Wednesday, March 03, 2004

Old Yeller

Radley approvingly posts a letter he found in the Washington Post, which says among other things:
The Feb. 14 editorial "Focus on Red-Runners" mentioned a Fairfax City study that apparently showed a 44 percent drop in one year in red-light running at five intersections with cameras. But it did not mention the results of a 2001 analysis by the National Motorists Association of a Fairfax County intersection. That organization found that red-light violations dropped 96 percent at the intersection when yellow light time was increased from 4 seconds to 5.5 seconds.

A 1998 study by the Insurance Institute for Highway Safety also found that 80 percent of red-light entries occur within the first second of the light turning red, indicating that inadequate yellow time is the major cause of red-light entries.
The implication, of course, is that the government should extend the length of yellow lights. Apparently something is “special” about the difference between 4.0 and 5.5 seconds, which makes those added 1.5 seconds especially useful in avoiding accidents.

I don’t buy it. The 96% reduction in red-light violations was almost assuredly a temporary effect, brought on by the fact that people adjust their driving behavior to the expected duration of yellow lights. People are willing to go through a yellow if they think it will last just long enough to get them about halfway through the intersection before it turns red. If the duration were permanently raised to 5.5 seconds, people’s unconscious timers would gradually readjust, and they would start blowing through yellows that under the 4.0 regime would have seemed stale.

And then, if someone performed another study on the effect of increasing the duration of yellows from 5.5 to 7 seconds, I’ll bet it would have the same effect: a dramatic but temporary reduction in red light violations, which could be used by the credulous to justify another increase in the yellow light time.

UPDATE: Radley responds, pointing out some studies that apparently show the driver adjustment effect I’m talking about is small. Unfortunately, the studies are unavailable on the web. Radley also gives more details about the 96% reduction:
In 2001, the National Motorists Association persuaded the Virginia DOT to lengthen the yellow light at one particularly egregious intersection in Fai[r]fax by 1.5 seconds -- from 4.0 seconds to 5.5. A camera installed at the intersection monitored the number of infractions. About 70 days after the yellow was increased, infractions fell from 52.1 per day [to] less than one per day -- or about 96%.

That was three years ago. NMA reports that infractions at that intersection have remained at about .80 per day in the three years since.
I’m still skeptical, because this could easily be the result of the change having been made at a single traffic light. My “internal timer” is not so finely tuned that it knows the typical length of all the different traffic lights, even the ones I go through regularly. I assess the expected length of a yellow light by reference to all the yellow lights I deal with. So it doesn’t surprise that extending the yellow on a single intersection might result in a substantial reduction in red-light infractions at that one light. But it doesn’t follow that extending the duration of yellows at all the intersections would have a similar effect, because people would then get used to the longer yellows. The conclusion that does follow is that extending the yellow at especially notorious intersections might be an effective strategy, because it makes use of internal timers that have been calibrated on the basis of shorter yellows.

I’ll concede (as I did in the comments box) that there might be an optimal duration for yellow lights, but I still stand by my argument that this experiment is very far from proving it. I also agree with Radley's broader point, which is that local governments seem more interested in maximizing revenue than preventing traffic violations.

No comments: