A speed limit makes sense because driving too fast or too slow puts other drivers on the road at risk, thus the decision how fast to drive can create an externality. But how should policy makers set the right speed limit? Engineers can weigh the costs of higher speeds (more accidents) against the benefits (getting places faster), and determine the optimal level. But in reality they are set at discrete levels that don’t vary nearly as much as the optimal speed on various lengths of road would appear to vary. Furthermore the optimal speed clearly depends on the preferences of the drivers, the current weather, and other factors that shift by hour of the day.

Variable speed limits, in contrast, present a more flexible, even Hayekian, way of setting the speed limit. One example is Interestate 80 in Wyoming, where sensors detect driver speeds, which are then used in an algorithm, along with weather conditions and other factors, to set speed limits that vary. An interesting article, via Radley Balko, provides more information on this road:

Driversâ€™ speeds are tracked by sensors embedded in the pavement and installed on markers alongside the highway.

However, thatâ€™s just one element the Wyoming Department of Transportation uses to calculate and set variable speed limits.

Other factors include weather, road condition and recommendations from Wyoming Highway Patrol troopers and department maintenance operators

According to Wikipedia, variable speed limits date back to at least 1965, with a road between Munich and Salzberg able to have a speed limit of 60, 80, or 100 km/h. These speeds were set by individuals who monitored traffic speeds, but today computers can do this automatically. This seems like the kind of law/technology that should would be more widespread, but speeding tickets mean government revenues. And at the local level, research shows that towns that are undergoing fiscal problems are more likely to issue traffic tickets (yes, traffic tickets are countercyclical). In addition, I think people suffer from a general fear of allowing safety to be at the mercy of algorithms.

But if it is true that the variance of traffic speed matters more than the average in determining the probability of an accident, then it would seem sensible to let speed limits vary with drivers perceptions of what is optimal, making adjustments for the externality of driving too fast. After all, a speed limit to far below the “natural level” probably creates more variance.