Furthermore, the role models and community leadership material has mostly been sufficiently upwardly mobile to have departed, leaving only the hopeless.

The exurban "slum" is probably preferable to the inner city one due to health advantages and affordable proximity to employment of types that tend to disperse to urban perimeters. If violent crime relates to socio-economic levels, it is the high levels of crime in inner cities that is the tragic anachronism in advanced economies; it does not surprise me at all if the exurbs have it.

Serious automobile accidents are ironically reduced when traffic congestion is so severe that no-one can get up to a speed where an impact will do much damage......!

Michael Lewyn's picture

Not relevant to the places I studied

I focused on areas (Buffalo and St. Louis) where there isn't much gentrification, where the exurbs are the wealthiest communities, and where traffic congestion isn't really an issue- less because I wanted to pick out such communities than because (1) I am familiar with them, (2) they have a lot of small suburbs (while more affluent cities often have annexed lots of territory).

I ran a similar analysis with Atlanta (which has a somewhat more affluent inner city), and there wasn't much difference except that the gap between inner and outer suburbs was smaller until you get to almost-rural areas. But I again focused on the more affluent side of suburbia, not on the lower-income southern suburbs (I think because Lucy's analysis was directed towards the choices of well-off families, and so I was elaborating on his work).

Build Your Own Paper Block City

Urban Fold is an all-inclusive kit that allows anyone to build the city of their dreams with a few simple folds.
building block set

NEW! Build the world you want to see

Irresistible block set for adults when placed on a coffee table or desk, and great fun for kids.
Book cover of Unsprawl

Unsprawl: Remixing Spaces as Places

Explore visionary, controversial and ultimately successful strategies for building people-centered places.
Starting at $12.95
Women's t-shirt with map of Los Angeles

City T-Shirts for the ladies!

Women's Supersoft CityFabric© Fashion Fit Tees. Now available in six different cities.