The Open-Source Software as a Mechanism to Allocate Attention
IKEDA Nobuo
Center for Global Communications,
International University of Japan
Open-source software (OSS) is a new name for the
program that used be called "free software" or “freeware”. It is
misleading to call it “free”, because it doesn't mean “no money”. You can make
money on it, but it can be freely copied and should be distributed with its
source codes (original codes written in programming language).
If you think such programs are unreliable
toys for nerds, you are wrong. Everyday you must be using OSS, namely the
Internet. Its protocol, TCP/IP, was developed by IETF (Internet Engineering
Task Force), opening the codes in “Request For Comments” format and e-mail discussions.
There has been a long tradition of OSS since UNIX, Emacs, and TeX. Last March,
Netscape Communications freed the source codes of Communicator 5.0 - nicknamed
as “Mozilla” - which millions of people are downloading, compiling, and
debugging.
It is a wonder that the international
standard of computer networks was neither built by the national project of
Information Superhighway nor “value added networks” of corporate titans, but
informal groups of volunteers, giving away codes free. They seem to be
remarkable exceptions for economists’ assumption that economic activities are
organized either markets or governments. Here I present some stylized facts
observed among them and try to explain them economically.
In
a neoclassical economy, goods are priced equal to their marginal costs to
produce. But there is a serious flaw in this mechanism when applied to
information. Because the marginal cost to copy information is practically zero,
it will be priced as zero, so nobody can recoup the fixed cost in competitive
equilibrium. The most popular prescription for this problem is the “intellectual
property right” (IPR) which creates monopoly to protect innovator's R&D
cost.
However, monopoly distorts resource
allocation, so the net social gain from IPR depends on whether the appropriation effect, in which imitation
reduces innovator’s incentives, is larger than the learning effect, efficiency resulting from information sharing.
This is not so much a theoretical as an empirical problem: historical evidences
from the Industrial Revolution suggest that the latter is far greater than the
former (Crafts 1997).
Learning effects are even larger in
software, because, in writing programs, all programmers learn from other source
codes and sometimes copy them in part. They can learn nothing from object codes
(executable programs). If they were all forced to write software from scratch,
the resulting efficiency loss from duplication would overwhelm the gains from
protecting IPR. Moreover, Raymond (1998) argues that a programmer's incentive
to innovate is an increasing function
of the number of copies of her works in the active communities of OSS, such as
Linux, a “clone” operating system for UNIX, and Apache, the most popular WWW
server software. Programmers “homestead”
cyberspace to write programs, for which they claim “ownership” of
originality. You can see long credits of programmers at the end of release
notes of OSS.
According to the property right theory
(Hart 1995), this kind of joint ownership is inefficient for physical assets,
but efficient for non-rival assets such as information. Taking into account
network externalities, the net benefit of protecting IPR is dubious for
software. If learning effects are greater than appropriation effects, it might
be socially efficient not to protect
IPR for software, as insisted by OSS proponents such as the GNU project..
OSS
needs not be “free”. Some are distributed with strict licenses such as GPL (GNU
public license), and Netscape Communications intends to sell server software
complementary to Mozilla. Linux has many commercial distributions, for which
user supports are guaranteed. Many programs are distributed as shareware, which can be freely
downloaded (without source codes) but should be paid for its author.
There would be more innovations to make
money without exclusive “copyright”, which ignores user's rights. It is
desirable to design different types of ownership for different types of assets
to balance owner’s and user’s benefits. This problem seems similar to the
well-known one in public economics: if a good is non-excludable and non-rival,
it is efficient to be supplied publicly. Similarly, my conjecture might be
stated as follows:
1. Appropriation effects are negligible than
very large learning effects for “platform” software such as operating systems
and compilers, so it is most efficiently supplied as OSS.
2. As for small utilities for specific use,
learning effects are not so large that it is suitable to supply them as
shareware .
3. Large applications such as DTP tools, for
which appropriation effects (R&D costs) are significant, had better be
supplied with various kinds of licenses tied to servers.
4. It is desirable to supply software commercially
only if its appropriation effects are great and learning effects are
negligible, e.g., business applications.
In more familiar terms in public
economics, 1 might correspond to pure public goods, 2 to “club goods”, 3 to “commons”, and 4 to
pure private goods. But the most significant difference is that “public” doesn't mean “government”.
Legal enforcement of property rights is only a subset - the most expensive one
- of much broader concept of ownership. OSS is usually supplied by non-profit
organizations, which are functioning as allocation mechanisms of information
without prices. If governments fail as well as markets, it may be safe to
supply public goods by neither of them.
This
“gift economy” is, however, no less competitive than the market economy. To win
the official standard at IETF, you have to write thousands of e-mails to
discuss, negotiate, and revise the codes with many engineers over the world.
This harsh and open competition is one of the reason that made the Net the
international standard. OSS is much more reliable and stable than commercial
software such as MS-Windows, because OSS is tested and debugged by millions of
users before its final releases.
Here engineers are not competing for
more goods but more attention,
according to Herbert Simon. Since information is so
abundant as to be free goods in cyberspace, the intermediaries to allocate the
scarce attention are important: search engines, directory services, and platforms
such as OSS, by which engineers and users can collaborate. Raymond regards Linux
and Apache as decentralized “bazaar” models of collaboration, which are much
more efficient than the older “cathedral” models commanded by central planners.
Prices are efficient mechanisms to induce
production, but they are in no way the only ones. Especially in intellectual
production, pursuit for attention (reputation) often dominates pecuniary
incentives. For example, if the authors of famous Black-Scholes formula for
option pricing had patented it instead of publishing, they could earn billions
of dollars. If, as Schumpeter argued, the crucial mechanism for innovation is
not the price but the creative
destruction by entries and exits, it is not surprising that the innovation
is highest in cyberspace where there are no sunk costs to deter entry and exit.
Crafts,
N.F.R. (1997) in D.M. Kreps and K.F. Wallis (eds.), Advances in Economics and Econometrics, Cambridge U.P.
Hart,
O. (1995) Firms, Contracts, and Financial
Structure, Oxford University Press.
Raymond,
E. (1998)
http://sagan.earthspace.net/~esr/writings/homesteading/
Varian,
H. (1995)
http://www.sims.berkeley.edu/~hal/pages/sciam.html