Traffic deaths, safety and suburbia, Part 2

Michael Lewyn's picture

A couple of months ago, I wrote a blog post comparing the safety of inner suburbs and outer suburbs. (See )

My post showed that (in least in the metropolitan areas I looked at) inner suburbs were safer than outer suburbs, because violent deaths from murder and traffic combined were lower in the former.

However, I devoted only a couple of lines to core cities (and, as some of the comments suggested, perhaps a few lines too many).(1)  It seems to me that comparing core cities to suburbs is not quite ideal, because typically, a regional central city contains a wide variety of neighborhoods- some very safe, some not so safe.   In an ideal world, one would resolve this problem by comparing individual neighborhoods or groups of neighborhoods to suburbs.  But this technique is not feasible because neighborhood crime statistics are not easily available for most cities (let alone neighborhood traffic fatality statistics). 

An alternative way to compare cities to suburbs is to look at "intown suburbs"- places that, although technically suburbs, are so close to downtown as to effectively be city neighborhoods. For example, in Atlanta, the city limits extend about 10 miles out from downtown in some directions, so Decatur (6 miles out) is an intown suburb.  If a region's intown suburbs have a lower "traffic + homicide" death rate than its outer suburbs, one can plausibly argue that  (parts of) cities truly are safer than outer suburbs.

I began my research with the following educated guess: low-income intown suburbs are more dangerous than wealthy outer suburbs, high-income intown suburbs less so.  Does data bear this out?

I chose Boston as a case study because Boston is one of the few cities with a mix of low-income and high-income intown suburbs.  So I focused on Chelsea (low) and Cambridge (higher).  Between 2005 and 2009, Chelsea had 4.6 murders per year per 100,000 people, and 4.9 traffic deaths, for a total violent death rate of 9.5 per 100,000.  More affluent Cambridge had 1.8 homicides per year and 2.1 traffic deaths, for a total violent death rate of 4 per 100,000.  An even richer (though more suburban) intown suburb is Brookline, which begins about four and a half miles from downtown Boston.  Between 2005 and 2009, Brookline had 0.4 murders per 100,000 per year, and 1.1 traffic fatalities, for a total violent death rate of 1.5 per 100,000 per year. Thus, the low-income intown suburb (Chelsea) was apparently more dangerous than the high-income ones (Cambridge/Brookline).  But how do both compare to real suburbs?

Since William Lucy's research (which led to my work)(2) focuses on the relative desirability of city and suburb, I chose one of the most desirable suburbs as a case study of an outer suburb: Newton (an inner ring suburb about 6 miles from Boston) and Sudbury (18 miles out, the region's fifth wealthiest suburb)(3).  Sudbury had 4.5 traffic deaths per year per 100,000 and 1.2 murders, for an overall violent death rate of 5.7- 40 percent more than Cambridge, and far more than Brookline. Sudbury's figures are fairly typical; in 2005-06 Middlesex County (which includes Sudbury and numerous other, mostly outer-ring, suburbs) had about the same level of traffic deaths.  (No countywide data were available for 2007-09).

 My discussion so far hasn't taken account of the distinction between homicides caused by strangers (and thus arguably more threatening to the public as a whole) and those caused by acquaintances.  Unfortunately, digging up statistics on this issue is difficult- not just because individual municipalities' statistics may not be easy to obtain, but because many homicides are unsolved (which means we don't know whether they were "stranger homicides").  However, nationally we know that about 12 percent of murders were committed by persons who the victim did not know, but that the murderer/victim relationship was unknown as to an additional 44 percent of homicides.(4) If you assume that only the 12 percent figure matters, the gap between Chelsea and Sudbury narrows quite a bit: Chelsea's violent death rate plummets to 5.5, only about 20 percent higher than Sudbury's.  But if you assume that 56 percent of murders involved strangers, then Chelsea's "stranger violent death rate" is 7.5, still about 50 percent higher than Sudbury's.

 Another city with both high- and low-income intown suburbs is Detroit.  Grosse Pointe, less than 6 miles from downtown Detroit, had zero traffic deaths between 2005-09, and only 3.6 homicides per 100,000 per year, for an overall "violent death rate" of 3.6.  By contrast, working-class Hamtramck, surrounded by the city of Detroit, had 7.5 murders per 100,000 and 2.9 traffic fatalities for a total violent death rate of 10.4.    

How does these results compare to those of "nice" outer-ring suburbs? One of the wealthiest suburbs in America, Bloomfield Hills, had no murders and 5.2 traffic deaths per 100,000- slightly worse than Grosse Pointe, but perhaps not significantly so since only one death was involved, artificially knocking up the rate for a very small suburb.  On the other hand, nearby Birmingham had only 2 traffic deaths per 100,000, even lower than Grosse Pointe.    Oakland County as a whole had 6.3 traffic deaths per 100,000 people in 2005-06; however, the county includes inner-ring as well as outer-ring, and middle-class as well as upper-class, suburbs.

 So my examples (admittedly the tip of a very large iceberg) back up the theory that lower-income urban places are less safe than suburbs.  They also suggest that well-off intown suburbs like Brookline and Grosse Pointe are as safe or safer than outer suburbs.  However, the data is unclear as to which adjective ("as safe" or "safer") applies.  

Having said that, I wouldn't pretend any of this is any more than a starting point for discussion.  A more comprehensive article would create a much larger database of (1) intown suburbs, (2) inner suburbs and (3) outer suburbs, and would discuss a broader demographic range of suburbs.




(1)One of the comments pointed out that I did not distinguish between homicides caused by strangers and those caused by acquaintances – for more on this, see my last paragraph on Boston suburbs.

(2) .  Lucy's data is mostly by county; but county-by-county data does not always distinguish inner from outer suburb, or outer suburb from rural area, because outside the largest metro areas a county can include a very wide range of places.  For example, Detroit's Oakland County borders Detroit, but contains a lot of suburbs that are quite far from the city.

(3)The four wealthiest (Weston, Dover, Carlisle and Sherborn) lacked crime and/or traffic data for all five years studied.  


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



Todd Litman's picture

Implications for Community Safety

Thanks Michael for this analysis. It contributes to an important discussion about what constitutes a truly safe community.

However, I think the data you present requires additional analysis to be useful for decision-making by either individual households searching for safe neighborhoods or planners looking for transport and land use policies that improve overall community safety and health.

This and other analyses indicate that high geographic murder rates are mainly explained by concentrated poverty, while traffic fatality rates increase with annual per capita vehicle travel. A typical household does not become poorer and more murderous by moving to an urban neighborhood, but will increase their annual vehicle travel and associated traffic fatality rates by moving to a more automobile-dependent suburb.

As a result, a typical household is likely to become overall safer by locating in a more central, multi-modal neighborhood than in an automobile-dependent suburb, and for society there are probably significant net reductions in total violent death rates if non-poor households shift from automobile-dependent sprawl to more accessible, multi-modal neighborhoods. Described differently, individuals looking for overall safety and health should choose wealthier urban neighborhoods, and avoid both impoverished urban and automobile-dependent suburban neighborhoods. Policy makers looking to increase overall safety should work to reduce both poverty and automobile-dependency.

For more information on the relationships between sprawl and traffic fatality rates see:

David Clark and Brad M. Cushing (2004), “Rural and Urban Traffic Fatalities, Vehicle Miles, and Population Density,” Accident Analysis and Prevention, Vol. 36, pp. 967-972.

Eric Dumbaugh and Robert Rae (2009), “Safe Urban Form: Revisiting the Relationship Between Community Design and Traffic Safety,” Journal of the American Planning Association, Vol. 75, No. 3, Summer 2009; at

Reid Ewing, Richard A. Schieber and Charles V. Zegeer (2003), “Urban Sprawl As A Risk Factor In Motor Vehicle Occupant And Pedestrian Fatalities,” American Journal of Public Health (

Todd Litman (2012), "Risk Versus Dread: Implications for Planners; or Let's Not Let The Terrorists Win," Planetizen ( ).

Todd Litman and Steven Fitzroy (2006), "Safe Travels: Evaluating Mobility Management Traffic Safety Benefits," Victoria Transport Policy Institute (; at

Wesley E. Marshall and Norman W. Garrick (2011), “Evidence on Why Bike-Friendly Cities Are Safer for All Road Users,” Environmental Practice, Vol 13/1, March; at

Todd Alexander Litman
Victoria Transport Policy Institute
"Efficiency - Equity - Clarity"

Michael Lewyn's picture

a few more thoughts

Does a household experience higher risk by moving to an urban neighborhood? Depends on the neighborhood and the risk- since if you move to a rougher urban neighborhood you might not be more murderous but you might be more likely to be exposed to that of others.

But the issue gives me an idea for Part 3 - maybe a comparison of working-class intown suburbs vs. comparable outer suburban neighborhoods (for example, Atlanta's East Point vs. Clayton County, Detroit's Hamtramck vs. Inkster, etc)

Todd Litman's picture

Determining How Risks Actually Change With Location

Yes, Michael that is why it is important to determine how location actually affects risks, for example, the degree that moving to an urban neighborhood actually increases crime risk, as opposed to reflecting confounding factors such as concentrated poverty. For example, non-acquaintance crime risk might increase if you move from a suburb to a poorer urban neighborhood, but acquaintance crime risk (being assaulted by family members or close friends) probably increases little. Since acquaintance crime appears to be the dominant type of risk, there is probably little actual net risk increase to most individuals who move to a typical urban neighborhood. To the degree that reducing the concentration of poverty through more income mixing (e.g., a poor child having more middle- and higher-income neighbors who provide positive roll models and employment connections) it may reduce total regional crime.

Todd Alexander Litman
Victoria Transport Policy Institute
"Efficiency - Equity - Clarity"

Michael Lewyn's picture


You wrote: "acquaintance crime risk (being assaulted by family members or close friends) probably increases little."

It depends. I think acquaintance crime risk is probably highest where you have the most acquaintances. Assuming that you have more acquaintances in your old neighborhood than your new one, if you are moving from a suburb TO a rough urban neighborhood your risk might not change so much (unless you are a teenager, in which case the neighborhood public school will give you many new acquaintances who will be in their most crime-prone years). On the other hand, if you are LEAVING the rough urban neighborhood, your risk may decrease dramatically (assuming that you have many acquaintances in the neighborhood you left behind).

Todd Litman's picture

Do More Acquaintances Really Breed More Crime?

I'm skeptical, Michael. This would be true if crime follows a disease distribution model, so having more contacts increases total crime, but I suspect that in most situations crimes-per-criminal are fairly constant so where people have more acquaintances the risk per contact is lower. For example, if there are ten criminally-inclined people in a neighborhood and a thousand non-criminals move there, the risk per non-criminal would be small, assuming that an inclination toward crime is not contagious. In fact, criminality may decline if morality is contagious, or if criminality is partly caused by poverty which is reduced by income mixing (e.g., if migration of gainfully employed people to an area that previously had concentrated poverty helps inspire morality or increases economic opportunity).

These questions are important because concentrated poverty and crime are to some degree self-fulfilling prophecies. In the past, suburban areas were attractive in part because they were considered "safe places to raise a family," because poor people couldn't afford their relatively high transport and housing costs, which created pockets of concentrated poverty in urban neighborhoods. This pattern is less consistent now that cities have more mixed incomes and poverty is suburbanizing, but many people still assume that cities are dangerous, based on incomplete and biased risk analysis, such as focusing on selected examples of urban neighborhoods with high murder rates while ignoring high traffic fatality and murder rates in suburban and rural areas. In fact, in Canada, homicide rates are higher in rural areas than in cities and small towns ( ), and in the U.S., rural Southern states have higher homicide rates than Northern urban states ( ), yet these risks seem to attract little attention compared with exaggerated fear of urban neighborhood crime.

Todd Alexander Litman
Victoria Transport Policy Institute
"Efficiency - Equity - Clarity"

One thing to consider on the

One thing to consider on the traffic safety side is the type of road network and walkability of the neighborhood and suburb. I live in Grosse Pointe Park and work in Detroit so I'm very familiar with your Detroit area locations discussed.

Grosse Pointe has no interstate and the roads all have traffic counts of less than 30k vehicles a day running through. It is very walkable grid city with an active pedestrian culture. Birmingham also has no interstate running through it with one road in excess of 30K vehicles a day. It is also a very walkable grid city with an active pedestrian culture. Hamtramck has I-75 running through a portion of it with some heavy industry traffic but otherwise no roads in excess of 30k vehicles a day with a very walkable grid city and active pedestrian culture.

Bloomfield Hills on the otherhand has no interstate, but has at least 3 major roads in excess of 30k vehicles a day with a traditional suburban style road network. Inkster (mentioned below) has no interstate but has at least 4 major roads in excess of 30K vehicles a day, heavy industry traffic and a traditional suburban style road network.

Detroit is hard to compare to these types of cities from a road standpoint. Detroit has traditional neighborhoods with walkable street grids and a pedestrian culture, as well as suburban style neighborhoods with no pedestrian culture.

Interesting topic though and there seems to be some promise in finding some way to slice some numbers to analyze this.

Brand new! Urban Grid City Collection

Each city has its own unique story. Commemorate where you came from or where you want to go.
Grids and Guide Red book cover

Grids & Guides

A notebook for visual thinkers. Available in red and black.
building block set

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.
Wood necklace with city map

City Necklaces

These sweet pendants are engraved on a cedar charm with a mini map of selected cities. The perfect gift for friends and family or yourself!