The Peltzman Effect on the Golden Gate

A safety barrier on the median was just installed on the Golden Gate Bridge; unintended consequences follow.

…in the days since the more secure movable median barrier was installed, the average speed of drivers on the approach from the north has jumped even though the speed limit was lowered from 55 to 45 miles per hour.

“We’re really seeing unreasonable speeds on the bridge, much faster than before,” said Priya David Clemens, a representative for the Golden Gate Bridge District. For whatever reason, including the possibility that drivers feel safer knowing a car won’t come barreling at them from the opposite direction, “we’ve noticed speeds going up,” Clemens said. “That’s why we asked the CHP to help us.”

More on the Peltzman effect.

Hat tip: Carl Danner.


In India the introduction of the medians has really helped reduce accidents on the highway. The worst accidents are those that are headon.So even if speeds go up overall fatalities may go down. No one can really say seat belts have increased the % of fatalities.

Is this really so bad? Getting more cars over the bridge means fewer traffic jams. The speed was 45 before the median, I see no reason it shouldn't be 55 now.

Capacity (volume) increases with speed only if the spacing between cars remains constant. If you follow the rules of the road, increasing your speed increases your spacing from the car ahead, so the actual number of cars on the bridge declines although the number exiting the bridge remains constant.

This is simple queuing theory.

And queuing theory notes that higher speeds often led to lower capacity if any bottleneck anywhere causes slow downs - congestion - which then ripple backward at speeds which increase with peak speeds, so the traffic goes from high speed to stopped traffic and then only slowly returns to open road that allows momentary return to high speed.

Kleinrock, who literally wrote the book on queuing theory, instructs students to find a high spot in a city at dusk and watch the traffic and how the brake lights ripple backward as congestion causes traffic to stop far from the point of congestion. This is the reason that after traffic finally resumes speed after rubber necking causes congestion you see absolutely no reason for the slow down and congestion.

The idea of unintended consequences is an important one - is there a taxonomy out there that breaks that general idea down to more specific effects like risk compensation? Some of the big ones are gaming the system / perverse incentives, regulatory capture, and regulatory arbitrage (sometimes a good thing), and side effects (too general?), but after that I run out of ideas.

Speaking of regulatory arbitrage, I heard an interesting private sector example recently. The CTO at a software company was known to periodically browse the code repository and do spot code reviews on developers' work. As a result, developers checked in their code less frequently.

Ok, I skimmed the Wikipedia page on unintended consequences. Other interesting related concepts include the Streisand effect (attempting to remove material drawing attention to it), the boomerang effect / psychological reactance, and blowback (originally a CIA term somewhat related to the boomerang effect). Oh, also cognitive dissonance! That one's a doozy.

The "shared space" entry that Wikipedia leads you to is interesting. My initial reactions are like those above - as either a driver or walker I like clear rules and physical obstacles between me and vehicles that could hit me. But it's surely true that when you perceive less separation you get more cautious and vigilant. At which point it seems like an empirical question which effect dominates and under what circumstances.

The barrier should have been transparent, like tempered glass or a metal mesh. Alternatively, it could be covered with small, nasty-looking spikes.

This reminds me of idea I had a couple decades ago -- car spikes. These would be hollow rubber spikes with a self-adhesive base that you'd stick to your car. They'd be harmless, because they'd just bend if they struck a pedestrian or another car. But other drivers wouldn't know that, so they'd increase the distance they'd keep from your car.

I should make up a batch. They'd be most effective when you're the only one whose got them.

Truckers have these for their lug nuts. I think they'd actually break off, not just bend. Idea is the same.

Outside of the Circus Maximus, what do the lug nut spikes do for you?

Accidents would plummet if every steering wheel had on it a steel spike aimed at the chest.

@Steve- you would make a good Public Choice theorist. As well as a population control advocate, not to mention being popular with trauma surgeons everywhere.

I wonder how driverless cars would respond under these changes. Would they possess sufficient AI to increase their speeds (as they should; see next paragraph) or would they read the new speed limit and slow down?

In any event, have we observed the Peltzman effect? The effect occurs when people take increased risks that lead to more injuries and deaths than would be otherwise expected. We do not yet have accident data on the effects of the new median to make this claim. Thus, far, speeding up seems like a good response.

Why is speeding up a good response?

It will not increase the number of cars crossing the bridge unless you drive with the same distance between you and the cars in front and behind.

Why is the ability to see cars in a separate lane coming to you determinant of the safe following distance you maintain based on speed?

Note that Sam Peltzman hypothesized that safety features would not lower the "death rate" or other metric used, but his hypothesis has been repeatedly falsified.

Traveling by car has become much safer over the past 50 years as thousands of safety regulations have been forced on cars, drivers, and road builders.

For flying, the greater the safety regulation, the lower the risk. Commercial passenger has the highest level of safety regulation and the lowest accident rate, and the accident rate increases as regulation decreases with commercial cargo, private air, and then experimental and exhibition air.

For a long time, driverless cars will follow the strategy of "minimum speed limit from all available inputs." If the speed limit is 45 MPH, they won't ever go over that.

Are higher speeds not safer now? Why are the higher speeds unreasonable? If you don't have any numbers, how do you know there's been an "uptick" in minor accidents at the toll booths? More importantly, there is no/zero/nada/bupkis reason to have tollbooths anywhere anymore; what they hell are they still doing there other than causing minor accidents?

Higher speeds are not much safer - the danger of the rare but catastrophic collision is now much lower, but the danger of collision with cars going the same direction is actually slightly higher (the lanes on the bridge are already substandard width, and the barrier takes away a few more inches); increased speed raises the risk of same-direction accidents especially since there is a need to slow near the end of the bridge that's not evident until you're close to it.

Same direction accidents are much less dangerous than opposite direction, so even if their risk increased, the overall danger would be lower than before.

The article states "An uptick in minor accidents has been seen at the toll booths" but if that has been the only effect why are the speeds "outrageous"?

Accidents, even minor accidents, cause congestion which reduces the carrying capacity of the bridge, thus higher speeds on the bridge followed by rapid deceleration at the end of the bridge results in reduced capacity than keeping speeds lower..

The entire Bay Area has toll booths. It's insane, especially for how highly trafficked all those bridges are. Really basic technology.

The Golden Gate Bridge has toll booths even though the toll collection was automated a few years ago. In true California fashion, you are supposed to feel bad for all the poor toll collectors and the loss of that human touch.

Peltzman claimed the use of seat-belts would cause an increase in accidents.

He specifically said that history showed that the accident rate had declined some 3% annually for decades but that it would now start increasing.

But the accident rate has continued to fall at a 3% rate while the use of seat-belts has risen from about 10% when Peltzman studied the matter to about 80% in recent years.

A chart of the accident rate and seat-belt usage looks like an X.

I don't think that's what Pelzman claimed. He claimed there would also be a counter effect that would negate some of the positive effect of the safety feature.

Also, I think that fact that the trendline for traffic safety was completely unaltered by a significant set of command-and-control style federal highway safety regulations supports Peltzman's general point.

@TMC -sounds like metaphysics if you say that's what Pelzman claimed, since you can never accurately measure such a counterfactual. Charlie below makes a more plausible claim, that safety has gone up despite more reckless driving due to seat belts, though even that claim can be rebutted by the charge of why drivers have not 'consumed the safety premium' by driving even more recklessly knowing there are safety barriers and air bags to save them? A lot of freakonomics is just bunkum to sell books.

Looks like speed bumps are coming to SF bridge.

Higher speeds can only be considered good if you limit your analysis to the people in the cars. As soon as you consider that people exist outside the cars, whether as pedestrains or people who live near the approaches to the bridge or taxpayers who will have to support higher infrastructure costs, higher speeds can no longer be seen as beneficial. For everyone outside the cars, higher speeds mean more danger, more stress, more noise, more pollution, more difficulty crossing the street and more traffic as other drivers pile into the street network, lured by faster crossing times.

Great post! Thanks for sharing your article. I enjoyed reading it. I will definitely share this to my friends.

If one day autonomous cars allow the increase of average speed in highways, somebody is going to quote Peltzman and say "robot cars are more dangerous than humans, look they drive faster.......FASTER".

Where is the cost benefit analysis of the barrier decision/speed limit decision? Sounds like that thanks to the increased speed, the decision may have been the correct one (speed increased) taken for the wrong reasons (losses from accidents would decrease).

We've had the same experience when the barrier was installed on the causeway from the mainland to the island. The barrier was installed for two reasons: to prevent head-on collisions when vehicles and other things wander into on-coming traffic and to prevent u-turns on the causeway (which are a cause of many accidents). The "other things" that wander into on-coming traffic include trailers (including trailers with boats on them) that come loose from the vehicle; indeed, that has been the cause of the worst accidents on the causeway, as the rogue trailer is like a missile. The causeway is about four miles long, and some objected to the barrier because it could make it more difficult for EMS to reach those injured in accidents.

“We’re really seeing unreasonable speeds on the bridge, much faster than before,” said Priya David Clemens, a representative for the Golden Gate Bridge District. For whatever reason, including the possibility that drivers feel safer knowing a car won’t come barreling at them from the opposite direction,

I'm not seeing this as unreasonable. It's logical to want to drive faster, the point of driving is to get from A to B and part of the cost is time. Driving faster lowers the cost, at the expense of safety (and fuel economy and nerves if you're not used to high speeds). If a center barrier reduces the chance of a car drifting into the wrong lane, then the safety cost of driving faster falls which means people should go faster.

A safety barrier wasn't there before? Hm, then again I haven't been over that bridge in three years or so. Before that, like ten. It's the Bay Bridge doing all the work 'round here.

The small business arranges the sale on behalf of the leasing organization and obtains the bulk of the sale proceeds.

Comments for this post are closed