Pecuniary externalities don't always have welfare significance

1/16/2014 04:59:00 PM
Of course, one problem with most arguments in favor of congestion pricing is that it needn't be true that a rich person's time is more valuable than a poor person's--it is hard to argue, for example, that a rich dad's visit to his son in the emergency room involves greater social value than a poor dad doing the same. To make this argument complete, you need to add that revenues from congestion pricing can theoretically be used to fully offset the harm to those too poor to afford the tolls.
Here's an article arguing that driver's don't pay enough to drive:
"Driving on roads should cost you more than it does now. According to a new report from the nonpartisan Tax Foundation, only half of state and local road spending is covered by tolls, user fees, and consumption taxes. The report reveals the extent to which driving is subsidized by general revenues, leading to congested roads and frustrated commuters."
This is a useful occassion to discuss the difference between a pigouvian negative externality--which causes economic inefficiency--and any old pecuniary externality, which is any situation where one entity imposes a cost on a different entity.

Pecuniary externalities needn't be associated with inefficiency. Pecuniary externalities that don't cause inefficiencies arise all the time in market contracts. Consider, for example, condo association fees: if one condo's roof starts leaking, the condo association will typically levy a fee on all it's members in order to fix that leak. That is a pecuniary externality because people other than the owner of that leaky roof are being forced to pay to fix his roof. But there's no loss of economic efficiency here--the optimal number of leaky roofs are being repaired. Pecuniary externalities arise all the time in finance: when a corporation screws up and ends up defaulting on its bonds, it has imposed a pecuniary externality on the bond owners, who have just lost a portion of their wealth as a result of the default. This too needn't be welfare-reducing, because when those bond owners invested in the bonds, they made sure to charge that corporation an appropriate risk premium. Because risky corporate behaviors drive up borrowing costs, the costs of those risks are fully internalized: this market engages in the optimum level of risk-taking, even though the bond-owners, not the corporation, pay the costs of any risks. Here's another example: non-drivers pay to build highways. There's a clear pecuniary externality here--drivers have imposed a cost on non-drivers--yet this too needn't be welfare-reducing if we build the optimal number of highways.

The general lesson here is that matters of efficiency are totally separate from matters of distribution. Who pays for something is merely a matter of distribution and has nothing to do with efficiency. The economic concept of efficiency depends solely on real variables: there exists an optimal quantity demanded and an optimal quantity supplied, and so long as we achieve those two quantities, who pays what is irrelevant. So keeping that in mind, there are two sides to the highway funding issue: 1) are we building the optimal number of highways (quantity supplied), and 2) are we driving the optimal amount on these highways (quantity demanded)?

Let's talk about supply first. Some might argue that having drivers pay more in user-fees, tolls, gas taxes, and license taxes, would incentivize state and county governments to build more highways. This is wrong for two reasons. First, state and local governments are always characterized by irrational choice structures due to the Arrow problem, so they needn't respond to marginal incentives. Second, even if we supposed that these polities did respond to marginal incentives, forcing drivers to pay more for highways does not create any coherent marginal incentive for them to build more highway. In order for this to be a marginal incentive, it would have to be revenue that is available to the state if and only if they build more highway--but in reality tax revenue is tax revenue, and the state can always raise these revenues regardless of whether they build more highway. Since we are assuming that polities are rational at the moment, in technical terms the choice of how much revenue to raise is "independent of irrelevant alternatives."

Ok, so forcing drivers to "internalize" the costs of highways does nothing to move the quantity supplied closer to the optimal level. What about quantity demanded? The answer here is "it depends." Congestion pricing does move us closer to the optimal level of quantity demanded. The idea here is that when too many drivers are on a road, each additional driver imposes a negative externality on all the other drivers in the form of slower traffic and increased likelihood of accidents. Note that these externalities aren't pecuniary! This has nothing to do with who's paying the cost of building and maintaining highways! Under congestion pricing we charge people more to drive during high demand hours, which causes some people to find alternative routs, carpool, take mass transit, or just drive at some other, less congested times of day. In other words, congestion pricing encourages more efficient use of existing infrastructure.

But to further illustrate where the article goes astray, consider: what is the optimal congestion price for driving on a highway at, say, 2:00AM? The answer is almost certainly $0. The reason is that at 2:00AM, there are so few cars on the road that adding an additional car has zero impact on driving speeds or likelihood of accidents. There is no efficiency gain to charging drivers a toll to use a highway during times when the highway is below-capacity. None whatsoever. In fact, there would be an efficiency loss because such a toll would discourage people from taking what would have been the most efficient transportation rout.

Thus, we've established that it is actually only optimal to charge tolls on highways during particularly busy hours, as a way to incentivize drivers to use the infrastructure more efficiently. But what if the revenue generated from these tolls is not enough to pay for the optimal number of highways? This is a very likely possibility. And the answer is that you shouldn't attempt to pay the rest through taxes solely on drivers. It is true that, say, a tax on drivers licenses would reduce the number of people getting drivers licenses somewhat, but this would actually cause inefficiency (if anything) because it prevents them from driving at times when the highways aren't congested and when driving really would be the most efficient means of transportation. Furthermore, in a world where we already have congestion pricing, a license tax cannot possibly take us closer to the optimal number of cars on the road because we already achieved the optimal number of cars on the road through congestion pricing! There's a generalizeable principle here: to acheive efficiency, the price must equal the marginal cost, not necessarily the average cost--hence, the fixed costs of building a highway should not affect the toll charged to use it.

Now, to see how absurd it is to argue that the cross-subsidy from non-drivers to drivers is inherently inefficient, just consider the Coase theorem: it would be equally valid to eliminate the inefficiency of congestion by paying a subsidy to drivers who choose not to drive during high-traffic times, for example. In general, forcing everyone to "internalize" the costs of the externalities they cause is a sufficient but by no means necessary condition for efficiency.

Ok, one final caveat: if you think that it is unfair to have non-drivers paying for things only drivers use on purely moral grounds, then there's an argument to be made that there could be a redistributive gain that offsets the loss of economic efficiency here. I will let the moral philosophers ponder that one, except to add that there is a different way to view the current system of highway funding: we tax rich people to pay for things that help poor people get to work everyday--there's a redistributive gain there too.