What Does “Zero Transaction Costs” Mean, Epistemically?
29 May 2007 at 12:54 am Nicolai Foss 6 comments
| Nicolai Foss |
What does the Coase Theorem require epistemically? To put it less mysteriously, what are the assumptions concerning agents’ knowledge that must be made for the Coase theorem to hold? Or, to rephrase it somewhat, what does zero transaction costs mean in terms of agents’ knowledge (an inquiry started by Carl Dahlman in this paper)?
In his retrospective (1988) discussion and assessment of the debate on the theorem, “Notes on the Problem of Social Cost,” Coase seems to imply that the Theorem requires omniscience. I think that Barzel makes the same inference in his Economic Analysis of Property Rights. In other words, the Coase Theorem holds iff all resource uses, current as well as future ones, are known by everyone.
Not all writers seem to agree with this interpretation, however.
For example, Israel Kirzner in Competition and Entrepreneurship argues that even if transaction costs were zero, there would still be a need for alertness. Per implication, and if Kirzner is right, the zero transaction cost condition does not require omniscience.
Makowski and Ostroy (2001) in their excellent 2001 JEL paper implicitly side with Kirzner, pointing out that while the Coase Theorem may break down in the presence of asymmetric information (the Bazaar Problem, etc.), it is consistent with Hayekian private knowledge (roughly, the knowledge conditions underlying the general equilibrium model, which — contrary to common conception — does not require omniscience).
I tend to favor the Kirzner/Makowski & Ostroy interpretation. I do so partly for pragmatic/heuristic reasons: I think that the setting underlying the Coase theorem is a much more useful benchmark for economic analysis than the general equilibrium model because it is more “loose” and therefore more flexible. That is, the CT does not make specific assumptions concerning the number of players, the structure of markets, the presence of a forward markets, the auctionarius etc. It is in a sense broader in scope. However, these advantages seem to me to pale if the Coase Theorem requires more extreme epistemic conditions than the general equilibrium model. No economic model should be founded on an assumption of omniscience.
My interpretation of the setting underlying the Coase Therem is that the zero transaction cost assumption implies a setting where players are fully informed about current resource uses (which is already an extreme assumption). However, they are not omniscient, since new resource uses may appear through the exercise of entrepreneurship. Given that transaction costs are zero, any externality problems associated with such new resources will, however, be instantaneously eliminated.
Entry filed under: - Foss -, New Institutional Economics.
Organizations and Markets 
1.
spostrel | 29 May 2007 at 2:46 am
Avinash Dixit and the late Mancur Olson had a paper where they argued that with large N, even with zero transactions costs there would still be a collective action problem. The nub of their argument was that “zero transactions costs” should be modeled as common knowledge of the payoffs and strategy sets along with frictionless ability to make (and receive) offers. Anything more than that, they claimed, expanded the meaning of “transactions costs” so far that it would be meaningless. Then they showed how you could still have free-rider problems with large N under their definition of zero transactons costs.
I think they were basically correct. But that position gets you in hot water with all the folks who want to use “zero transaction costs” to mean “all externalities extingushed by bargaining.” The former sounds more scientific than the latter, but I think that’s more of an appearance than a reality. I would prefer a definition of TC that is less tied to the results and more tied to the inital conditions of the problem.
2.
Mark | 29 May 2007 at 7:50 am
It all depends on how you define transaction costs. If information problems are deemed a transaction cost, then by assumption Coase would argue that without such problems there is no need for markets.
The fact that we have ANY frictions whatsoever – much less to the extent we actually have – requires market institutions to try to overcome. And as Coase also (rightfully in my opinion) argued, the fact that we have some frictions that are not overcome via markets does not mean that government has any better chance of rectifying via force.
3.
Joe Mahoney | 29 May 2007 at 10:27 am
The Coase “Theorem” is typically expressed as: In the absence of transaction costs, the initial setting of liability rules does not matter for economic efficiency.
That said, when I teach “The Problem of Social Cost” I emphasize that in the world we live in, of positive transaction costs, the initial setting of liablity rules DOES matter for economic efficiency . Focusing on the world of positive transaction costs was the agenda of Coase (1960). Coase over the years has expressed his frustrations that the profession has focused on the (ideal) case, in which he was trying to move economists to think beyond.
As a footnote, the Modigliani-Miller (1958) Theorem states that in the absence of transaction costs (no taxes, no bankruptcy costs) the Debt-Equity ratio does not matter, and the financial economists tend to focus on expressing the result this way.
Once again, when I teach this material, I emphasize that in a world of positive transactions costs, the debt-equity ration DOES MATTER for economic efficiency.
It is curious that in both cases in the choice of focusing on intellectual nourishment here and now or intellectual pie in the sky that economists choose to focus on latter — that is, on the world of zero transaction costs. Coase, I am certain, does not approve.
4.
spostrel | 29 May 2007 at 2:50 pm
The zero-transaction-cost benchmark is an interesting one. You can’t understand the role of transactions costs without thinking about the world without them. Looking at problems without transactions costs is no silier than looking at the kinematics of falling bodies while ignoring friction. Sometimes abstraction is the best way to see what’s going on.
Positive transactions costs, on the other hand, are a bit like bounded rationality–no one knows exactly what they entail, what they look like, etc., and there are lots of theoretical choices. So they can become an all-purpose ex post rationalization for any observed outcome. If someone would write down a transactions cost function and empirically show under what circumstances it fit, we would be far ahead of the game. Right now, however, discussions of transaction costs are often far more pie-in-the-sky than analyses that abstract away from them.
5. ~Tim Miano Seeks Job | 29 May 2007 at 5:04 pm
The Relationship Between Price & Utility
It will always maximize utility to satisfy the higher ranked preference of both A and B for any pair of disparately ranked preferences (presuming no transaction costs).
6.
Paolo Mariti | 30 May 2007 at 5:57 am
spostrel’s comment above is well taken indeed. From much of literature one all too often gets the feeling that transaction costs are transportation costs from ignorance to omniscience.
The very few things – but of great relevance – we can say ( and formally prove ) are that TC reduce gains from trade:redistribution of commodities in accordance with people’s preferecnes is hindered and specialization in production is decreased.
As to building transaction cost functions one could start from the idea that they depend on a) the number of parties involved in a transaction, b) the number of distinct commodities per transaction ,c) the nature (or characteristics) of commodities, d) the frequency of echange (consumers and firms do not trade continuously) ,e) the volume of commodities traded.
To write down a testable cost function along all this is obviuosly difficult but not impossible.