Economic Experiments in Internet Infrastructure By Shane Greenstein

1
Economic Experiments in Internet Infrastructure
By Shane Greenstein

• Designing Cyberinfrastructure for Collaboration
  Innovation
• National Academies, January 29-30, 2007

2
Prologue

• Thanks for giving me the opportunity to
  participate.
• My experience in the last decade focused on
  analyzing economic conduct in commercial markets
  related to the IP stack.
• Try to offer conceptual thinking.
• Essay draws illustrations from my experiences in
  access markets, which I know in the most detail.
  The arguments and observations are meant to be
  more general.

3
Why different experiences integrating innovative
advance into goods services?

• What is being integrated pooled? Innovation.
• Origins of ideas? Why in need of integration?
• Useful to consider economic experiments
• Yields lessons for participants in commercial
  Internet markets. Can be acquired only through
  tangible market experience. Cannot be learned in
  lab. More coming.
• Contrast w/traditional ladder, a sequential
  process, where inventions go from lab to
  commercialization.
• Applies to some RD, but not all innovation
  policy.
• E.g., DARPA funds RD in packet switching. Later
  and after many steps the invention finds its way
  into commerce.

4
Directed experiments

• Firms invest additional dollars/time/effort with
  intent to learn information valuable for own
  business.
• Learn about customer responses to new, then
  refine products, operations, etc...
• Usually while also operating on-going business.
• Both large and small firms undertake these.
• Common in first decade of commercial Internet.
• What was an ISP contract? Changes many times.
• What line of business and services for an ISP?
• Who does it? Cable firms contract w/_at_home for
  service?
• ISPs explore tailoring service to needs of local
  community.

5
Undirected experiments

• Collections of actions by firms and users results
  in unanticipated learning about value.
• Interplay of one participants actions w/
  another.
• Narrow motivation for invention, but events lead
  to unanticipated broader application, defying
  earlier forecast.
• A feature since Bells improvement to
  telegraphy
• Example IEEE committee 802.11 founded in mid
  1990s to develop wireless Ethernet for business.
• Unanticipated growth Firm-user interaction,
  redesign efforts, growing interest, enthusiastic
  uses in new locations, etc., blossoms into
  Wi-Fi for popular hot-spots.
• Which then motivates further directed
  experimentation e.g., Intel initiates its
  strategies to promote Centrino.

6
Theory of the economic experiment cycle

• Cycle of commodification accumulation
• As lessons spread, then a decline in value for
  single firm b/c loss of differential advantage
  from uniqueness.
• Know-how is accumulation of such common lessons.
• Dispersed technical leadership.
• Firms in many locations learning, growing
  know-how.
• Explains why some lessons spread quickly and
  explains why lessons in different locations build
  on one another.
• Set of lessons becomes larger than any single
  firm would have or could have developed on its
  own.
• Often with few restriction on later use.

7
Practically speaking.

• Codified know-how tends to spread quickly.
• Many channels for spreading of technical
  lessons...e.g., how does a new 56K modem bank
  technology work?
• Near-technical lessons spread as faste.g.,
  what fraction of dial-up users crowd modem banks
  at peak hours?
• This cycle can support extraordinarily rapid
  technical progress in communities of
  users/vendorse.g., among dial-up ISP firms b/w
  1994 and 1999, or among Wi-Fi equipment firms
  providers/users b/w 1999 2004.
• Non-codified business lessons spread more slowly.
• More barriers to spreading complex business
  lessons e.g., What labor policies save costs? Or
  e.g., What costly actions lead to highest
  customer loyalty? And on on.

8
Broad policy principles for economic experiments

• Policies shape innovative outcomes
• By increasing/decreasing incentives for
  experiments
• By increasing/decreasing integration of lessons
  into use
• Encourage Re-use of lessons infrastructure.
• Use re-use of lessons infrastructure for
  multiple innovative opportunities
  experimenters.
• Avoid discouraging initiatives from entrepreneurs
  and/or incumbents.
• Avoid policies that encourage lots of
  entrepreneurship without also encouraging
  integration into use.
• Policies that overly determine direction of
  experimentation.

9
Institutions that encourage discourage
emergence of economic experiments

• Institutions supporting confrontation of
  different opinions/forecasts about likely value
  from innovation
• Standards committees that deal with anticipatory
  designs
• Transparency, as in IETF and W3C, facilitate
  diverse views
• Breakdowns over technical disagreement less
  likely than one over assessment of business
  prospects. E.g., Issue in 56K modem war arise
  from control of design ? affects royalties,
  prestige, influence over upgrades.
• Encourage sharing of accumulated lessons.
• More than just encouraging the trade press.
• Equipment firms pass lessons on to buyers, etc.
• Apache community a channel for sharing lessons.

10
Institutions that encourage discourage use of
economic experiments

• Large firms can (and do) contract for
  experiments.
• Large firms afford access to more source of info.
• Broad array of support instruments (e.g., Intels
  actions).
• VC-mediated channels a partial substitute for
  small.
• Best VCs have access to short term information
  about rate direction of technical change. Help
  their young firms.
• Entrepreneurs depend on regulatory rules that
  mediate small/large firm negotiations in presence
  or absence of contracting.
• E.g., Computers II III fostered a heterogeneous
  BBS market and dial-up ISP market.

11
Open questions about policy principles

• Sweeping statement about IP rules? No.
• Apache clearly beneficial for building know-how.
• Clear IP facilitates efficient contracting over
  technology transfer b/w private parties, very
  useful.
• According to this framework, the key question in
  a standards committee what set of IP rules
  facilitate confrontation of views and
  accumulation of lessons?
• Sweeping statement on firm size/numbers? No.
• Large numbers of firm are better than fewer.
• More diversity among firms is better than less
  diversity.
• Better funded firms are better than less.
• Large better funded implies fewer firms less
  diversity.

12
Open questions about policy principles

• Depend on contracting for all experiments between
  existing stake-holder and young experimenter? No.
• Often not possible before demonstration of an
  experiment, but often possible after a
  demonstration. E.g., You-Tube.
• Contracting in advance potentially not in
  experimenters interest when revelation about an
  experiment to another firm in advance undermines
  probability of success.
• Default rules for contracting b/w large/small
  firms shapes returns after innovation, which
  shapes ex ante incentives.
• The consequence from spread of non-technical
  lesson in wiki-environments? Positive.
• Facilitates sharing of technical and
  near-technical lesson.
• The promise of aiding the spread of business
  lessons.

13
Epilogue

• Thank you for your attention.