Thursday, December 20, 2012

The Coase Theorem, Landsburg and Pigou

In response to a New Yorker article by Elizabeth Kolbert, Steve Landsburg recently wrote a post on externalities that caught my attention, not necessarily for what it said, but for what it did not say.

For background, here's what Kolbert said (the part that is quoted by Landsburg):
A man walks into a bar. He orders several rounds, downs them, and staggers out. The man has got plastered, the bar owner has got the man’s money, and the public will get stuck with the tab for the cops who have to fish the man out of the gutter.
…..
The man pulls into a gas pump. He sticks his BP or Sunoco card into the slot, fills up and drives off. He’s got a full tank; the gas station and the oil company share in the profits. Meanwhile, the carbon that spills out of his tailpipe lingers in the atmosphere, trapping heat and contributing to higher sea levels. As the oceans rise, coastal roads erode, beachfront homes wash away, and, finally, major cities flood. Once again, it’s the public at large that gets left with the bill.
Here is Landsburg's response:
Coase’s key insight is that all of these externality problems are fundamentally symmetric. The question is never “how do we stop A from harming B?” but instead “should we let A harm B, or should we let B harm A”? A consequence of that symmetry is that no abstract principle — including the abstract principles that guided Pigou and still guide Kolbert — can possibly be used to guide policy. Any purely abstract argument for preventing harm from A to B is an equally good argument for preventing harm from B to A.
Kolbert seems to believe that Pigou settled this question 100 years ago. So he did, just as Newton settled the issue of absolute space. But we now know that Pigou was wrong (although his insights laid the indispensable foundation for later, better insights).
Landsburg is right.  An important insight in Coase's fantastic article (The Problem of Social Cost) is that externalities are fundamentally symmetric, but he didn't just stop with that insight.  

Coase continued the article by proposing an alternative solution: let those who are involved with the externality bargain with one another. If they could bargain at no cost and the terms of bargaining were well understood, Coase deduced that there would be no externality problem because the all parties would realize what the efficient outcome is, and would agree to a system of side payments to support that outcome.

This "no cost bargaining" ==> efficiency result is known to economics students today as the Coase Theorem.  Despite its apparent conceptual simplicity, the Coase Theorem is one of the hardest concepts to get students to appreciate.  This is because the Coase Theorem is about cause and effect -- IF bargaining is costless, THEN agents will bargain to an efficient allocation.  Students often forget the IF, and just jump straight to the conclusion ... even in scenarios when bargaining is expensive.

This mistake by students is not what Coase had in mind.  In fact, Coase spent most of his article describing the contrapositive of the Coase Theorem -- if there is an inefficient allocation of resources due to an externality, then there must be high cost bargaining.  In this environment, how policy is enacted matters a great deal.  This was Landsburg's point in a follow up post on carbon taxation in which he provides some numerical examples on how policy can matter.

Because it gets at some subtleties in the analysis of externalities, I was glad that Landsburg followed up on his original post. On the other hand, Landsburg's follow up post assumes away a role for transaction costs is silent about the role of transaction costs (making an implicit assumption; see comments below), and the costs of coordinating the interests of (b/m)illions of stakeholders.  In the real world setting of carbon taxation, these costs are of first order importance.  There isn't a single laundromat that absorbs the brunt of the carbon externality of a single plant.  Carbon pollution affects billions of people (some positively, some negatively, but I think most science suggests a negative effect on average).

On the other side of the externality, industries that emit carbon are much easier to identify, to regulate, and to tax.  Thus, from a practical standpoint, it makes sense to start with policy that affects these industries.  If we think that too much carbon is emitted, a carbon tax is a simple policy that pushes the quantity of carbon emitted "in the right direction" without having to coordinate a billion-person negotiation.  If you don't like a carbon tax, an alternative that gives industries and out-of-industry stakeholders more say is Cap and Trade, but that is more complicated, and it is more politicized.

Landsburg and I agree on one thing.  There are no simple answers, especially once political considerations come into play.

2 comments:

  1. On the other hand, Landsburg's follow up post assumes away a role for transaction costs, and the costs of coordinating the interests of (b/m)illions of stakeholders.

    In fact, the numerical examples in the second post all (implicitly) assume very high transactions costs --- because they don't even consider the possibility that producers will pay pollution victims to move away, or that victims will pay polluters to cut back. This is perhaps an unrealistic assumption for a steel mill and a laundromat, but it seemed to me to be the right assumption, given that we were ultimately going to apply all this to the problem of carbon emissions, with, as you say, billions of affected people.

    ReplyDelete
    Replies
    1. I'm glad you responded, and even happier that you responded to this point!

      I had thought about the implicit assumption behind the high transaction costs, but my post was already starting to get long and unfocused. Thanks for clarifying that point.

      Delete

Please feel free to share your ideas about this post in the open forum. Be mindful that comments in this blog are moderated. Please keep your comments respectful and on point.