Naive [stupid] Models and Economics

Newmark's Door has a good example of ignorance in the application of computer simulations to economic phenomena. The link he provides is to an article which shows that if the only way people determine which restaurant to patronize is probablistically based on how many customers are already there, then whoever gets the first patron will get many more patrons, while the other suppliers languish.

That is a very big "if". I hope people do not try to extrapolate the results of this rather dumb (oops -- incomplete) model to the market for computer software.

A better model would assume that most customers play a repeated game with the restaurants having favourites but sampling others now and then. It would also assume that if customers find something that satisfies them more than the current popular restaurant, in terms of quality, service, price, etc. [geez, where are these economic variables in most such simulations?], then customers will respond to these incentives by shifting their patronage.

There's a reason some schools call it "Price Theory" and not microeconomics.

Let's face it: consumers are not robots responding to only one variable. We may take popularity as a signal of quality, but we also look at other variables or sources of information, and we especially ignore popularity as a measure of quality if we have repeated samplings ourselves.

Do you think this doesn't happen with software because people get locked into bad technology? Contrast the following:

• CPM -- MSDOS
• Lotus -- Excel
• Wordstar -- WordPerfect
• WordPerfect -- MSWord
• Internet Explorer -- Firefox
• Windows XP -- Linux

If a better product comes along, more people are more likely to use it (okay, there hasn't been a stampede to Firefox and Linux, but they are gaining market share). Despite the cries and fears of our being locked into inferior technologies by market failures, there sure has been a lot of change in the computer industry over the past 25 years.

For a rigorous, yet highly readable, treatment of this problem within the context of computer software, see Winners, Losers, & Microsoft by Stan Liebowitz. And for a good selection of work on the topic in general, see this collection of works by Liebowitz and Margolis, which includes their famous QWERTY article.