Episodes of collective
invention
Peter B. Meyer
U.S. Bureau of Labor Statistics
25 March 2004
at ESSHC
(1)
(2)
Outline
Definition of collective invention
Historical cases
(3)
Bessemer steel in US
Microcomputers
Networks and institutions
(4)
Modeling ideas
Clubs
Professional associations
Journals
Job turnover
Search for production function
Subsidy by hobbyists
Definition of “collective invention”
• Process in which improvements or
experimental findings about a production
process or tool are regularly shared.
• Allen (1983) defined it to apply to profitseeking firms only, and distinguished
collective invention from formal research
and development.
• This is a social learning process, by which
people learn about technological
opportunities, working collectively,
outside hierarchies
• There is evidence that productivity
growth is high during or after collective
invention periods.
Cases
Previous literature:
• Cleveland district (Britain) iron makers,
1850s-1870s (Allen 1983)
• steam engines after 1800 (Nuvolari 2001)
• free or open-source software
Here:
• Bessemer steel in US 1866-1881
• microcomputers in 1970s
Others:
• steam engines before 1718 (Thurston
1878)
• aeronautics in late 1800s (Couch, Jakab)
Networks
Networks appeared through which the
experimenters communicated. Central
figures usually organized any institution.
- Organizations
- professional associations, like American
Institute of Mining Engineers
- Bessemer patent pool
- clubs, like the Homebrew Computer Club
- Journals
- AIME’s Transactions
- several in aeronautics
- Lean’s Reporter in post-1800 steam engines
- Job turnover
- as in Silicon Valley in the case of personal
computers and software
- also observed in the steel case
Why model it
economically?
• Economic outcomes pose problems for analysis here:
- Time and resources are applied to technology
development: who chooses this and why?
- Something of value -- a new technology – may be
produced. How did it happen?
• A model might generalize over these cases for
predictions and evaluating counterfactuals.
– Why didn’t the invention appear somewhere else first?
– Could technological development be faster or slower?
– Were many alternatives were considered to the design of
the modern personal computer, or was the result very
path dependent?
– How will free and open-source software projects evolve?
• A model might avoid details of technology
– even where path dependence is historically significant.
• Compatible microeconomic theory is new – Harhoff,
Henkel, and von Hippel (2002), Saint-Paul (2003).
Model of search for
technology
In search models, participants usually sample
random information and consider changing
behavior, e.g. search for a job, search for an
applicant for a job, or search for mate.
Search for technology models usually have profitseeking participants investing optimally in
research and expecting useful findings.
A search model of collective invention would be
different:
-- no balanced equilibrium level of searching but
rather strategies, perhaps in Nash equilibrium
-- positive feedback: more searching produces more
success which draws in more searching
-- players can exit by starting firms
-- collective search may expire and be replaced by an
industry
-- begins with cooperation, may end in competition
Search is for production function y=f(X,T) where y
is hoped-for output, X is set of known inputs, but
technology design T has unknown dimensions.
Players share information through network.
Why did they share?
• Recognition, prestige, fame (for
individual or employer)
• Employment opportunities
• Impractical or costly to keep secrets (e.g.
job turnover cases)
• New technology could raise value of
assets (e.g. local iron ore in Allen, 1983)
• Employer’s own technology might
improve through paybacks, alliances, or
side effects
• To establish desirable standards (especially
in software, for market power or to make a particular
feature universal)
• For employer to evaluate employees
• Fun, exciting
• Virtuous
How to model incentives?
• One structure is “career concerns”
– the person wants employment opportunities,
recognition, and other private gains
– hypothesized in some IO writing about opensource, and makes sense on the Web
– but this presumes a superstructure for recognition
exists, thus assuming away part of the answer
– returns to this activity are often low. some
participants are treated as crazy.
• Another structure: assume the player wants the
technology to advance
– then players share for the same reason they
experiment
– the effect they want is to change the environment
– historically, some of the experimenters say they
want this
Modeling idea
-- Participants are for private reasons willing to pay
to experiment with this technology: “hobbyists”.
-- Treat individuals and organizations as the same.
Some individuals may then start firms.
-- These players subsidize the public search.
-- Might they optimally choose to share their
findings to advance the technology of the whole
network?
Stages in simulations
Period
t-1
Period
t+1
Period t
Production
decision
Use
decision
Sharing
decision
Adoption
decision
Production
decision
n
If other player offered me tech info,
shall I adopt it?
Shall I
produce?
y, n
y
n
Shall I
produce?
Receive revenue
Does Nature give
me new tech info?
n
y
Shall I use it?
n
y, n
y
Shall I share it with other
players?
y
Do other players
adopt it?
Hypotheses
• These help collective invention:
– freedom to assemble
– freedom of the press
– availability of the Web
• Collective invention declines
when profitable enterprises
stabilize.
• Productivity gains can be
attributed to the process
Conclusion
• There is a list of examples in which
collective invention by the definition here
was significant
• We can abstract from the certain common
elements, and maybe model them:
– search
– networks for sharing technology
– subsidy by experimenters
• It continues alongside formal R&D
– in many free / open-source software projects
simultaneously
– potentially in other fields where designs can
be represented electronically on the Web
• computer hardware; architecture; mechanical
designs; art.