|
|
Jim, I appreciate learning a little about random walks in 'real options'
theory, which seems both to affirm the reality of a certain sort of path
dependence and to demonstrate very substantial limits to its effects.
You contrast the limits to 'purely random activity' in these models to the
results of Brian Arthur's models featuring 'UNLIMITED increasing returns.' I
think it fair to note that Brian quite explicitly pointed out that his models
apply to cases of 'unbounded increasing returns'-- or at least cases with bounds
high enough that they don't matter. Paul David also acknowledged the point, I
think already in the QWERTY paper. It is, of course, important to follow up this
qualification more than either Brian or Paul did.
Apparently I did not make myself fully clear in calling it a mistake to
distinguish between 'strong' and 'weak' path dependence 'on the basis of
whether the outcome is--compared to a meaningfully available counterfactual--
inefficient.' By 'meaningfully available' I in fact meant what you expressed
more precisely: 'sufficient information .. available to allow a more efficient
choice'. So there is little disagreement between us about what this 'strong'
path dependence might be supposed to mean.
My point, though, is that there is no sense in which this sort of path
dependence, whether it exists or not, is necessarily 'stronger' than other path
dependence. Perhaps the best way to explain this is by considering a case where
path dependence appears to be very weak: namely your example from options
markets. As you note, the allocation in this case can vary from the neoclassical
solution, but only within narrow limits. By contrast, the path dependence that
(perhaps) sets an industry on one technological trajectory rather than another--
Fred's point--or that (perhaps) leads to the industrialization of one region
rather than another--Joshua's point--is 'strong' indeed in its effects, whatever
the relative efficiency of the outcome and the ability of agents to foresee (and
act on the basis of) this relative efficiency. (Footnote: Paul and Brian would
probably want to point out that only the latter path dependence is truly
non-ergodic, while constrained random walks are ultimately ergodic.)
Path dependence is, even by simple definition, first of all about how allocation
happens, and intensifying adjectives should reflect this, simply for clarity.
Still, I largely agree with the substance of your remarks. Availability of
information surely plays a role in whether (or how) allocation processes are
path dependent, and it is indeed relevant to ask whether anyone could have made
better choices than those made. And I do wish that Paul David would explain his
current position on QWERTY in response to the questions that have emerged.
Douglas Puffert
University of Munich
|
|
|
 Send comments and questions to admin@eh.net
|
||