« The NYT and Sudan | Main | More on Buses »

Tuesday, March 21, 2006

Report Report Report 1: Buses

Many thanks to Gari N Corp for recommending this Slate article as the first recipient of my Report Report Report, wherein I compare academic reports with the journalism which reports on them.

At first, I thought the Slate article would get high marks. I wasn't convinced by Gari's quibbles, and what's more the Slate author, Austan Goolsbee, is an economics professor in his own right, rather than an innumerate journalist.

After reading the underlying research, however, I have to give the Slate article a C–. The research simply doesn't say what Goolsbee says it says, and in fact the government of Santiago, where the research took place, has drawn exactly the opposite conclusions to Goolsbee.

The underlying research here is a paper called "The War for the Fare", by Ryan Johnson, David Reiley, and Juan Carlos Muñoz. Follow my link to find it: Goolsbee unhelpfully links to an NBER page which will charge you $5 to download it.

According to Goolsbee, the paper shows that if cities like Chicago moved to an incentive pay system, rather than paying their bus drivers a fixed hourly rate, that would improve productivity, "raise efficiency and save tax dollars."

Goolsbee frames the question simply: why do Chicago bus drivers sit in traffic on Lake Shore Drive, when they could use surface-street shortcuts instead? His answer: they wouldn't benefit, financially, by doing so. And if you look at what happens in Santiago, where drivers are paid per passenger rather than per hour,

Paying by the passenger leads to significantly shorter delays. Give them incentives, and drivers start acting like regular people do. They take shortcuts when the traffic is bad. They take shorter meal breaks and bathroom breaks. They want to get on the road and pick up more passengers as quickly as they can. In short, their productivity increases.

Does paying by the passenger lead to significantly shorter delays? It depends on what your definition of "significantly" is. The Santiago study does find that

a passenger waiting for a bus 30 km away from the start of a route can expect to wait 15% longer, on average, when the driver is paid a fixed wage rather than per passenger.

Does Chicago have 30km bus routes? I don't know. On shorter routes, the effect will be smaller. Here's the scatter chart:

As you can see, in aggregate the per-passenger buses do have slightly shorter wait times than the fixed-wage buses. But the difference isn't huge.

Do the per-passenger buses "take shortcuts when the traffic is bad"? On that subject – the key subject of Goolsbee's article – the report he cites is completely silent: there's no mention of shortcuts at all. Do they "take shorter meal breaks and bathroom breaks"? Yes.

Goolsbee omits, however, most of the downside to paying drivers on a per-passenger rather than a per-hour basis. Here's what he says:

Not everything about incentive pay is perfect, of course. When bus drivers start moving from place to place more quickly, they get in more accidents (just like the rest of us). Some passengers also complain that the rides make them nauseated because the drivers stomp on the gas as soon as the last passenger gets on the bus. Yet when given the choice, people overwhelmingly choose the bus companies that get them where they're going on time. More than 95 percent of the routes in Santiago use incentive pay.

Do drivers paid on a per-passenger basis get into more accidents? Yes. The report shows that they have 10.03 accidents per million kilometers travelled, compared to just 5.98 accidents per million kilometers travelled on the Chicago-style buses. That's a huge difference, which can't be shrugged off by saying that they are "moving from place to place more quickly". The pay-per-passenger system means that drivers have a very strong incentive to overtake the bus in front of them, and pick up all of the passengers which the bus in front would otherwise get. So they are likely to drive more aggressively, and less safely.

As for feeling nauseated, the report finds that

Once a bus finally does stop, the driver quickly gets the bus moving at full speed, often in complete disregard for the stability or comfort of the passenger. These rapid stops and quick accelerations can occur for the entire duration of the trip.

This causes not only nausea, of course, but also injuries:

if all buses in Santiago had the same number of accidents per km as those with drivers paid a fixed wage, we estimate that it would save 55 lives per year. It would also eliminate 227 serious injuries, 210 less serious injuries, and 1,293 light injuries.

And these numbers actually understate the likely truth, because the per-passenger bus companies are less likely to report minor accidents than the per-hour bus companies.

In economic terms, then, moving from a per-hour system to a per-passenger system is a tradeoff. On the one hand, people will wait slightly less for a bus, and that saved time is worth money. On the other hand, they will be killed and injured in bus accidents more frequently, something which carries a significant economic cost of its own.

There is another reason why the pay-per-passenger system is flawed, which Goolsbee doesn't mention:

When a passenger is waiting at a stop alone, sometimes the driver won’t stop because the opportunity cost of the time spent picking up that passenger is greater than the income from the fare. In fact, often times a single passenger waiting will have to wait for several buses or until more passengers arrive at the stop.

A public service such as a bus system should treat all customers equally, but moving to incentive pay gives drivers an incentive not to.

And Goolsbee is downright wrong when he says that "when given the choice, people overwhelmingly choose the bus companies that get them where they're going on time." The citizens of Santiago are not "given the choice" between bus services running on a per-passenger or per-hour basis. Some bus routes pay one way, other bus routes pay the other. You take the bus which goes to your destination.

Does the report conclude, at least, what Goolsbee says it concludes? Not really:

We find our most interesting result to be the fact that per-passenger compensation yields more regular spacing of buses, and hence lower expected waiting time for passengers, than does fixed-wage compensation. However, this clearly does not mean that per-passenger compensation is a superior system. We find significant costs to the per-passenger system as well. In particular, per- passenger compensation exhibits a much higher incidence of accidents and much lower passenger comfort.

So to Goolsbee's point that "more than 95 percent of the routes in Santiago use incentive pay". Here's what we find in the report:

In 2006 Santiago will complete a dramatic overhaul of its bus system. The plan, called Transantiago, will replace the current system of disorganized owners with a dozen or so large companies. Partially influenced by conclusions of this research, drivers will all be paid a fixed wage.

Not exactly the outcome you might have expected from reading Goolsbee's article, eh?

UPDATE: David Reiley emails:

I do think the wait times are quite economically significant ($38,000 worth of lost time per day!), and I'm disappointed that you don't mention what we think is the most important point in our article. We think that there are significant gains to be had in improving bus spacing, and we're interested in finding ways to do that without promoting accidents or discomfort. We think that it's well worth exploring what types of driver incentives will matter, and that's why we speculate about GPS-based incentive systems that give drivers a reason to try to cooperate with each other (as opposed to competing in a way that causes traffic externalities). I would prefer it if your Report Report Report reflected this.

Since this is a Report Report Report and not a Report Report, I did miss out the stuff about GPS-based incentive systems. But I'm happy to oblige.

At the end of Reiley's report, we find this:

With a full implementation of GPS technology, drivers could have information on the location of other buses at all times, not just on corners where sapos work. Drivers could have a real-time display showing the locations of other buses both in front of and behind them, enabling them to respond with adjustments to the spacing. GPS technology opens up a whole new realm of contractual possibilities. For example, one might pay drivers a bonus based on the continuous-time average spacing between their bus and other buses on the route, thus providing drivers with appropriate incentives to minimize passenger waiting time. Such systems could improve quality of passenger service in cities like Santiago, and could be potentially even more useful in the cities of developed nations.

At some bus stops in central London, there are now signs which tell passengers how long it's going to be until the next bus arrives. The technology for these static signs was expensive, but they are certainly useful. What Reiley is proposing is a system which puts that kind of data than into a computer display which is easily readable by bus drivers. I don't know whether that technology exists, how reliable it is, or how expensive it is. Neither, I suspect, does Reiley.

I do suspect that bus drivers' unions would react badly to any attempt to pay some drivers more than others, especially when bonuses would be based on something as abstract as "continuous-time average spacing". There could be unintended consequences, too, such as drivers not stopping at bus stops if they're too close to the bus behind them, or being much less willing to help out wheelchair-bound passengers because the amount of time involved in getting those passengers on and off the bus would throw off their bonuses.

It's also worth noting that bus drivers already have a lot of demands on their attention: collecting fares, navigating traffic, keeping an eye on the passengers, telling tourists when it's their stop, etc. Would trying to keep up with dynamically-generated GPS maps impact the other aspects of their job?

It's clear, then, that there would be a significant number of costs associated with a scheme like the one Reiley proposes. Maybe a pilot scheme somewhere could try to determine whether the benefits in decreased wait time might outweigh those costs. But I suspect that simply adding more buses on busy routes might be an easier and cheaper way of decreasing average wait times.

In any case, Reiley's plan involves a significant up-front expense: at the bare minimum it involves kitting out every bus with a computer and GPS system, as well as setting up a centralised system for calculating and paying bonuses. It's a far cry from the quick fix that Goolsbee implies is possible.

UPDATE 2: I've had an email exchange with Austan Goolsbee, some of which might be of broader interest. Said Goolsbee:

And then at the end I must wonder if you have gone crazy: "In 2006 Santiago will complete a dramatic overhaul of its bus system. The plan, called Transantiago, will replace the current system of disorganized owners with a dozen or so large companies...drivers will all be paid a fixed wage." Are you under the impression that replacing competing companies with a few giant firms and then removing all incentives to the drivers is a good idea? It is a recipe for monopolization. Prices will rise and delays will get worse. You are not talking about a change that the consumers are driving. You are talking about a change being imposed by the government. Why would you expect this to raise consumers' value when the evidence shows exactly the opposite?

I replied:

As for the Transantiago plan, again that was taken straight from the report (footnote 23). It seems to me that its introduction could present you with a prime opportunity to test your hypothesis about incentive pay. Wait for a little while, and then see whether prices have gone up, delays have gotten worse, or passengers are less happy. But in my experience, Chilean government officials are generally the most economically sophisticated government officials I've ever met. (Did you know that every single minister in the new government has a postgraduate degree?) And they were very aware of the findings of this report when they made their decision.
If you look at what Chile is doing with toll roads, you'll know that it remains a leader in terms of bringing private-sector discipline to government functions. (And, of course, it was the first country ever to embark on a major privatization program.) I would look very carefully at the Transantiago plan before dismissing it as "being imposed by the government" and therefore likely to decrease consumers' value.

And Goolsbee wrote back:

This bus decision is a federal govt decision or a city govt decision? The federal government people predominantly have Ph.D.s from here at U. Chicago (and indeed I taught several of them Ph.D. Public Finance). My impression was that this was more of a city level thing.
No question that I would predict that this new system will lead to more delays and the concentration of ownership to higher prices. We will agree to revisit this issue in a year and see if it was born out. I am prepared to admit that the incentives didn't work if that doesn't happen.

Someone remind me to ask about Santiago's bus prices in a year's time!

Posted by Felix at 12:32 EST

Comments

Forgive me if this is a fundamental issue addressed by pro-market economists, but does it make good economic sense (and increase safety) to expect a bus driver to exercice the arbitrage skilz of a bond trader? You can argue that unionization leads to smaller margins, which causes capital to retreat to where labor is cheaper, but this isn't real possible with bus drivers. Otherwise, I can't see how the tradeoffs of inequitable service and increased danger (I rode the Sphinx buses in Jersey for several weeks once; they were far more dangerous than the NJ Transit buses they competed with, and they weren't any faster, since the trade off convenience, meaning I got picked up closer to where I was staying was offset by additional stops elsehwere) are worth the cost.

Congestion is far more debilitating to bus travel than driver incentive. Productivity rules can be written into a flat rate contract (bad union contracts are as much more an effect of politicans acting in self interest, thus undermining the check and balance relationship).

All these geegaws to improve service -- GPS, signs, computers? There is a city in Belgium (Denmark maybe?) that reduced congestion by removing all traffic signals and indicators.

And, to frame this entire discussion, if we really want to get into market solutions to traffic, it should be noted that (in the US, at least), the cost of driving is more heavily subsidized by the government than perhaps any transit system worldwide. There are plent of market-based solutions to speed up buses without even bothering with this, and none of them require any knowledge of what oppotunity cost is.

Posted by: 99 at 16:47 EST, March 21, 2006

Glad you found it fruitful. Turns out he was - just about - comparing apples with apples.

With respect to the Transantiago plan, I expect that, like the toll roads programme, the MOP (federal public works ministry) will be in charge. The reason why Chile and the UK moved in such similar ecconomic directions during the 80s, I always assumed, was the strong lead taken by central government in policy.

Stress again, though, on the assumed, much of the quibbles I raised were the results of scattered impressions I was left with the few times I've been to Chicago and Santiago, rather than any in-depth reading.

You wonk, Felix.

Posted by: Gari N. Corp at 20:43 EST, March 21, 2006

As I think I might have hinted at in the piece, surely the best way to improve bus service would be to, well, increase the number of buses. In London, they've done that AND introduced congestion charging -- and bus use is skyrocketing because now buses come frequently and move relatively quickly. Fiddling around at the margins with incentive pay could be a way of avoiding tackling the big issues, such as congestion.

Posted by: Felix at 21:21 EST, March 21, 2006