Still Measuring Congestion Wrong

National congestion rankings leave traffic-watchers more bufuddled than ever.

How is the Texas Transportation Institute like Britney Spears?

Oops, they did it again.

Yes, the Texas Transportation Institute released their annual rankings of traffic congestion last week.  And for the second time in a decade, the report has undergone a complete change in methodology—one that’s rendered its previous congestion estimates moot.  And for the umpteenth time in a row, they’ve championed a measure of congestion that gets the relationship between congestion and commute time exactly backwards.

seattle flickr rutloThose with long memories will recall that, prior to 2004, the TTI’s Urban Mobility Report  ranked Seattle as having some of the worst congestion in the country.  But then, their 2004 report introduced a new methodology, showing that Seattle-area traffic wasn’t anywhere near as bad as they’d reported before.  The newspapers at the time reported that congestion had “eased” (see, e.g., the chart to the right, from the Seattle Times).  But that simply wasn’t true.  TTI researchers had simply fiddled with their methods—but that fiddling was enough to flip the local congestion debate completely on its head.

Now it’s 2004 all over again.  The Institute has jettisoned the old traffic models used in its 2009 report in favor of more detailed data from traffic analysis firm INRIX—and lo and behold, the national congestion rankings have gone topsy-turvy yet again.  For example…

  • In the 2009 report, Chicago ranked 13th in the country, with 41 hours of traffic delay per peak hour commuter per year.  But in this year’s report, Chicago ranks #1 for worst congestion in the country, with 70 hours of delay.
  • Detroit did the opposite; its tally fell from 52 hours of congestion delay per commuter in 2009 to 33 hours in 2010.

The thing to remember here is that these changes are not real!  Chicago’s congestion didn’t skyrocket, and Detroit’s didn’t plummet.  For the most part, it was the methods that changed, not the congestion itself.  But the city-by-city rankings are so different now that it’s almost as if the Texas Transportation Institute is repudiating its own previous work—or, worse, that they’ve admitted that people shouldn’t have been paying any attention to the city-by-city congestion rankings they’ve been publishing year in and year out.

Which raises a question:  why start paying attention to them now?

The problems with Detroit and Chicago’s rankings are really just the tip of the iceberg.  Across all 15 of the largest urban areas in the TTI report, the 2009 figures just don’t correlate that well with the 2010 figures. The 2009 vs. 2010 delay estimates for the next-smaller tier of cities are a little better correlated.  Still, it’s clear that the new methods represent a massive break from TTI’s previous estimates.

One response to all of this is to say that the authors are admitting that they got the analysis wrong in previous years, but they’ve learned from their previous mistakes.  But I’m just not sure of that.

Consider, for example, their consistent claim that cities that build more roads get less congested.  (You can see their charts to this effect in this pdf.) But they’ve made the same exact claim year after year after year— even after methodological changes have wreaked havoc on their city-by-city rankings.  And look at the list of 14 cities that most clearly “demonstrate” that road growth slows congestion.  Can you really generalize from the experiences of Poughkeepsie, NY or Provo, UT, or from the emptying urban centers of Cleveland and New Orleans, to draw conclusions about road building across the rest of the urban US?

Or consider the findings from INRIX itself, which publishes its own congestion scorecard.  By INRIX’s reckoning, congestion fell by 33 percent from 2007 through 2009, as the double whammy of gas prices and a sour economy took the edge off rush hour.  Similarly, the Washington State Department of Transportation estimated that congestion fell by 20 percent over the same period.  But TTI shows that total congestion delays fell by just 11 percent over the period, while its “travel time index”—the report’s most widely used measure of congestion—fell by just 8 percent.

But perhaps the biggest reason to question the Urban Mobility Report’s usefulness is its continued reliance on the deeply flawed “travel time index.”  I’ve mentioned this before, and the redoubtable Joe Cortright has thoroughly critiqued this problem (along with many others) with last year’s UMR report.  But in a nutshell, the “travel time index” rewards sprawling cities with long commutes.  For the geeks out there, the index represents the ratio between commute times on congested vs. uncongested roads.  So if you’ve got a 10 minute commute that’s delayed 5 minutes by heavy traffic, you have an index of ((10+5)/10) = 1.5; but if you’ve got a 20 minute commute that’s delayed by the exact same 5 minutes, you have an index of ((20+5)/20) = 1.25.  Even though the delays are identical, the travel time index makes the longer commute look like the less “congested” one.

This problem isn’t just theoretical: Joe’s got a great real-world comparison based on the new figures that shows how this works:

According to the UMR, Portland [Oregon] has a worse traffic problem than Nashville, with a Travel Time Index of 1.23. and 36 hours of delay per year per traveler, compared to Nashville, which has a Travel Time Index of 1.15 and 35 hours of delay.  But these data also mean that the average peak traveler in Nashville has to spend a total of 268 hours per year commuting compared to the commuter in Portland who travels only 193 hours per year.  So the commuter in Portland travels 75 fewer hours annually because of shorter travel distance, due in large part to less sprawling development patterns.

In my view, these sorts of problems represent a systemic bias in favor of cities with longer commutes and more sprawling land use patterns—a bias that renders the “travel time index” somewhere on the spectrum between baffling and harmful.

Which brings me back to the similarities between TTI and Britney Spears.  Just like Britney, the Texas Transportation Institute produces popular products year after year after year—and yet, just like a Britney song, there’s a lot less there than meets the eye.

We are a community-supported resource and we can’t do this work without you!



  1. Alan Durning says:

    Clark,My daughter and I think you should back off on Britney. Her songs are catchy, and no one bases billions of dollars of transportation infrastructure investment on them.In conclusion, Leave Britney Alone: (language alert).Alan

  2. Japhet Koteen says:

    Definitely check out/link to the CEOs for Cities critique of the flawed methodology. Very helpful FrontSeat, CNT, and Sightline can develop a quick web based congestion assessment using google maps and traffic data. Travel time required to reach basket of services (employment, services, etc.) that policy makers can use to assess actual mobility on a more holistic scale, with an added factor for potential to reduce traffic via transit (i.e. sensitivity to disruptions in road capacity or collisions, etc.). Keep rockin’ it. Like Britney.

  3. Japhet Koteen says:

    oh. you already got the CEO’s for cities link in there. . . never mind.

  4. Michael Weiser says:

    Since most people drive by themselves, and 98% of people commute by themselves or carry only the driver and passenger, the entire passenger sides of cars are unnecessary.The point of commuting is transporting people yet the result of our commutes is lugging around 3 or 4 empty seats on every trip.Commuter Cars, Lumeno Smera, Nissan and GM are all building and developing ultra-narrow cars which should effectively double the lane capacity of our roads and bridges. See, for example.For buses, check out this link which demonstrates a bus design that allows for traffic to flow underneath where the passengers sit. studies have their place, but it’s best to concentrate on the solution. The manufacturing and driving of ultra-narrow cars and straddling buses make the most sense for easing traffic congestion.

  5. RegionalGuy says:

    Are the raw data available to permit running comparisons using the latest, or all, of the models for each of the years being discussed? That would clearly show the points you are making.

  6. Georgie Bright Kunkel says:

    Headline language is incorrect.”Wrong” should read “Incorrectly.” It takes an old school teacher to catch grammatical errors.

  7. Levin Nock says:

    It is useful to consider the question in terms of mobility for cars vs. access for people. Do cars need to move around without delay (as this metric measures), or do people need quick, convenient commutes and other errands? overview of better performance metrics for transportation networks is available here:

Leave a Comment

Please keep it civil and constructive. Our editors reserve the right to monitor inappropriate comments and personal attacks.


You may add a link with HTML: <a href="URL">text to display</a>