1
Networks of Tinkerers:
a Model of Open-Source Innovation
Peter B. Meyer BLS brown bag seminar,
March 15, 2006
2
IntroductionHobbyists have developed important technology.
open source software Linux; email processing; Web servers/browsers
personal computers Homebrew Club of hobbyists, circa 1975
airplanes – a clearly documented case
Goal here: try some assumptions about the hobbyists / tinkerersshow they would share information in networksexplore assumptions a bit
3
Aircraft developmentsFrom the 1870s to the 1900s there is a lot of “open-source” innovation in aircraft
1800 – George Cayley and many others try aeronautical experiments
starting 1870s, several aeronautical journals
1893 Octave Chanute’s Progress in Flying Machines
1903 – Wrights fly
4
Assumptions for micro model Assume there are motivated tinkerers
We do observe this Assume they have a way to make
“progress” defining progress carefully
Assume total technological uncertainty No market is identifiable so no R&D / competition
The tinkerers would share information
5
The Tinkerer
U t 0
tat
Tinkerer has activity/hobby A. (for “aircraft” or “activity”)
Tinkerer receives positive utility from A of at per period.• a0 is known• later choices and rules determine at
β is a discount factor between zero and one (assume .95) applied to future period utility.Net present expected utility:
6
Tinkering rules Tinkerer may invest in ("tinker with") A The agent thinks that tinkering this
period will raise all future period payoffs at by p units each time period. p stands for a rate of progress, which is
subjectively experienced by the agent We assume p is fixed and known to the
agent Example: .07
7
Tinkering decision
p p 2 p 3 p 4 p1
Tinkerer compares those gross benefits to the cost which is 1 utility unit
Tinkerer weights estimated costs and benefits
Benefits from one effort to tinker equal p in each subsequent period.
The present value of those payoffs is:
8
Rates of Progress
p 1
Progress must meet the criterion above for tinkering to be worth it
Progress is subjective
There are not many tinkerers working on this activity who can make this much progress.
More issues later.
9
Payoffs from endless tinkering
Payoffsperiod 0 period 1 period 2 period 3 period 4 …
-1 p p p p(pβ/[1-
β] )
-1 p p p β()
-1 p p β β ()
-1 p β β β()
-1 . . . a0-1 a0-1+p a0-1+2p a0-1+3p a0-1+4p a0-1+tp
a01 1
1 p1 2
Present value of all that at time
zero has a closed form:
10
A network of two tinkerers
U0 a01 1
1 p1 fp21 2
Case of two tinkerers with identical utility functions p1 and p2 – subjective rate of progress Their innovations are useful to one
another Tinkerers form a network Present value of expected utility:
11
Subgroups of occasional tinkerers
A group of slow-progress tinkerers might agree to work together to generate progress rate p.
Then the group acts like a single “tinkerer” in terms of its output
and also in its incentive to join other groups There are something like economies of scale here;
it’s a positive sum game. So Wilbur and Orville Wright could be one tinkerer maybe also:
Boston-area group All readers of a certain journal Kite people, together, as distinguished from balloon
people
12
“Progress” is subjective
13
Standardization and Specialization
U0 a01 1
1 cs p1 f2p2
1 2
Only the fraction f (between 0 and 1) of experiments by player two are usable to player one
Suppose for a cost cs player one can adjust his project to look more like the other tinkerer’s project
And, that this would raise the usable findings to f2
That’s standardization Present value of utility from standardizing is:
14
Standardization and Specialization
Key comparison is above Player one benefits more from
standardizing if, ceteris paribus: the other tinkerers are producing a large flow of innovations p2; the cost of standardizing cs is small; the gain in useful innovations from the others (f2-f) is large.
Same logic supports specialization These are technology phenomena, not requiring
market processes
p2f2 f1 2
cs
15
Searching and Joining Suppose there is a cost to joining the network
costs of subscribing, paying attention it’s worth the cost to a tinkerer if
the cost (cj) is low he values future outcomes a lot the others are producing a lot of progress (p) their progress is useful to him – f is high enough
Suppose there is also a cost to searching for new members Chanute wrote book others published journals Then the search costs affect innovative output (Web has
effect) There is a role for a special effort to expand the network Paper does not model this
16
Entrepreneurial Exits At a few points there was tension:
Ader “drops out” in 1891 Langley keeps secret wing design after
1901. (Chanute shares it anyway.) Wrights stop sharing as much in late
1902 After some perceived of breakthrough Jobs and Wozniak start Apple
they hire Homebrew club people as employees Red Hat becomes a company
17
Example exit: Clement Ader’s Eole
It traveled 50 meters in uncontrolled flight in 1891 French military thought it would be useful. Ader didn’t patent outside France because it would
expose details. Chanute criticized this choice. Ader “drops out” from prior communication links.
18
Entrepreneurial exits from network
If a tinkerer has an insight into how to make a profitable product it may be worth leaving the network
conducts directed R&D becomes an entrepreneur enters economic statistics
19
Conclusion This process may be important
explaining the rise of industrial countries a long time ago with open source software, now
I do not know of other models of it Key assumptions:
technological uncertainty (no clear product and market) motivated tinkerers some way to make progress some way to network
Search and matching costs take some more thinking An industry can spring out of this, not well modeled
yet