Competition and Innovation: An Inverted U Relationship

advertisement
Competition and Innovation:
An Inverted U Relationship
Presentation to HM Treasury May 2004
Philippe Aghion (Harvard & UCL)
Nick Bloom (CEP, LSE)
Richard Blundell (IFS & UCL)
Rachel Griffith (IFS & UCL)
Peter Howitt (Brown)
Contact Details:
n.bloom@lse.ac.uk; 020 7955 7286
1
Evidence from conventional wisdom, theory and empirics on
the impact of competition appears contradictory
•
Conventional wisdom provides mixed views
“Schumpeterian effect” – the carrot:
“....anti-trust discourages innovation.......” (Bill Gates and lawyers, frequently)
“Competition effect” – the stick:
“... 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 production...” (Nickell, 1996)
•
Economic theory often supports the Schumpeterian effect
•
But the empirical evidence in mixed
– Early studies in 1940s/50s found a negative “Schumpeterian” relationship
– Scherer (1965) found a mixed relationship with an inverted-U form
– Recent work focusing on linear effects find a positive relationship:
Geroski (1995), Nickell (1996) and Blundell, Griffith and Van Reenen (1999)
2
Empirical results
Model
Policy implications
3
Basic Relationship Appears to be an Inverted-U
•
Using UK data (details to come) we estimate a flexible Kernel estimator (a local
‘moving average’) and find an inverted-U in the raw data
•
Firms are quite evenly distributed across the up, flat and down parts of the curve, so
linear estimation would
beregression,
misleadingbw = .025, k = 6
Kernel
x=(1-Lerner), y=CW Patents
Peak occurs around
high
10
0.85 – i.e. 15%
price cost mark-up
8
Patents
6
4
2
low
.83
low
.85
.87
.91
.89
Grid points
.93
Kernel with Gaussian weights
Competition
.95
high
4
Measure innovation using patents from US PTO
•
UK firm level accounting data 1968-94 cover all stock market firms
•
Measure innovation by matching to US Patent Office data
– Choose US data as easily available electronically from 1968-1999 and PTO
where all major international inventions are patented
– Hand matched patents to UK firms so includes overseas subsidiaries
– Weighted patents by citations received to measure innovation “quality”
– Also confirm results using an alternative R&D measure of innovation
•
Final Data set has
– 330 firms, 4500 observations, 60,000 patents, 200,000 citations
•
Bloom and Van Reenen (2001, EJ) use this data set and demonstrated patents
play a powerful role determining market value and productivity of UK firms
5
Measure Competition Using Lerner “Profit Index”
•
Traditional measures based on market share
– But problem defining the product & location market, particularly in open economies
– For the UK international markets important – i.e. Glaxo has 10% UK NHS market
but 70% “market” defined by sales of UK listed firms
•
So we use the Lerner Index and assume MC
AC

P  MC
P  Q  TC
P  AC


P
PQ
P
•
Mean of value of the Lerner is around 0.1, minimum is 0.21 and maximum is 0.02
6
Confirm this inverted-U shape is robust by re-estimating
quadratic model with industry and time controls
•
Kernel moving average data descriptive but can not control for different industry
and time effects
•
So use an exponential `Poisson style’ model because patent counts are skewed
with many zeros, taking a quadratic form and control for industry and time
effects
•
Again find strong inverted U-relationship which is significantly different from a
linear specification
7
Exponential quadratic with year and industry dummies
Peak occurs around
0.88 – i.e. 12%
price cost mark-up,
and in middle of the
data
high
Each point is an industry year
observation, indicating most
industries are clustered around the
peak innovation level
Patents
Estimating a straight
line predicts a
positive impact in
line with Nickell
Outer feint lines provide
the point-wise 95%
confidence interval
low
low
high
Competition
8
Finally need to check cause and effect
•
Competition drives innovation but the reverse is also true, typically higher
innovation acts to reduce competition, i.e.
• Innovative firms obtain patents and reduce competition
•
Firstly, we include time and industry dummies to remove much of the spurious
correlation between competition and innovation
•
Secondly, we instrument changes in competition using the large number of
competition changes that have occurred in the UK since 1970:
• The 1992 EU single market program
• Thatcher and Major era privatisations
• Remedies imposed on industries by MMC
•
Policy instruments have a significant negative impact on our Lerner measure
9
After identifying causal impact still find inverted-U
shape, but shifted rightwards as would expect
Controlling for cause
and effect shifts the
peak to around 0.95 –
i.e. 5% price cost
mark-up, and at the
upper end of the data
high
Patents
low
low
high
Competition
10
Empirical results
Model
Policy implications
11
In the paper we model competition and Schumpeterian effects,
which together generate an inverted-U shape
•
In our model:
– 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
– Overall obtain a balanced inverted U-shape
high
Innovation
low
low
high
Competition
12
There are other potential explanations for the inverted-U
•
For example, managerial effort versus cash-flow
– At low level of competition the principal-agent problem of motivating
managers is paramount so higher competition helps innovation
– At high level of competition managers work harder but as competition
increases further firms lack finance for innovation
high
Positive
management effort
effects dominate
Negative cash
flow effects
dominate
Innovation
low
low
Competition
high
13
Empirical results
Model
Policy implications
14
Competition has other positive welfare effects
(2) Competition may also improves
productivity by forcing more rapid
adoption of innovations – the
“McKinsey” hypothesis
Welfare
high
(3) Higher
competition
also reduces
monopoly
pricing
deadweight
loss, directly
improving
consumer
welfare
(1) Baseline
relationship
between
competition
and innovation
low
low
high
Competition
15
So overall higher Competition likely to be positive,
particularly in low-competition industries
•
In industries with low or moderate levels of competition increasing
competition is unambiguously good
•
U-shape suggests greatest gain from focusing on industries with lowest
levels of competition
•
At higher levels of competition policy less obvious
– results suggests reducing competition may promote innovation
– But competition has other positive effects and so probably always
beneficial
•
Low or moderate competition could indicated by high average industry
price-cost mark-ups
16
Download