Gonçalo Pacheco-de-Almeida (HEC Paris)�Peter Zemsky (INSEAD)
Some Like It Free�Innovators’ Strategic Use of Disclosure to Slow Down Competition
Empirical Motivation
– GM vs. Toyota launching a new line of alternative fuel SUVs
– $1 million revenue loss per day for a $10,000 car (Clark, 1989)
– HP vs. IBM introducing a new series of computer models
– $1.1 million revenue loss per day for the HP 930 computer (Waldman, 1986)
– Intel vs. AMD developing the next generation of microprocessors
– $733,000 revenue loss per day for semiconductor plant (Salomon et al., 2006)
But is faster technology development always better?
Empirical Motivation
Empirical Motivation
– Intel’s 386 cost $200 MM and took 48 months to develop (FTC, 97)
– A two-week schedule compression would have increased costs by $8.6M (Mansfield, 88)
Empirical Motivation
– Intel’s 386 cost $200 MM and took 48 months to develop (FTC, 97)
– A two-week schedule compression would have increased costs by $8.6M (Mansfield, 88)
This tradeoff is called Time Compression Diseconomies
Costs ($B)
0
10
20
30
40
50
0
1
2
Costs
Development Time
Industry Average Refinery
(Oil and Gas Journal)
(30 Months, $0.23B)
Mansfield’s (1988)
Estimate
(Months)
Source: Pacheco-de-Almeida, Hawk, and Yeung (2010)
With Spillovers
(Acceleration-Cost Elasticity = 2.7)
Time Compression Diseconomies
Research Questions
– Industry patterns of technology diffusion, sustainability, performance
– Firms’ internal timing decisions with imitation and spillovers
– Technological resource accumulation with complexity, diminishing returns
R &
D
Stochastic
(Reinganum, 81)
Deterministic
Research Questions: Levels of Analysis
– Industry patterns of technology diffusion, sustainability, performance
– Firms’ internal timing decisions with imitation and spillovers
– Technological resource accumulation with complexity, diminishing returns
Industry
Firms
Resources
IO
RBV
2
1
3
Agenda
Model Overview
– Development depends on time compression diseconomies, firm capabilities
– Imitative development: follower waits and imitates the leader
– Concurrent development: both firms develop the resource in parallel
Phase I: Pre-Pentium
Phase II: Post-Pentium
– Firms maximize the NPV of revenues minus development costs
Model: Revenue Flows
T1
T2
0
Time
π00
π01
π10
Δ1
π11
Δ2
Revenue Flows
Model: Revenue Flows
Model: Development Costs
Discount Rate
Development Time
Development Effort
− Leader: Z1 = K
(1− d1)
Complexity
Capabilities
− Follower: Z2 = (1− d2) K
(1− s)
Spillovers under Imitative Development
Model: Development Costs
– Both firms develop the technology,
– Follower not much more capable than leader,
Firm-Level Analysis
Optimal Development Times
0
10
5
0
Single Firm Development Time
Cost
Present Value of Revenues
Development Time
Optimal Time
20
10
Profits
Revenues and Costs
Costs
Concurrent Development Times
Condition: Zi is non-observable and Δ2 is not too much lower than Δ1
Leader firm
Follower firm
Imitative Development Times
Leader firm
Follower firm
Condition:
Industry-Level Analysis
Competitive Dynamics
Technology Diffusion and Sustainability
0
30
20
10
0
1
0.75
0.25
0.5
Leader in Concurrent
Follower in Concurrent
Leader in Imitative
Follower in Imitative
d1 = d2 = ½, K = 19, r = .1, {π10, π11, π00, π01} = {4, 3, 2, 1.9}
Sustainability in Imitative
Spillovers (s)
Time-to-Market
Proposition: The leader is faster to market under concurrent development than under imitative development.
N.B.: There is no clear ordering for the follower.
Proposition: The sustainability of the leader’s competitive advantage is greater under imitative than concurrent development if and only if spillovers are sufficiently small.
Technology Diffusion
and Sustainability
N.B.: The leader has a period of competitive advantage under both imitative and concurrent developments.
Performance and Endogenous Spillovers
0
1
0.75
0.25
0.5
22
21
20
Leader in Concurrent
Follower in Concurrent
Leader in Imitative
Follower in Imitative
Optimal spillovers for leader
Optimal spillovers for follower
Firm Profits
Spillovers (s)
The Limits to Imitation
Free Revealing
Performance and
Endogenous Spillovers
Proposition: Suppose there is imitative development.
(i) The leader’s profits are maximized at s = 0.
(ii) The follower’s profits are maximized at some s < 1
if π10 is large relative to π11.
Proposition: Suppose the leader commits to certain level of spillovers s and then the follower chooses concurrent or imitative.
For any parameters (K, di, r, π00, π11), there exist values of π10 and π01 such that the leader optimally sets s > 0.
The Limits to Imitation
Free Revealing
Proposition: Suppose the leader is at least as capable as the follower (i.e., d1 ≥ d2).
(i) Under concurrent development, the leader has higher
profits than the follower.
(ii) Under imitative development, the leader has higher
profits if and only if spillovers are not too high.
Performance and
Endogenous Spillovers
Proposition: In a new market, an increase in the time compression capabilities of the leader increases the relative profits to the follower of imitation.
Resource-Level Analysis
Microfoundations of Technology Development
Model Microfoundations
Knowledge
Complexity
=
K
(1 ─ )
Spillovers
s
(ct)
Investment
∫
0
T
dt
Accumulation
α
Diminishing Returns
(1 ─ )
Capabilities
d
Model Microfoundations
=
K
(1 ─ )
s
(ct)
∫
0
T
dt
α
(1 ─ )
d
C(T)
=
ct*(T)
e –rt
∫
0
T
dt
α r
(1 ─ α)(e rT α / (1-α) ─ 1)
(1- α) / α
=
(1 ─ s )
1/α
K
1/α
(1 ─ d )
1/α
when α = ½ (for tractability)
=
r Z
2
e − 1
rT
Discussion
| | |
Hypothesis
► Spillovers are detrimental to innovators because R&D rents are not appropriated
: Free Revealing
Classical View
Free Revealing View
Free Revealing View
Managerial Implications
► Innovators should conceal IP to reduce diffusion if spillovers are uncompensated
Hypothesis
► Spillovers may be beneficial to innovators when losses in R&D rents are compensated by:
• gains in reputation | • market enlargement
• standard setting | • cross-spillovers
Managerial Implications
► Innovators should reveal IP to induce diffusion if spillovers are compensated
(Arrow, 1962; Spence, 1984; Scherer, 1984; Mansfield, 1985; Levin et al., 1987; Cohen et al. 2000; Harhoff et al., 2003)
(Allen, 1983; D’Aspremont and Jacquemin, 1988; De Fraja, 1993; Harhoff, 1996; Lim, 2000; von Hippel and von Krogh, 2003)
Hypothesis
► Spillovers are detrimental to innovators because R&D rents are not appropriated
BUT may persuade rivals to wait and imitate, thereby reducing competitive pressure
Managerial Implications
► Innovators should reveal IP to induce delayed diffusion if spillovers are uncompensated
(Pacheco-de-Almeida and Zemsky, 2010)
Technology diffusion decreases profits
Technology diffusion increases profits
Technology diffusion decreases profits
Discussion
| | |
Hypothesis
► Spillovers are detrimental to innovators because R&D rents are not appropriated
: Free Revealing
Classical View
Free Revealing View
Managerial Implications
► Innovators should conceal IP to reduce diffusion if spillovers are uncompensated
Hypothesis
► Spillovers may be beneficial to innovators when losses in R&D rents are compensated by:
• gains in reputation | • market enlargement
• standard setting | • cross-spillovers
Managerial Implications
► Innovators should reveal IP to induce diffusion if spillovers are compensated
(Arrow, 1962; Spence, 1984; Scherer, 1984; Mansfield, 1985; Levin et al., 1987; Cohen et al. 2000; Harhoff et al., 2003)
(Allen, 1983; D’Aspremont and Jacquemin, 1988; De Fraja, 1993; Harhoff, 1996; Lim, 2000; von Hippel and von Krogh, 2003)
Hypothesis
► Spillovers are detrimental to innovators because R&D rents are not appropriated
BUT may persuade rivals to wait and imitate, thereby reducing competitive pressure
Managerial Implications
► Innovators should reveal IP to induce delayed diffusion if spillovers are uncompensated
(Pacheco-de-Almeida and Zemsky, 2010)
Examples
Intel post-Pentium; IBM pre-09; GM pre-71
Examples
Intel pre-Pentium; IBM post-09; GM post-71
• time-compression diseconomies |• innovation timing is critical (Moore’s Law) | • duopoly
Discussion
: Free Revealing
• time-compression diseconomies |• innovation timing is critical (Moore’s Law) | • duopoly
Discussion
: Free Revealing
Free Revealing
► Intel shared chip designs
► AMD waited and imitated
Classical View
► Intel refused to share designs
► AMD started parallel development
“We don’t have any barriers for competitors, just a few speed bumps that people have to go around” – Intel lawyer
Negative effect on Intel’s profitability
Discussion
| | |
Hypothesis
► Spillovers are beneficial to imitators because development costs and time are reduced
Classical View
The Limits to Imitation View
Managerial Implications
► Imitators should maximize absorbed spillovers by investing in absorptive capacity, co-locating with leaders
(Arrow, 1962; Spence, 1984; Scherer, 1984; Mansfield, 1985; Levin et al., 1987; Cohen et al. 2000; ; Cohen and Levinthal, 1990; Dierickx and Cool, 1989)
Hypothesis
► Spillovers are beneficial to imitators because development costs and time are reduced
BUT too high spillovers may slow down innovators’ development and, thus, imitation
Managerial Implications
► Imitators should limit the know-how that they can freely appropriate from innovators: the dark side of absorptive capacity
(Pacheco-de-Almeida and Zemsky, 2010)
: Limits to Imitation
Discussion
| | |
: Empirical Testing
Hypothesis 1 (Outgoing Spillovers):
Free revealing by innovators Delayed technology diffusion
Increased innovator performance
Hypothesis 2 (Incoming Spillovers):
Limits to appropriation by imitators Faster technology diffusion
Increased innovator performance
Spillovers: patent citations, technology sharing agreements, geographical co-location, secrecy measures from Carnegie Mellon University (CMU) survey; absorptive capacity measures (e.g., R&D investments)
Diffusion: publicly available information on project development times in industries such as semiconductors, oil and gas, automotive, and consumer electronics or imitation lags from CMU survey
Performance: focus on publicly traded companies (Compustat, etc)
Data:
Conclusions
Conclusions
Back-Up Slides
30-Year-Old Theoretical Debate
| Technology development (Reinganum, 1981; Ruiz-Aliseda and Zemsky, 2006) | Technology Adoption (Fudenberg and Tirole, 1985) |
Assumptions | | |
Equilibrium | | |
Outcomes | | |
► Acquired externally (from suppliers)
(*) Credible responses to observable actions
► Unique subgame perfect equilibrium
(*)
► Rent equalization
(1) Late joint adoption OR
(2) First mover but preemptive threats
► Costs fall over time due to supplier efficiency
► Developed internally (by firms)
► No rent equalization but first-mover advantages
N.B.: Δ2 cannot be too much lower than Δ1
► Costs fall over time due to TCD
► Unique subgame perfect equilibrium
(*)
► No imitation, spillovers, or capabilities
► Focus on imitation, spillovers, capabilities
Other Differences
Performance and Endogenous Spillovers
0
2
-1
1
0.75
0.25
0.5
Profits Leader − Profits Follower in Imitative
Profits Leader − Profits Follower in Concurrent
0
1
Profit Differential
Spillovers (s)
25
20
15
10
5
0
10
5
0
Cost
Present Value of Revenues
T* = ∞
Inimitability
Single Firm Development Time
Development Time
Revenues and Costs
Model Microfoundations
10
20
0
0
10
20
0
0
Speed Capabilities
Diminishing Returns
Costs
Development Time
Development Time
Costs