Ranking congestion reports

There are now a slew of reports that rank congestion. The grand-daddy of these reports is the Urban Mobility Report from Texas A&M University’s Texas (A&M) Transportation Institute (TTI). Texas A&M also performs a separate analysis of congestion for the Federal Highway Administration. Some of the highway GPS data vendors (TomTom, Inrix) put out their own reports. Maybe I have missed some.

These reports are useful for comparisons across cities (which city is most congested?) and ideally for comparisons across time (is city X getting more or less congested over time). It is more difficult at a national level, because then you need to weight the congestion by those who experience it, with recognition that many people experience very little congestion or are not in a monitored place.

Congestion in Sydney

Reading the Urban Congestion Report we find that congestion increased from 2007 pre-recession to 2015, the travel time index (cleverly abbreviated TTI) rising from 1.26 to 1.37 over time time period on the roads considered. Congestion is getting worse. The hours of congestion dropped however from 5:20 to 4:33.  Congestion is getting better.

Reading the Urban Mobility Report the travel time index rose over the same years from 1.21 to 1.22. Congestion is getting worse. Delay per commuter remains at 42 hours. Congestion is unchanged.

TomTom does not as far as I can tell produce a national average. [Disclosure, TomTom has a partnership with the Accessibility Observatory, their disaggregate data are inputs to our National Accessibility Evaluation].

Inrix says the average commuter wastes 50 hours per year in congestion, but does not report a Travel Time Index. We can try to reverse engineer one. At  250 days of commuting, this Inrix estimate of  50 hours is actually 1 hour per week or 12 minutes per day, or 6 minutes in each direction. A commute of say the national average of 25 minutes or so each way and implies a freeflow time of 19 minutes. 25/19 = 1.31. But 6 minutes is not really that much (certainly it is a cost), and averages can be misleading, it is more likely something like 3 minutes for 4 days a week and 18 minutes 1 day. Unreliability is an issue.

In contrast the TTI estimate of 42 hours is 10 minutes per day or 5 minutes each way, which gives us an estimate of 1.2. This is lower than the 1.37 TTI they report. The estimate also differs by 20% from that implied by INRIX. Recognizing different methodologies, this is still a lot given they are presumably using the same data.

We can compare individual cities as well. I will pick two that I have familiarity with: Minneapolis and Washington for the most recent year.

TTI data is the same from the two reports here (though it differs above). I am not clear why it should be identical.

INRIX doesn’t produce a travel time index (ratio of freeflow to congested time) so I will construct it from the 75 hours of congestion, 250 commute days per year, and a one-way commute time for the metro area of Washington DC of 30 minutes. 75 hours of congestion implies 9 minutes of congestion per commute, or a 21 minute freeflow time in DC. 30/21 = 1.47.

TomTom produces a congestion level. I will add 1 to that to get the Travel Time Index equivalent. It is reported for Morning and Evening peak separately. I average the two

        TTI        INRIX        TOMTOM

MSP 1.39       X                 1.35

DCA 1.59       1.47           1.50

The general trends are similar, DC is not surprisingly more congested than MSP. The published numbers are describing the same thing with different methodologies and slightly different sources.

So in terms of most congested for these two cities, TTI comes in first and INRIX comes in third. However nationally, INRIX says we are more congested than TTI.

TomTom doesn’t report details on every city, but they profile selected cities, for instance Pittsburgh. They say PGH has 81 h of extra travel per year. But that is not just “commuting”.  INRIX does not report PGH. TTI says 39 hours per commuter per year. These numbers are not directly comparable either. Most travel is not work travel, and some of that experiences congestion. 81 h implies 13 minutes per day for a 365 day year.

Someone with more time and/or funding could do this analysis more systematically.

One additional problem with the GPS-based data sets is the sparseness of data in previous years, and the short-term time trend available. The problem with loop detectors is that more-or-less they only cover freeways. Also older data is likely suspect due to the lack of detector deployments in the 1980s and 1990s.

Many state DOTs have their own congestion reports (e.g. Minnesota). These are undoubtedly more accurate locally, but cannot be compared between places in other states due to differences in methodologies.

Now, I certainly don’t believe congestion reduction is an appropriate goal for transportation, but it may be a means to achieve a more appropriate goal like accessibility. There are many ways to reduce congestion, I have identified 21 strategies in an earlier post.

The measures are usually place weighted, so don’t account for trips, and cannot address the question if 30 miles at 30 miles per hour is better or worse than 60 miles at 60 miles per hour, both of which take an hour, the latter of which results in twice as many vehicle miles traveled (but may also have higher accessibility). For that you need more information.

Still, people like these indices, and they could be useful as a time series indicator of whether traffic is worsening or not. But given the high variability between sources, my advice is to remain skeptical.


Further reading: