Research and Development

Chapter 22: Research and 1

Development
 Introduction
• Technical progress is the source of rising living
standards over time
• Introduces new concept of efficiency
– Static efficiency—traditional allocation of resources to
produce existing goods and services so as to maximize
surplus and minimize deadweight loss
– Dynamic efficiency—creation of new goods and
services to raise potential surplus over time

Chapter 22: Research and 2

Development
 Introduction 2
• Schumpeterian hypotheses (conflict between static
and dynamic efficiency)
– Concentrated industries do more research and
development of new goods and services, i.e., are more
dynamically efficient, than competitively structured
industries
– Large firms do more research & development than small
firms

Chapter 22: Research and 3

Development
 A Taxonomy of Innovations
Product versus Process Innovations
• Product Innovations refer to the creation of new goods and
new services, e.g., DVD’s, PDA’s, and cell phones
• Process Innovations refer to the development of new
technologies for producing goods or new ways of
delivering services, e.g., robotics and CAD/CAM
technology
• We mainly focus on process or cost-savings innovations
but the lines of distinction are blurred—a new product can
be the means of implementing a new process

Chapter 22: Research and 4

Development
 A Taxonomy of Innovations 2

Drastic versus Non-Drastic Innovations

• Process innovations have two further categories
• Drastic innovations have such great cost savings that they
permit the innovator to price as an unconstrained
monopolist
• Non-drastic innovations give the innovator a cost
adavantage but not unconstrained monopoly power

Chapter 22: Research and 5

Development
 Drastic versus Non-Drastic Innovations
• Suppose that demand is given by: P = 120 – Q and all
firms have constant marginal cost of c = $80
• Let one firm have innovation that lowers cost to cM = $20
• This is a Drastic innovation. Why?
– Marginal Revenue curve for monopolist is:
MR = 120 – 2Q
– If cM = $20, optimal monopoly output is:
QM = 70 and PM = $70
– Innovator can charge optimal monopoly price ($70)
and still undercut rivals whose unit cost is $80

Chapter 22: Research and 6

Development
 Drastic versus Non-Drastic Innovations 2
• Now consider the case if cost fell only to $60,
innovation is Non-drastic
– Marginal Revenue curve again is: MR = 120 = 2Q
– Optimal Monopoly output and price: QM = 30; PM = $90
– However, innovator cannot charge $90 because rivals
have unit cost of $80 and could under price it
– Innovator cannot act as an unconstrained monopolist
– Best innovator can do is to set price of $80 (or just
under) and supply all 40 units demanded.

Chapter 22: Research and 7

Development
 NonDrastic Innovation: Q M
< QC
DrasticDrastic
Innovation: Q M
> Q
vs. Non-DrasticsoInnovations
C
innovator cannot3charge
so innovator can charge monopoly price PM
because
Innovation is drastic
M if monopoly output Q at MR = new marginal
monopoly price P without M

c’ exceeds the competitive output Qrivals canmarginal

C at old undercutcost
thatc price
constraint
$/unit = p $/unit = p

PM
c c
PM c’

Demand Demand
c’
MR MR
QC Q M Quantity QM Q C Quantity

Chapter 22: Research and 8

Development
 Innovation and Market Structure
• Arrow’s (1962) analysis—
– Innovative activity likely to be too little because innovators
consider only monopoly profit that the innovation brings and not
the additional consumer surplus
– Monopoly provides less incentive to innovate that competitive
industry because of the Replacement Effect
• Assume demand is: P = 120 – Q; MC=
Initial $80. QisisYellow
Surplus initially 40.
Innovator lowers cost to $60 and can sell all 40 units at P = $80.
Triangle--Social Gains
• Profit Gain is $800–Less than Social Gain
$/unit from Innovation are Areas
120 A ($800) and B ($200)
But Innovator Only
80 Considers Profit Area A
60 A B ($800)

40 60 120 Quantity

Chapter 22: Research and 9

Development
 Innovation and Market Structure 2
• Now consider innovation when market structure is monopoly
– Initially, the monopolist produces where MC = MR = $80 at Q =
20 and P = $100, and earns profit (Area C) of $400
– Innovation allows monopolist to produce where MC = MR = $60
at Q = 30 and P = $90 and earn profit of $900
– But this is a gain of only $500 over initial profit due to
Replacement Effect—new profits destroy old profits
$/unit
120 Monopolist Initially Earns
100 C Profit C—With Innovation it
90 Earns Profit A—Net Profit
80
A Gain is Area A – Area C
60 Which is Less than the Gain
to a Competitive Firm
Demand
MR
20 30 60 120 Quantity

Chapter 22: Research and 10

Development
 Innovation and Market Structure 3
• Preserving Monopoly Profit--the Efficiency Effect
• Previous Result would be different if monopolist had to worry
about entrant using innovation
– Assume Cournot competition and that entrant can only enter if it
has lower cost, i.e., if it uses the innovation
– If Monopolist uses innovation, entrant cannot enter and
monopolist earns $900 in profit
– If Monopolist does not use innovation, entrant can enter as low-
cost firm in a duopoly
• Entrant earns profit of $711
• Incumbent earns profit of $44
– Gain from innovation now is no longer $900 - $400 = $500 but
$900 - $44 = $856
– Monopolist always has more to gain from innovation than does
entrant—this is the Efficiency Effect

Chapter 22: Research and 11

Development
 Competition and Innovation
• The incumbent/entrant model just discussed seems closer in
spirit to Schumpeter’s ideas than Arrow’s analysis.
• Dasgupta and Stiglitz (1980) come even closer by directly
embedding innovation in a model of Cournot competition
– Profit for each firm: i = P(Q) – c(xi)qi – xi
– Here, firm’s unit cost falls as the firm engages in R&D activity
– What is the equilibrium?
– Define x* as the optimal R&D level of each firm
– From Chapter 9, we know that
P  c ( x*) si

P 
–But with n symmetric firms si = 1/n, So we have
 1 
P 1    c  x * Output Condition
 n 
Chapter 22: Research and 12
Development
 Competition and Innovation 2
• How much should x* be?
– The usual marginal calculations apply. Every increase in
x costs $1. The benefit is the cost reduction this brings,
c(x)/x, times the number of units q* to which this cost
reduction will apply
dc xi  R&D Condition
 qi  1
dxi
– Both the Output Condition and the R&D Condition must
hold simultaneously in any equilibrium
– One obvious implication of the R&D Condition is that the
R& D effort of any one firm will fall as the number of n
firms increases because this will decrease the output of
each firm

Chapter 22: Research and 13

Development
 Competition and Innovation 3
• Making n endogenous means allowing firms to enter until
they no longer have an incentive to do so
• This will occur when firms earn zero profit after allowing
for R&D costs. Defining n* as the equilibrium number of
firms, the Output Condition then implies:
P
P  c x * 
n *
• Substitution into the R&D Condition then yields:
n* x* 1
  Industry R&D as Share of Sales
P  Q * Q * n *
• Industry R&D effort declines as n* rise, i.e., as industry
becomes less concentrated—fairly strong theoretical
support for Schumpeterian Hypothesis
Chapter 22: Research and 14
Development
 Competition and Innovation 4
• But empirical support for Schumpeterian view is mixed
– Need to control for science-based sectors (e.g., chemicals,
pharmaceuticals, and electronics) and non-technology based
sectors (e.g., restaurants and hair stylists)—R&D much more
likely in science-based sectors regardless of firm size
– Need also to distinguish between R&D expenditures and true
innovations. Common finding [e.g., Cohen and Klepper (1996)],
is that large firms do somewhat more R&D but achieve less real
innovative breakthroughs—e.g., Apple produced the first PC
– Market structure is endogenous. Innovations might create
industry giants (e.g., Alcoa) not the other way around.
• Bottom Line: Validity of Schumpeterian hypotheses
is still undetermined

Chapter 22: Research and 15

Development
 R&D Spillovers and Cooperative R&D
• Technological break-throughs by one firm often “spill
over’ to other firms
– Spillover is unlikely to be complete but likely to arise to
some extent
– We can model this in the Dasgupta Stiglitz world by now
writing a firm’s unit cost as a function of both its own and its
rival’s R&D
• c1 = c – x1 - x2
• c2 = c – x2 - x1
• To obtain solution, need also to assume that R&D is
subject to diminishing returns, e.g., r(x) = x2/2.

Chapter 22: Research and 16

Development
 R&D Spillovers and Cooperative R&D 2
• In this setting, response of firm 1’s R&D to firm 2’s
R&D depends on size of spillover term .
– When  is small, R&D expenditures are strategic substitutes
—the more firm 1 does the less firm 2 will do
– When  is large, R&D expenditures are strategic
complements—the more firm 1 does the more firm 2 will do
• However, determination of whether R&D efforts are
strategic substitutes or strategic complements is not
sufficient to determine what happens when there are
spillovers

Chapter 22: Research and 17

Development
 R&D Spillovers and Cooperative R&D 3
• Let Demand be given by: P = A – BQ
– Let ci = c – xi – xj;
– Each firm now chooses both production qi and research
intensity xi
• General Solution is: C C 2 A  c  2   
x1  x2 
9 B  2 2   1   
•To illustrate, consider two cases
– First case: Low Spillovers;  = 0.25
– Second case: High Spillovers;  = 0.75

Chapter 22: Research and 18

Development
 quilibrium is for both firms to
the high level of research
R&D Spillovers and Cooperative R&D 4
y (x = 10). Why? When degree
Pay-Off Matrix for  = 0.25
The that
overs  is small, firm know
can do R&D knowing that it
most of the benefits. Since this Firm 1
advantage the rival, each firm
avoid being left behind by doing
R&D itself. Low Research High Research
Intensity Intensity
Low Research
Intensity $107.31, $107.31 $100.54, $110.50

Firm 2
High Research $110.50, $100.54 $103.13, $103.13
Intensity

Chapter 22: Research and 19

Development
ash Equilibrium is for both firms to
oose the low level of research
ensity (x = R&D Spillovers
7.5). Why? When degreeand Cooperative R&D 5
spillovers  is large, a firm knows
The Pay-Off Matrix for  = 0.75
at it will benefit from technical
vance of its rival even if it doesn’t do
y R&D itself. So, each firm tries to Firm 1
e-ride off its rival and each does little
&D itself. Low Research High Research
Intensity Intensity
Low Research
Intensity $128.67, $128.671, $136.13, $125.78

Firm 2
High Research $125.78, $136.13 $133.68, $133.68
Intensity

Chapter 22: Research and 20

Development
 R&D Spillovers and Cooperative R&D 6
• MORAL of the foregoing analysis is that the Outcome of
non-cooperative R&D spending depends critically on the
extent of spillovers.
• What if R&D spending is cooperative?
• R&D cooperation can take two forms:
– 1. Do R&D independently but choose x1 and x2 jointly to
maximize combined profits, given competition in product
market is maintained.
– 2. Do R&D together as one firm, e.g, form a Research Joint
Venture. That is, effectively operate as though the degree of
spillovers is  = 1, again though, continue to maintain
product market competition.
• The two types have very different implications.

Chapter 22: Research and 21

Development
 R&D Spillovers and Cooperative R&D 7
• Consider first the case of coordinated but not
centralized R&D using our generalized demand and
cost equations
– Total R&D spending now rises unambiguously as 
increases.
– To see this note that given our earlier demand and cost
assumptions, and given the fact that x1 and x2 are chosen to
maximize joint profits, the optimal values for x1 and x2 are:
2 A  c 1   
x1  x2 
9  21   
2

Chapter 22: Research and 22

Development
 R&D Spillovers and Cooperative R&D 8
• This solution for the profit-maximizing research level
under cooperation is unambiguously increasing in  but
this is a good news/bad news story.
• The good news is that for the high spillover case ( >1),
the free- riding problem is no longer an issue and firms
now do more R&D
• The bad news is that for the low spillover case ( < 1),
there is no longer a fear of being left behind by one’s rival.
So in this case firms do less R&D which means costs (and
consumer prices) are higher than without cooperation.

Chapter 22: Research and 23

Development
 R&D Spillovers and Cooperative R&D 9
• What about a Research Joint Venture?
– As noted, this effectively changes  to 1.
– For our demand and cost equations, it can be shown that:
4 A  c 
x1  x2 
9B  8
– This is clearly more R&D than occurred with simple coordination
for any given value of 
– As a result, it leads to lower costs and more output to the benefit
of consumers 3 A  c 
q1  q2 
9B  8
– Profits are also higher. Thus, in the presence of spillovers,
Research Joint Ventures are unambiguously beneficial.

– The only trick is to make sure that cooperation is limited to

research and does not extend to other dimensions of competition
Chapter 22: Research and 24
Development
 Empirical Application: International
Spillovers in R&D
• Perhaps we can get evidence on the extent of R&D
spillovers by looking at technology spillovers between the
same industry across different countries
• This is the strategy employed by Keller (2002)
• Consider output in industry i and country c
– ln Qci = (1 – )n Kci + ln Lci
– Or: Total Factor Productivity in country c and industry i is:
– TFPci = ln Qci – (1 – )n Kci + ln Lci

Chapter 22: Research and 25

Development
 Empirical Application: International
Spillovers in R&D 2
• Using bars to denote average industry i levels across the entire
sample, Keller measures the relative factor productivity in
industry i country c then as follows:

 
Fci  ln Qci  ln Qi  1   ci   ln K ci  ln K i    ci ln Lci  ln Li 
• This equation measures relative productivity at a point in time.
• Keller measures Fci for each of 12 industries in 14 countries
over the years 1970 to 1995.
• He thus has productivity observations that vary across space and

time Fcit
• He seeks to explain the observed changes in productivity in each
industry/contribution over time on the basis of domestic and
Chapter 22: Research and 26
foreign R&D Development
 Empirical Application: International
Spillovers in R&D 3
• Keller divides his 14 countries into two groups
– Engines of technical change: France, Germany, Japan, UK and US
– Nine Others: Australia, Canada, Denmark, Finland, Italy, the
Netherlands, Norway, Spain, and Sweden
• He asks how relative factor productivity in industries in
each of the nine depends on their own R&D and the R&D in
the 5 engines of technical change countries, the G5
• In particular, Keller supposes that the relevant technical
base for any of the 12 industries in any of the nine countries
is a function of its own R&D and R&D in the G5

Chapter 22: Research and 27

Development
 Empirical Application: International
Spillovers in R&D 4
• Let Scit be the amount of R&D done up to year t in country c and industry i.
• Keller says that this contributes to a nation’s relative factor productivity
according to the following equation

  Dcg 
 ln  S cit     S git e 
  gG 5 
• Here  measures the productivity impact of (the log) of all
R&D relevant to the domestic industry’s relative productivity
• That relevant R&D is comprised of two parts
• Domestic R&D Scit contributes directly to productivity

• R&D in the G5 contributes less directly if  is less than one and

the effect is weaker for countries farther away as measured by
distance D if  is positive
Chapter 22: Research and 28
Development
 Empirical Application: International
•
Spillovers in R&D 5
Keller (2002) uses nonlinear estimation techniques to estimate the
parameters , , and  jointly after controlling for country-specific and
time-specific effects. His basic results are
Coefficient Estimate Std. Error
 0.078 (0.013)
 1.005 (0.239)
 0.843 (0.059)
• All these effects are statistically significant.
– Own country R&D raises productivity by roughly 7.8 percent if there is
no G5 R&D
– G5 R&D is 84.3 percent of domestic R&D but the 1.005 value for 
implies that his effect declines rapidly as the G5 country is farther away
and disappears entirely once that distance reaches 400 miles
• However, once Keller allows for differential impacts depending on
which G5 country does the R&D the spillover lasts up to 1200 miles.
Such extensive spillovers between nations suggests that R&D
spillovers between firms are probably significant

Chapter 22: Research and 29

Development