We all intuitively know that a crash — or even a disabled vehicle or a piece of interesting debris on the roadside — can disrupt the flow of traffic. Cambridge Systematics has suggested that as much as 25% of congestion in the U.S. is due to “incidents.” Some of this due to physical loss of highway capacity (e.g., a blocked lane). But how much is actually due to drivers slowing down to rubberneck? (the dreaded “gaper’s block,” as they say in the Midwest)

Previous estimates have been based largely on models or “macroscopic” calculations. But an interesting bit of research — one version is here — by a Dutch team (Victor Knoop, Serge Hoogendoorn, and Henk J. van Zuylen, at the Delft University of Technology), to be published as part of this conference, has examined the “microscopic behavior” of drivers at crash sites, as well as quantified the actual loss in traffic capacity due to rubbernecking.

How did they do it? The authors write: “The observation team waited at the Traffic Management Center in the centre of the Netherlands until an accident had occurred somewhere nearby, after which it flew with the helicopter to the accident location. From the moment of arrival, traffic operations for both directions were recorded. From the other side of the road, the incident was visible but there was no physical obstruction.” They point out, by the way, that the helicopter was high enough to itself not be a cause of traffic disruption.

As pictured above, a light truck overturned near the city of Apeldoorn. One lane (in the ‘bottom section’ in the photo above) was closed for the sake of emergency response. Not surprisingly, given that a lane was missing, and two lanes had to merge, the researchers a large drop in “outflow capacity” in the section closest to the overturned van.

But what is striking is the airborne researchers observed a sizable effect — roughly a 50% capacity reduction — in the opposite bit of highway, in which no lanes were blocked. The video of this is rather amazing: At roughly the position that affords the best view of the crash, the traffic begins to bunch, as it does in stop-and-go congestion, even though the road ahead is otherwise clear. One vehicle, the “leader” is essentially slowing to look at the incident, creating in essence a backward “shock wave” that everyone else drives into. Once in the slowing wave, their curiosity is no doubt aroused and so they too continue to creep along for a look. At the exact point of the wreck, the traffic again begins to accelerate.

This raises the idea of the magical “first driver” in a traffic jam you may have wondered about as a child — the idea that there must be some driver at the head of even the longest traffic jam, and if only they would get going, or hadn’t slowed to begin with, things wouldn’t be so bad for us, etc. It also raises Thomas Schelling’s description of rubbernecking in the classic book Micromotives and Macrobehavior: Because they’ve already been made to wait because of everyone else’s look, each individual driver feels they too have the right to look themselves. Each person’s five seconds adds up to incremental delays (“it is a bad bargain,” he wrote). If everyone could only agree not to look, the congestion actually would not form.

There were a number of other interesting observations made. On the section of highway without any physical obstruction, the speed of vehicles in the left lane drops faster than those in the right (the authors speculate that left-lane cars are probably going faster to begin with; that they may be closer to the event and thus more distracted; and that there are more slower-moving trucks in the right lane).

Another curious phenomenon is that as the lead vehicle accelerates past the point of the crash, the time that the following driver begins to also accelerate varies more widely than under the “normal” conditions of heavy congestion. It is as if some drivers are simply more distracted by the sight of the crash, or perhaps some drivers instinctively behave with more caution in the immediate aftermath (the presence of emergency personnel may also affect driver behavior). All drivers react more slowly than in typical conditions, but some react much more slowly.

In any case, this is fascinating research that suggests the traffic impacts of rubbenecking are greater than those found in previous work.

This entry was posted on Tuesday, March 10th, 2009 at 4:18 pm and is filed under Drivers, Etc., Roads, Traffic Engineering, Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.