Are Transaction Costs a Distraction?
| Nicolai Foss |
Yes, says Harold Demsetz in a paper, “Ownership and the Externality Problem,” which was published in 2003, but which I only read recently (there does not seem to exist an online version; the paper is chpt. 11 in this book).
Consider the steel mill and the laundry of the Traditional Externality Tale. The two firms could merge, in which case externalities per definition would be absent. This, of course, only substitutes (additional) management costs (the costs of reduced specialization) for the transaction costs of market exchange. The former may exceed the latter in which case specialization is preferable, but then externalities emerge.
In fact, Demsetz argues, management costs may be positive and transaction costs zero, making merger inefficient and allowing externalities to exist, even with positive transaction costs. Conversely, transaction costs could be positive, but management costs low, in which case externalities may disappear even if transaction costs are positive. Demsetz concludes that there “… simply is no reason to proclaim a special role for transaction costs in the externality problem except for the fact that, if we insist on separate ownership [as Coase does according to Demsetz], positive transaction cost creates the problem of choosing between two alternative assignments of rights” (2003: 296). Transaction costs are insufficient for the existence of externalities.
One critique of Demsetz’ reasoning — which also applies to his much better known paper, “The Theory of the Firm Revisited” — is his dichotomy between “management costs” and “transaction costs.” Modern theory makes no such distinction, referring instead to internal and external transaction costs. Moreover, in Demsetz management costs and transaction costs seem to live separate lives: Management costs can be positive while transaction costs are zero. That seems odd, however: If separate firms engaged in a bargaining situation can learn the costs and benefit schedules from pollution under zero transaction costs, it does not seem reasonable to claim that a unified firm cannot learn the same schedules for its divisions (i.e., management costs are positive), as Demsetz would seem to have it.
(Fred McChesney has a nice discussion of Demsetz’ critique of Coase here that also makes the above critical point, among many others).