Transit Ridership and Observation Bias

One of the problems that afflicts any public service as widely used as transportation is that everyone has an opinion. In fact, everyone *is* an expert on their own commute. The problem is the generalization from anecdote to data (data is not the plural of anecdote). Just because someone understands their own travel patterns doesn’t mean that individual understands everyone else’s.

Elements of Access: Transport Planning for Engineers, Transport Engineering for Planners. By David M. Levinson, Wes Marshall, Kay Axhausen.

An assumption, satirized in The Onion, is that transit is a solution to transportation problems, because other people will take transit.
While evidence is thatPublic transit ridership up in U.S., by 32% since 1995 (to the highest levels since 1957), which is explained in part by high gas prices, and in part by the huge investment in rail transit in the past three decades, this number is still overall quite small. It should be noted that Vehicle Miles Traveled grew 24% in the same period (and has been flat the past several years, so the increase is slightly faster than overall demand (of course, comparing trips to miles isn’t really right, since distances change. Total transit mode share in the US is on the order of 5% for work trips (depending on how you measure) according to this article A Closer Look at Public Transportation Mode Share Trends by Polzin and Chu, but that it only carries about 1% of total passenger miles traveled in the US.
If you tell people, even transportation professionals, that transit carries only 1% of travel in the US, they are usually surprised. Why?
The answer in part lies in an observation bias. To illustrate: imagine there are two buses, one carries 49 people, one carries 1 rider. The average ridership is 25 passengers per bus. (49+1)/2 = 25.
However, the perceived average from riders would be (i.e. the rider-weighted average) is (49*49 + 1*1)/50 = 48.04 passengers per bus. The mis-estimate by using an on-board observer weighting rather than a systems weighting is nearly 100%. (Lest you think this is a straw man example, consider dead-heading commuter trains, with 500 passengers inbound in the morning and very few or none (if the agency truly deadheads) outbound)
The same mis-estimate occurs on highways, where congestion is over-estimated because more people experience congestion than its absence. No one is there to observe a truly empty road.

One thought on “Transit Ridership and Observation Bias

  1. I think it make perfect sense that transit would carry a very small amount of passenger-miles; after all, isn’t that the hope? That transit encourages a more dense, efficient form of development that allows commuters to live closer to their workplace? I don’t think that comparing the passenger-miles of transit and other modes is even really very elucidating. I also wouldn’t say that 1% passenger-mile figure is very surprising at all. Instead, I’d say it’s a reflection of the efficiency of those transit trips!


Comments are closed.