This response to Mike Hollinshead is a little tardy but it took me
awhile to reacquaint myself with what I have at hand. In the
meantime, others have had more sensible things to say but I'll post
this anyhow.
> ...from what I can determine in Waldrop's book about Santa Fe,
> they can't break free from the strait jacket of conventional
> thinking and terminology and the "traditional" economic problems.
I don't know the "Waldorp's book" or its date. Did I miss an earlier
reference?
> The person at Santa Fe who did the most interesting work in
> economics was the biologist Stuart Kauffman who built network models
> in which life (including economic life) bootstrapped itself into
> existence. When Kauffman explained his model to [Brian] Arthur, all
> the latter could do was to talk about the fact that it led to
> increasing returns, which is interesting and important but no news to
> biologists, and immediately smothered the results in conventional
> economic terminology and thinking.
Arthur went to Santa Fe in April of 1987 and he's (what I
would call) a mathematical economist so it's not surprising that, at
least initially, he would couch his thoughts in economists' terms.
In his intro to _Increasing Returns and Path Dependence in the
Economy_ (1994), Arthur recounts,
In March of 1987 I went to my old University, Berkeley, to have
lunch with two of it's most respected economists. What was I
working on? Increasing returns. "Well, we know that increasing
returns don't exist", said one. "Besides, if they do", said the
other, "we couldn't allow them. Otherwise, evey two-bit
industry in the country would be looking for a handout."
I think it's to Arthur's credit that he has been thinging critically
about increasing returns and has made the effort to beat up
Kauffmann's rather difficult models. His articles intended for
economists are regretably impenetrable (at least for me) because of
their allusion to presumably well known statistical and economic
paradigms. (E.g. priors, convexity, standard Poincare-index
topological arguments, Verdoorn's "law", Walrasian dynamics or
threshold capital-labor ratio.) But those not intended for such an
exclusive audience reflect an attempt to model what it is that
actually happens in the world rather than a defensive attempt to
keep all the ex cathedra assumptions of neo-calssical economics
together without soiling them.
One of his chief points is that, given the possibility of multiple
equilibria, sensitivity to intial condiditions, path dependence and
increasing returns, there is a potential to become locked into an
economic choice that is non-optimal (or even pessimal [1]) even
when each individual choice by each player is "rational". One of his
examples:
Consider the case of the person who has the choice of practicing
law or medicine each year. Each activity pays more, the more
previous experience has been accumulated. Suppose that rewards
to practicing law rise rapidly with experience but then flatten
out; and those to practicing medicine are small initially but
eventually surpass those of law. According to the theorem,
whichever acivity the person chooses, he will continue to choose
thereafter. If he has a high discount rate [2], he will choose
law. And this choice will at all stages continue to be rational
and superior to the alternative first-year payoff as a doctor.
Yet there may be *regret*, in the sense that after N years in
the law, an equivalent time served in mediceine would have paid
more at each time into the future. Self reinforcement can lock a
single rational economic agent in to one activity, but not
neccessarily the one with the best long term potential.
He observes elsewhere:
If an economic system is locked in to an inferior local
equilibrium, is "exit" or escape to a superior one possible?
There is rarely in economics any mechanism corresponding to
"annealing" (outside injections of energy that "shake" the
system into new configurations so that it finds its way randomly
to a lower cost one.)
This is a concept that's familiar to chemists as an energy well and
to various kinds of systems analysts as local optima in hill
climbing models or gradient descent in optimization problems.
I hypothesize that economics as a whole suffers from just this kind
of lockin. Players adopted the constellation of basic ideas that we
generally associate with Adam Smith, making approximately rational
decisions all along the way. Two hundred years ago, the implicit
assumptions were more or less correct: many small players of about
the same size, power and access to information. The ideas of "free
markets" and "self interest" were appealing because of consonance
with notions of the natural rights of natural persons.
The world has changed. Five thousand corporate players and a few
hundred individuals have disproportinate power by six or eight
orders of magnitude. (Six is the difference between me and Bill
Gates. I'd guess eight between, say, Merrill Lynch and the Senegalese
school system.) The social return -- the solution to the problem of
wealth distribution -- from the economic system is no longer
acceptable for a number of reasons that Jay and others have
catalogued. But at each step, continuing with it is the most
profitable for most of those powerful players. If it appears not to
be working, their answer is more of the same: stamp out economic
heresy, hew more closely to the orthodox dogma and all will be well
in the end. Trust us.
The orthodoxy doesn't propose oligarchic control of the world and
it's economy by a few hundred or thousand excessively powerful
corporations. But the Inquisitors and Cardinals of that orthodoxy
are quickly coming to the conclusion that such control is neccessary
and desireable in order to shore up the truth against heresy.
(Witness Robert Bork and the "law and economics" movement.)
Brian Arthur's models (and those of Kauffmann, other complexity
folks and even C. H. Waddington) lead me to believe that an
annealing episode is way overdue. We are frittering away natural
resources, indeed the whole substrate of the biosphere, and hundreds
of millions of lives in a frantic effort to maintain a historically
successful algorithm for hill climbing while sitting on the apex a local
maximum that is now suicidally far from a global optimum.
Kauffmann visualizes this sort of problem as a landscape -- you
chose x and y in order to move up hill to an improved z. (Of
course, this is just a simplifying visualization of a "landscape"
over an n-dimensional space. A hill or local optimum on such a
landscape corresponds approximately the energy well of the chemist
with the sign of z reversed.) He shows that on "rugged" landscapes
-- those that approximate messy real-world situations -- random
jumps to get off of dead-end local optima have a poor probability of
improving things. If we want to change how the economy runs and how
influential economics is done, we need to calculatedly change some
of the n parameters that determine the shape of the landscape.
The other alternative is annealing: shake up the "energy surface"
sufficiently for the system to bounce over the surrounding barriers
and onto some other place on the surface. That sounds to me like
revolution with it's attendant chaos and uncertainties. We can't
afford that kind of global revolution any more than we can afford
more of the current orthodoxy.
- Mike
---
[1] Nice coinage. See, for recreational reading, The story of Mel,
in _The Hacker's Dictionary_, aka _The Jargon File_.
[2] I would be grateful if Mike H. or someone who knows these things
would send me privately a concise explanation of what "discount
rate" means in this context.
---
