Subject: Re: Mass. Bill to Block OpenDocument Format Standard
From: "Stephen J. Turnbull" <stephen@xemacs.org>
Date: Wed, 09 Nov 2005 13:16:07 +0900

>>>>> "Marshall" == Marshall W Van Alstyne <marshall@MIT.EDU> writes:

    Marshall> At 10:58 PM 11/6/2005, Stephen J. Turnbull wrote:

    >> The Battle of the Sexes was solved in the Garden of Eden when
    >> Adam conformed to the pure strategy equilibrium of doing it
    >> Eve's way.  There's really nothing new or hard here.  The point
    >> is that "network externalities" are not a game-theoretic
    >> concept; they are a decision-theoretic concept.

    Marshall> Hadn't introduced GT here but the subject having been
    Marshall> broached, I'll take the bait for 2 different claims (the
    Marshall> second jumps ahead in your text) :)

    Marshall> 1) Game Theory does matter to standards with network
    Marshall> externalites, not just decision theory.  The reason is
    Marshall> that expectations in game coordination are critical.

Of course it does, and for the reason you give.  My claim is that you
don't need a worked-out theory of network externalities to understand
the fundamental problem.

    Marshall> Folks might not only coordinate on standards of
    Marshall> differing quality (as judged by diverse criteria), they
    Marshall> can fail to coordinate at all.

Unlikely in a game of pure coordination, but an ancient theorem in the
Battle of the Sexes.  You don't need a weatherman to know which way
the wind blows, and you don't need a worked-out network externality
theory to understand why these coordination failures occur.  You do
need a weatherman to predict where the hurricane is going to land, but
he can't stop the rain.  Nor can the externality theory show the way
to guaranteed coordination success, although it can help to understand
the gains to be had from coordination once achieved.  Does that make
what I'm trying to get at any clearer?

I'm not sure it's the right point to make, I'm definitely open to
discussion about that.  The reason I choose this point to emphasize is
that technologists typically speak as though they believe that the
coordination failures are due to mistakes by management.  In fact,
they're an inherent feature of an equilibrium.

    Marshall> In your example, they guy always coordinates by giving
    Marshall> into the gal! ... which is of course better than
    Marshall> coordination failure :)

Exactly.  But is there anybody here who would be willing to delegate
to Microsoft or the Chinese government without objection?  There
_will_ be coordination failures (at a minimum, delays).

    Marshall> Network externality theories are only 20 years old
    Marshall> (Rohlfs wrote 1 obscure paper 10 yrs before Katz &
    Marshall> Shapiro).

They've only been called that for 20 years but the economics goes way
back, and the ideas were a huge influence on the 1970s deregulation
movements in the airline, telecomms, and electric power industries.
Cf "subadditive costs and multiproduct monopoly", aka the "contestable
markets" literature.

The specific "customer utility is an increasing function of the number
of other customers" mathematical model is relatively new, and has a
number of interesting twists compared to externalities between nodes
in production.  But all you really need to write down the Battle of
the Sexes as the "fundamental strategic model" of these markets is
that qualitative, verbal statement.  And that was analyzed by
Leibenstein [1950], "Bandwagon, Snob, and Veblen Effects in the Theory
of Consumer Demand", Quarterly Journal of Economics.

What network externalities in consumption do is to provide a
"functional" basis (Leibenstein's term) for the "bandwagon" effect,
which in his discussion was motivated as "mob psychology".
Nevertheless, the mathematical model (implicit in his case) is the
same as the one that results from modern network externality theory.

    >> So let me respell that as "Any standard is better than no
    >> standard.  Decide quickly."  Classic example: AFAIK today
    >> everybody agrees that SONY Beta was "better" than VHS

    Marshall> Here's I'd echo Ben on the QWERTY story that it's not
    Marshall> nearly so simple as you may have heard.

Sorry, I should have specified that I meant "quality == as perceived
by a consumer who gets one VTR and all the videos he wants to watch
for free".  What I meant _is_ that simple (of course, I may have it
simply wrong :-).

    Marshall> Along these lines, Liebowitz & Margolis also make a very
    Marshall> good point that most technology standards are adaptable.
    Marshall> They're not like meters or kilograms, so if new features
    Marshall> are needed the standard can usually adjust to include it
    Marshall> -- or poach it from the competing std.

Sure, but (and this is a new side of recent models) that can make the
installed base useless if the whole industry jumps to the new standard
together.  The standard adapts, but the installed base can't.  Vendors
happy, customers can live with it (and next year they'll be glad they
did).

A recent example of that (described in technologist terms, which I am
relaying but not competent to judge) is the move by graphics card
vendors to the PCI Express technology.  According to reviews on
tomshardware.com, you just ain't gonna need it, AGP is more than
sufficient.  But in a couple years, all the mobos will be PCIx, and
consumers will-nilly must follow with their graphics cards as they
upgrade hardware for other reasons.

A lot of people still harbor hopes that "software" can become a
panacea, but it can't (that's one thing that the research into network
externalities has taught us).  So you end up with a "punctuated
equilibrium" style of evolution.

What the L&M point about adaptabililty and the reductionist Battle of
the Sexes model tell us is that the conflict among standards is a
distributional issue, but that we can hope to achieve a dynamic
efficiency improvement by shortening the discussion periods (ie,
overlapping them with the early adoption of a "suboptimal" standard).
Where the detailed network externality theories come in is that they
can give us a handle on how long a "suboptimal" standard should be
allowed to dominate, and how "suboptimal" is should be allowed to be.
Again, I deemphasize that at this time because I believe that people
still hope that coordination failures are mistakes that we can learn
to overcome.

IMHO, YMMV, of course.  But even you are too optimistic:

    >> But suppose we can find another way to "encourage" conformance
    >> to a specific standard?  It will have the same effect.

    Marshall> "Encouraging conformance" just provides the coordination
    Marshall> mechanism!

No, it's not mere coordination.  A coordination mechanism is when in
response to a pothole in the sidewalk I gently push you to the left
and my daughter to the right.  Everybody happy.  "Encouraging
conformance" is when I force you all the way into the weeds on the
left so my daughter doesn't have to step in the mudpuddle on the
right.  Are you still happy?  :-)

The point is that gentleman that you are, you probably _are_ happy,
when you come to think of it, but you would not have responded
instantaneously to a gentle pressure---and there would have been a
coordination failure.

In the situations we're talking about, though, firms can become 100%
bankrupt, and (even more likely) people can become 100% unemployed.
Gentility tends to come in second in those cases.

    >> To me, the question is "who do you trust with enough compulsory
    >> power?"

    Marshall> Great question. Markets are generally great at decision
    Marshall> making except in cases of externalities and coordination
    Marshall> failures.  So, answers?

I'll take the state and local governments (SLG).  The Feds are too
big, the vendor consortia too conducive to "conspiracies to raise
prices" etc, individual customers are too small to be effective or
foresighted[1] with too restricted views to be socially efficient.
The SLGs have their own problems, but we can at least _hope_ for
competition among jurisdictions to point out viable solutions.

Dismal-scientifically y'rs,


Footnotes: 
[1]  This is not a statement about their intelligence, rather their
understanding that their purchase decision does not affect the market,
so they optimally respond to actual status, and not the desired high
level, of diffusion.

-- 
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
               Ask not how you can "do" free software business;
              ask what your business can "do for" free software.