R. A. Hettinga on 14 Mar 2001 12:53:09 -0000

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

[Nettime-bold] network effects

--- begin forwarded text

Date:         Wed, 14 Mar 2001 01:18:57 -0600
Reply-To: Law & Policy of Computer Communications
Sender: Law & Policy of Computer Communications
From: Stan Liebowitz <liebowit@UTDALLAS.EDU>
Subject:      network effects

I apologize for the length of this posting, but I am trying to cover a lot
of territory.

I have seen some, but not all, of the messages that occurred a few weeks ago
having to do with network effects. The ideas are not that complicated but
one does need to take some time in understanding them and also learning what
the terms mean. My webpage http://wwwpub.utdallas.edu/~liebowit/netpage.html
has links to most of the articles I have written (with Margolis) on the
subject. As far as qualifications, Margolis and I are considered at least
among the leading authorities on the subject. That, of course doesn't make
what we say correct. But we can not be dismissed as cranks or crackpots.
Besides the journal articles (which were edited at Yale, Chicago, U of
Washington, Berkeley and Harvard) we have written the entry for 'network
externality' in Palgrave's dictionary of Law and Economics (Macmillan), the
Encyclopedia of Law and Economics (Edward Elgar), and the forthcoming
Handbook of Telecommunication Economics (Elsevier). Each of these entries is
between 5000 and 10000 words, and each was solicited by these well-respected
publications. For those of you who can't abide the idea that the Independent
Institute might publish good work, a collection of our papers on these
subjects (edited by Peter Lewin) is being published this year by
NYU/Macmillan press. Prior to my current position at UT Dallas, I had been
on the faculty at Rochester and Chicago, and except for a now completed
three year stint as associate dean of academics (i.e. research), have always
been in research. Margolis has always been in research, except for his
current stint (third year) as the chairman of his economics department. I
shouldn't have to tell you this but the level of discourse has descended to
such a low level that questions about these facts arose.

There has been so much misinformation written, and such hostility that I
thought I would make a few points to start:

1. Network effects are real and often important. Fax machines and telephones
are two markets where network effects are clearly central. (These, by the
way, are known as direct network effects). So Margolis and I never claimed
that network effects didn't exist or weren't important.
2. The theories that we were critical of claimed the network effects
locked-in products and standards that were clearly inferior to alternatives.
Most importantly, even when the costs of switching was low enough that it
made sense to switch, these theories claimed that coordination failures
among consumers would keep the switch from being made. [e.g., we all prefer
beta, but none of us think anyone else is going to switch from VHS, so we
all buy VHS. If we could all coordinate our behavior, we would all buy
3. The lock-in stories have played a large role in the case against
Microsoft, first with Gary Reback's tirades and then in the justice
department briefs and the judge's decision.
4. This concern with locking-in consumers has led some Internet companies to
foolishly go for market share at all costs. This bad strategy is in part
responsible for the current meltdown in Internet stocks. In fact, our
evidence shows that consumers will make a change to a better product if one
comes along. Our evidence also shows no support for a 'tipping point'
although it has often been theorized. Instead superior products gain market
share at a fairly constant rate.
5. We have also argued that network effects are probably less important than
more traditional economies of scale for most software products, and that in
some cases network effects would be unimportant for software products.
(network effects are indirect here, meaning that consumers do not get direct
benefits just because someone is part of a network, but benefit in indirect
ways, such as easier transferal of files). There is no good evidence
measuring the importance of indirect network effects, which of course
doesn't mean that they are not be important in some cases. The four or five
studies trying to estimate these effects are all seriously flawed.
6. AOL's instant messaging has powerful direct network effects. It is in
AOL's interest to keep its standards proprietary as long as consumers don't
object. But if a better system comes along, or if consumers want more
compatibility, AOL will have to change (just as all the online services
found that they needed to provide Internet connectivity when the Internet
became popular) if it wants to avoid being dislodged. If consumers are
content with AOL's product as it stands, there is no problem, except for
AOL's competitors. Antitrust law is not meant to help competitors, but to
help consumers. I am not aware of any evidence indicating that the
government should force AOL to open up its system. Obviously, I disagree
with Microsoft on this issue and think they are being hypocritical.

Now I have some responses to claims that I found in the discussion. The tone
of some of these claims makes it difficult to respond in an evenhanded
manner, but I am trying to keep the conversation at a somewhat adult level.

Kevin Coates claims that this lock-in theory isn't central to the case, but
he is mistaken. The application-barrier-to-entry is a network effect, and it
is central to the judge's thinking. Mr. Coates says: "The authors look at
network effects arguments, and try and analyze whether Microsoft's leading
products lead as a result of their superior quality. They make some
interesting points, but how is that crucial to the specific anti-trust
allegations leveled against Microsoft?"                  The answer is that
it is crucial to the government's case. The fact that Navigator fell to
Internet Explorer is at the heart of the government's case. Our evidence
shows that it can be explained by the quality difference, whereas the
government claimed it was due to Microsoft's restricting Netscape's ability
to get their product to market. Knowing which interpretation is correct
allows one to decide much of the case.

Wayne Jones states: "This reinforces why I brought up AOL as an example. AOL
is a pretty good example of the early stages of de facto lock-in via network
effects (Windows being an example of late-stage lock-in where a platform
shift would be near impossible for many applications developers).  Some on
the forum seem to doubt the antitrust relevance of network effects, even if
such a phenomenon exists. The relevance is this: once first-mover status has
been secured and customers sign on to a given protocol, the number of
customers using that protocol becomes a barrier to entry because new
customers will be more inclined to use a protocol that gives them access to
other customers than to go with a relatively un-established protocol (Beta
being a really tired example of this. LPs, in an odd way, are another
example, since consumers would not buy a record player to play new
content...it simply isn't being released any more)."
The problem with this is the use of the term lock-in. We are all locked-in
to eating and breathing, but that is not a bad thing. Lock-in as it is used
in this literature is bad because a majority of consumers are locked-in to
using products that they really would rather not be using. New entrants will
find that they have to have better products to dislodge an incumbent. But so
what? In this hypothetical, it doesn't matter which product is used and if
new entrants have problems it is of little matter to the economy. AOL
instant messaging is, however, a good example of a market with strong
network effects.

Mr. Hallam-Baker requires more time since he has written so much and his
dislike of David Theroux is apparently so strong that he apparently has no
interest in a real discussion of the issues.
He is correct that Qwerty and Betamax are not the leading examples of
network effects, but they are by far the leading examples of lock-in. the
QWERTY story, brought to economics by Paul David (although Brian Arthur
claims David took it from him) is particularly famous. These lock-in
examples are factually incorrect.
For whatever reason, Mr. Hallem-Baker feels perfectly free to bluster away
with misinformed statements: "Microsoft won because of superior reviews?
Come on, how stupid do you think we are?  So if a company pays for glowing
reviews that would be okay?"                                             If
he had tried looking at what we had written before spouting off, he would
have realized that in many cases we found the product with the better
reviews to be the company that was not the bigger advertiser: Quicken when
Managing Your Money was the leader; Quark, when Pagemaker was the leader;
Excel, when Lotus was the larger firm; and so forth. If Mr. Hallem-Baker
discards all information found in advertising supported magazines that is
his business, but there is strong evidence that advertising doesn't buy the
better review.

I do agree with his views that just about any good argument should be
understandable by the layman, but it might take a little effort on the
reader's part. Our book is written for just such an audience, as he would
have realized if he had bothered to look at it. But we at least have to
understand what we mean by terms such as network effect and lock-in before
there can be a useful discussion.

His quoting from the Economist misses the point. Lock-in is inefficient when
we as a group fail to choose the better product because the group cannot
coordinate its behavior. It doesn't require irrational behavior on the part
of any individual-in fact each individual is assumed rational.

He states: "I read the L-M evidence and came to a quite different
conclusion. They discounted all evidence pointing to the fact that a
re-organized keyboard could be more efficient as 'biased' but do not point
to [I think he means 'discount'] independent trials that found QWERTY
better. The most obvious explanation for this is that the obvious costs of
retraining the typists compared to the possible advantage meant that nobody
thought it worth bothering to even *try* to break the QWERTY dominance."
If he had read our paper he would know that many organizations were
considering making just such a change and based their decision on a study
run by the General Services Administration in the 1950s. That study
concluded that there was no advantage in switching. The main evidence to the
contrary was a Navy study that in fact was biased. It did not treat the
Qwerty group the same as the Dvorak group and this bias increased the
reported advantage of the Dvorak group. We don't even know if the control
group was a fair control. And the study appears to have been conducted by
the creator of the alternative keyboard, who was then a Lieutenant Commander
in the Navy in charge of such things. Other studies that we tended to
discount did not have any control group and we don't take them too seriously
because it is well known that typing speed will increase for any format if
you put typists in (re)training programs. Without a control group such
examinations are meaningless.

Seth Finkelstein reports on a debate in the Economic History (not our field)
discussion group where we were asked to provide the starting article so that
other could debate the concept of path dependence. He quotes Richard
Rosenbloom as debunking our Beta/VHS story. Although Richard disagrees with
our conclusion that alliances didn't matter, he doesn't disagree with our
more important conclusion that Beta wasn't better than VHS. I still do not
agree with his claim that alliances were the key. He was also probably
annoyed that we didn't cite his article with Cusomano where they claim that
VHS won largely because Sony didn't have as powerful a group of firms
following its lead as did VHS. We should have cited their article, but the
fact is that we based our analysis on a book that I believe to be at least
equally well researched as their article. It is also the case that the size
of the coalition is itself a function of which standard looks more likely to
win on the merits. VHS's longer playing time was the only serious difference
between the formats and was responsible for RCA going with VHS when it might
otherwise have gone with Sony.
Finkelstein also gives some links to recent works of Paul David whose Qwerty
story we debunked in 1990. Although David is a well-known economic
historian, there is nothing in his articles that questions our dismantling
of his QWERTY story. Instead he questions our definitions of path dependence
because he is concerned that it is becoming too popular in the economics
profession. His anger is so great (the QWERTY piece is by far his most
famous), and his arguments so distorted (his major claim is that we don't
follow chaos theory closely enough), that it isn't surprising that he hasn't
been able to publish these articles now going on five years. Go ahead and
read them, but make sure you read ours too. Judge the work, not which
university we are at.

Finally Finkelstein (what is the source of his anger, anyway?) claims that
we are merely Panglossians, believing that this is the best of all possible
worlds. I wonder what is the source of his claim? He won't find it in our
papers, which I doubt that he read. Probably it is something he took out of
Paul David's screed. Why he would read our critics and not bother reading
our papers I do not know. We do believe markets make mistakes. But the large
types of mistakes that David talks about (turning down investments in
keyboards that provide a return on investment of 2200%) doesn't seem to us
very likely to happen.

Brad de Long says that we believe the burden of proof lies with those who
believe QWERTY worlds exist [actually he says we don't believe in network
effects, but he knows better and presumably just carelessly typed]. He is
correct. We do believe they need to come up with some examples. But our
reasons are not hidden or inexplicable. QWERTY economics is a new theory.
The old theory generally worked pretty well. QWERTY aficionados believe we
should have the government act on the basis of these theories. All we are
asking is a single piece of support for the theory. The other side says that
it should be presumed that QWERTY instances happen and that Margolis and I
need to prove they never do. I view this as analogous to cold fusion. Is it
asking too much to request that the proponents provide at least one instance
where it works? I think not. The alternative, that anti-cold fusion
proponents prove that cold fusion never works is unreasonable. Asking us to
prove that a QWERTY-type lock-in never has occurred is an impossible task
although asking the other side to find but a single case should not be too
difficult a task. The fact is, if they had even a single case, [QWERTY, or
Beta, or whatever] they wouldn't be objecting to this fairly trivial burden
of proof, would they? This demonstrates quite nicely that even Paul David
doesn't believe they yet have a reasonable instance of a QWERTY world.
And speaking about how the world looks DeLong states: "But while that is
important from the perspective of intellectual originality, it seems beside
the point as far as understanding the world is concerned..."  If I
understand him, Brad is interested in theories that talk about how the world
actually works. This is exactly what I am interested in, which is why I
believe that so much that passes as pure theory is useless. But surely, how
can a theory that has no supporting empirical evidence be useful in telling
us how the world runs?

Stan Liebowitz
Professor of Managerial Economics
University of Texas at Dallas
972-883-2807, fax 972-883-2818

For Listserv Instructions, see http://www.lawlists.net/cyberia
Off-Topic threads: http://www.lawlists.net/mailman/listinfo/cyberia-ot
Need more help? Send mail to: Cyberia-L-Request@listserv.aol.com

--- end forwarded text

R. A. Hettinga <mailto: rah@ibuc.com>
The Internet Bearer Underwriting Corporation <http://www.ibuc.com/>
44 Farquhar Street, Boston, MA 02131 USA
"... however it may deserve respect for its usefulness and antiquity,
[predicting the end of the world] has not been found agreeable to
experience." -- Edward Gibbon, 'Decline and Fall of the Roman Empire'

Nettime-bold mailing list