Competition and Innovation:
An Inverted-U Relationship
Philippe Aghion (Harvard & UCL)
Nick Bloom (CEP, LSE)
Richard Blundell (IFS & UCL)
Rachel Griffith (IFS & UCL)
Peter Howitt (Brown)
The fact
• The theories of industrial organization
typically predict the negative relationshiip
between competition and innovation
• a positive effect – i.e. Geroski (1995 OUP),
Nickell (1996 JPE) and Blundell, Griffith and
Van Reenen (1999 RES), Mohen and ten Raa
(2002, WP)
The two effects
• “Competition effect”:
– “... from Adam Smith to Richard Caves: the
belief that competition is good, rests on the idea
that competition exerts downward pressure on
costs, reduces slack and provides incentives for
efficient organisation of production...” (Nickell,
1996 JPE)
• “Schumpeterian effect”:
The framework of this paper
• The new fact : an inverted-u relationship
use a panel data and find a robust inverted U-shape between
competition and innovation (patenting)
• Combining agency models with Schumpeterian models
– At low levels of competition the “competition effect”
dominates leading to a positive relationship
– At high levels of competition the “Schumpeterian effect”
dominates leading to a negative effect
The structure
• The empirics research
• The model
• Concludes
The data
• UK firm level accounting data , the firms list
on the London Stock Exchang.
• Our sample includes all firms with names
beginning A to L plus all large R&D firms
• An unbalanced panel of 311 firms spanning
seventeen industries over the period 19731994.
Measure of innovation
• Average number of annual patents taken out by
firms in an industry.
• Weighted patents by citations received to
measure innovation “quality”
Measure of Product Market Competition
• Traditional measures based on market share
– But problem defining the product & location
market.
– For the UK international markets important – i.e.
Glaxo has 7% global market but 70% share of UK
market as defined by sales of UK listed firms
• So we use the Lerner Index
Our competition measure is the average of this
across firms within the industry
• A value of 1 indicates perfect competition,
while values below 1 indicates some degree of
market power.
• The entire sample of Stock Market Listed
firms in each industry.
Passion regression(p527-p531)
• 因变量是计数变量（count varible），非负整数
值。确保y的预测值也总是正数，将其期望值模型
化为一个指数函数。
E ( y | x1 , x2 ,..., xk )  exp G( x1 ,..., xk )
P( y  h | x)  exp[  exp( x )][exp( x )] / h!
h
• MLE(QMLE):不管泊松分布成立与否，仍然可以得
到待估系数的一致和渐进正态的估计量
• Estimate the key moment condition E[P|C] =
exp(g(C))
– P is the patent count, C the competition measure, and g(.) a
flexible function
– We use an exponential `Poisson style’ model
– g(.) is non-parametrically approximated using a quadratic
spline-function (see Ai and Chen, 2002 Econometrica)
• To allow for industry and time variables effects these
are parametrically included in addition to yield a final
estimating equation E[Pit|Cit] = exp(g(Cit) + Xit’b)
Endogeneity
• PMC may be endogenous as higher patenting firms
may gain higher rents
• Firstly, we include time and industry dummies
– changes in competition identify changes in
patenting
Secondly, we instrument changes in competition using
the large number of competition changes that have
occurred in the UK since 1970:
– Differential changes in competition across industries
following the 1992 EU single market program
– Changes in competition following major privatizations
– Change in completion following structural and behavioural
remedies imposed on industries after a Monopolies and
Mergers Commission
Results
The inverted-U shape is robust
• Five –year average
• R&D expenditure
• Each of top four innovating industries
Assumptions
• Sturcture assumption:
• The economy contains many industries, with (for
simplicity) two firms, which are either:
– “neck-and-neck” as firms have the same technology
– “leader-follower” as firms have different technologies
Technonlogy is step-by-step
Innovation depends on difference between postinnovation
and preinnovation rents for incumbent firms
The model
• A logarithmic instantaneous utility function
• A continuum of intermediate sectors
• Doupolists in sector j
• Max xAj + xBj st pAj*xAj +pBj*xBj =1
• Each firm produces using labor as only input, a
constant –returns production function, and take the
wage rate as given.
• The unit costs of production cA and cB of the two
firms in an industry are independent of the quantities
produced.
• let k denote the technology level of duopoly firm i in
some industry j . One units labor generates an output
flow equal to
• The state of an industry is then fully characterized by
a pair of integers (l,m)
• For simplicity,we assume knowledge spillovers between
leader and follower in any intermediate industry are such
that the maximun sustainable gap is m=1.
• Two kinds of intermediate sectors in the economy:
– Leveled or neck-and-neck sectors, m=0.
– Unleveled sectors, m=1.
R&D cost is
in units of labor ,with a passion
hazard rate n. ----innovation rate or R&D intensity
• Leader firm moves one technological step ahead with a
hazard rate n
• a follower firm can move one step ahead with hazard rate
h, even if it spends nothing on R&D ,by copy the leader’s
technology . Then,a follower firm moves ahead with a
hazard rete n+h
• They do not collude when the industry is unlevel.
• Each firm in a level industry earns a profit of 0 if the
firms are unable to collude. In Bertrand competition,
each farm has maximun profit is
.
•
is also the incremental profit of an innovator in a
neck-and-neck industry, also indicates PMC
Escape competition
• Under low competition “neck-and-neck” firms earn
moderate profits, yielding little gain from innovation,
so
– “neck-and-neck” firms undertake little innovation
– leading to an equilibrium with mainly “neck-andneck” industries
– so increasing competition raises innovation as
“neck-and-neck” firms increase innovation
Schumpeterian effect
• Under high competition “neck-and-neck” profits are
low, so the rewards to innovating to become a leader
are high, so:
– “neck-and-neck” firms undertake a lot of
innovation
– leading to an equilibrium with mainly “leaderfollower” industries
– so further increases in competition lower the
profits for followers to innovate and become
“neck-and-neck”, reducing innovation
Schumpeterian effect vs escape-competition effect
• On average, an increase in product market
competition will thus have an ambiguous effect on
growth.
• The overall effect on growth will thus depend on the
(steady-state) fraction of leveled versus unleveled
sectors.
• But this steady-state fraction is itself endogenous,
since it depends upon equilibrium R&D intensities in
both types of sectors.
Conclusions
• the competitioninnovation relationship takes the form
of an inverted-U shape.This result is robust.
• Extend the current theoretical literature on step-bystep innovation to produce a model that delivers an
inverted-U prediction.
• the equilibrium degree of technological neck-andneckness among firms should decrease with PMC
• the higher the average degree of neck-and-neckness
in an industry, the steeper the inverted-U relationship
between PMC and innovation
• All innovations equal in the model
• Lerner index ? More profit doesnot mean
less competitive.
• Which is the cause? Innavation or
compitition ? What determints the initial
conditions
Some puzzles
• Why the hazard rate is difference between
a laggard frim and a firm in neck-and-neck
industry? h?
• FOC