Travel Time to Work
The Census Bureau has a new publication about commuting in the US as of the 2009 American Community Survey. There isn’t much change from 2000 that’s mentioned, but one table of commute time piqued my interest. This is figure 8, on page 11, showing mean travel time to work by mode of transportation and time of departure to go to work.
It is well-known that commutes on transit take much more time than commutes in a car, but the breakdown based on time of departure defied my expectations. I thought transit trips take the least amount of time at rush hour, when frequencies are the highest, and the most amount of time late at night; car trips should be the opposite. Since people with very long commutes would leave earlier, perhaps around 7 to get to work at 9, the longest commutes by car should be for people leaving to go to work between 6 and 7 or between 7 and 8.
However, it turns out that by all modes, late-night commuters travel the longest. Trip-to-work time declines monotonically with departure time until the 9 am-noon category, where it’s at a minimum (my guess is that people in this category are disproportionately academics and other flex-time professionals who live close to work). This is close to expectation for transit, but for cars, it’s weird: why would people take longer to drive when the roads are clear than when they are congested?
I’d reject an explanation based on leaving very early to work. Insignificantly few people travel 3 hours to work. Instead, the only answer I can think of is that the groups of people who travel to work at rush hour and late at night are of different social classes, and this is reflected in commute times. Late-night commuters are usually low-paid service workers at gas stations, casinos, etc.; those also live on the wrong side of their metro area or in low-income exurbs, and need to drive considerable distances to the favored quarter.
Observe that the long late-night commute trend is the sharpest for carpoolers, who in general skew poor and nonwhite: eyeballing the graph, carpoolers who leave for work between midnight and 5 am travel twice as much time a those who leave between 9 am and noon, compared with a 50% time premium for single drivers (who are wealthier) and transit riders (who tend to work in CBDs, which are equally accessible from all directions).
Obviously, I’m going on partial data here; anyone with access to fuller data could potentially trivially refute my theory. If the theory is true, then two things will be observed in a fuller set of data. First, the late-night time premium will be the longest in large metro areas with favored and unfavored quarters, such as Los Angeles and Washington, and shortest in metro areas with less edge city-favored quarter development, such as New York. And second, if the departure time is broken into more granular categories, then the peak travel time will be at night rather than very early in the morning.
Pedestrian Observations 
Maybe you have cause and effect backwards. People with long commutes depart earlier to arrive on time, people with short commutes can travel later.
This is possible, but look at the figure again. The peak travel time is for people leaving between midnight and 5 am. Either they travel to work 3-4 hours, or they just start work at very different hours from everyone else – and in the latter case, they would avoid traffic entirely and should take less time to travel.
When I got the email notification of this post, the first time I read it I thought it said “Time Travel To Work”, and was wondering how they managed to get a city bus up to 88MPH… 🙂
It involves trolleybuses with 1.21-gigawatt motors…
I suspect that people who work ‘off-hours’ may be less sensitive to distance since there is less congestion on the roads. Thus they may actually be travelling longer distances as well as having longer travel times.
For example, there was a period when I was ‘commuting’ from Jerusalem to Tel Aviv for work. It actually was not such a bad commute: as long as I got to TA before 7:00 am, the entire trip was barely an hour. If I got there any later though it would take well over an hour just to get from Ben Gurion to the central bus station. And there was not much to do about getting back up to Jerusalem in the evening….
Alon, one night-time long carpool commute explanation is quite simple: people starting shifts in the middle of the night are usually working in rigid shift-scheduled jobs with precise entry and exit times. These jobs are usually odd, and almost never office-flexible, like people working in a water treatment facility, as a 24/7 store cashier etc.
Under those circumstances, it is usually easier for workers to organize carpools: the jobs don’t pay well, and traffic is light, making extra detours to take more passengers per car feasible.
This reminds me of when I would vacation on Mid-coast Maine. After being ‘burned’ by weekend/rush traffic, I planned ahead. I would rise in Brooklyn at 4AM Friday and be on the road at five. I always planned to get though Hartford by 8Am rush. After that I was coasting.
More recently I took notice of NYPD employees nearby my home sitting in their cars at 4AM, waiting to go in to work. I learned they all lived in the lower Hudson Valley and were able to choose work shifts early in the AM with early afternoon departure(they worked clerical-type jobs, no public interface). They then could live in their dream home way out of NYC but legally. They often worked this out with their mates that they would work different parts of the day(one at home aas needed for the family). If they could keep it up over the years it had great benefits: both can work, no relocation, kids stay at same ‘good’ schools, retire where they live.
Your discussion is based entirely upon Figure 8.
If you look at Figure 10, there is a breakdown of departure time by occupation. In this figure you will see that the workers departing in that 12 midnight – 5 am period are primarily employed in farming/fishing, production/transportation, and armed forces. And to a lesser extent construction/extraction. Not so much services, as you hypothesize, and sales or management. So maybe the longer and earlier commutes have more to do with the location of work sites on the fringe of urban areas, and the need to arrive at those types of jobs earlier in the day. Another factor could be longer signal timing and less frequent transit service.