What Causes Long Commutes?
March 23, 2009
According to the Census Bureau, New Yorkers have the longest average commute of anyone in America. New York City also has the most comprehensive transit system in America, the most daily riders by a long shot, and the highest percentage of residents who commute via transit in the country. I've seen arguments that the extensive use of transit in NYC is the cause of these long commutes, that transit is slower and less efficient in getting someone from point A to point B, and therefore policy should that attempts to expand transit will only make commutes worse (read more time-consuming). While I can't argue with the first two points, I take major issue with the third.
Using data from the Census's American Community Survey, take a look at the relationship between commute time and transit usage for America's 50 biggest metro areas:
The relationship is similar if I plot the relationship between commute time and individuals who commute by means other than driving (ie. transit, walking, bicycling, etc.):
If you're quick to assume that correlation means causation, you could make the case that transit does equate to longer commute times. But look what happens when I plot commute times against the Texas Transportation Institutes's congestion index:
It seems just as reasonable to suggest that long commute times stem from congestion problems as much as they do from transit inefficiencies. In which case, policy that seeks to reduce commute times should primarily target congestion-reduction. Additionally, well-designed transit systems could serve as an excellent means of reducing congestion. Yes, it is an expensive option, but given the choice between new lane miles and new transit lines, policymakers should opt for the latter. The beauty is that the relationship between drivers and congestion isn't necessarily linear; even a small decline in drivers can result in a big decline in congestion.
New transit lines would also decrease some of the inefficiency inherent in these existing systems; and even if commute times are still longer, one thing that people who drive everywhere don't seem able or willing to comprehend is that not all commutes are created equally. A 30 minute commute via car, train and bike are all significantly different. Driving alone means that you're spending an hour of your day, well, driving alone. An hour round-trip commute via train or bus gives you the opportunity to spend some time reading the newspaper, magazine, book, or sleeping. An hour-long commute via bike is an hour of exercise per day, and I'm not sure there are many who couldn't use more exercise. So while longer commutes are worse than shorter commutes, regardless of mode, they're not necessarily equally worse.
Using data from the Census's American Community Survey, take a look at the relationship between commute time and transit usage for America's 50 biggest metro areas:
The relationship is similar if I plot the relationship between commute time and individuals who commute by means other than driving (ie. transit, walking, bicycling, etc.):
If you're quick to assume that correlation means causation, you could make the case that transit does equate to longer commute times. But look what happens when I plot commute times against the Texas Transportation Institutes's congestion index:
It seems just as reasonable to suggest that long commute times stem from congestion problems as much as they do from transit inefficiencies. In which case, policy that seeks to reduce commute times should primarily target congestion-reduction. Additionally, well-designed transit systems could serve as an excellent means of reducing congestion. Yes, it is an expensive option, but given the choice between new lane miles and new transit lines, policymakers should opt for the latter. The beauty is that the relationship between drivers and congestion isn't necessarily linear; even a small decline in drivers can result in a big decline in congestion.
New transit lines would also decrease some of the inefficiency inherent in these existing systems; and even if commute times are still longer, one thing that people who drive everywhere don't seem able or willing to comprehend is that not all commutes are created equally. A 30 minute commute via car, train and bike are all significantly different. Driving alone means that you're spending an hour of your day, well, driving alone. An hour round-trip commute via train or bus gives you the opportunity to spend some time reading the newspaper, magazine, book, or sleeping. An hour-long commute via bike is an hour of exercise per day, and I'm not sure there are many who couldn't use more exercise. So while longer commutes are worse than shorter commutes, regardless of mode, they're not necessarily equally worse.
It would be interesting if we you could plot dedicated right-of-way transit alone. Buses have to fight traffic just like everybody else of course. Or maybe incorporate bus use in one of the charts some other way.
Also, in the cases of New York and Washington, there is a big difference between transit and commuting in the urban core and the far-flung super-low density suburbs and exurbs.
If this study does regions, it is skewed by those who live in the Shenandoah and commute to Washington and live in the Poconos in Pennsylvania and commute to New York.
I think the study would look different if you only compared people inside the Beltway in Washington and in the five boroughs and just across the Hudson in New Jersey. Those cores are very transit-rich. Compare those places to places without cores like Los Angeles.
The other thing to be noted is that the more economically vibrant the city, the more congested. This graph seems to be somewhat correlated to economic vitality.
Gleb, good suggestion. I will look into whether or not this is possible.
Cavan, these plots use Census defined Metropolitan Statistical Areas. So yes, they do include the far flung suburbs of both Washington and New York.
I'd be interested to recalculate these plots using different geographies, but I'm not sure how to appropriately compare different geographies. For instance, when you say Los Angeles, should that only include Los Angeles County? For that matter, maybe I should ditch comparing MSAs altogether and instead look at only the primary counties. Of course, in bigger metros, this could require including multiple counties and incorporated areas, but it could lead to a different result.
The problem is many studies have shown that transit does little to reduce congestion (ex: Stopher, 2004). More transit tends to lead mainly to more total trips, displacing very few drivers. While improving mobility may be a good of its own, this will not reduce the travel times shown in your third graph.
Even more significantly, studies of individual commuters show that transit riders have on average almost double the trip time of passenger car riders. The high travel time shown for New York isn't mainly or only from drivers stuck in traffic, its from the extremely lengthy trips of mass transit commuters (around 50 minutes). For the same reason, European countries with extremely high transit usage also show extremely long commutes (as does Japan).
There is also the problem that transit systems (especially rail) tend to focus commercial development in a central business district, requiring more monolithic commute patterns, and drawing more non-transit passengers to the same area (see: downtown Manhattan), and, therefore, yes occasionally, increasing congestion as opposed to those areas with more dispersed employment.
And finally you could say that the ride on transit is better because people can read etc, but obviously plenty of people would say that they enjoy the feeling of control in a car, the ability to listen to the radio, and that the stress of traffic is still better than the stress of making transfers and dealing with strangers pushing you out of the way to board. In other words, the benefits to an individual on either mode are debatable and in any case are internalized in peoples' decision as to which mode to take.
Good post Rob.
Nice post. Nice to see someone actually use some statistical techniques to back up their comments. I'm sure with a little more work, maybe not just by you, some more light will be shed on the subject. I'm fairly certain of one thing though, in the "chicken or the egg" context, availability of transit does not cause congestion. If anything, it minimizes it. Just count the number of cars on the road with one person in it, and you'll get your answer. Transit has nothing to do with those most of those people. Unless they get picked up and dropped off at their doors, they aren't going to use transit.
Steve C.