Vanishing Automobile update #28

Analysis of the 2002 TTI Congestion Data

Issued June 21, 2002, updated June 23 and June 28

The Texas Transportation Institute (TTI) released its 2002 annual survey of urban congestion yesterday. Unlike previous years, which often introduced new measures of congestion and revised earlier data, this year's survey merely tacks 2000 data onto the 1982 through 1999 data previously released. Thus, much of my analysis of last year's survey and its implications for transport planners still applies. The survey does add several new urban areas.

TTI has developed several congestion measures over the years. It originally started using the "roadway congestion index." While it still calculates that index, it now prefers the "travel time index," which measures the additional time people waste driving in congestion. It also measures the actual hours of delay, fuel wasted, and the cost of congestion based on the value of people's time plus the cost of fuel.

Here are a few highlights:

I've previously commented on limitations with the Texas Transportation Institute data (see p. 388 of The Vanishing Automobile or update 13). In a nutshell, the data are probably not reliable for comparing two or more urban areas in any one year, but are more reliable for comparing changes in congestion among urban areas over time.

Some agencies and planners were particularly critical about the data this year. The Washington State Department of Transportation announced that it was withdrawing its financial support for the program because it was upset that TTI data portrayed Seattle as being so congested (it is). Transportation planners in Los Angeles were similarly upset about LA being represented as the most congested urban area in America (it probably is).

More significantly, the Surface Transportation Policy Project announced that it would not put out a companion report putting its "spin" on the data "because of doubts about the validity of the data in accurately measuring congestion." The real problem may be that the data simply do not support the group's preconceived notions.

Recognizing the limitations of the data, here are a few more observations on 2002 results. You should be able to download TTI's entire data set from the TTI web site. You can also download Excel files summarizing the data, showing TTI's calculations of 1982 and 2000 congestion indices and the raw data showing 1982 and 2000 miles of driving and road mileages.

Is Portland Congestion Growing Fastest or Third Fastest?

Last year's report indicated that Portland had the fastest growing congestion from 1982 to 1999 using the "travel time index," which measures the amount of time a journey would take during rush hour compared with the same journey in uncongested conditions. In other words, a 30-minute trip at midnight would take 45 minutes during rush hour if the travel time index was 1.50.

This year's report shows Portland as having the third fastest-growing congestion after Los Angeles and the Twin Cities. It turns out that the institute has completely recalculated travel time indices for all cities and all years. Last year, Portland's 1999 index was 1.65; this year, it is only 1.35. The institute says the changes are due to recalculations of travel speeds in the various urban areas.

Since the travel time index and other measures of congestion are based on calculations, not actual measurements of congestion, they need to be taken with a few grains of salt. The raw data are a little more reliable, as they are estimates made by state and local highway departments based on traffic counts.

Growth in Per Capita Driving

The raw data indicate that per capita driving grew by an average of 33 percent in the 75 urban areas reported. The highest growth rates of per capita driving were:

St. Louis    86%
Pensacola    84%
Portland     83%
Albany       78%
Birmingham   75%
Atlanta      72%

Thus, smart growth and light-rail have not had much of an effect on driving in Portland or St. Louis.

On the other hand, Houston, which enjoyed a 79-percent increase in freeway lane miles, had only a 26-percent increase in per capita driving. Perhaps Houston driving is tempered by the fact that many of the new "freeways" are actually tollways.

Growth in Driving vs. Growth in Road Mileage

A recent report from a smart-growth group disputes the claim that urban road construction has not kept up with urban traffic growth. But these data indicate that major highways have failed to keep up with congestion.

In the 75 areas reported by TTI, total driving grew by 71 percent but total road miles grew by only 33 percent. The claim that total road miles is the wrong measure is valid since most driving is done on only a few roads. In the 75 areas reported by TTI, 40 percent of all driving is done on the freeways and another 24 percent is done on the principle arterials. Yet the table below shows that driving on both freeways and principle arterials grew far faster than the lane miles of such roads available.

Percent Growth in Driving and Lane Miles
             Driving      Lane Miles*
Freeways       106            40
Arterials       64            36
All roads       71            33
* Centerline miles in the case of all roads

Of course, if there is a surplus of freeway and arterial capacity, there is no need for new construction to keep up with driving. But TTI data indicate that most of the urban areas in TTI's study were already suffering congestion (as measured by the travel time index) in 1982, when the study began. Thus, there was no surplus in road capacity.

Another way of dealing with congestion that could reduce the amount of needed new road construction is to charge congestion tolls. Such tolls would not only reduce peak demand but would provide funds to build any new roads that might still be needed.

Does New Road Construction Induce More Driving?

This raises the question: Does building roads lead people to drive more? We can estimate the correlation between the growth in roads with the growth of driving in several ways, including:

R represents the correlation between two sets of data. An R of 1 is perfectly correlated, for example (1, 2, 3, 4) and (2, 4, 6, 8). Two sets of 76 random numbers can easily generate Rs as high as 0.1, so an R that is less than 0.1 isn't any better than random.

Thus, building more freeways does correlate with (and may contribute to) more freeway driving. But freeway construction has less of a correlation with total driving. Freeway construction and total road construction have virtually no correlation with per capita driving.

Other studies that claim to find that road construction "induces" more driving may have actually measured something else, such as increased driving due to population growth, income, or other factors.

Thoreau Institute | Vanishing Automobile | Vanishing Automobile Updates