Key to safer roads is identified in California study


Philip Langdon

Issue Date:

Thu, 2009-01-01

Page Number:



A review of fatalities in 24 cities shows that safety grows as street networks become denser.

Transportation researchers Wesley Marshall and Norman Garrick fed the facts from more than 130,000 vehicular crashes into their computers in recent months, hoping for a systematic answer to a life-and-death question: How can America’s streets and roads be made safer?

Highway departments have typically focused on “finding the most problematic locations and fixing those roads or intersections,” say Marshall and Garrick of the University of Connecticut’s Center for Transportation and Urban Planning. But the conventional approach doesn’t go far enough, the two researchers assert. They felt it was time for “a more comprehensive approach to road safety that takes into account the complete street network.”

Consequently the two gathered data on nine years of road safety records for 159 California cities of 30,000 to 150,000 population, and ultimately zeroed in on 24 medium-sized cities with some of the best and worst crash frequencies.

Their conclusion: The most unsafe cities in California, in terms of traffic fatalities, are the newest ones — those developed primarily since 1950. The cities with the fewest fatalities, by contrast, are those with significant portions built before 1950.

The newer cities tend to have more “dendritic” networks — branching, tree-like organizations that include many cul-de-sacs, limiting the movement of traffic through residential areas. They also don’t have as many intersections. The pre-1950 cities, on the other hand, tend to be more grid-like, giving motorists many more routes to choose from.

For several decades, traffic specialists believed a tree-like hierarchy of streets was superior because it made residential neighborhoods quieter and presumably safer. But an American Society of Civil Engineers (ASCE) study cited by the UConn researchers points out that more-connected street networks tend to reduce travel speeds. That’s important because even a small reduction in speed can boost safety — mainly by reducing the severity of the accidents.

A recent report from Europe found that when average vehicle speeds drop by just 5 percent, the number of injuries drops by 10 percent and the number of fatalities falls 20 percent. Extensively connected street networks may not have fewer crashes over all, but the crashes that occur are less likely to leave someone dead.

The ASCE study also concluded that street networks containing many cul-de-sacs increased travel demand on arterial roads by 75 percent and on collector roads by 80 percent, compared to a gridded street design. That, too, may help explain the higher fatality rate associated with the street networks that became prevalent after 1950.

The 24 cities
The safer cities identified by Marshall and Garrick were municipalities such as Santa Cruz, San Mateo, Berkeley, La Habra, and Danville. The more dangerous cities were places such as Turlock, Rialto, Victorville, Antioch, and West Sacramento. The safer cities had roughly twice the population density of the more dangerous cities — 5,736 people per square mile, versus 2,673 per square mile.

In the period from 1997 through 2005, the safer cities experienced 3.1 fatal crashes per year per 100,000 population, while the more dangerous cities suffered 10.1 fatal crashes per 100,000 population — a death rate more than three times as high.

New urbanists often identify “connectivity” as a critical factor in street design. Marshall and Garrick suggest that the focus should be defined a bit differently. They say that what matters most, from the perspective of reducing deaths in traffic accidents, is how many intersections there are in a given land area. The more dangerous cities had 41 percent fewer intersections per square mile. (In the lingo of Marshall and Garrick, this measure is “real intersection density.”)

A street network with low intersection density might have fewer than 81 intersections per square mile. A street network with high density — which is much safer — might have more than 225 intersections per square mile. With a high number of intersections per square mile, fatalities decline and the frequency of severe injuries decreases, too. Communities with many small blocks — a description that fits many older cities and new urbanist developments — tend to have more intersections. It should be noted that not even the safer cities have a uniformly high intersection density. That group of cities includes newer, poorly connected areas — just significantly fewer of them  than the “less safe” cities.

Municipal connectivity requirements
“We need to be much clearer about what we want in a transportation network,” Garrick, the director of the UConn center, said in presenting a preliminary report on the findings to a CNU Transportation Summit in Charlotte in November. (He and Marshall provided more complete findings to New Urban News this January.)

Communities influenced by new urbanist thinking have begun requiring connectivity in their street networks, mainly because they want to make it easier for people to walk to their daily destinations. Marshall and Garrick are trying to help communities do a better job of this — by making finer distinctions in how connectivity and network density are defined and measured.

In a paper the two are preparing for CNU, they note that some municipalities — including San Antonio, Texas; Cary, North Carolina; Orlando, Florida; and Middletown, Delaware — measure connectivity by counting the number of “links” (road segments between intersections) and dividing that by the number of “nodes” (or intersections).

Connectivity, as indicated in a high ratio of links to nodes, appears to help reduce fatalities and severe injuries, according to the California study. But intersection density or street network density, rather than the link-node ratio, seems to be a better measure of safety.

One problem of both these measures is that they’re hard for people to understand intuitively. Neither the link-node ratio nor the street network density calculation is easy for people to translate into actual street patterns. As Marshall and Garrick admit, “most of the existing street network measures are either misunderstood and misused or tough to interpret and difficult to explain to the general public.”

Marshall and Garrick are continuing their quest for the best ways of identifying and measuring well-performing street networks — at both the neighborhood level and the regional level. If these can be identified and communicated fairly simply to the public and to local officials, communities will have a better chance of obtaining street networks that are safe — and presumably pleasurable as well.