Re: Path dependence and costs of changing paths.

From: John Conover <>
Subject: Re: Path dependence and costs of changing paths.
Date: Mon, 14 Dec 1998 08:26:40 GMT writes:
> Stan Liebowitz had this to say about that:
> "Finally... there is the claim that path dependence
> enhances the value of economic history. We think not. If historians
> only record facts and dates, then by pinpointing the "small" event
> that caused a market to go astray, they might be thought to have
> achieved their highest calling, presuming that they can identify, ex
> post, these small events. If, however, historians provide explanations
> for events, then the strong form of path dependence would make their
> job empty, since path dependence claims that small, unpredictable,
> events are responsible for final outcomes, eliminating the ability of
> historians to describe truthfully the grand forces at work in the
> economy. By analogy, do you think weathermen are of greater or lesser
> value in a world where weather is controlled by the flapping of
> butterfly wings, or by larger and more predictable forces? While we
> will all presumably agree that self-interest is an inappropriate basis
> for making scientific judgments, we at least should understand that
> path dependence is not likely to usher in a golden age for economic
> historians."

I'm not sure I understand the argument, but it seems that the
epistemological issue, with both path dependency and increasing
returns, is the concept of what Brian Arthur termed "locked-in by
historical events," (,) implying that the
success of an innovation is determined by lottery as opposed to a
rational process, (the QWERTY keyboard is often cited as an example.)

It would seem that if the success of an innovation is determined by
lottery, we would expect innovation to have random properties-and if
it is path dependent, the path would depend on the cumulative history
of the innovation's success, (ie., it would be a random walk.) If
these two things are true, industrial metrics could be used to provide
empirical insight.

It might be reasonable to look at the run lengths of innovation in
high technology industries, through inference, by looking at market
dynamics. If the run lengths tend to follow a erf (1 / sqrt (t)) type
of scenario, it might provide supporting evidence, one way or the

I analyzed various segments of the electronic industries for run
length, using data available from the US DOC, (which is pitifully
limited,) and the graphs are available at:

FWIW ...



John Conover,,

Copyright © 1998 John Conover, All Rights Reserved.
Last modified: Fri Mar 26 18:53:40 PST 1999 $Id: 981214003911.12734.html,v 1.0 2001/11/17 23:05:50 conover Exp $
Valid HTML 4.0!