|
|
Doug Puffert wrote:
>...Jerry Dwyer said he thought that path dependence must
>mean more than just that 'Where we are today depends on where we were
>yesterday,' since any autoregressive process (including a random walk)
>satisfies such a definition and few if any would argue against the reality
of
>such processes. Perhaps few would so argue, but where have they treated the
>reality of such processes in economic theory (as opposed to ad-hoc
>econometrics, which of course takes ample note of autoregression)?
Random walks are central to "real options" theory and this theory offers a
useful corrective to both simplistic neo-classical models and simplistic
path dependence models (see Dixit and Pindyck, "Investment Under
Uncertainty" for an introduction).
Despite the mathematical nature of this theory (I was initially skeptical
that anything so sophisticated could have practical value, but it does), it
does offer some insights:
A standard result is that if the random walk strays too far from the
neo-classical equilibrium, then opportunities for profitable economic action
arise. E.g., if price increases beyond a threshold, then new firms will
enter. In other words, although the neo-classical equilibrium is NOT an
accurate description of reality, there are also limits to purely random
activity.
This contrasts with Brian Arthur's models featuring UNLIMITED increasing
returns. In the context of the current discussion, Deirdre McCloskey has
already made the point that David avoids the question of "How much?"--was
QWERTY so inefficient that it could have been profitably improved upon? My
earlier point was that the information costs of switching might well
outweigh any benefits.
Also, although the random walk gives rise to inefficient choices (relative
to a neo-classical "ideal"), usually there is little anyone--social planner
or private agent--can do about this. Faced with uncertainty and limited
information, social planners typically end up making the same choices as
private agents. A very relevant question is whether sufficient information
existed (or could have been inexpensively obtained) for anyone to choose
something signficantly better than QWERTY.
The relevant difference between "weak" and "strong" path dependence is NOT
that one leads to efficient outcomes and the other does not. In all
liklihood, both lead to outcomes that are inefficient when compared to some
unrealistic neo-classical ideal. The relevant difference is whether
sufficient information was available to allow a more efficient choice. With
weak path dependence such information is not available and hence economic
choices are more or less random. Like the weather, these choices may be
economically meaningful ("economically meaningless" is an overstatement),
but there's not a whole lot one can do about it.
On the other hand, strong path dependence argues that the threshold was
crossed and the wrong decision was still taken--this is a much more
difficult case to make and David does not really address it.
Jim Bessen
Research on Innovation
|
|
|
 Send comments and questions to admin@eh.net
|
||