economic efficiency of free software

Here's a naive economic argument that spending money on free software
is more efficient than spending it on proprietary software.  Perhaps
there are some good ideas in here, but I don't know enough about
economics to be able to judge.

90% profit margins induced by copyright-based monopoly ultimately mean
that less than the ideal amount of effort goes into improving software,
and less than the ideal number of users benefit from the software.  

The second is obvious --- users for whom the value of a Word upgrade is
less than its price simply won't upgrade, so even though the economy has
already allocated the human effort to make the improvements to Word,
only a fraction of the eligible users benefit from the upgrade.  Perfect
price discrimination would close this gap.

The first is less obvious, and perhaps requires lower friction than
presently exists in the market.  The underlying intuition is that if
there are 5 million Word users each paying $100, but only $50 million of
that $500 million actually goes into engineering effort to improve Word,
then the extra 90% is market friction.  The users would be much better
off if they could give $50 million directly to the Word developers,
perhaps through an administrative organization like the United Way that
added on 30% for administrative costs.  As long as anybody can make
those improvements to Word, the administrative-cost premium is limited
by market competition --- as indeed it was for Cygnus.

In this scenario, of course, there appears to be a prisoner's-dilemma
situation leading to a free-rider problem, leading to underproduction of
Word improvements.  Cygnus's solution was to promote its service brand
by pointing to its technical staff, who were the actual maintainers of
gcc, gdb, binutils, etc., and they did this successfully for many years.
If this had not adequately mitigated the free-rider problem, gcc et al.
would have rapidly fallen behind the competition: SparcWorks, SGI's
compiler, Compaq's compiler, etc.  Instead, the Cygnus products remained
comparable to their proprietary competitors --- clearly superior in some
areas, clearly inferior in others --- and they still lead the embedded
market, several years after Cygnus made some bad decisions and destroyed
itself.

So, while there is a free-rider problem, in many markets, it turns
out not to cost as much as the alternative frictions imposed by a
government-granted monopoly on a particular piece of software.

The frictions are actually pretty similar: in both cases, Word is
producing at least $500 million worth of value, but in both cases, only
a fraction of that goes into improving the product.  In the proprietary
case, the remainder is a monopoly rent, and the monopoly spends as
little as possible to maintain its monopoly rent; in the free case, the
remainder accrues to the users of the software.

It follows that if you want to develop a proprietary chunk of software
on speculation, you should persuade a possible future monopolist to
invest in creating it, in the anticipation of future monopoly rents; and
if you want to develop a free chunk of software on speculation, you
should analogously persuade a lot of possible future users to invest in
its creation, in the anticipation of future value derived from running
the software.

In a world of rational agents and perfect information, of course,
getting investment from users is easier than in the real world.  So is
getting capital from VCs, but we have traditions that make that easier.

I believe you're correct that VC funding is no longer going to produce
innovative software.  The incentives aren't there.  But I think the
incentives to produce innovative software are as strong as ever; the
money flow is just different.

The two effects I mentioned earlier --- more efficient allocation of money
due to elimination of monopoly rents (which we trade for the free-rider
problem, which is a smaller inefficiency in some cases), and more users
--- have a multiplicative effect that means that a dollar spent to fund
an improvement to a piece of free software is a far more efficient
allocation of resources than the same dollar spent to buy a license for
proprietary software.