Inform pedestrians, not drivers

From Sacha Kapoor and Arvind Magesan (pdf):

Most empirical studies on the role of information in markets analyze policies that reduce asymmetries in the information that market participants possess, often suggesting that the policies improve welfare. We exploit the introduction of pedestrian countdown signals – timers that indicate when traffic lights will change – to evaluate a policy that increases the information that all market participants possess. We find that although countdown signals reduce the number of pedestrians struck by automobiles, they increase the number of collisions between automobiles. We also find that countdown signals caused more collisions overall. The findings imply welfare gains can be attained by revealing the information to pedestrians and hiding it from drivers. We conclude that policies which increase asymmetries in information can improve welfare.

Hat tip goes to @m_sendhil.


Drivers are probably using the information to jump their light a fraction of a second before it turns green.


Drivers approaching a green that is counting-down to expiry accelerate, trying to beat the light.

In my hometown, I use the pedestrian signals (without countdown) to aid in hypermiling in my Prius. If I'm coming up on a red light and the pedestrian signal is still green, I'll probably have to stop. But if the pedestrian signal has just turned red, I'll be getting a green momentarily, so I'm better off trying to maintain as much speed as I can into the light.

Of course, all of this is bullshit brought on by the traffic "engineers" that my city scrapes from the bottom of the barrel. It's not an arterial, there's nobody waiting at the crossroad, and the f**king light still turns red. Or one blue-haired grandma who is too afraid to make a right turn on red with a quarter mile of space between her and the nearest car and 15 cars need to stop in both directions so she doesn't crap in her Depends.

Yeah, but in this scenario, you have to use phrases like "hypermiling", and I'm pretty sure no one wants to sound that pretentious.

Usually, there is some lag time between the moment the counter reaches zero and the moment the light turns from green to yellow, with further lag time from yellow to red. Do these lag times pose an information asymmetry? Or rather, are both drivers and pedestrians both unaware of the precise durations of these lag times? In any case, I would expect both drivers and pedestrians to be acting opportunistically. Also, this interaction between drivers and pedestrians might be illuminated by the Coase theorem: if making a deal were costless, what would they agree to ex ante as to right of way?

My question is whether the change in automobile collisions is directly dues to the drivers of vehicles reading the countdown timers, or whether the countdown timers change pedestrian traffic in such a way that it makes automobile-automobile collisions more likely.

Here's a somewhat complicated story that might have some bearing.

Pre-countdown-timers, pedestrian traffic tends to follow a front-loaded distribution: pedestrians hurry to cross the street when they could walk at a normal speed and still cross the street in time, and other pedestrians do not cross when they believe they might be marginal, instead waiting for the next signal.

Post-countdown-timers, pedestrian traffic is more normalized across the time interval.

A driver waiting to turn right in the post-countdown-timers world has a longer time to wait as a steady stream of pedestrians monopolizes the crosswalk. This does make it less likely that the driver hits a pedestrian than a gappy distribution would! But it means that when finally the last of the pedestrians goes through, he is impatient and dealing with an imminent light change, and accelerates hard. He has been focused on the pedestrians and misses a car making an unprotected left (similarly stalled until the pedestrian flow is past), or an erratically driving car in the oncoming direction, or whatever. No driver is directly observing the countdown timer.

I don't know how likely this is, but I'm having an equally hard time constructing a plausible story in my mind why drivers who directly observe a coutndown timer become more prone to hitting other drivers.

It is hard to selectively inform pedestrians?

No, an electronic voice would be heard by pedestrians near the traffic signals but not by drivers.

Don't think the deaf lobby will let you get away with that.

There goes Big "Huh?" again, always getting in the way of asymetric information schemes designed to increase social welfare. Fuckin Nancy Pelosi.

DOES NOT MATTER: pedestrians and/or drivers will be on their stinking cellphones anyway and already won't be paying strict attention to other information relevant to traffic flow, pedestrian or vehicular. Same goes for cyclists and skateboarders, et al.

This finding goes against my intuition and anecdotal experience in Chicago. I see many fewer drivers go through on red than without the timers. The timers in Chicago are reasonably easy for more than 1 car in line to see.

The only thing I can think of is similar to red light cameras, there are some people who are so risk averse that they come to a stop earlier than the person behind them anticipates, leading to higher numbers of rear end collisions.

If you dissect their data I suspect a lot of the "collisions" are relatively minor fender benders. OTOH it is somewhat hard to have a "minor" car versus pedestrian incident.

"I see many fewer drivers go through on red than without the timers."

I strongly doubt that you are capable of making an accurate comparison of the frequency of these events.

I note from table 2 in the article that the effect is an increase of just over 5% (magnitude of increase, NOT the significance level - page 9) and is estimated by a regression model. It is possible that the effect expands/decays over time or varies by locality.

I'm in Chicago, and I'm not calibrated enough to make an accurate comparison, either but subjectively I'd agree with BillD. But there's a confound -- there are a lot of "red light" cameras here and it is possible that the visible presence of one upgrade (pedestrian countdown) acts as a signal for the likely presence of another upgrade (red light camera), thus encouraging better driver behavior.

"a policy that increases the information that all market participants possess"

Drivers and pedestrians at an intersection are a market?

All the worlds market. :-)

Authors trying to shoehorn interesting topic into standard economic theory to increase chances of publication in econ journal.

I think it would be interesting to compare the results of a game where producers enjoy an informational advantage over consumers and one where asymmetry is reversed. Then compare both to a game where information is uniform.

Interesting, but not about "asymmetric information" as it is usually meant. Both pedestrians and driers are "transacting with" the lights, not mainly each other. It is not that drivers are not disclosing information to pedestrians or vice versa

Depends on the city, I think. Here in Boston, pedestrians and drivers will routinely jay walk and/or run a red if they think they'll avoid a collision — it isn't really about the lights.

Now that I'm thinking about it, bicyclists might be even more interesting, because they are by far the least concerned about a light, in and of itself. (I know, I bike.)

1. I had always thought that the side blinders around traffic lights were (at least partially) intended to prevent cars stopped at a light from trying to anticipate when their lights will turn green based on when the light from the moving traffic turns yellow or red.

2. Related to above: I am surprised that the authors assume that pedestrian traffic signals provide information to all. Drivers stopped at a light simply do not always see when a walking signal turns red (and, in turn, when his light is likely to turn green), because one cannot always see the walking signal change. (This depends on the design of the intersection, the width of the road, which lane the driver is in, etc.)

3. I would have liked to see some control for the design of the intersection.

I thought side blinders were either for sun-glare reduction or when an unintended stream of traffic might mistake lights for itself.

We are installing countdown timers in Brisbane that count down the time of the flashing red man. So far no incidents attributed to the timers. It will be interesting as we continue to roll out approx 140 intersections in total with countdown timers.

Comments for this post are closed