Content uploaded by Jeroen Hinloopen
Author content
All content in this area was uploaded by Jeroen Hinloopen on Aug 22, 2014
Content may be subject to copyright.
WORKING PAPER SERIES
Cooperation in the Classroom:
Experimenting with Research Joint Ventures
Michelle S. Goeree
Jeroen Hinloopen
WP 2004-08
500 East 9th Street
Claremont, CA 91711
Phone: (909) 607-3203
Fax: (909) 621-8249
Cooperation in the Classroom:
Experimenting with Research Joint Ventures
Michelle S. Goeree and Jeroen Hinloopen 1
December, 2004
Abstract
This paper describes a classroom exercise that illustrates the investment incen-
tives facing firms when technological spillovers are present. The game involves
two stages in which student “sellers” first make investment decisions then pro-
duction decisions. The classroom game can be used to motivate discussions of
research joint ventures, the free-rider problem, collusion, and antitrust policy
regarding research and development.
JEL Classification:
Keywords: classroom games, research and development, research joint ventures,
technological spillovers
1Goeree is at Claremont McKenna College (email: msgoeree@mckenna.edu) and
Hinloopen is at the University of Amsterdam (email: j.hinloopen@uva.nl). Address
for correspondence: 500 E. Ninth Street, Claremont, California, USA. We wish to
thank Jacob Goeree for helpful comments.
“Innovation is an activity in which “dry holes” and
“blind alleys” are the rule, not the exception.”
Jorde and Teece (1990)
1Introduction
In 2002, firms in the U.S. spent over $300 billion on research and development
(R&D)— nearly 3% of GDP.1The U.S. is far from unique in this dimension,
in fact many countries spend substantial amounts on R&D each year. Un-
derstanding firms’ incentives to invest and the role for governmental policy is
an important part of any course in Industrial Organization, and is discussed in
many applied courses including regulation, antitrust, game theory, and law and
economics.
Investment in R&D is distinct from capital investments along two dimen-
sions. First, there is a public good aspect associated with R&D investments.
That is, the use of an innovation does not diminish its availability to other firms.
This suggests it may be welfare improving to make the innovation available to
other firms. In addition, the understanding obtained from investment spills
over freely to others. These technological “spillovers” reduce firms’ incentives
to invest since rivals may benefit from their costly innovation.2
The literature has proposed several solutions to remedy the public good
and externality problems associated with R&D. One proposed solution is to
allow cooperation in investment decisions, research joint ventures (RJV).3The
coordination of R&D decisions improves incentives to invest when spillovers are
high. However, RJVs may increase the likelihood of coordination in the final
goods market by member firms.
In the 1980’s governmental authorities responded to these theoretical find-
1Source: www.nsf.gov/sbe/srs/
2For more on this topic see Spence (1984), d’Aspremont and Jacquemin (1988),
Kamien et al.(1992), and Hinloopen (1997).
3According to the National Cooperative Research Act, a RJV is defined as “any group
of activities, including attempting to make, making or performing a contract, by two or
more persons for the purposes of — (a) theoretical analysis, experimentation, or systematic
study of phenomena or observable facts, (b) the development or testing of basic engineering
techniques, (c) the extension of investigative finding or theory of a scientific or technical nature
into practical application for experimental and demonstration purposes..., (d) the collection,
exchange, and analysis of research information, or (e) any combination of the [above].”
2
ings by relaxing antitrust and competition policy as it pertains to cooperation in
R&D. Currently, RJVs are permitted in Europe, the US, and Japan. However,
the US legal system makes a distinction between joint R&D decisions and joint
production decisions, where the latter is illegal (Jorde and Teece, 1990).
In this paper we present a classroom game that illustrates the investment
incentives facing firms, particularly in markets where there are high technolog-
ical spillovers. It involves a two-stage game in which student “sellers” make
investment decisions in the first stage and quantity decisions in the second stage.
R&D investments reduce the seller’s second stage costs of production.
We present various treatments, which can be implemented individually or
together, depending upon classroom time constraints. The treatments demon-
strate the free-rider problem in investment decisions when technological spillovers
are present, how RJVs can be used to alleviate incentive problems, and the
temptation to coordinate on final good production. The observed decisions
in this experiment can be used to stimulate discussions of spillovers, research
joint ventures, tacit (or explicit) collusion, and competition policy regarding
investment in innovations.
2Overview
The experiment is a two-stage game in which sellers choose quantity to invest
and quantity to produce. There are three variations of this classroom exper-
iment and the decisions sheets for each treatment are provided in Appendix
B. Each of the treatments can be conducted alone or in combination with any
other treatment, depending upon classroom time constraints and the economic
concepts you wish to emphasize.
In the baseline treatment (treatment 1) cooperation among sellers in R&D
investment decisions is not permitted and a seller’s own investment decision
determines her costs of production (ie. there are no investment spillovers).
Treatments 2 through 4 are best used in combination with the baseline treat-
ment and can be divided into two groups: those which permit coordination in
R&D decisions, and those which do not. For a summary see Table 1 below.
3
Cooperation in R&D
Not Permitted Permitted
Investment Not present Baseline Treatment 1 Treatment 2
Spillovers Present Treatment 3 Treatment 4
Table 1: Treatment Summary
In the second treatment sellers are permitted to discuss investment decisions
(but not final goods production). However, as in the baseline treatment, the
seller’s investment decision alone determines her costs. The second treatment
illustrates the effects of research joint ventures when investment spillovers are
not present. You may observe (“illegal”) collusion in quantity decisions aris-
ing from the contact sellers make during R&D decisions, which provides an
opportunity to discuss anti-trust policy with regard to collusion and RJVs.
In the third treatment sellers are not permitted to cooperate in R&D de-
cisions, but a seller’s cost is determined by the total amount of investment
undergone by both sellers (i.e. there are full investment spillovers). In this way
Treatment 3 illustrates the free-rider problem in investment decisions arising
from investment spillovers.
The final treatment allows for cooperation in R&D investments under full
spillovers. This treatment illustrates both the benefits of RJVs in that they
alleviate the free-rider problem and the drawbacks of RJVs in that they provide
the potential to (illegally) collude in quantity decisions which arises from an
occasion to discuss strategies with the competition while in the “research lab”.
While each of the treatments can be conducted individually, it is most in-
formative to discuss results relative to the baseline treatment case. Since the
time necessary to perform all treatments exceeds a typical classroom allotment
we provide a summary of some of the possible treatment combinations, related
economic concepts, and time allocation (which includes time for discussion) rec-
ommendations in Table 2 below.
4
Treatments Economic Concepts Time
1 two-stage game 30 minutes
1 and 3 spillovers
free-rider problem
inefficiency in R&D investment
treatment 1 - 20 minutes
treatment 3 - 30 minutes
1 and 2 RJVs
collusion treatment 1- 20 minutes
treatment 2 - 30 minutes
spillovers
efficiency of RJVs
collusion
treatment 1 - 20 minutes
treatment 4 - 40 minutes
1 and 4
Table 2 : Treatment Recommendations
3 Procedures
In preparation for the experiment you will need to make copies of the instruc-
tions, copies of the decision sheets in the appendices, and set up a record table
on an overhead transparency. The record table should contain one column for
the investment decision and one for the quantity decision with enough rows for
about five periods. In each cell you will record the decision made by each seller
in each period. You will need two assistants, one to help gather decision sheets
and the other to help record results. If you plan to conduct treatments 3 or 4
you should make copies of the table in Appendix C, which gives cost reductions
for large total R&D investments.
Prior to beginning the experiment ask two or three students to assist with
the experiment. Ask the assistants to distribute the instructions, and, while
they are doing so, count the students and determine how many markets and
sellers you wish to have. We found that three markets with two sellers in each
works well, for a total of six sellers. Rather than increasing the number of
5
sellers for larger classes, we’ve found that allowing students to work in pairs,
where two (or three) students represent one seller, decreases the amount of time
necessary to conduct the experiment and enhances understanding. Number the
decision sheets accordingly, with a different number for each seller. To prevent
sellers in the same market from interacting, the groups of sellers should be
seated in separate areas of the room. For example, for an experiment with six
sellers, arrange three on one side of the room (say sellers numbers 1-3) and the
remaining (numbers 4-6) on the other. Once the sellers are arranged, distribute
the decision sheets and read the instructions outloud. The students may have
questions at this point, but try to keep discussion to a minimum.
After the first round of decisions, ask an assistant to collect the decision
sheets. At this time the sellers in each group are randomly assigned to a
market. To assign sellers randomly to a market for an experiment involving 3
markets with 2 sellers, for example. Shuffle the decision sheets for the first three
sellers then randomly choose one of the sheets — the chosen seller is assigned to
the first market, the second seller drawn is assigned to the second market, etc.
Shuffle the decision sheets for the other three sellers (numbers 4-6) and assign
markets analogously.
4 Discussion of Results
Table 3 gives a summary of results from a number of classroom experiments
conducted in Industrial Organization courses at the University of Amsterdam4.
Comparing treatment 3 to treatment 1, we see that when spillovers are present
but sellers are not permitted to cooperate in R&D decisions, R&D investment
levels fall relative to the benchmark treatment. You can begin a discussion in-
volving these treatments by asking students if they choose different investment
levels in the different treatments and why. They may not immediately realize
that spillovers in investment decisions result in a free-rider problem, but a way
to reach this understanding is to ask some leading questions. For instance, you
4An excel file with which the Nash equilibrium values can be calculated for the various
treatments is available upon request.
6
could ask those students who invested less in the final periods why they reduced
their investment levels. Typically the students were either responding to low
investment levels of their rival (due to the rival free-riding) or trying to free-
ride offthe rivals investments. The students should reach this understanding
quite naturally, and if they have had some game theory courses may relate the
situation to that of the prisoners in the prisoners dilemma game.
R&D Investment Production
Average Outcome Nash Average Outcome Nash
Treatment 1 13.30 18 6.25 6
No cooperation, no spillovers
Treatment 2 11.56 4 5.84 6
Cooperation, no spillovers
Treatment 3 6.91 2 5.84 6
No cooperation, spillovers
Treatment 4 18.42 9 5.58 6
Cooperation, spillovers
Table 3 : Results from Classroom Experiment
Students may be interested to discuss if this is really an issue in R&D, that is,
are there spillovers in reality and how would they arise? This type of discussion
leads naturally into a lecture on R&D incentives. You can then ask students
if they could think of ways to alleviate this problem. Is there a role for the
government? Should collusion in R&D be legal under certain circumstances?
The classroom discussion can then progress to the benefits of RJVs and the
prevalence of them in various industries such as the semiconductor industry
and the mining and manufacturing industry.5
Treatment 4 allows sellers to form RJVs, as can be seen from Table 3, in-
vestment decisions increase since the free-rider problem is eliminated. However,
these levels are lower than those with no spillovers since each seller can benefit
5See Link (1996) for an overview research joint venture filings.
7
from the rival’s investment decisions. Discussion can again begin by asking
sellers if they choose different investment levels in treatment 4 and why. You
could also focus on a seller who invested less than in the baseline treatment and
ask why they choose to do so. Since cooperation is permitted each seller can
reduce her own investment and benefit from the others investment decision. If
you had the time to conduct treatment 3 prior to 4, then the benefits of RJVs
will be evident. This is illustrated in Figure 1 where the average per-period
R&D investment levels are shown for treatments 1, 3 and 4 as they emerged
during experiments conducted during a summer school of the Economics Net-
work for COmpetition and REgulation (ENCORE)6.
R&D investment levels
0
2
4
6
8
10
12
12345
period
Non-cooperative R&D, no spillovers Non-cooperative R&D, full spillovers Cooperative R&D, full spillovers
Figure 1: R&D Results from ENCORE summer school
But even if there was no time to conduct treatment 3, you can ask students if
they would have invested the same amount if they didn’t know what their rival
was doing. If the free-rider problem isn’t evident, you could single out a student
and ask her what she would do if her rival invested 25 and why. Regardless of
her answer, you can ask the class if anyone would have made a different decision.
6Participants in the ENCORE summer school were mostly government officials of the Dutch
antitrust authority or the Dutch Ministry of Economic Affairs.
8
Once you have finished discussion surrounding the benefits of RJVs, you can
discuss potential drawbacks. You may have observed (“illegal”) cooperation
in production levels. Students will typically respond honestly to the question,
“were any of you deciding jointly how much to produce?” You may consider
telling the students before you run the final round of the experiment. If output
is set jointly then some students may “defect” by increasing their final round
output level. We observed final round defection in the form of increased out-
puts in the majority of the experiments we conducted. Figure 2 contains the
average per-period output decisions corresponding to the results in Figure 1.
It suggests indeed that in treatment 4 in some markets production levels were
jointly set.
Production levels
4
5
6
7
8
12345
period
Non-cooperative R&D, no spillovers Non-cooperative R&D, full spillovers Cooperative R&D, full spillovers
Figure 2: Output results from ENCORE summer school.
These results can motivate discussion of anti-trust policy and the fine line be-
tween encouraging investment while prohibiting collusion on output. To illus-
trate this trade-offyou can discuss total welfare results under different treat-
ments, as in Table 4.
9
Consumer Surplus Total Surplus
Average Outcome Nash Average Outcome Nash
Treatment 1 78.13 82.65 128.36 128.57
No cooperation, no spillovers
Treatment 2 68.30 63.28 123.63 119.53
Cooperation, no spillovers
Treatment 3 68.30 63.28 136.64 123.05
No cooperation, spillovers
Treatment 4 62.35 82.65 136.09 146.94
Cooperation, spillovers
Control setting: cooperative output 50.00 125.00
Cooperation, spillovers
Table 4: Results from a classroom experiment.
5FurtherReading
This paper describes an experimental game designed for use in undergraduate
classrooms. The experiment can be used to teach concepts related to research
joint ventures, spillovers, collusion, and government antitrust policy regarding
investment in innovation. Our classroom paper is in the spirit of many others
used as an aid in teaching topics common to industrial organization such as
predatory pricing behavior, asymmetric information, rent seeking behavior, and
auctions.7Wells (1991) and Williams and Walker (1993) provide surveys of the
classroom experimental literature.
In addition to classroom experiments, numerous laboratory experiments ex-
plore issues related to industrial organization. Suetens (2004) investigates
whether R&D cooperation leads to product market collusion for different levels
of spillovers. She finds that the degree of price collusion in the final goods
market is higher when research joint ventures are permitted. For a survey of
7See, for instance, Bergstrom and Miller (1997), Goeree and Holt (1999), Holt and Sherman
(1999), and Capra et al. (2000) as well as the ongoing Classroom Games column in the Journal
of Economic Perspectives edited by Holt.
10
the experimental industrial organization literature see Holt (1995) and Chapter
7 of Davis and Holt (1993).
11
A Instructions
In this experiment there are many independent markets in which the same good
is exchanged. Each of you is a seller in one of the markets for a series of periods.
You will be randomly paired with another seller, so in each market there will be
two sellers. Each of you has 10 widgets to sell. For each widget that you sell
you will incur a cost of 25. In each period, you can decide to make an investment
which will lower your costs for that period. Every unit investment results in
a cost reduction of the square root of the number of units invested. So if you
decide to invest 4 units your production costs will be reduced by √4=2.Each
unit you invest costs 1. You can invest at most 25 units and your investment
decision must be positive and an integer (i.e. you cannot invest half a unit).
The following table illustrates the cost reduction (and resulting cost incurred
from producing one widget) from each investment level.
invested reduction cost invested reduction cost invested reduction cost
1 1.0 24.0 10 3.2 21.8 19 4.4 20.6
2 1.4 23.6 11 3.3 21.7 20 4.5 20.5
3 1.7 23.3 12 3.5 21.5 21 4.6 20.4
4 2.0 23.0 13 3.6 21.4 22 4.7 20.3
5 2.2 22.8 14 3.7 21.3 23 4.8 20.2
6 2.4 22.6 15 3.9 21.1 24 4.9 20.1
7 2.6 22.4 16 4.0 21.0 25 5.0 20.0
8 2.8 22.2 17 4.1 20.9
9 3.0 22.0 18 4.2 20.8
The costs you pay per widget sold are determined in the following manner:
cost =25−cost reduction =25−√units invested
Once you have made your investment decision, write the number of units you
wish to invest on your decision sheet, in the appropriate column for the current
period. After all sellers have made their investment decisions we will collect
your decision sheets and write the investment decisions on the blackboard.
When all of the decisions have been recorded on the blackboard we will
return the decision sheets to you. You will then be asked to choose the number
of widgets you wish to sell. All widgets that you sell will be sold at the same
price. You can sell at most 10 widgets and the number of widgets offered must
be positive and an integer. You must write the number of widgets you selected
on your seller decision sheet, in the appropriate column for the current period.
After all sellers have chosen the number of widgets to sell, the decision sheets
will be collected and the quantity decisions will be written on the blackboard.
When all of the production decisions have been recorded on the black-
board we will return the decision sheets to you. The price you earn for each
widget sold is 40 minus the total number of widgets sold by all sellers in your
market. So if you decide to sell 4 widgets and the other seller in your group
decides to sell 3 widgets then the sales price is: sales price=40−4−3=33.
When all sellers have made their quantity decisions the trading period ends and
you can calculate your earnings for the period. Your earnings are determined
in the following manner:
earnings = (sales price ×quantity sold) - (unit cost ×quanity sold) - units invested
You can use the attached table to calculate your earnings. Any questions? We
will begin by having each seller choose an investment level, which you should
record on your decision sheet.
BDecisionSheets
Seller Number : __________________
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
1) units you invest (max 25)
2) cost reduction (from table)
3) cost per widget (from table)
4) quantity widgets you offer (max 10)
5) total widgets offered in your market
6) price = 40 - (5)
7) price x quantity = (6) x (4)
8) cost x quantity = (3) x (4)
9) profit = (7) - (8) - (1)
10) cummulative profit
Decision Sheets for Treatments without Investment Spillovers (1 and 3)
Seller number: _____________
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
1) units you invested (max 25)
2) total units invested in market
3) cost reduction (from table)
4) cost per widget (from table)
5) quantity widgets you offer
6) total widgets offered in your market
7) price = 40 - (6)
8) price x quantity = (7) x(5)
9) unit cost x quantity = (4) x (5)
10) profit: (8) - (9) - (1)
11) cummulative profit
Decision Sheets for Treatments with Investment Spillovers (2 and 4)
14
C Large R&D investments
units invested cost reduction cost units invested cost reduction cost
1 1.0 24.0 26 5.1 19.9
2 1.4 23.6 27 5.2 19.8
3 1.7 23.3 28 5.3 19.7
4 2.0 23.0 29 5.4 19.6
5 2.2 22.8 30 5.5 19.5
6 2.4 22.6 31 5.6 19.4
7 2.6 22.4 32 5.7 19.3
8 2.8 22.2 33 5.7 19.3
9 3.0 22.0 34 5.8 19.2
10 3.2 21.8 35 5.9 19.1
11 3.3 21.7 36 6.0 19.0
12 3.5 21.5 37 6.1 18.9
13 3.6 21.4 38 6.2 18.8
14 3.7 21.3 39 6.2 18.8
15 3.9 21.1 40 6.3 18.7
16 4.0 21.0 41 6.4 18.6
17 4.1 20.9 42 6.5 18.5
18 4.2 20.8 43 6.6 18.4
19 4.4 20.6 44 6.6 18.4
20 4.5 20.5 45 6.7 18.3
21 4.6 20.4 46 6.8 18.2
22 4.7 20.3 47 6.9 18.1
23 4.8 20.2 48 6.9 18.1
24 4.9 20.1 49 7.0 18.0
25 5.0 20.0 50 7.1 17.9
References
[1] d’Aspremont, C. and A. Jacquemin (1988) “Cooperative and Noncoopera-
tive R&D in Duopoly with Spillovers,” American Economic Review, 78(5):
1133-1137.
[2] Bergstrom, T. and J. Miller (1997) Experiments with Economic Principles,
New York: McGraw Hill.
[3] Capra M. et al (2000) “Predation, Asymmetric Information and Strategic
Behavior in the Classroom: An Experimental Approach to the Teaching of
Industrial Organization” International Journal of Industrial Organization
18(1): 205-225.
[4] Davis D., and C. Holt (1993) Experimental Economics. Princeton: Prince-
ton University Press.
[5] Goeree J. and C. Holt (1999) “Rent Seeking and the Inefficiency of Non-
Market Allocations,” Journal of Economic Perspectives 13(3): 217-226.
[6] Hinloopen, J. (1997) “Subsidizing Cooperative and Noncooperative R&D
in Duopoly with Spillovers,” Journal of Economics, 66(2): 151-175.
[7] Holt, C. (1995) “Industrial Organization: A Survey of Laboratory Re-
search.” In Kagel, J. and A. Roth, eds., Handbook of Experimental Eco-
nomics. Princeton: Princeton University Press, 349-443.
[8] Holt, C. and R. Sherman (1999) “Classroom Games: A Market for
Lemons,” Journal of Economic Perspectives, 205-214.
[9] Jorde, M. and D. Teece (1990) “Innovation and Cooperation: Implications
for Competition and Antitrust,” Journal of Economic Perspectives,4(3):
75-96.
[10] Kamien, M.I., Muller, E. and I. Zang (1992) “Research Joint Ventures and
R&D Cartels,” American Economic Review, 82(5): 1293-1306.
[11] Link, A. (1996) “Research Joint Ventures: Patterns from Federal Regist e r
Filings,” Review of Industrial Organization, 11(5): 617-628.
[12] Spence, M. (1984) “Cost reduction, competition, and industry perfor-
mance,” Econometrica, 52: 101-121.
[13] Suetens (2004) “Does R&D Cooperation Facilitate Price Collusion? An
Experiment,” mimeo, University of Antwerp
[14] Wells D.(1991) “Laboratory Experiments for Undergraduate Instruction in
Economics,” Journal of Economic Education, 22: 293-300.
[15] Williams, A. and J. Walter (1993) “Computerized Laboratory Exercises for
Microeconomics Education: Three Applications Motivated by the Method-
ology of Experimental Economics,” Jour nal of Economic Educa tion, 22:
291-315.
