INFORMATION TO USERS - Open...
Transcript of INFORMATION TO USERS - Open...
Market accessibility and the entry decision:A theoretical and experimental examination.
Item Type text; Dissertation-Reproduction (electronic)
Authors Kruse, Jamie Lynette Brown.
Publisher The University of Arizona.
Rights Copyright © is held by the author. Digital access to this materialis made possible by the University Libraries, University of Arizona.Further transmission, reproduction or presentation (such aspublic display or performance) of protected items is prohibitedexcept with permission of the author.
Download date 01/05/2018 18:35:22
Link to Item http://hdl.handle.net/10150/184619
INFORMATION TO USERS
The most advanced technology has been used to photograph and reproduce this manuscript from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are intypewriter face, while others may be from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. These are also available as one exposure on a standard 35mm slide or as a 17" x 23" black and white photographic print for an additional charge.
Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.
U·M·I University Microfilms International
A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor, M148106-1346 USA
313/761-4700 800/521-0600
Order Number 8907968
Market accessibility and the entry decision: A theoretical and experimental examination
Kruse, Jamie Lynette Brown, Ph.D.
The University of Arizona, 1988
U·M·I 300 N. Zeeb Rd Ann AIbor, MI 48106
MARKET ACCESSIBILITY AND THE ENTRY DECISION:
A THEORETICAL AND EXPERIMENTAL EXAMINATION
by
Jamie Lynette Brown Kruse
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF ECONOMICS
In Partial Fulfillment of the Requirements For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
1988
THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE
As members of the Final Examination Committee, we certify that we have read
the dissertation prepared by ____ J_a_m_1_·e __ L~y_n_e_t_t_e __ B_r_o_wn ___ K_r_u_s_e ________________ _
entitled Market Accessibility and the Entry Decision: A Theoretical --------------------~-------------------------------------------
and Experimental Examination
and recommend that it be accepted as fulfilling the dissertation requirement
for the Degree of Doctor of Philosophy ---=~~~~~~--~~------------------------------
Date
Date
Date
Date
Date
Final approval and acceptance of this dissertation is contingent upon the candidate's submission of the final copy of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.
Dissertation Director Date )
STATEMENT BY AUTHOR
This thesis has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under the rules of the Library.
Brief quotations from this thesis are allowable without special permission, provided that accurate acknowledgement of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgement the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.
APPROVAL BY THESIS DIRECTOR
This thesis has been approved on the date shown below:
V. L. SMITH Professor of Economics
Date
DEDICATION
I dedicate this Dissertation to my daughters, Casey and
Gina, who have inspired me in my work at the same time as they have
helped me to remember what is truly important in life.
iii
PREFACE
This Dissertation focuses on the effect entry has on the
organization of markets. Economic theory is certainly a tool of
the examination but the theory is tempered by test in a controlled
laboratory setting. The presentation is in five chapters. The
first chapter introduces the subject matter. The second, third,
and fourth examine specific areas of industrial organization,
namely, contestabi1ity, capacity constrained price competition, and
their predictive power when compared with other paradigms. In the
fifth and final chapter, I summarize and interpret the results. A
complete accounting of all experiments performed is contained in
the Appendices.
I wish to acknowledge the support and assistance of the
faculty of the Department of Economics. I wish to especially thank
Vernon Smith, Mark Isaac, David Conn, Elizabeth Hoffman, Brian
Binger, Bill Sharkey, Stan Reynolds, Glenn Harrison, Ron Oaxaca,
Kevin McCabe, Stephen Rassenti, and Shyam Sunder for suggestions
and ideas that they freely shared. Finally, I would like to
express my gratitude to my husband, Ed, and daughters, Casey and
Gina, for the love and support that has been so freely given.
iv
TABLE OF CONTENTS
LIST OF TABLES . . . .
LIST OF ILLUSTRATIONS.
ABSTRACT ....
1. INTRODUCTION
Page . vii
.viii
.xi
1
2. CONTESTABILITY IN THE PRESENCE OF AN ALTERNATE MARKET: AN EXPERIMENTAL EXAMINATION . . 9
Sustainability and Entry. Laboratory Tests of the Contestable Market Hypothesis Experimental Design . . . . . . . Alternate Market Treatment. . . . Human Subject Buyers vs. Computer Simulated Demand. The Profit Function . Results
.11
.13
.18
.20
.21
.23
Alternate Market .25 Human Subject Buyers vs. Simulated Demand. .28 Profit Polygon. .29 Tacit Collusion. .31
Conclusions. . . . .33
3. BUYER QUEUING RULES IN BERTRAND-EDGEWORTH DUOPOLIES. .36
Literature of Bertrand-Edgeworth Games. Bertrand-Edgeworth Duopoly Model ... Queuing Rules and Nash Equilibria . .
;Experiments and the Nash Equilibrium. Duopoly Queuing Rule Experiments. Results and Discussion. Conclusions . . . . . . . . . .
4. CONTESTABILITY AND EXCESS CAPACITY IN BERTRAND-EDGEWORTH
.37
.40
.43
.50
.51
.58
.64
OLIGOPOLY. . . . . . . .66
Contestable Natural Duopoly Bertrand-Edgeworth and Excess Capacity. Excess Capacity in Experiments ... Excess Capacity Experimental Design . .
v
.66
.68
.71
.72
vi
TABLE OF CONTENTS--Continued
Page
Results of Three Seller Natural Duopoly Experiments .74 Bertrand-Edgeworth Triopo1y Experiments .78 Conclusions .83
5. CONCLUSIONS. .84
APPENDIX A: SELLER PROFIT SPACE .87
Sellers' Profit Space of Chapter 2 Experiments. .88 Sellers' Profit Space of Chapter 3 Experiments. .89
APPENDIX B: AUXILIARY INSTRUCTIONS.
Seller Price Restrictions(Chapter 2). Seller Instructions(Chapter 2). Seller Cost Tab1e(Chapter 3) ..
APPENDIX C: EXPERIMENTAL RESULTS.
LIST OF REFERENCES . . . . . . . .
.90
.91
.92
.93
.94
114
LIST OF TABLES
Page Table
1. Experimental Design and Treatment Pairs .......... 19
2. Price, Profit and Quantity Sold by the Successful Seller in PLATO Posted Offer Experiment P037. 25
3. Theoretical Predictions ..... 57
vii
LIST OF ILLUSTRATIONS
Page Figure
1. Single Product Sustainable Natural Monopoly. 10
2. Buyer Marginal Valuations and Seller Costs 15
3. Profit as a function of Price. 24
4. PLATO Experiment P194. 26
5. Mean Prices, Alternate Market Treatment. 27
6. Mean Prices, Human Subject Buyers vs. Computer Simulated Demand. . . . . . . . . . . . . 28
7. Mean Prices, 25 Cent vs. 1 Cent Increment Rule 31
8. PLATO Experiment P206, Trading Prices. 32
9. PLATO Experiment P2l0, Trading Prices. 33
10. Residual Demand of Random and Value Queues 43
11. Profit Possibilities Under the Value Queue Assumption. 45
12. Profit Possibilities Under the Random Queue Assumption 48
13. Mixed Strategy Distributions Under the Random and Value Queue (capacity - 1/2). 49
14. Efficient Duopoly Market Price 52
15. Average Cost and Demand Curve. 53
16. Contingent Profit Functions for Value Queue Treatment. 54
17. Price Reaction Functions Under the Value Queue. 56
18. Price Reaction Functions Under the Random Queue. 57
19. Contingent Proft Functions for Random Queue. . . 58
viii
ix
LIST OF ILLUSTRATIONS--Continued
Page
20. P245 Random Queue Duopoly 59
21. P248 Random Queue Duopoly 59
22. P249 Value Queue Duopoly. 60
23. Mean Prices with 95% Confidence Interval for Random and Value Queues . . . . 61
24. Actual Distribution of Prices, Random and Value Queue 62
25. Density of Price Observations, Value and Random Queue 63
26. Nash Price Distribution for Different Capacity Levels 69
27. P282 Random Queue, Three Sellers, Alternate Market. 74
28. P279 Value Queue, Three Sellers, Alternate Market. 75
29. Density of Price Observations, Three Seller Excess Capacity Case. ...... . ...... . 76
30. Pooled Mean Prices with 95% Confidence Intervals, Three Seller Excess Capacity Case ........ . 77
31. Average Cost and Demand Curve, Natural Triopoly 80
32. P326 Natural Triopoly . . 81
33. Actual Distribution of Prices, Random Queue 83
34. Experiment P179 95
35. Experiment P18l 96
36. Experiment P194 97
37. Experiment P206 98
38. Experiment P207 99
39. Experiment P2l0 .100
40. Experiment P211 .101
x
LIST OF ILLUSTRATIONS--Continued
Page
4l. Experiment P236 . 102
42. Experiment P237 103
43. P249 Value Queue. 104
44. P250 Value Queue. 104
45. P251 Value Queue. 105
46. P252 Value Queue. 105
47. P242 Random Queue 106
48. P245 Random Queue 106
49. P247 Random Queue 107
50. P248 Random Queue 107
5l. P276 Random Queue, 3 Sellers, Alternate Market. 108
52. P275 Random Queue, 3 Sellers, Alternate Market. 108
53. P280 Random Queue, 3 Sellers, Alternate Market. 109
54. P282 Random Queue, 3 Sellers, Alternate Market. 109
55. P284 Value Queue, 3 Sellers, Alternate Market 110
56. P286 Value Queue, 3 Sellers, Alternate Market 110
57. P278 Value Queue, 3 Sellers, Alternate Market: 111
58. P279 Value Queue, 3 Sellers, Alternate Market 111
59. P322 Random Queue, 3 Sellers. 112
60. P326 Random Queue, 3 Sellers. 112
6l. P327 Random Queue, 3 Sellers. 113
62. P328 Random Queue, 3 Sellers. 113
ABSTRACT
The role of market accesibility and entry is the central
theme of this dissertation. Two theoretical models of oligopoly
theory are examined in a controlled laboratory market setting.
Experimental testing of Contestability Theory is extended beyond
the natural monopoly case and a "safe haven" provides subjects with
a viable alternative to the "contestable" market. Evidence
reported supports the conclusion that the contestable market is
robust to the introduction of the alternate market.
The theory of Bertrand-Edgeworth duopoly is explored from a
game theoretic perspective with special attention to buyer queuing
rule assumptions. Experimental evidence underscores sensitivity of
market outcomes to the queuing rule adopted.
The presence of excess capacity relative to market demand
tends to push theoretical Bertrand-Edgeworth equilibria toward
competitive levels. This result is substantiated by experimental
evidence and is independent of the queuing rule assumed. The
similarity between the Bertrand-Edgeworth excess capacity case and
a contestable natural monopoly market is investigated. The
presence of excess capacity/potential entrants is shown to exert
more downward pressure on observed laboratory market prices than
the presence of additional competitors alone. This result is at
xi
odds with the traditional Structure-Conduct-Performance Paradigm.
A herfindahl index calculated from experimental results has almost
no power to predict market outcome.
xii
Chapter 1
INTRODUCTION
The pursuit to identify factors that determine the
characteristics of markets and industries has been an integral part
of EconDmics from its very beginnings. Adam Smith in reflecting on
the self-seeking efforts of individuals, recognized the role that
the market plays in channeling these individual efforts to the
common good of society. Smith characterized five conditions of
competition:
1. The rivals must act independently, not collusively. 2. The number of rivals, potential as well as present, must be sufficient to fa1iminate extraordinary gains. 3. The economic units must possess tolerable knowledge of the market opportunities. 4. There must be freedom (from social restraints) to act on this knowledge. 5. Sufficient tim2 must elapse for resources to flow in the directions and quantities desired by their owners.
During the 19th century, the use of mathematical reasoning
led to a different, essentially structural concept of the efficient
competitive market. The mathematics of the day required a far more
restrictive definition of "competition". For the competitive model
to "work" mathematically, the competitive seller and buyer must
treat price as a parameter. In other words, each market
1
2
participant assumes that his or her own quantity decision has TIQ
effect on the market price. This requires that individual
quantities be an infinitesimal proportion of the aggregate quantity
traded of a homogeneous good. We are left with the requirement
that the perfectly competitive market have an infinite number of
sellers and buyers(thus making it logically impossible to find a
single example of perfect competition in our finite world). This
is traditionally contrasted with a model of the profit maximizing
single-seller market(monopoly). The monopoly has a decidedly less
attractive outcome from a welfare perspective. Over the decades
economists have attempted to devise models of market organization
that capture the salient features of the industries we actually see
operating in our world, namely markets whose buyers and sellers
number one or more and fewer than infinity.
The goal of this dissertation is to explore a subset of the
models of oligopoly theory. I will confine my investigation to two
main branches of oligopoly theory that have received considerable
attention in recent times. They are Contestable Market Theory and
Bertrand-Edgeworth Oligopoly Theory. I will examine the
implications of each of these theories, highlight their common
ground, and experimentally test hypotheses implied by the theories.
William Baumol chose to use his 1981 address as outgoing
President of the American Economic Association as the vehicle for
launching what he called "An Uprising in the Theory of Industry
Structure". Writings by Baumol as well as Elizabeth Bailey, John
Panzar and Robert Willig (1982) present a mathematically rigorous
representation of the contestable market case. Contestabi1ity
Theory is rich in the array of cost functions and product
configurations that it covers. Central to the Theory of
Contestable Markets is the disciplining effect that a potential
entrant can have on the incumbent(s) in a market. If the
3
potential rival can enter and exit cost1ess1y, ilas access to the
same production technology as the incumbent and has identical
access to buyers, then the equilibrium outcome that ensues from
this perfectly contestable market will be efficient. Among the
desirable properties of the equilibrium outcome are zero economic
profits, Ramsey optimal prices, efficient production and market
structure and an avoidance of cross subsidies in pricing, even for
the case of monopoly. Contestabi1ity theory serves to bridge the
gap that was created between mathematical analysis of a competitive
market and Smith's conceptualization of competition.
The most dramatic results of Contestabi1ity Theory are
apparent in the single product natural monopoly industry. By
definition, the natural monopoly firm's minimum efficient scale
exceeds market demand even at prices equal to minimum average cost.
According to the theory, the only sustainable price is that one
associated with zero economic profits. Any other price would allow
a potential competitor to profitably undercut the incumbent. The
successful entrant can costlessly exit before the incumbent can
negate the entrant's profit through retaliatory or even predatory
4
pricing. Chapter 2 is devoted to laboratory tests of
Contestability in the natural monopoly setting. This study builds
upon previous work by Coursey, Isaac, and Smith(1983) and Coursey,
Isaac, Luke, and Smith(l984). Three treatments are tested in a
controlled natural monopoly setting with two sellers. Treatment 1
requires that sellers must incur a 50 cent opportunity cost when
they choose to post a price in the natural monopoly market. This
embodies the notion of a normal rate of return. Previous
experiments assumed a zero opportunity cost of entry. Contestable
market theory is shown to be robust to a change in entry
opportunity cost. Treatment 2 tests the effect on market outcome
of (possibly strategic) human subject buyers vs. demand-revealing
computer simulated buyers. Markets with human subject buyers
demonstrate a faster and more sustained convergence toward
efficient price levels than markets with computer simulated demand.
Treatment 3 restricts sellers' price offers to be divisable by 25
cents. This greatly simplifies the subjects' price decision space.
Additionally, it negates the nonconcavity problem that exists in
the profit relation in other treatments and previous studies.
Experiments that invoke the 25 cent increment rule show more price
volatility and some interesting cases of tacit collusion.
The formal presentation of Contestable Market Theory is a
relatively new development in Industrial Organization, whereas
Bertrand-Edgeworth(hereafter B-E) Oligopoly Theory stems from
initial formulations dating back to the 1800's. B-E firms choose
5
prices and are restricted in the quantity they will produce. This
restriction is embodied either by a rising marginal cost schedule
or by a capacity constraint. When we focus on the single product
case, we can identify a commonality between the predictions of
Contestability and Bertrand-Edgeworth Models. For example, in a
Bertrand-Edgeworth duopoly in which either seller has sufficient
capacity to satisfy demand at any profitable price, the efficient
price will obtain in equilibrium. This equilibrium prediction
corresponds with the efficient price prediction of a contestable
natural monopoly. The two models utilize markedly different
assumptions, thus, the overlap of predicted results creates the
interesting question of which assumptions are behaviorally
significant.
Joseph Bertrand(1883) first proposed a duopoly model of
firms whose decision variable is price for comparison with Augustin
Cournot's(1838) quantity choi~e duopoly model. Bertrand criticized
Cournot's pioneering work on the basis that it lacked a plausible
explanation of price formation. In Bertrand's model, producers
simultaneously and independently choose prices. Demand is
allocated to the lower-price seller who then produces the quantity
demanded at the stated price. Given constant marginal cost,
Bertrand showed that the only equilibrium must be at the price
which equals marginal cost. The motivation for this result lies in
the fact that either seller is capable of producing enough to
satisfy the market. Hence, the lower-price producer serves the
6
entire market and the higher-price producer sells nothing. If a
producer chooses any price greater than marginal cost, its opponent
will undercut by £ and earn a positive profit.
Francis Edgeworth's 1925 critique of Bertrand recognized
that, except in the case of constant marginal cost, there are
serious existence problems when firms produce a homogeneous good.
Edgeworth modified the original Bertrand model to introduce the
possibility of limited capacity. Edgeworth's firms have either
constant marginal cost up to some fixed capacity or strictly
increasing marginal costs. He characterizes a set of demand
conditions in which one firm cannot serve the entire market.
Edgeworth suggests that prices might be expected to fluctuate over
a range whose upper and lower bounds depend on the two firms'
respective capacities. Appealing to a loosely dynamic approach,
Edgeworth envisioned several rounds of price cutting until
prevailing market price hits a trigger price. Given this trigger
price, the opponent firm finds quoting a high price more profitable
than undercutting the trigger price. Without the threat of being
undercut, the low price firm will subsequently raise its price.
This is then followed by another round of price cutting. The high
price and the trigger price, as described, form the upper and lower
bounds of the Edgeworth range.
Game theorists now recognize that Edgeworth described a
case in which no pure strategy equilibrium exists. There is,
however, a Nash equilibrium in mixed strategies. Bertrand-
7
Edgeworth games, in which capacity-constrained firms choose prices,
have commanded considerable recent interest. Chapter 3 introduces
the Bertrand-Edgeworth Game and demonstrates the sensitivity of
it's theoretical predictions to the choice of buyer queuing rule.
A group of laboratory experiments is reported in which two
different buyer queuing rules are compared. The Value Queue
assumes that single unit buyers queue up with highest-reservation
price buyers first. This is analogous to allocating demand by
running "down the demand curve". The second queuing rule tested,
called the Random Queue, assumes that single unit buyers are
randomly arranged in the queue. All other variables are controlled
across the two treatments. The results under the two queuing rule
assumptions are significantly different, with the random queue
markets sustaining considerably higher prices.
In B-E models using either queuing rule, both firms'
individual capacity and total "market" production capacity as well
as their relationship to aggregate demand, determine the nature of
the Nash equilibrium prediction. In some caSes, B-E models run
contrary to the traditional Structure-Conduct-Performance Paradigm
that has been central to industrial organization policy and
empirical work. Under this paradigm, causality starts with the
market structure representing some measure of the size of the firm
or coalition of firms relative to the entire market. The relative
size of the decision making body then determines the effectiveness
of behavioral strategies or conduct. Market performance as
8
measured by allocative and production efficiency is the end result
of the causal flow. This causality implies an inverse relationship
between the number of firms and the pricing efficiency of the
market. Chapter 4 reports experiments designed to gauge the effect
of number of sellers and capacity/demand relationships on market
outcome. The results are compared to discover which of the
determinants described above exerts a stronger influence on pricing
efficiency. Markets in which there is sufficient productive
capacity that the highest price firm will sell nothing if the other
firm(s) price efficiently show strong convergence to a
competitive-type equilibrium. Also, as the number of firms
increases from 2 to 3, market prices tend to be lower. However,
the presence or absence of over-capacity is the dominant effect.
Chapter 5 will provide a brief concluding statement of the
contribution of this body of research to the fields of industrial
organization and experimental economics. I propose directions for
future research that stem from results reported in the following pages.
CHAPTER 2
CONTESTABILITY IN THE PRESENCE OF AN ALTERNATE MARKET: AN EXPERIMENTAL EXAMINATION
The theory of contestable markets as developed by William
Baumo1, John Panzar and Robert Wi11ig(1982) describes a market with
perfect freedom of entry and exit. Though the theory covers a
broad range of cost structures and production configurations, this
chapter is devoted to tests of contestabi1ity theory as applied to
single product sustainable natural monopolies. In a sustainable
natural monopoly industry, minimum efficient scale exceeds market
demand. Figure 1 depicts the cost-demand relationship
characteristic of the single product sustainable natural monopoly.
The application of Contestable Market Theory to the natural
monopoly industry yields striking results. A single natural
monopolist can produce and sell more efficiently than any
multiple; however, in the absence of potential entry, there is no
incentive for this firm to do so. If only one firm does exist,
that firm has the greatest potential for profitable exercise of
monopoly power. In other industries, the possibility of losing all
of one's customers to active rivals serves to discipline a firm's
pricing behavior away from monopoly levels. According to
Contestabi1ity theory, the disciplining force that pulls market
prices toward efficient levels may come from potential as well as
9
10
active rivals. Thus if entry and exit are costless, even an
unregulated monopolist has an incentive to price efficiently.
Price
Demand
Average Cost
o Output
Fig. 1. Single Product Sustainable Natural Monopoly
In this chapter, I will discuss the concept of
sustainability, previous entry treatments, and the effect of sunk
costs on equilibrium outcome. A survey of the initial experimental
tests of Contestability Theory is followed by the description of my
experimental design and discussion of the results.
11
Sustainability and Entry
The theory of sustainabi1ity was developed by Baumo1,
Bailey, and Wi1lig(1977) and Panzar and Wi11ig(1977). A natural
monopoly is said to be sustainable if there exists a price and
output such that entry by rival firms is unattractive, while all
demand is satisfied and revenues cover total costs of production.
Definition: A natural monopoly with cost function C and market
demand D is sustainable if there is a price p and output q such
that
(i) q - D(p),
(ii) p.q - C(q),
(iii) p~q'<C(q') for all p'<p and all q'<D(p').
In other words, a profit maximizing potential rival sees no
opportunity for positive profits due to entry so long as the market
incumbent chooses a sustainable price and quantity. A sufficient
condition for sustainabi1ity is that average costs of production
fall as output expands. By producing the total market demand, the
monopolist can achieve an average cost that is at least as low as
any potential rival. However, if average costs first fall and then
rise with output, the natural monopoly need not be sustainable as
demonstrated by Sharkey(l982). So by restricting ourselves to
natural monopolies represented by Figure 1, we will examine the
predictions of Contestabi1ity Theory for sustainable natural
monopolies. As stated previously, the Theory of Contestable
Markets relies on cost1ess entry and exit. If some barrier makes
12
entry or exit costly, then the nature of the sustainable price
quantity pair changes, moving further away from the efficient pair.
Early arguments by Bain(1956) and Sylos-Labini(1962) assume that
price is uniquely determined by market demand and a potential
entrant expects the incumbent to maintain preentry production
levels. Under this scenario, the scale economies characteristic of
a natural monopoly would constitute a barrier to entry.
Demsetz(1968) argues that if a potential entrant could replace the
incumbent, then high fixed costs that create economies of scale do
not constitute barriers to entry. In contrast, sunk costs do
create entry barriers even under the Demsetz assumption as
demonstrated by Spence(1977). For any given project or expenditure
of funds, the sunk cost is defined as the difference between the ex
ante opportunity cost of the funds and the value that could be
recovered, ex-post, if the project is terminated.
The need to sink funds into a new enterprise, whether into
physical capital, advertising, or anything else imposes a
difference between the incremental cost and the incremental risk
that is faced by an entrant and an incumbent. Obviously, entry
yields a nonnegative expected profit only if the profits weighted
by the probability of success outweigh the unrecoverable entry
costs that will be lost in the event of failure. Thus for any
industry characterized by positive sunk costs, if entrants face a
nonzero probability of failure then the sustainable price quantity
configuration involves a positive profit to the incumbent(s).
13
Laboratory Tests of the Contestable Market Hypothesis
Previous experimental testing of contestability theory
focuses on a sustainable natural monopoly industry. Coursey,
Isaac, and Smith (1983) produced the initial laboratory tests of
the Contestable Market Hypothesis under zero sunk costs using a
posted offer mechanism on the PLATO computer system. The PLATO
posted offer program allows buyers and sellers sitting at
individual terminals to make exchanges for a maximum of 25 periods.
Each buyer (seller) has a record sheet which shows a marginal
valuation (cost) associated with each unit that she consumes
(produces). Buyers have capacity to buy up to five units of the
homogeneous commodity. Sellers in the contestability experiments
have a capacity to produce and sell up to ten units. The sequence
of events in a single period of a PLATO experiment is as follows:
1. Subject sellers are asked to post a selling price and a maximum quantity (number of units in stock). The price and quantity chosen may be changed any number of times at the subject's discretion.
2. Once satisfied with the price and quantity chosen, the subject seller presses an "offer box" on the touch sensitive screen to irrevocably place his(her price on the market.
3. When all sellers have made a price quantity offer, all sellers' prices are displayed on the terminals of all buyers and sellers. The quantity offered remains private information.
4. Buyers are randomly assigned to a queue and are allowed to purchase one at a time.
5. As each buyer's turn comes up, she may choose not to buy or may choose to purchase units by touching the "price box" on the screen of the seller from whom she wishes to purchase.
6. When buyers have purchased all of the u~its a given seller has in stock, an "out of stock" display repla,:es that seller's price box.
7. After all buyers have had an opportunity to pULchase, the current trading period is closed. Each seller's (buyer's) record sheet is updated to record the period's sales (purchases) and associated subject profit.
8. A new pe~iod begins.
The costs parameterized in my experimental study and in
14
previous work are those of a natural monopoly. Figure 2 shows the
marginal and average costs of a typical seller and the aggregate
demand which comprises individual buyers' marginal valuations.
Appendix A contains sellers' profit space for all feasible price
and quantity combinations. Either of two identical firms can
accommodate the entire market demand at the competitive quantity of
10 units. In the event that both firms match prices, substantial
losses will occur at prices close to the competitive range.
CIS found support for the strong version of CMH in four of
their six duopoly experiments. The remaining two experiments
showed significant convergence of price and quantity toward, but
did not enter the competitive range. CIS also compared the results
of their six laboratory duopolies with four laboratory monopoly
experiments. All experiments utilize human subject buyers on the
demand side of the market. The markets with two sellers show
qualitatively more "competitive" results than the single· seller
markets. Clearly, the presence of additional potential sellers
causes markets to tend toward competitive price levels. This
seminal work was followed by Harrison(1984), Harrison and
McKee(1984) and Coursey, Isaac, Luke and Smith(1984).
$6 r- Demand
$5
$4
$3
$2
$1
I
'--""'"1. •• _. _. I
• .. -.. -·-· .. ·:·:..:.:.:.·:.·:t .. ~~-~ ... -.. ~ ! -. ---"1. - - - - - to ____ "' ____ _ ................... • ~ .1------.-.. -.~.---~- . -----
Average Cost ............ ···L .. _· .. t ......... .
Marginal Cost JI""-'-'"
$o~------~------~I------~--------~------+--------024 6
Quantity 8 10
Fig. 2. Buyer Marginal Valuations and Seller Costs
Harrison(1984) noted several areas in which the original
Coursey, Isaac, and Smith experiments depart from the Theory of
Contestable Markets as described by Baumol, Panzar and
15
Willig(1982). The lively discourse between the authors raised an
issue that is pertinent to the entire field of experimental
economics. Every experimental economist must decide "where to
start" in a study. We are equipped with some 200 years worth of
theory intended to explain the workings of the world. This theory,
16
adheres in varying degrees to the standards of "good" economics.
It should capture the driving forces which organize the economic
behavior of decision making entities in addition to being logically
consistent and mathematically complete. Often, a theorist must
trade-off salient features of a model to gain ~athematical
tractability or certain restrictive assumptions are drawn for the
sole purpose that they facilitate a determinate solution. The
issue becomes whether the experimental economist should perform a
strict test of the theory as stated using the same information
states, etc. where possible or test the predictions of the theory
in a richer environment similar to one in which policy based on the
theory will be applied. I propose that both types of experiments
are important and should be performed and labeled accordingly.
The original set of contestability experiments reported by
Coursey, Isaac and Smith attempted to capture the essence of the
theory in a more general laboratory market setting than that
described by the theory. Harrison's criticism of this study
focused on the sequence in which information is revealed to
subjects stating that the fact that the incumbent and potential
entrant must post prices and quantities simultaneously puts the
Coursey, Isaac and Smith operationalization at odds with the
theoretical models. Harrison reports a set of experiments in which
the incumbent and potential entrant move sequentially. Using what
Harrison calls the Bertrand/Nash assumption, the potential entrant
knows the incumbent's posted price before making the entry(pricing)
17
decision. Harrison reports the results of four experiments which
he calls "an exact operationalization of the theoretical model
proposed by BPW"(Baumol, Panzar and Willig). Three support the
strong version of the Contestable Market Hypothesis as described by
Coursey, Isaac, an~ Smith. The subjects in the fourth experiment
use the sequential information to coordinate a profitable
collusive arrangement. Harrison also reports tests of
contestability with three potential sellers in a market that will
accomodate a single seller and finds that increasing the number of
potential entrants puts additional downward pressure on price.
Nevertheless, Harrison did not provide an "exact
operationalization" of Baumol, Panzar and Willig's theoretical
model either. Panzar and Willig(1977) state:
Since the entrant's products or services are assumed to be perfect substitutes for those of the monopolist, a potential ambiguity arrises if the entrant were to set his price for service i, p.e, equal to the monopolist's price,
m ~ Pi ;i.e. how would the demand, Qi , be split? The natural treatment would require the entrant to actually beat the monopolist's price before he receives any customers. Thus e m when, foreexam~le, we say p S P , the reader may take us to mean p S P - E.
Harrison handles matched prices by randomly choosing a seller to
service the market at the (matched) price which does not reflect
what Panzar and Willig clearly state.
Harrison and McKee(1984) contrast contestability with
various regulation schemes and offer their tests of the
sustainable natural monopoly market using computer simulated demand
for comparison with the Coursey, Isaac and Smith experiments that
18
used human subject buyers. Such a comparison is inappropriate
since Harrison and McKee lacked adequate facilities to perform a
controlled test of the effect of human subject buyers vs. computer
simulated demand.
Coursey, Isaac, Luke, and Smith extend testing of
contestability theory to a market characterized by sunk costs. As
the theory predicts, the laboratory market price, quantity, and
efficiency are not as close to the competitive range as the
outcomes of markets with zero sunk costs. They discuss the effect
of human subject buyers vs. computer simulated demand but used
parameters that differed in absolute terms across the treatment so
did not provide a clean test.
Experimental Design
The present experimental study focuses on three areas of
interest which are briefly described below and expanded in sections
2, 3, and 4. An alternate market with certain payment is
operationalized in the first treatment of the experimental design.
This "safe haven" affords a place from which hit-and-run entry can
occur and establishes $0.50 per period as the normal (competitive)
profit that is the opportunity cost of entry. A second question
addressed in this series is that of the effect on market outcomes
of human subject buyers as compared to demand fully revealed by
computer simulation. The third treatment attempts to test the
effect of restriction of seller offers to those divisible by $0.25
19
on market outcomes. The described $0.25 increment rule restricts
the sellers' decision space and concavifies the otherwise jagged
profit relation. Nine experiments are conducted under a partial
factorial design. Table 1 identifies the experiments by their
PLATO Posted Offer number and describes the treatments exercised in
each.
Table 1. Experimental Design and Treatment Pairs
Experiment Alternate Simulated Offer Market Demand Restriction
CIS Mean NO NO NO
P237 NO YES NO
P236 YES NO NO
P1B1 YES YES NO
P194 YES NO YES
P179 YES YES YES
P211 NO YES NO
P207 YES YES NO
P210 NO YES YES
P206 YES YES YES
20
Alternate Market Treatment
The essence of the "story" of a contestable market is that
an incumbent seller is only safe from hit-and-run entry if he keeps
his price offers so low that the resulting profit yields a rate-of
return that is consistent with that being earned in other
industries. The Coursey, Isaac, and Smith, Harrison and McKee and
Harrison experiments implicitly assumed a normal rate-of-return of
zero.
The alternate market treatment allows potential entrants
access to a safe haven with a risk-free normal profit of $0.50.
Potential entrants must forego the $0.50 opportunity cost to enter
the "contestable" market with its associated risks. For the case
of a natural monopoly with two potential sellers, there is no
equilibrium in which two sellers simultaneously serve the market
profitably. In fact, if price matching market splits do occur,
given the parameters used, substantial losses may be sustained by
both sellers. The Coursey, Isaac and Smith(1983) experiments, with
a high incidence of profit destroying market splits, can be
interpreted as examples of destructive competition. A stable
equilibrium must, therefore, involve one seller choosing not to
enter a competitive bid. A common characteristic of the Coursey,
Isaac and Smith(1983) experiments is that they closely resemble a
special case (n-2) competitive market where "possession" of the
market can change nearly every period. In theory, if profits in a
21
competitive market fall short of a normal return, some of the
participants will choose to pull their resources out of production
of the good and seek another industry in which at least a normal
rate-of-return is possible. The alternate market treatment is
constructed so as to provide experimental sellers with the option
of choosing to leave the contestable market in favor of a safe
haven in which a normal rate-of-return is defined as $0.50 per
period. A supplemental instruction sheet is given to sellers
informing them of the alternate market and the rate-of-return
available in that market. The instruction sheet includes a table
in which subjects record each period the alternate market is used.
A sample of the instruction sheet used for this treatment is in
Appendix B. The procedure for entering the alternate market is to
post a price of $0.00 and quantity in stock of 0 units. Buyers
would then receive an "out of stock" message on their screens from
this seller. The action required to enter the alternate market
involves hitting the same number of keys as to post a price to the
contestable market. Experiments P18l, P206, P207 and P236 offer
sellers the option of an alternate market. These experiments are
paired with P237, P2l0, P2ll, and are also compared with the CIS
results.
Human Subject Buyers vs, Computer Simulated Demand
Coursey, Isaac, Luke and Smith (1984) present the first
comparison of human subject buyers with computer simulation of
22
demand. The comparison of human subject buyers vs. computer
simulation programmed to fully reveal demand has been a topic of
discussion in Coursey, Isaac, Luke and Smith(1984) and in Harrison
and McKee(1984). Coursey, Isaac, Luke and Smith recognize the
effect of under revelation of demand on industries with decreasing
average costs. Buyers can effectively discipline the seller by
withholding demand. In the natural monopoly case, this hits
sellers in their most profitable units. In the Coursey, Isaac and
Smith design a one unit under revelation of demand reduces profit
as much as $1.20 per period. In the Coursey, Isaac, Luke and
Smith study, six experiments are conducted with human subjects as
buyers and six with computer simulated demand. PLATO simulation
of demand is struc tured such that, from the sellers' point of view.,
the time frame is the same whether human experimental subjects or
computer simulation is used to create the demand side of the
market. The only difference, of course, is the possibility of
human error and the potential for strategic withholding of demand.
In the Coursey, Isaac, Luke and Smith study, contrary to
expectations, the presence of human buyers results in a stronger
degree of convergence to a competitive price range. All 12
experiments use the same parameters in terms of deviation from the
identified competitive price. However, the absolute competitive
price is $3.28 for all six human buyer experiments and one of the
simulated demand experiments, whereas the remaining five with
simulated demand use an absolute competitive price of $4.38.
· .
23
Harrison and McKee(1984) criticize the use of human subject
buyers advancing the argument that strategic buyer behavior is
something that we can and should control for in laboratory
experiments. This argument begs the issue. Experimental evidence
reported here and in Coursey, Isaac, Luke and Smith demonstrates
that the presence of human subject buyers does make a difference.
The fact that the theory does not consider strategic behavior of
buyers is a weakness of the theory that the experimental evidence
simply points out. Harrison and McKee lack the hardware (and
software) to run an unbiased test of the effect of human subject
buyers. The present treatment is designed to offer such a test.
In my experimental series, two human subject buyer experiments are
performed, Pl94 and P236. These are paired with Pl79 and Pl81
respectively. Also, P237 is paired with the means of the original
Coursey, Isaac and Smith experiments. All other relative and
absolute variables are the same between pairs.
The Profit Function
Recent contestable market experiments (Coursey, Isaac, and
Smith(1983), Coursey, Isaac, Luke and Smith(1984), Harrison and
McKee(1985), Harrison(1984» have all had a common design feature,
namely a marginal cost structure that yields decreasing average
cost over the relevant range. Tha relationship between profit and
price does not approximate a smooth function in the experimental
setting of PLATO. The combination of this cost structure with a
24
step function induced demand curve leads to a profit relation as
characterized in Figure 3. Given that profit maximization remains
the assumed motivation for sellers (and buyers), this profit
relation could confound subject sellers in their search. Consider
the trading periods 7-12 in CIS (1983), P037 reported in Table 2.
Though prices were relatively stable (within a range of $.10),
profit levels fluctuate widely for prices which deviate only a few
cents from $4.99 (which yields a local profit maximum of $2.92).
SaUer Profit $5~------------------------------------------~
$31----
$2 ./' ............ /. __ ..
$1 .-- .. --.- -- -.--... ----.------------.-... -.---.-.... ---.-.. ----... ---.. -.... --.----... -. -.-.--.. - .-.-
$O~+-----~--------~------~--------~----~~
o 0.5 1 1.6 2 2.e
Price in Deviation from AC< 10) Fig. 3. Profit as a Function of Price
A method of making the profit function concave is
operationalized. This treatment restricts the prices that sellers
are allowed to offer by requiring that they be divisible by 25
25
cents. Sellers are given an additional sheet of instructions
explaining the offer restriction. Appendix B contains a copy of
the auxiliary instructions. The basic struccure of this laboratory
market is the same as that in Coursey, Isaac, and Smith, with
identical parameters in most of the experiments. The rest of the
experiments use parameters that shift the profit function down by
approximately $.40. Posted offer experiments P179, P194, P206, and
P210 include the seller offer restriction. These are paired with
P181, P236, r207, and P211 respectively. All prices are reported
in deviation from AC(10).
Table 2. Price, Profit and Quantity Sold by the Successful Seller in PLATO Posted Offer Experiment P037.
Period Price Quantity Profit
7 5.05 7 2.10 8 4.99 8 2.92 9 5.00 7 1.25
10 4.95 8 2.60 11 5.00 7 1.25 12 5.00 7 1.25
Results
Alternate Market
The presence of the alternate market in which participants
can earn some positive profit with certainty is effectively
incorporated in sellers' decisions in the PLATO market. The
alternate market is offered in six experiments and is used an
average of eight periods during the course of a typical experiment.
It also affords a safe haven from which entrants can hit-and-run.
26
Fig~re 4 shows an example of hit-and-run entry by seller B.
Seller B uses the alternate market extensively throughout the
course of the experiment and in period 11 enters the market,
undercuts the incumbent and escapes to the alternate market again.
Also in period 17, seller B leaves the safe haven and undercuts the
incumbent to take the market. Prices in the experiments without
the alternate market are reported in deviation from AC(lO).
Prices for the experiments that include an alternate market are
reported in deviation from [AC(lO) + $0.50/10] to reflect the
$0.50 opportunity cost.
S8
$6
Pm $4
$3
$2
$1
$0
Trading Prloe
-
1
I!
'- i- '-
8
I!
0 I!
E E 0
0
- I! ~ 0 ,Ii - E - -
-- - '--- '--'- '-
10 18
Trading Period 20
III A Seiling Price 0 B Seiling Prtce o Uneuooeuful Otfer
E denotes eXIt to Alternate Market
Fig. 4. PLATO Experiment P194
Price In deviation. from Min At; 27
Pm Pm
$1.0 _._--_ .... _--_ .. _----_ .. _--_ .. _ ... __ .. _-_._ ........... _ ... _-.••.... _--_._--_._-_ .......
'0.8
$0.0 o
~.~ /~'i\. ~*~~2~ "Ii r . .
'-..",
6 10
Trading Period 16
--*- AI.mats Marlatt -Eo- No Alternate Marklet
Fig. 5. Mean Prices, Alternate Market Treatment
Po
20
Mean prices of the experiments in the two treatment groups
are illustrated in Figure 5. Eight experiments are paired
according to the alternate market treatment variable and are tested
using a t test for two dependent samples. With a t statistic of
1.597, the hypothesis that the mean prices for the two treatments
are equal cannot be rejected at the 5% significance level. A
Wilcoxon nonparametric test for paired samples yields similar
results. An F test on the population variances yields a test
statistic of 1.378. We cannot reject the hypothesis that the
population variances are equal.
28
Human Subject Buyers vs. Simulated Demand.
Strategic withholding of demand by subject buyers can
seriously erode the profits of sellers, particularly if their cost
structure is characteristic of a natural monopoly. When seller
strategies are the area of interest, simulation of demand makes an
increase in the number of replications possible with the same total
money allocation for subject payments. This experimental series
yields three pairings for evaluation of the effect of simulation of
demand versus human subject buyers. Figure 6 contains the mean
prices resulting from the two treatment groups.
Price In deviation from Min AI;
Pm Pm
".0 .. __ .................. _, .. _._ .... _ .... _ .. _._.H ..... ____ .. __ .. __ -_._ ... _ ............... __ .. __ .-.. _ ...... __ ..... __ .. _-_._- ....... --.. ---.---.-.. - ... -...... - ..
$0.0 '--_____ ....L..-_____ .....L... _____ --'-_____ ---'Po
o 6 W W ~
Trading Period --*- HumllUl Subject Buyer. -&- S4mulatttd Demand
Fig. 6. Mean Prices, Human Subject Buyers vs. Computer Simulated Demand
29
We can reject the null hypothesis that mean prices are the
same across the two treatment groups in favor of the alternate
hypothesis that prices are lower with human subject buyers. A t
test for dependent samples is used with a 5% significance level.
Also, the Wilcoxon nonparametric test for paired samples yields the
result that we" can reject the hypothesis that the prices arise
from the same distribution.
Profit Polygon
The method used to smooth the profit function, restriction
of seller offers to those divisible by $0.25, has some additional
effects on seller behavior which were not foreseen before
experiments were actually performed. First of all, the restriction
on prices simplifies the seller's decision space. Previously, over
the range of prices from $0.00 to $2.50, a seller had 250 different
price offers that could be chosen. This decision space shrinks to
10 possible price offers under the price restriction treatment. In
and of itself the simplification of the decision space is not a
particularly bad consequence of the treatment. In fact, it may
help subjects sort out strategies in especially complicated
experiments. However, when the cost structure is that typical of a
monopoly, there is a potential problem. Under the decreasing
average costs of a natural monopoly, if sellers match prices and
share the market, this can result in substantial losses ($2.00 or
more per period for both sellers) when prices are near the
30
competitive range. The reduction of allowable price offers
increases the probability of sellers matching prices by a factor of
25. Since there are market forces pushing seller price offers
toward the competitive price, the probability of loss generating
splits is even higher than splits at other prices. Subject seller
reaction to these loss producing splits is hard to predict. The
reaction to suffering a loss can result in sellers offering higher,
lower, or the same price in the following period. Examples of all
three of these reactions can be found in the reports of individual
experiments contained in Appendix C. A different cost structure
which does not produce losses at matched prices may be the place to
ascertain the effect of smoothing the profit function.
The mean prices of experiments with the price restriction
compared to those without are shown in Figure 7. Note the $0.25
increment group prices show more "noise" than its comparison group.
The hypothesis that mean prices for experiments with the 25 cent
increment rule are the same as the prices of experiments without
this restriction cannot be rejected on the 5% significance level.
Also, a Wilcoxon nonparametric test yields the result that we
cannot reject the hypothesis that the prices arise from the same
distribution.
31
Prloo In deviation from MIn ~
Fmr---------------------------------------__________________ ~pm
".0 . -' .. --....... -.~~.-.---.. ---.-.-.~ ..... -.-------... -.. -. __ ..... _---_ .......... __ .. _._._-_ .......... .
$0.15 t----'\:------:f---
$0.0 L---------L...-------I _____ ..L-____ --L---1Po
o 10 18 20
Trading Period -60- 1 Cent Increment ~ 28 Cent Increment
Fig. 7. Mean Prices, 25 Cent vs. 1 Cent Increment Rule.
Tacit Collusion
The simplification of the price decision space by the $0.25
increment rule allows sellers to identify alternative strategies.
P206 and P2l0 in Figures 8 and 9 both illustrate examples of the
two sellers managing to work out collusive arrangements that
increase profit. P206 has an alternate market treatment and $0.25
increment rule. The sellers work out a market swapping arrangement
of alternating high and low prices in periods 9 through 15, then
the same market swapping arrangement is maintained by taking turns
dropping to the alternate market for the remaining periods of the
experiment.
32
Prloe Offer $8~-----------------------------------------------'
1------~~~~~~--~--~~_,r_----------------------_1Pm
o
~.--------------------------------------------------~~ ... Alt. Mkt. A B B B I!J A B A '8~~~~~~~~~~~~~~~~~~~~~~~
1 10 15 20 25
Trading Period o A Price or .. r ~ B Price Offer
Fig. 8. PLATO Experiment P206, Trading Prices
P210 also has the $0.25 increment rule but no alternate
market. In this case, seller B experiences very low profits during
the initial periods of the experiment. She then attempts to
signal, starting in period 5, but is ignored by seller A. In
period 16 she starts giving very strong signals and seller B
finally responds. This arrangement is not as smoothly cooperative
as the collusive arrangement generated in P206. Se1~er B tends to
occasionally "cheat" on the arrangement.
33
Price Offer .10~~---------------------------------------------'
$9
$8
$7
$3ULJ-~L-~~~-L~J-~~~~-L~~~~~~~
10 16 20 26 1
Trading Period A Price Offer - B Price Offer
Fig. 9. PLATO Experiment P2l0 Trading Prices
Conclusions
The goal of this experimental study is to test two
questions of experimental methodology and examine seller
performance in a contestable market with positive opportunity cost
of entry. For the case of seller price restrictions the series
yields some interesting results. The restriction of sellers'
offers which shrinks the decision space may allow sellers to more
easily look for alternative strategies like the collusive
arrangements discussed in section 4 of the results. Future
34
research with seller price restrictions should include other cost
structures in which matched prices are not so damaging to seller
profits.
The results from this series in comparing human buyers with
simulated demand complement those found by elLS (1984). The
presence of human subject buyers tends to discipline the sellers.
Prices converge toward competitive levels more quickly. This
implies that experimental results that converge to competitive
levels under simulated demand are robust. The addition of human
buyers would tend to reinforce the forces already present.
Secondly, these results emphasize the necessity that complete
theoretical models include both sides of the market.
The addition of an alternate market which induces a
positive opportunity cost of entry is an important contribution.
The safe haven provides a base from which hit-and-run entry is
observed, yet the adjusted market price outcomes are not
significantly different from those in the zero opportunity cost-of
entry case.
The next step for contestabi1ity should be an examination
of market performance under other cost structures and a multiple
market setting. A proposed design will give each experimental
subject seller access to two PLATO terminals. Each terminal will
display the seller's cost structure in a particular market. There
will be a natural monopoly, a potential duopoly, and a safe
alternate market with various entry requirements. The natural
35
monopoly has decreasing average costs over the relevant range so
that any single seller can service the entire market. In the
potential duopoly, firms have identical classic textbook U-shaped
average cost curves. The relationship of firms' costs to demand is
such that two firms can efficiently accommodate demand at a
sustainable price close to competitive levels. Entry conditions
for these two PLATO markets are affected by the rules governing
payments in the safe haven. Issues of cost complementarity and
economies of scope will be accessible under the mUltiple market
design.
Chapter 3
BUYER QUEUING RULES IN BERTRAND-EDGEWORTH DUOPOLIES
This study evolved from what was intended as a set of
baseline experiments initiated in February 1986 to be used for
comparison with a more complicated mUltiple market design (see
IIExperimenta1 Tests of the Contestable Market Hypothesis in a
Multiple Market Setting: A Research Proposa1,11 December 1985).
After the first baseline experiments yielded unexpected results,
more attention was given to the question of buying urder. The
PLATO posted offer market allows purchasing by one buyer at a time
and randomly assigns buyers to the queue. Independent of my
experimental exploration of the effect of buyer queuing,
theoretical developments in oligopoly modeling were also moving
toward a realization of the importance of buying order.
In this chapter, I will discuss recent contributions to B-E
oligopoly theory with particular attention to the buyer queuing
rules assumed. My presentation of a basic model of B-E duopoly
includes a description of the three classes of Nash equilibria ,that
arise from different capacity/demand relationships. This is
followed by an analysis of buyer queuing rules and their effect on
Nash Equilibrium predictions of the model. A description of B-E
experimental work is followed by my experimental design and
discussion of results.
36
37
Literature of Bertrand-Edgeworth Games
A majority of recent papers on B-E Oligopolies utilize
either of two main queuing rule assumptions. The queuing rule
specifies the order in which buyers whose reservation values
represent the demand curve are allowed to purchase one or more
units. One queuing rule, hereafter the value queue, allocates
demand by "marching down the demand curve". Under the value queue,
highest reservation price units go to the lowest-price seller.
There are two constructs on the microeconomic structure of demand
that may be used to generate the value queue. First, consider a
society of n (homogeneous) consumers with identical downward
sloping demand curves. Each customer must be limited to purchase
the same fraction (k./n) of the low-price firm's capacity. l.
Another approach is to consider market demand as the summation of
inelastic demands of heterogeneous customers, each wishing to
purchase one unit of the good provided the price is below his/her
reservation price. Under this construct, the value queue is
generated by ranking the buyers according to their reservation
value and allowing highest reservation values to be first in line.
The random queue has also been used extensively in the
literature. For the homogeneous customer case, the random queue
would occur if customers are allowed to make unlimited purchases on
a first-come first-served basis. The low-price firm's stock could
well be exhausted before all customers have an opportunity to
purchase. The random queue will arise in the heterogeneous
38
customer case when prospective buyers are randomly assigned to the
queue. The low-price firm would serve a random sample of the
population.
Shubik(l959) discusses the structure of a Bertrand
Edgeworth game and three candidate queuing rules which give rise to
different contingent demand for the high-price firm. Martin
Beckmann(l965) chose one of Shubik's suggested rules, a random
queue, and explicitly calculated mixed strategy equilibria for an
example with two symmetric capacity-constrained firms and
normalized linear demand. For explanatory purposes, identify the
amount, a as market demand when p-unit cost c. Beckmann showed
that when each firm's capacity is 1/4 a, the B-E Nash equilibrium
is in pure strategies at the monopoly price. When each firm has
capacity equal to at the pure strategy B-E Nash equilibrium is at
the efficient price (equal to unit cost)-. No pure strategy
equilibrium exists for capacities between 1/4 a and a. For this
case, Beckmann solves for the mixed strategy equilibrium
distribution. Levitan and Shubik(l972) use the value queue in a
model with symmetric firms and linear demand. Additionally, they
look at the case in which at least one firm has sufficient capacity
to supply the entire market. The value queue assumption simplifies
the process of solving for mixed strategy distributions. Levitan
and Shubik show that a pure strategy Cournot solution exists for
identical capacities of 1/3 a and demonstrate the pure strategy
"price-unit cost" equilibrium when each firm has a capacity equal
39
to a. For capacities between 1/3 a and a, only mixed strategy Nash
equilibria can be found.
Osborne and Pitchik(1986) prove existence and uniqueness
for a more general model that includes the one described by Levitan
and Shubik. They permit firms with different capacities and assume
only that demand is continuous and decreasing. Their propositions
include one that equilibrium prices are lower the smaller is
demand relative to capacity. Dasgupta and Maskin(1986b) establish
general existence of mixed strategy equilibria in the Bertrand
Edgeworth model using existence theorems for discontinuous games
developed by Dasgupta and Maskin(1986a). Maskin(1986) presents
existence results for models with production in advance that do not
require symmetry and permit cost functions that are continuous,
nondecreasing, concave or convex, and equal zero when there is no
production. Maskin(1986) also proves existence when production is
to order for strictly convex cost functions. Kreps and Scheinkman
study a two stage game and demonstrate that if firms choose
capacity first and prices in the second stage that a Cournot pure
strategy equilibrium obtains (assuming a value queue). Davidson
and Deneckere(1986) show that the Kreps and Scheinkman(1983) result
crucially depends on the the queueing rule adopted and describe a
numerical method of calculating for the mixed strategy price
distribution in the two stage game with a random queue assumption.
Allen and Hellwig(1986) concentrate on market perf~rmance
in Bertrand-Edgeworth games when the number of sellers becomes
40
large. With their random queue assumption, though the mixed
strategy distributions converge toward the efficient price as the
number of sellers gets large, there is still a positive probability
that some seller will name the monopoly price. Vives(1986) shows
that the supports of the mixed strategy equilibrium do converge to
the competitive price as the number of firms gets large under the
value queue (which he calls a surplus maximizing rule for reasons
not clear to me).
Bertrand-Edgeworth Duopoly Model
The following model represents a simple version of the
Bertrand-Edgeworth game that is similar to those presented in
recent studies of capacity-constrained price-setting firms.
There are two firms and no potential market entrants. Firm
i has capacity ki and kl~k2>0. The total market capacity is
k-kl +k2 . Each firm can produce the same good at the same constant
unit cost, c~O up to its capacity. Given the prices of all other
goods, the aggregate demand is represented by the mapping D:R+~R+.
Let p denote the excess of price over unit cost and let S-[-c,oo].
Define d:S~R+ by d(p)-D(p+c); d(p) is the aggregate demand for the
good when its price exceeds the unit cost by p.
The function d is assumed to be continuous and decreasing
on (-c,PO) where a PO exists such that d(p)-O for P~PO>O and d(p»O
for P<PO' This implies that profit, pd(p) attains a maximum on S.
41
Customers will try, as far as possible to buy the product
at the lowest price. If Pi<Pj' customers buy from firm i until its
supply (k.) is exhausted, then turn to firm j. Assume that when 1
sellers match prices, demand will be divided according to each
seller's capacity. Now we may represent the profit of firm i when
it sets price PiES and firm j sets PjES by
I Li(Pi,ki) if Pi<Pj
hi(Pi,Pj,ki,kj)- ~i(p,ki,k)if Pi-Pj-P
Mi(Pi,Pj,ki,kj)if Pi>Pj
(3.1)
The function Li gives firm i's profit when i has the lowest price.
In this case, firm i will either exhaust all demand at Pi or it
will sell its capacity, ki . Thus
(3.2)
The function ~i represents firm i's profit when i and j match
price. If the price, p, that both firms choose is such that d(p)~k
then both firms will sellout their respective capacities.
Otherwise each will sell some proportion of its capacity. Thus
Firm i's profit in the event that it chooses a price that is higher
than Pj is given by Mi' The nature of Mi depends on the queuing
rule that is assumed. The queuing rule controls the composition of
42
the residual left after the low price firm has sold out its
capacity. Since the possibility of entry is excluded, the high
price seller is in essence a monopolist over the residual demand.
If the residual demand is rich in high-reservation-price buyers,
the firm's profit opportunities are much greater than if only low-
reservation-price buyers are left.
Depending on the relationship of aggregate capacity(k) to
demand, we can identify three regions of interest in terms of Nash
equilibria.
(a) a If aggregate capacity(k ) is less than or equal to the
capacity that a profit-maximizing monopolist would choose, given
demand, then there is a pure strategy Nash equilibrium with both
firms choosing pm, the monopoly price. (For some queuing rules,
this type of pure strategy equilibrium may be supported by
a aggregate capacities larger than k ).
b (b) There is aggregate capacity(k ) and individual capacities k i
and kj
such that both firms will find the residual demand so thin
that maximizing profit over the residual demand is less profitable
than matching prices. The pure strategy equilibrium will be at the
efficient price p-unit cost.
(c) The third region corresponds with aggregate capacity between
a b k and k. In this region, there is no pure strategy Nash
equilibrium. However, a mixed strategy Nash equilibrium exists and
can be determined in many cases. The degree of difficulty
43
associated with solution for the equilibrium distribution of prices
depends on the queuing rule chosen.
Price
d(p)
Fig. 10. Residual Demand of Random and Value Queues
Queueing Rules and Nash Equilibria
The choice of queuing rule affects a seller's profit in the
event that he/she has chosen the highest price. To see this, I
will describe the M.(p.,Pj,k.,k.) function for the value queue and 1 1 1 J
random queue assumption and show how the M. function effects the 1
nature of the Nash equilibrium.
Under the value queue, the highest reservation-price units
go to the lowest-price seller. The residual demand left to firm i
in the event that firm j offered a lower price is given by d(p.)-k. 1 J
44
~nd is represented by GO for the case of linear demand shown in
Figure 10. Note that under this queuing rule assumption, sometimes
called the parallel rationing rule (Maskin, 1986b), the residual
demand curve is determined by a parallel shift of the original
market demand curve inward by kj
(the low price firm's capacity)
Therefore, the high price firm's profit function depends on its
own price choice and the other firm's capacity, which is
parametric. The high price firm's sales will range from 0 to it's
capacity, ki' and will be determined by the residual de manu
function between these two restrictions, thus,
Figure 11 illustrates a duopolist's three possible profit
(3.4)
functions described by (3.1) using the value queue for the case of
linear demand (q-A-p) and symmetric capacities A/2. The linear
portion of Li corresponds with those prices, Pi' such that i sells
out its capacity. The curved portion of Li pertains to the case
where prices are higher than P(ki ). In this case, d(Pi)<ki so firm
i completely satisfies demand leaving no residual buyers for the
higher price seller. Notice that all three coincide at P(k). P(k)
is the price at which demand will accomodate both sellers at their
chosen capacities, kl and k2 '" !.~re.}s no excess capacity at p (k)
nor is there unsatisfied demand.
45
Pro fit (J)
1 2 P(k.) 3 5
Firm i's Price
Fig. 11. Profit Possibilities Under the Value Queue Assumption
Let ~ be the set of cumulative distribution functions in S.
Define Fjf~ such that Fj(p)-prob(pj~p), Define ~ as min supp
Fj(p). This implies that Fj(~)-prob(pj~~)-O. Also define p as max
supp of Fj(p), so Fj(p)-l. The expected profit of firm i for any
price PifS given that firm j chooses prices according to the
distribution Fj is
v Pi v v hi (Pi,Fj ) - J~ Mi (Pi)dFj(Pi ) - Mi (Pi)Qj(P i ) + ~i(Pi)Qj(Pi)
+ fP L. (Pi)dF. (Pi) Pi ~ J
(3.5)
where Qj(Pi ) is the size of the atom (if any) in Fj at Pi' Under
46
v the value queue assumption, the Mi function may be factored out of
the integral since Pj is not an argument of Miv.
v v fPi v fP hi (Pi,F.)=Mi (p.) dFj(Pi)+a·(Pi)[~·-Mi (p.)]+L.(p.) dF.(p.) J 1 E J 1 1 1 1 p. J 1
1
v v - M. (p.)F.(p.) + Qj[~.-M. (p.)]+L.(p.)[l-F.(p.)]
1 1J 1 111 11 J 1 (3.6)
For each capacity pair (k1 ,k2), we can study the price setting game
H(k1 ,k2) in which the strategy set for each firm is ~ and the
payoff function for firm i is h .. A Nash equilibrium may now be 1
defined for H(k1 ,k2) as the pair of distributions (F1,F2)E~x~ such
that for i-1,2
h.(F,F.) :S h.(F.,F.) for all Fd 1 J 1 1 J
(3.7)
The method of solving for the Nash equilibrium relies on the
property of the equilibrium that every price in the support of F. 1
must yield the same expected profit given F .. If some prices J
yielded a higher expected profit than others then (Fl ,F2) would not
be an equilibrium because i could play a strategy which used only
the prices that yield the higher expected profit and reap a larger
payoff. This characteristic is used to "back out" the mixed
strategy solution to the Bertrand-Edgeworth game with value queue.
Since M. is not affected by p., we can identify the maximum of M. 1 J 1,
the maximizer, Pi*' and the associated profit, ITi*' p.* is the max 1
supp of F. since i knows that its opponent will undercut any price 1
greater than Pi* with probability 1. The min supp of Fi is the
price E such that
47
Li(E)-kiE-ni*-hi*(Fi,Fj)
Any price lower than E insures a payoff lower th: '~i*. If Fi and
Fj are atomless, direct solution for F is possible by setting
hi(Pi,Fj)-Eki ·
pki - hi(Pi,Fj ) - Mi(Pi)Fj(Pi) + Li(Pi) [l-Fj (Pi»)
- Li(Pi ) + Fj(Pi ) [Mi(Pi)-Li(Pi »)
- Pi(ki + Fj(Pi)[d(Pi)-kj-ki)
ki(Pi)-P F.(p ) -J i ki+kj-d(Pi)
Solution of the B-E game with random queue is not so
(3.9)
straight forward. The random queue is generated by letting the
fraction of consumers left to the higher priced firm be a random
sample of the consuming population. The expected sales by the
higher-price firm are given by d(P i ) [l-kj /d(Pj ») for Pi>P j .
Refering to Figure 10, the possibilities resulting from a random
assignment of buyers to the queue could range from GD to HF. The
expected residual demand is HD under the random queue.
Firm i's profit given that Pi>Pj under the random queue is
(4.0)
Firm i's expected profit under the random queue is represented by
the following equation.
+ IP L. (p. ) dFj (p . ) Pi ~ ~ ~
functions. r Several Mi
r that the maximum of Mi
curves are drawn for different Pj's.
m m is p for all PjE[P(k),p]. The mixed
48
(4.1)
Note
strategy equilibrium is obtained by solving a set of differential
equations. Davidson and Deneckere(1986) and Beckman(1965) provide
such solutions in each of their models.
Proflt(J)
P(k) 1 2 Pm 3 4 6
Firm i's Price Fig. 12. Profit Possibilities under the Random Queue Assumption
Davidson and Deneckere(1986) assert that the random queue
leads to the best contingent demand curve for firm j. This is true
only when we have a population of identical consumers. If we
49
consider the interpretation of aggregate demand as the summation of
the demands of heterogeneous consumers, then the "best possible"
residual demand curve would be that which allows the low price firm
to sell k. but leaves the maximum number of high value buyers to 1
firm j. (HF in Figure 10)
Davidson and Deneckere(1986) prove that the Cournot-Nash
equilibrium found by Kreps and Scheinkman(1983) is not robust to
any other queuing rule. Krep~ and Scheinkman found a pure strategy
equilibrium in the price stage of a two stage game when Cournot
capacities were selected in the first stage. Davidson and
Deneckere(1986) discuss the same game using a random queue and find
Nash equilibrium in mixed strategies only.
F(p) 1.2
Value Queue
,,/!-~~ Random Queue
1.0
0.8
0.6
0.4
0.2
;I' ;1<"
I
.12 .2 .3
Price .4
Fig. 13. Mixed Strategy Distributions Under the Random and Value Queue (capacity - 1/2)
.5
50
The price distributions in Figure 13 serve to illustrate
the difference in the Nash prediction due to the different queuing
rules assumed. The price distribution functions are the mixed
strategy Nash predictions for the case of constant marginal costs,
linear demand, and symmetric capacities of one half the market
demand at price equal to marginal cost. Firms in the value queue
environment will choose prices equal to or below the min supp of
the random queue distribution with a probability of 0.8. The
experiments that I report desc~ibe a model in which a pure strategy
Nash equilibrium exists when the value queue is used but the Nash
equilibrium exists only in mixed strategies under the random queue.
Experiments and the Nash Equilibrium
Ketchum, Smith and Williams(1984) and Kruse, Rassenti,
Reynolds and Smith(1987) provide initial results as to the
predictive strength of the Nash equilibrium in a laboratory
environment. Ketchum, Smith and Williams(1984) examine two designs
of posted offer markets. In design I, the competitive equilibrium
price is also a pure strategy Nash equilibrium. The competitive
equilibrium price in design II experiments is not a Nash
equilibrium in pure strategies. The resulting market outcomes are
significantly different thus lending support to the Nash
equilibrium concept. Kruse, Rassenti, Reynolds and Smith(1987)
test a four-seller Bertrand-Edgeworth model that is of sufficient
simplicity that an explicit solution of the model is possible.
Kruse, Rassenti, Reynolds and Smith(1987) reject the exact
prediction of the risk-neutral mixed price strategy Nash
equilibrium. Using a Komogorov-Smirnoff test, they reject the
hypothesis that the actual price distribution matches the
theoretically predicted mixed strategy price distribution.
51
However, the Nash prediction does provide the correct qualitative
ranking of the price distributions for the capacities parameterized
and predicts better than all alternative hypotheses. The
experiments I report in the following pages represent a design in
which sellers have U-shaped average cost curves. Due to this
richer design, and discreteness of the parameterization, I can
identify the pure strategy Nash equilibrium for the value queue,
and the min supp and max supp of the mixed strategy Nash
equilibrium for the random queue. The parameterization does not
provide for a tractable solution for the form of the mixed
strategy price distribution.
Duopoly Queuing Rule Experiments
The purpose of the experiments reported in this section is
to determine the effect of buyer queuing rules on market outcomes.
Eight experiments were conducted involving two sellers with
identical cost structures trading in the Plato Posted Offer
Institution with simulated demand. Figure 14 illustrates the
theoretical model I test. Each seller has a U-shaped average cost
curve with minimum at X*. Demand will just accommodate two sellers
52
at the efficient price(P ) that corresponds with AC(X*). Thus the c
efficient outcome involves both sellers producing X* and receiving
P for their output. c
$/unlt
I I
... AC(X)
Pc
Quantity
Fig. 14. Efficient Duopoly Market Price
The laboratory interpretation of the theoretical model is
shown in Figure 15. Each seller has a discrete units version of a
"U-shaped" average cost curve, with minimum at four units. The
market clears with each firm selling 4 units at a price up to $0.15
greater than minimum average cost [ACi (4)]. This yields each
seller a profit of $0.60 per period. In treatment I, fictitious
buyers are randomly queued. Treatment group II buyers are queued
from highest to lowest values (value queue).
$/unlt In deviation from AC(4) $2.5
$2.0 .__--....
$1.5
$1.0
$0.0 ................ II~T?~~
ACl(QI)
1 2 3 4 ! 8
Quantity 7 8
Fig. 15. Average Cost and Demand Curve
53
9
Each subject seller is given a table for calculating total
and average costs from the marginal cost schedule given on the
Plato screen. Subjects are told that demand will be simulated
and, in the event of matched prices, each subject will sell about
half of the quantity demanded at that price. A copy of the
instruction sheet is contained in Appendix B.
Entry costs for additional sellers are assumed to be
infinite. Consequently the duopolists operate in a market with
zero threat of further entry. Each experiment lasts twenty trading
periods, but the end period is not announced. Under the
competitive prediction and assuming zero opportunity cost of
subject time, sellers would cut price to the minimum of average
54
cost, each selling 4 units. The competitive price, quantity and
profit for this experimental design are [$0.00, 8, $0.00].
Collusion by sellers to maximize joint profits would result
in a shared monopoly outcome. A sophisticated version of the joint
profit maximum would result if the two sellers take turns acting as
a monopolist and charge a price of $1.15. This enables the firm
with the low price of $1.15 to sell 4 units and earn a profit of
$4.60. Sellers would have to coordinate a high and low price
pattern to achieve this outcome which yields average profit of
$2.30 per period. A high degree of cooperation is necessary to
achieve this result. Sellers must willingly shut themselves
completely out of the market every other period.
Profit
I 4.6
. '
.15
. ~ ' .
. ' ... L/(pl) .. . - .
\ \
Ir' \ • .' " I . -
\
I~ - _ .... _ .... ~,
,001 ~ . ~. 1111) III) I)))) I) III); 1111; III!!!! I! I!!!! i!!!!!!!!!! 'L:...l
Fig. 16.
p m
Price Contingent Profit Functions for Value Queue Treatment
55
A less sophisticated shared monopoly would result if
sellers match prices at $0.90 and split the market, with an
expected profit of $1.38 per period. The "jaggedness" of the
profit function makes $0.90 the collusive matched price profit
maximum. All of these solutions require cooperation since either
seller can increase profit by undercutting.
So far all hypothetical outcomes result in a uniform price
for this homogeneous product. Therefore the ordering of buyers has
no effect on the predictions. However, if we look at the profit
outcome of one seller's price decision contingent upon the rival's
decision, other strategies become apparent. The contingent profit
functions L. and M. for each seller in the experimental duopoly ~ ~
are illustrated in Figure 16 for the value queue. Notice that the
maximum of the M. function is at the price, $0.15, such that ~
M.($0.15) - L.($0.15). Thus, the Bertrand-Edgeworth Model predicts ~ ~
the pure strategy Nash equilibrium price of $0.15. The price
reaction functions for both firms are illustrated in Figure 17.
The pure strategy Nash equilibrium of $0.15 can be identified as
the intersection of reaction functions.
Firm J'8 Prloe $2.0
$US
$1.0
$0.5
.15
1'-- --- ____ .-&.--..
$0.0 0.15
--------~--------~----------~-----------~
0.5 1 1.6 2 2.5 Firm its Price
Fig. 17. Price Reaction Functions Under the Value Queue
The price reaction functions and contingent profit
functions for the random queue are shown in Figures 18 and 19
56
respectively. Notice that the reaction functions in Figure 18 do
not intersect so no pure strategy equilibrium exists at any price
under the random queue. When its opponent chooses a price of
$0.22, either firm's best reply is to choose the monopoly price,
$1.15. The prices $0.22 and $1.15 form the minimum and maximum in
the support of the equilibrium mixed strategy distribution for
this treatment. The equi1ibrilli~ profit is $0.86.
52.0
$1.5
$1.0
50.6
.22
57
Firm 1'8 Price
$O.O~--------~-------+--~--------~--------~~--------~
o .22 0.6 1 1.6 2 2.6
Firm i's Price
Fig. 18. Price Reaction Functions Under the Random Queue
The Nash equilibrium predicts that value queue treatment
experiments should reach a stahle market price of $0.15. In the
random queue experiments, th~re should be TIQ observations below
$0.22 and more price dispersion. Table 3 summarizes the
theoretical predictions for this study.
Table 3. Theoretical Pr.edictions
Price Quantity Profiti/period Efficiency
Competitive $0.00 8 $0.00 100%
Collusive Sophisticated $0.15 4 $2.30 74% Matched Price $0.90 5 $1. 38 64%
Nash Value Queue $0.15 8 $0.60 100% Random Queue [$1.15,$0.22] [4,7] $0.86 [50%,98%]
Profit
4.6
. LI(pJ)
.....•
Pc a Pm
Price Fig. 19. Contingent Profit Functions for Random Queue
Results and Discussion
The results in this section are based on eight
experiments: four posted offer duopoly markets with a random queue,
and four posted offer duopoly markets with a value queue. Contract
prices are plotted as solid circles and unaccepted offers are
denoted by open circles. Two random queue duopoly experiments are
shown in Figures 20 and 21. In P245 illustrated in figure 20, two
sellers actively compete until period 5. Seller A then offers a
price $1.00 higher than seller B (overshooting what would be the
profit maximum). This market shows no signs of converging to
either the cooperative or the competicive solution. In Figure 21,
prices appear to be cycling and come back up in response to price
signals in periods 4, 5 and 8. Note that each time prices reached
about $0.20 a signal results.
59
Price
I ;;t 3 ~ 5 0 7 ~ 9 10 1/ I~ 13 1'1 /E" lIP 17 It Iq dO 0
0 3.00
Pm • • • •• •• i-- -- -- - - ~ 1-- - -- - - f- - - - -- f-- - - I- -
•••• • ~ ... •
•• •• . ... "
, .. ••• .... • ••• '" •• , .. - OM ~ .. :e ... •••• 1-' .- .... . ...
~ ...... ..... •••• ~ ~-.... -Pe -- -- - f--- i-- - - - - - - -- f-- -- - -- - I- - --
I-- -- f-- -,- - - - 1-- -.-1-- - - 1-- - -- 5 - 1-- -6 7 7 7 5 4 5 6 6 6 5 7 7 4 6 6 5 7 6
Fig. 20. P245 Random Queue Duopoly Price
I :l 3 'f 5 ~ 7 g 9 10 II I~ 13 11 15 110 17 18 1'1 OXJ 0
•
Pm , 1-- - r--- - - -- -- - -- -- - - 1---- I- - 1-- -- -- -..
•• .. .... .... .. , ... u • . ... . .. •• .. ~ ... ~ .. -- ... - 1-" ... ... ... ~ . .... .. ... .... ... .. . . ... ... ~
t!!~- fvu ~.!... -- -- ---- -- -- - - -- - i- - - - - - - -Pe --- - - t- - -- - I- --- - --'- 1-- - - - -- - - r-- -- -
5 7 7 5 4 7 7 5 6 7 6 7 6 7 5 7 7 7 7 7
Fig. 21. P248 Random Queue Duopoly
60
Figure 22 charts the results of a duopoly experiment under
the value queue. This experiment shows the strongest and fastest
convergence to the Nash equilibrium of $0.15 with 8 units traded.
Price
J O!.. j 1 .s ~ 7 g q 10 II IJ. I) '/'1 15 II:; /7 Ig /9 .;1.0
0
Pm
1-- -- -- -- -- -- ---- f-- - -- ---- - - -- -- - -- ---
.... H"--- . .. ~.~~ H _":. .. -- ._- Ht, - ... -.. _ ..... .... ~ .. -- ~'H _,NN I- tIH-- ._. -'-- -- - - -- -- - - -- i- f-- -- - - - - -- -- -- -- ---5 7 7 7 7 7 7 8 8 5 8 8 4 8 5 7 8 8 8 8
Fig. 22. P249 Value Queue Duopoly
The pooled mean prices under the treatments of a random and
value queue are ~~ported in Figure 23. In comparing the results
with the theoretical predictions, we can reject the hypothesis that
the mean price is at the zero-opportunity-cost competitive price of
$0.00 for both the random queue and value queue with t statistics
of 35.17 and 14.43 respectively. Likewise, we can reject the $1.15
sophisticated collusive outcome with t statistics on the order of
-46.52 and -38.57. Also, the collusive matched price joint profit
maximum prediction of $0.90 can be rejected with t statistics of
-28.76 and -27.05 for the random and value queue respectively.
61
PrIce In devIatIon from AC(4)
$08
1 Q
$0.6 [
$0.4
$0.2
$O.O~------~------~--------~-------~------~
o 6 10 16 20 Trading Period
Fig. 23. Mean Prices with 95% Confidence Interval for Random and Value Queues
What is apparent in the figures is borne out by
statistical tests. Using the Wilcoxon nonparametric test for
25
matched pairs, we can reject the hypothesis that the prices for the
treatments came from the same distribution at the 1% significance
level. A t test for dependent samples yields a test statistic of
7.92, again allowing rejection of the hypothesis that the mean
prices are equal in favor of the alternate hypothesis that ranqom
queue duopoly mean prices are greater than value queue duopoly mean
prices at the 1% level of significance.
62
1.0
..---------................ t •• ~-:- ~ _ •• -...,. ~ ..
r----'/ 0.8
................................ ;. 0.8 ..... .,. ,
/ ./ ... _ ..... . 0.4
/ 0.2 ...... / ................................. / .. .
/ ... .. o.o~/~'-L~L-~~--L-~~--L--L~--~~--L-~~--~~~--~~
.15 .5 1.0 1.1 5
Prices Value Queue Random Queue
Fig. 24. Actual Distribution of Prices, Random and Value Queue
Figure 24 shows the distribution of price offers for
periods 16 through 20 in the four random queue and four value queue
experiments. As expected, the value queue experiments have a
greater incidence of low price offers. The random queue prices are
more dispersed and tend to be higher. In fact of the 80
ob:;;ervations from four twenty-period random queue duopoli.es. there
are llQ prices observed at $0.15 or below.
The frequencies of observations from all periods is
illustrated in Figure 25. The median price for value queue
experiments is $0.15, with the p.d.f. skewed to the right. The
median price of the random queue experiments is $0.50. More weight
63
is placed on the monopoly price in the random queue experiments as
would be expected under the Nash prediction. The mean income per
period in the value queue treatment is $.636 which is not
significantly different from the Nash prediction of $0.60. Random
queue sellers did better than the mixed strategy Nash equilibrium
predicts with average profit of $1.03 per period.
Frequency 0.30-
0.26 -
0.20-
0.16 -
0.10 -
0.05 - L. 0.00--0 .1 .2
.. Value Queue . g Random Queue
J Ie -- ~.J .3 .4 .5 .6 .7 .8 .9 1.0 1.1 1.2
Price
111m] ~
1.3 1.4
Fig. 25. Density of Price Observations, Value and Random Queue
This set of experiments supports the conclusion that the
order of buyers does have a significant effect on market outcomes.
Random queuing of buyers substantially changes the structure of the
Nash prediction. Under a value queue, two of the four experiments
attained the Nash equilibrium of [$0.15, 8, $0.601 with the
64
remaining two outcomes approaching the Nash equilibrium. The
random queue group shows higher price variability and no trading
within the $0.00 to $0.15 range. Mean prices tended to be higher,
although no collusive arrangements are detected.
Recall that these experiments characterize a market that
has the potential for an efficient outcome in which eight units are
traded at a normalized price of up to $0.15. The experiments
reported in this section show that the efficient outcome is not
robust under different buyer queuing 'rules in a duopoly setting.
Conclusions
The accumulation of evidence in this chapter shows that the
impact of different queuing rule assumptions in Bertrand-Edgeworth
models is not trivial. The value queue has often been adopted as a
mathematical convenience in numerous Bertrand-Edgeworth models
including a high proportion of the theoretical pieces cited here.
The theoretical results from these models do not generalize to more
believeable queueing rules. These experiments show the magnitude
by which market outcome can change with a change in the ordering of
buyers.
The Microeconomic constructs that must be used tv generate
the queuing rules studied in this chapter are unsatisfactorily
restrictive. A fruitful avenue for future research is to explore
the equilibria that arise from the value queue and it's mirror
image, the "best" queue. These form the boundaries of all possible
65
buyer arrangements, thus could be used to define the limits of
equilibrium outcomes. Using this approach, we can identIfy those
capacity/demand configurations whose equilibria are robust to
changes in the queuing rule and those that are not. When the
queuing rule does make a difference, the range of equilibrium
outcomes can then be fully described.
The strength of the unique Nash prediction continues to
amass evidence in it's favor. Future directions for further
research include tests of· models in which parameterization permits
the pure strategy uniform price prediction, the pure strategy
nonuniform price prediction, and the mixed strategy prediction.
Such an experimental model will provide a strong test of Nash
theory's ability to organize market outcome.
CHAPTER 4
CONTESTABILITY AND EXCESS CAPACITY IN BERTRAND-EDGEWORTH OLIGOPOLY
My goals in this fourth chapter are: 1) to extend the tests
of the Contestable Market Hypothesis initiated in Chapter 2 to the
case of a contestable natural duopoly, 2) to show that my
contestable duopoly experiments may also be interpreted as the
excess capacity case of the Bertrand-Edgeworth model, 3) show the
common ground occupied by both theories and 4) to report
experiments that separate the effect of excess capacity from
experimental results due to an increase in the number of sellers.
Contestable Natural Duopoly
William Baumol(1982) describes contestable market
equilibria in which the endogenously determined number of sellers
could be 2, 3, etc, " ... suppose two firms can produce the output
vector at a total cost lower than it can be done by one firm or
three firms or sixty or six thousand. Then we say that for the
given output vector, the industry is a natural duopoly." The two
firms in a contestable natural duopoly are forced to price
efficiently due to the presence of potential entrants that will
capitalize on any profit opportunity afforded them by the market
incumbents. Using Baumol's arguments, we can fully describe a
contestable market equilibrium for the case illustrated in Figures
66
67
14 and 15 of Chapter 3 when there are more than two potential
competitors. 'TIle contestable Natural Duopoly prediction
(irrespective of queuing rule assumption) in terms of number of
active sellers, market price, and profit is [2, Pc' 0].
Sharkey(1982) proposes two games as interpretations of a
contestable market. In the first game of n firms with U-shaped
average cost curves, he describes the payoff functions, hi' in the
following way:
For the vector P-(Pl' .... 'Pn) of prices stated by all firms,
hi<P) - 0 if Pi>Pj for some j-l, .. ,n
hi(P) - Pi D(Pi)/k - C[D(Pi)/k] if p ~~; for all j-l, ... ,n where ~is the number of firms choosing the lowest price.
Sharkey finds·that his game generally does not have a pure strategy
Nash equilibrium. This is due to the way in which total demand is
equally shared by all the lowest price firms.
Sharkey then proposes a second game which he calls a
Bertrand-Nash model in which firms are only willing to supply those
outputs which yield at least zero profit. This capacity constraint
also opens the possibility that firms will sell at different (low)
prices. Sharkey also found it necessary to queue the firms based
on some priority ranking. In the case of matched prices,
customers, after a complete search, could purchase only from the
lowest price firm in the front of the queue until its supply is
exhausted. Sharkey's queue of firms, which takes his model closer
68
to previous contestability models serves the same purpose as Panzar
and Willig's(l977) assumption that entrants must undercut the
incumbent by at least £ to gain customers.
Utilizing his second model. Sharkey(l982) proves a set of
propositions for the natural monopoly case. His arguments
generalize to the natural duopoly case.
Proposition 1. If P is an equilibrium price vector for a
sustainable natural duopoly and ~3. then ~i(P)-O for i-l •...• n.
A criticism of Sharkey's characterization of equilibrium is that it
requires at least two firms choose prices for the natural monopoly
case. three firms for the natural duopoly case. and so on. This
does not coincide with the contestable market that Baumol describes
but does lead us toward the excess capacity case of Bertrand
Edgeworth competition.
Bertrand-Edgeworth and Excess Capacity
Levitan and Shubik(1972) and Beckmann(1965) characterize the
variations in Nash pure and mixed strategy predictions as the
symmetric capacities of two firms vary. A common result (though
they use different queuing rules) is that when firms have
sufficient capacity to shut their competitor out of the market at
the efficient price. the Bertrand pure strategy equilibrium (p-c)
emerges. Figure 26 shows the predicted Nash Equilibrium price
distributions under the value queue assumption for capacities
69
ranging from k-A to k-A/3.(where A is the quantity demanded at the
efficient price)
F(p) 1.2
1.0
0.8
0.6
o.~
0.2
o ,1
/ Value Queue A8&umed
~-- .-----;--- ,,-
,- I .'
/ / ;_k-.5A
.2 .3 .4
Price
Fig. 26. Nash Price Distribution for Different Capacity Levels
Osborne and Pitchik(l986) study comparative statics for
capacity variations by one or both firms. Restricting themselves
to the value queue and constant marginal costs, they completely
characterize the Nash equilibrium predictions for a full range of
capacity levels(not necessarily symmetric) and very general demanU
condi tions . They show that the endpoints of the suppo'rts of the
equilibrium strategies are nonincreasing in the capacity of the
larger firm. The endpoints of the supports of equilibrium
strategies increase as the capacity of the smaller firm decreases .
•
70
Quite intuitively, the equilibrium approaches the monopoly outcome
as the smaller firm's capacity converges to zero.
Shubik(1959) shows that as the number of firms is increased
in a Bertrand-Edgeworth market, if ~i~ (where a is defined as
demand at price equal to minimum average cost) and ki~a/n, then the
value of the game approaches that of the competitive equilibrium.
Also, the probability distribution of prices shifts toward the
lower end of the range as n increases. Allen and Hellwig(1984)
show that for a random queue, as the number of firms increases,
Nash equilibria (in pure or mixed strategies) converge in
distribution to a (perfectly) competitive price. However the
supports of the mixed strategy distribution do not converge with
positive weight on the monopoly price even as n~. Vives(1986)
shows that as the number of sellers increase, the supports of the
equilibrium strategies do converge to the unique competitive price
if the value queue is assumed.
My interest is not in the number of firms in the carket but
rather in the level of excess capacity. The papers previously
cited do provide some predictions. For the value queue,
Vives(1986) proves the following proposition for n fir.ms with
symmetric capacities, k , and constant costs. n
Proposition 2.Given n firms in the market and restricting attention
to symmetric equilibria, if k ~a/(n-l) then each firm charges a n
zero price.
This result follows from the fact that firm i can be completely
shut out of the market when the other n-l firms choose a lower
price and thus are capable of supplying the market quantity Q.
71
When we plot the Mi(Pi'~i~j kj ) function, it is a horizontal line
at the zero profit level so the maximum(not unique) will coincide
with P(K) which(in this case) is zero. For the random queue, given
sufficient excess capacity that the highest price firm is thrown
out, the same arguments can be used to predict the Nash equilibrium
price of zero.
Excess Capacity in Experiments
The preceding discussion underscores the fact that a large
proportion of the experiments reported by Coursey, Isaac and
Smith(1983), Coursey, Isaac, Luke and Smitp(1984), Harrison
McKee(1984) and Harrison(1985) cau be interpreted as tests of
Bertrand-Edgeworth competition when there is sufficient capacity to
shut one or more sellers out of the market. Recall that the bulk
of these experiments conform with the zero economic profit
prediction.
Kruse, Rassenti, Reynolds and Smith(1987) report a set of
experiments expressly designed to test the N8.sh prediction in a
Bertrand-Edgeworth(four seller) oligopoly setting. The
relationship of aggregate capacity to demand is a treatment
variable in this study. The population of simulated buyers is held
constant across three levels of aggregate capacity. Treatment I
72
oligopolies have no excess capacity. In fact, there is
insufficient aggregate capacity to satisfy demand at a price equal
to constant marginal cost c. Treatment II experiments test the
case where there is a slight undercapacity at the price equal to c.
Treatment III sellers have an aggregate capacity that is larger
than demand at the competitive price. In this treatment, the
capacity is still not sufficient to shut any seller out of the
market at the competitive price. No pure strategy Nash equilibrium
exists for the capacity-demand relationships characterized by the
three treatments. Kruse, Rassenti, Reynolds and Smith(1987) found
that the risk neutral mixed strategy Nash prediction correctly
organized the qualitative differences in the observed price
distributions arising from the three treatments. As aggregate
capacity increases relative to demand, low prices are observed
with a higher frequency. The experiments I report in this chapter
focus on markets in which there is sufficent excess capacity to
shut the highest price seller out of the market as compared with
markets in which there is no excess or under capacity.
Excess Capacity Experimental Design
Recall the model of the duopoly experiments described in the
previous chapter. lbe treatment operationalized in this study
makes the natural duopoly tested in Chapter 3 "contestable". This
is accomplished by adding an additional potential seller. Instead
of two, there are three potential sellers with identical U-shaped
73
average cost curves. Harket demand will accomodate two sellers at
the competitive price. If all three sellers match price at the
competitive level, they will incur losses. A positive opportunity
cost of entry is induced by providing subjects with a safe haven
that may be chosen instead of the PLATO natural duopoly market.
The return from occupying the safe haven is $0.50 per period.
The experiments reported in this section are parameterized
specifically to reflect the case described by Baumol in which two
firms can provide the output demanded at a total cost lower than
can be done by one or three or sixty firms. The contestable market
hypothesis predicts two firms actively selling at a price equal to
minimum average cost and earning a normal rate of return ($0.50)
each period. The third seller chooses the alternate market in
equilibrium and enters the natural duopoly market only in response
to prices above competitive levels. This prediction is independent
of the queueing rule chosen.
One can identify a sophisticated collusive outcome which
requires that the three sellers take turns charging the monopoly
price of $1.15 in the PLATO market and choosing the alternate
market otherwise. This collusive outcome yields an expected profit
of $1.87 per period. A matched price joint profit maximum yields
an expected profit of $0.70 per period when each seller chooses
$1.15 every period. Of course, cheating on either of these
arrangements yields a higher profit.
Om ..
Pe
74
A qualitative prediction of the Bertrand-Edgeworth model is
that prevailing prices are lower in the presence of excess
capacity. For both the value queue and random queue case, the pure
strategy Nash equilibrium is the same as the contestable market
prediction of [Price, Quantity, Profit]-[$0.125, 8, $0.50]
Price I .;t 3 'f 5 {p 7 8 9 10 1/ J,2 IS 1'1 15 16 17 18 If ~ CJ./ d.o1 .:13 .2'1 $
E E E EE E E E £ E. E £. E E E e. E
- - - 1--- - - -- - I-- - -- - - -- - r- f- -- - - - -- I--- -c ....: .. _ ..
~ , -- - 0
~ -:. ... - ~- -..:. ~ ... _ ... - ~ ~ ~- _ .. be!!: -~ -- I-- -- ~ ~-- - r- --
5- - - -- I-- - - - - -5- 6 "6 f-;.- -.- -7 7 1--
7- -
7 7 7 8 8 4 7 7 7 7 7 7 6 7 7 8 8 7
Fig. 27. P282 Random Queue, Three Sellers, Alternate Market
Results of Three Seller Natural Duopoly Experiments
Eight experiments are reported: four with a random queue and
four with a value queue. Figures 27 and 28 contain representative
experiments for each of the queue treatments in the natural duopoly
design. The random queue natural duopoly shown in Figure 27
attains a relatively stable configuration with one seller in the
alternate market and two sellers in the PLATO market for 11 of the
75
last 12 periods. The prices chosen by the two sellers in the
"duopoly" market hover around $0.15 which is a few cents higher
than the Contestable/Bertrand-Edgeworth prediction. This may be
due to the riskiness of the PLATO market relative to the safe
haven. The value queue contestable natural duopoly experiment
shown in Figure 28 exhibits a stable market configuration at the
market price around $0.125.
Price 1 .:t j 'i ~ lD 7 9 9 10 11 I:J.. 13 1'1 15 16 17 IS Iq ';0 c}'1 ~ ~3 ,).'1 .;IS"
t: E: E E £ E E t: £ E E £ £ £ £ E
c
Pm
-1-- - 1-- -- - - .- -- -- - -I- - - 1-- I- - -- -- - - 1-- -- --
- c • - . -l-
• . . I- ..
.~ .~ - ~~ i-= ~ ..
I-'!!" . -~ I-~ \-'" j;.;'-' \- .' -- I- I-- l- I- ..... J-,-. -00 - 1""""-..... '"" ,.... -Pe
-1-- - -- - 1-- - t- -- \-- - I- --- -- - - 1-- 1-- 1-- 1-- 1-- \-- - -
6 6 7 7 7 7 5 7 7 8 6 6 4 6 7 8 7 7 8 8 8 8 8 7 8
E denotes exit by a seller to the alternate market
Fig. 28. P279 Value Queue, Three Sellers, Alternate Market
Figure 29 shows the frequency of price offers made by all
sellers in the final five periods of the experiments. For both
queues, observations greater than $0.10 but less than or equal to
$0.20 predominate. Notice, however, that more random queue offers
are observed in the higher price ranges. None of the theories
76
predict the observed difference in equilibrium. One factor that
might explain the behavioral difference in markets with identical
equilibrium predictions is that the random queue still yields
higher profits than the value queue for offers above the
equilibrium price.
Frequency 0.8 -
0.6 -
0.4 -
0.2 -
0.0-- L
III Value Queue g Random Queue
~jJi -.mL-EI --0 .1 .2 .3 .4 .6 .6 .7 .8 .9 1.0 1.1 1.2 1.3 1.4
Price
Fig. 29. Density of Price Observations, Three Seller Excess Capacity Case
The pooled mean prices with 95% confidence intervals are
plotted in Figure 30 for the two queuing rules. Clearly, the
laboratory outcomes allow us to reject both of the collusive
predictions. The markets for both queuing rules seem to be
converging toward the Contestable/Bertrand-Edgeworth prediction.
Though the two queueing rules yield results that are more similar
77
for the excess capacity case than when there is no excess
capacity, they still diverge. According to a Wilcoxon
. nonparametric test, we can reject the hypothesis that the mean
prices arose from the same distribution. The t test for dependent
samples yields a t statistic of 5.56. We may reject the hypothesis
that the mean prices from the two treatment groups are the same in
favor of the alternative that the random queue mean prices are
greater than the value queue mean prices.
o Random Queue v Value queue
Price In deviation from AC(4) $1.0
$0.8
$0.6
$0.4
$0.2
$0.0 L.... ___ -L-___ ...L-___ ..l-. ___ .L-__ ---l ___ --'
o 6 10 16 20 26 30
Trading Period Fig. 30. Pooled Mean Prices with 95% Confidence Intervals,
Three Seller Excess Capacity Case
When the random queue contestable du.!:lpoly results are
compared with the random queue duopoly experiments reported in
Chapter 3, the resulting t statistic is 3.19. We may reject the
hypothesis that the mean prices are equal in favor of the alternate
78
hypothesis that the contestable duopoly mean prices are lower than
the (uncontested) duopoly prices.
When we compare the results from uncontested and contestable
natural duopolies under the value buyer queue, we cannot reject
the hypothesis that the prices arose from the same distribution at
the 5% level of significance. The t statistic of 1.56 conforms
with the results of the Wilcoxon nonparametric test.
One alternate explanation of the results reported in this
section has not been clearly eliminated. Introducing another
potential competitor to the natural duopoly market obviously
increases the number of sellers from two to three. A viable
alternate explanation, in concert with the traditional structure
conduct-performance paradigm, is that the number of competitors has
increased and that, in and of itself, would generate a more
competitive solution.
Bertrand-Edgeworth Triopoly Experiments
The structure-conduct-performance paradigm has been a
mainstay in the study of industrial organization. Under this
research program, a strong causal link has been welded between the
number of sellers in a market and the prices that prevail. The
inverse relationship between the number of sellers and industry
wide profit levels is accepted as a stylized fact in this branch of
industrial organization. A vast body of empirical research is
devoted to exploring the Structure-Conduct-Performance linkages in
79
various industries. Empirical tests require quantification of
industry structure using some measure of the number of firms
present and their size relationship. One well accepted summary
measure is the Hirschman-Herfindahl index, often called the
Herfindahl index and given by:
H-~ s 2 i
where si is firm i's quantity share of total market sales. H
varies from 0 to 1 and tends toward 1 as the number of sellers
decrease and/or are more unequal in size. Thus, one would expect a
positive relationship between H and prevailing market price.
If we accept the hypothesis of the Structure-Conduct-
Performance Paradigm, this would serve as a suitable explanation
for the result reported in the previous section. The experiments
reported in this final section are designed to allow separability
between the effect of increasing the number of sellers and the
introduction of excess capacity.
Smith(198l) reports tests of the effect of increasing the
number of sellers from one to two in a posted offer market. To
isolate the effect of increasing the number of sellers, Smith
replicates both sides of the market by cloning each subject's cost
or value parameters. Merket prices fell with an increase in the
number of sellers from one to two. Smith's experiments represent a
pure scale change while holding the capacity/demand relationship
constant.
80
In the final treatment, I make a pure scale change in the
size of the market, keeping per capita income opportunities
constant. To scale the market up from two sellers to three, a
clone, with cost identical to the other two sellers is added to the
supply side of the market. Since all buyers are not identical,
scaling up the demand side of the market requires that new values
be added to the aggregate demand curve which are equal to the
expected value of one half of the original demand curve. This
preserves the profit-to-price relationship that existed in the two-
seller duopoly experiments. Figure 31 illustrates the costs and
values parameterized in this treatment.
$/unlt $3,Or $2.5
$2.0
$1.5
$1.0
$0.5
Demand
1 2 3 4 6 6 7 8 9 10 11 12 13 14 16
Quantity Fig. 31. Average Cost and Demand Curve, Np.tural Triopoly.
81
If sellers match prices at the shared monopoly price of
$1.15. they can each sell two units for a profit of $1.30 per
period. The capacity/demand relationship is such that no pure
strategy Nash equilibrium exists. The supports of the mixed
strategy prediction are the same as in the duopoly case(as per
Allen and Hellwig(1986» but we would expect higher probability
weight on the lower prices in the support. The minimum and maximum
prices in the support are $0.22 and $1.15 respectively. A
significant positive relationship between the Herfindahl index and
market price would lend support to the Structure-Conduct-
Performance Paradigm.
Price
1 c 2 3 '( 5 b 7 8 9 10 /I 12 13 1'( 15 If) 17 lie 19 20
Pm . ,
c .. .... _. - .... c 0 - c - 1--- -.... ........ .. oJ .. oJ _ . -1-.-- ... .... .... -_. l- I-- . .. ....... . .. -...
I-,... .....
1--"- ~ ..... r--Pe
-- --I-- 1-- - -- -- --- I-- -- -- -- -- -- -- I-- t- -- -- -- - -5 8 8 8 7 9 6 8 8 8 8 8 8 8 9 8 9 8 8 8
Fig. 32. P326 Natural Triopoly
82
Figure 32 shows a representative experiment from this
treatment group. The price pattern in this experiment is
reminiscent of those found in the two-seller random queue duopoly
experiments reported in Chapter 3. Comparing the mean prices of
the random queue duopoly and triopoly experiments with no excess
capacity yields a t statistic of .059. We cannot reject the
hypothesis that the mean prices are equal. The Wilcoxon
nonparametric test supports this conclusion.
The distributions of observed price offers for the duopoly
and triopoly with no excess capacity and the natural duopoly with
excess capacity are shown in Figure 33. Notice that the triopoly
prices are qualitatively more like the distribution of prices from
the two-seller natural duopoly experiments than the three-seller
excess capacity experiments. Apparently, excess capacity
influences the "competitiveness" of the market outcome more than
the number of competitors does. Finally, a Herfindahl index was
calculated for each period of the twelve experiments. A mean
Herfindahl index was calculated for each experiment. A linear
regression was calculated using the mean Herfindahl index
as the independent variable and mean market price as the dependent
variable for the twelve experiments. The coefficient on the
Herfindahl index of .199 is not significant with a standard error
of .507 and a t statistic of .392 . 2 The R for the regression is
. 015 indicating that the Herfindahl index has practically no
predictive power in the controlled laboratory setting of these
experiments.
F(p) 1.2 r
I 1.0 _~ ............ ~~~~ ... _. " ... ~".'='~'~ fJ· ~" .
. / e---~' 0.8 .. ...... ........... >/' ",.- . 'r" ... .
.... l-· .-0.6
1
....... /.~~./.. . ... r·-:~.·<" .' III
0. 4 r···· .. /'-"', ,.' .r··· • /'if /' ... .. ... .. ...... . . ..... -.. -- ... -......... - .. ---. - .
0.2 1-.. --1 ..... - ... /- ............ .... -;1-.. -.......... ................. -....................... -.. -...... -................................... -................... _- ...... .
0.0 I. j tJ,' e.=£.~.' .16 .6
Prices 1.0 1.16 1.6
83
. ~.. 3 Firma No Exceaa 3 Firma Exc8ea Cap. -e- 2 Firma No Excesa
Fig. 33. Actual Distribution of Prices, Random Queue
Conclusions
The experiments reported in this chapter served to "close
the loop" initiated in Chapter 2. The presence of potential
entrants or excess capacity facilitates a competitive-like outcome
even in markets with very few sellers(only one or two active
sellers). Contestability/excess capacity is shown to have a
stronger influence than does increasing the number of competitors.
A naive structure-performance linkage that does not include some
measure of excess capacity or potential entrants has little
predictive power.
CHAPTER 5
CONCLUSIONS
The oligopoly models examined and tested in this study were
initially developed in different centuries. Joseph Bertrand's
1883 model of firms that compete through price has been enriched by
the contributions several generations of economists. This
enrichment of Bertrand and eventually Bertrand-Edgeworth models
epitomizes a progressive research program that moves the science
of economics forward. Contestable Market Theory also builds on
ideas that have existed since Adam Smith. However, the credit for
formalization of the Theory of Contestable Markets must go to
Baumo1, Bailey, Panzar, and Willig for their work in this decade.
This study utilizes laboratory testing throughout as a
fundamental tool to facilitate better understanding of economic
phenomenon. The results of those laboratory tests indicate the
strengths and the weaknesses of the theories tested. Contestable
Market theory does have predictive power at least within the realm
in which it has been tested so far. Some of the assumptions
necessary to make Contestabi1ity theory work analytically are not
required for it to work behaviorally. This is demonstrated by the
first set of experiments performed by Coursey, Isaac, and
Smith(1983). My experiments which induce a positive "normal rate
84
85
of return" through an alternate market complement the results that
preceded them. This indicates that the direction of future
research should extend further into the richer cost and product
environments that Contestability Theory covers.
The behavioral significance of active(human) demand vs.
passively revealed demand is not captured by any of the theories I
examine. The results of my experiments clearly indicate that some
equilibria may be quite sensitive to actively strategic buyers.
The competitive equilibrium for the contestable natural monopoly is
a case in point. This indicates that future theoretical research
should formalize the effect of strategic buyers in market models.
I suspect that the characterization of equilibria will change very
little for some market models, but could be vastly different for
others.
Bertrand-Edgeworth Game-theoretic models are sensitive to
the fairly restrictive assumptions presently being made about the
allocation of demand to sellers with different prices. The
experiments reported in Chapter 3 show the magnitude of differences
that are possible. Game theorists should move toward a more
general characterization of demand allocation or at least
recognize the limitations on the models they presently propose.
Further experimental research should explore proposed game
theoretic equilibrium concepts with particular attention to the
Nash Equilibrium.
Excess capacity is an important determinant of market
performance. The limited tests reported in Chapter 4 show that
excess capacity is more important than the number of active
This
86
Market
competitors in pushing markets toward efficient outcomes.
information should be used in prescribing public policy.
mechanisms that utilize free access to excess capacity can
potentially produce efficient results in lieu of regulation.
Future research should move toward developing and testing such
mechanisms.
My contribution to Contestable Market Theory and Bertrand
Edgeworth Oligopoly has been made primarily through the device of
laboratory tests. Those controlled laboratory tests provide the
yardstick by which the efficacy as well the deficiencies of the
the~ries can be measured. My experiments indicate several fruitful
directions that the next stage should take. In a sense, this
final chapter summarizes important results reported in this body of
work but also provides an introduction to the next phase of my own
research program.
88
SELLERS' PROFIT SPACE FOR CHAPTER 1 EXPERIMENTS
Price 2.36 2.11 1.86 1. 61 1.36 1.11 0.86 0.61 0.36 0.11 Range 2.12 1. 87 1.62 1.37 1.12 0.87 0.62 0.37 0.12 -0.13
Q(P) 1 2 3 4 5 6 7 8 9 10
Profit q.-1 1.24 0.99 0.74 0.49 0.24 -0.01 -0.26 -0.51 -0.76 -1.01
1 1.00 0.75 0.50 0.25 0.00 -0.25 -0.50 -0.75 -1.00 -1.25
q -2 2.23 1. 73 1.23 0.73 0.23 -0.27 -0.77 -1. 27 -1. 77 i 1. 75 1.25 0.75 0.25 -0.25 -0.75 -1. 25 -1. 75 -2.25
q.-3 2.97 2.22 1.47 0.72 -0.03 -0.79 -1.53 -2.28 1 2.25 1. 50 0.75 0.00 -0.75 -1.50 -2.25 -3.00
q.-4 3.46 2.46 1.46 0.46 -0.54 -1.54 -2.54 1 2.50 1.50 0.50 -0.50 -1.50 -2.50 -3.50
q.-5 3.70 2.45 1.20 -0.05 -1.30 -2.55 1 2.50 1.25 0.00 -1. 25 -2.50 -3.75
q .... 6 3.69 2.19 0.69 -0.81 -2.31 1 2.25 0.75 -0.75 -2.25 -3.75
q -7 3.43 1.68 -0.07 -1. 82 i 1. 75 0.00 -1. 75 -3.50
q -8 2.92 0.92 -1.08 i 1.00 -1.00 -3.00
q -9 2.16 -0.09 i 0.00 -2.25
q -10 1.15 i -1. 25
89
SELLERS' PROFIT SPACE FOR CHAPTER 2 EXPERIMENTS
Price 1. 90 1.65 1.40 1.15 0.90 0.65 0.40 0.15 Range 1.66 1.41 1.16 0.91 0.66 _0.41 0.16 -0.09
Q(P) 1 2 3 4 5 6 7 8 Profit q -1 1.15 0.90 0.65 0.40 0.15 -0.10 -0.35 -0.60 i 0.91 0.66 0.41 0.16 -0.09 -0.34 -0.59 -0.84
q.-2 2.30 1.80 1.30 0.80 0.30 -0.20 -0.70 1 1.82 1.32 0.82 0.32 -0.18 -0.68 -1.18
q -3 3.45 2.70 1. 95 1.20 0.45 -0.30 i 2.73 1.98 1.23 0.48 -0.27 -1.02
q.-=4 4.60 3.60 2.60 0.60 -0.40 1 3.64 2.64 1.64 0.64 -0.36
q.-5 3.25 2.00 0.75 -0.50 1 2.05 0.80 -0.45 -1.70
91
SELLER PRICE RESTRICTIONS
In this experiment, seller prices have been restricted to those divisible by $0.25. If you enter a price that is not divisible by $0.25, then the machine will automatically round your price to the nearest $0.25. You may either use that price or select a different one.
For Example:
If you choose a price of $1.87, your offer will appear as $1.75. You will then be asked to choose a quantity or you may change you price by simultaneously pressing shift and edit.
If you choose a price of $1.88, your offer will be entered as $2.00 and the same procedure as outlined above follows.
92
ALTERNATE MARKET INSTRUCTIONS
As a seller you may choose to enter an alternate market instead of the one on your screen, for which you will receive $0.50 per period. You may do this by selecting a selling price of 0.00
and quantity of O. At the end of the experiment, you will be paid the gross profit from the record sheet on your screen plus total earnings from the alternate market listed in the table below.
Period
1
2
3
4
5
6
7
8
9
10
Participated in Alternate Market
Total periods in Alternate
Market
________ X $0.50 -
Period
11
12
13
14
15
16
17
18
19
20
Participated in Alternate Market
Total Alternate Market Earnings
# of Units Additional Sold Cost
I
2
3
4
5
SELLER COST TABLE
Total Cost
Per Unit Cost (total cost/# units)
93
IN THIS EXPERIMENT, THE DEMAND FOR YOUR PRODUCT WILL BE SIMULATED. IN THE EVENT THAT YOU MATCH PRICES, YOU WILL EACH SELL APPROXIMATELY 1/2 OF THE TOTAL QUANTITY SOLD IN THAT PERIOD.
$.8,-----------------------,
I
Pm $4 Pc
10 15
Trading Period Cd Seller A, _ Seller 8 Offer
Fig. 34. Experiment P179
20 26
R Alt MH
95
96
$6,----------------------,
$6
Trading Period DSeller A. _ !?eller 8 • (lffer 1( .A.I t Mk t.
Fig. 35. Experiment P18l
97
$8.---------------------------------------~
:~ Pc
$2 r'~ .. ~
$0 1 ~
r n
. ~~ l I
~ IIII1
10 1':- 20
Trading Period :)f !9r
Fig. 36. Experiment P194
;f,6,--------------------,
,f.;6 •
Pm' $4 Pc
$2
. ~. ri . i~. ~I' ~i ! I ! I II . ~ II II
~1 ! II 1 t~ t I I
5
I I I I
10 15 20
Trading Period o S811~r A M:::<?lIer 8 • (lffl?r tt to.1 t MI< t.
Fig. 37. Experiment P206
98
10 115
Trading Period
Fig. 40. Experiment P211
101
I I I I . '. I
, "I ~ I M ~
~ ~ I ~ ~ ~ ~ I ~ ~ R· I t\
102
$8~----------------------------------~
10 15
Trading Period
Fig. 41. Experiment P236
RI I I
I n I : I
II II r I
20
103
:f.8r------------------,
I $6 . I
P . . $4 .
I ~ n
. Pc
~ ~ .
" ~ .
.. I $2 ~ I I . . I ,. I . I t\ I I I
..
III I I I \ \. $0
.
I~ 1·. . . . rn I
~I~I II~I tJ Ii I .
t\i ! II ~n · '5 10 115 20 213
Trading Period
Fig. 42. Experiment P237
PC'ice
1.00
.15
.00
Price
1.00
• 15
.00
J <L J 'I S ~ 7 S q
- - -- - -- - - -- -
-f--I'":": _": b:-: r-~ r-":
- - - - - - - 1--S 7 7 7 7 7 7 e 8
Fig, 43.
1 ~ 3 'I C, . IP 1 II 9 10 10·'"
• - - - -- -- -I- -
- .. r' - .... . .. .... 'M' I- r--..
h' .: -.... - - -- - - -- - - -rS -i--
~ 1-4 - 1-.-5 6 6 7 6 7 7
Fig. 44.
10 /I I:" IJ Ii( 15 Ib 17 Ii
- --- I- - -- - -
'r-' i'--1-" I-- I-r-~
~ 1-- - re- - -..: ---e 8 4 S 7 8 8
P249 Value Queue
1/ I~ 13 1'/ 5 i/: '7 18 19
- - - l- .- - - -- -
.. ... -- ~ 1;-... - ~! ... 1--... . ... ~ .. ~ .. ~ ... - 1-- I-S -1-- 1--
7 7 7 4 5 5 7 8
P250 Value Queue
19
-
.10
---
Trading Period
-I-----8
,;0
-
....... 1--
8
8 Quantity Traded
Trading Period
Quantity TC'aded
104
Price
1.00
. 15
.00
I
.
....
-~
d 3
1- -
----
56
I
'/ S 10 7 s 9 10 /I Id.. /3 I'{ IS 1(" 17
,- - r - - ,- - - - -- - - - -
:--N ... ... ... .... "N ... .... "- .. --... . - _.- .. - ..u . .. .....
- -- -- -- - - - - - - -- -- --- - '7 ~ -,- '5 7-6 6 6 7 7 7 7 7 7 7
Fig. 45. P251 Value Queue
5 10 1\ la 13 11 15 Ii
If 19
,- -
_.-I--- ri-7
d ~o
<)0
-
-:."
-7
Trading Period
105
Quantity Traded
d. 3 Ij I; 1 g q 110 17 Trading Price Period
1.00
.15
.00
-.... _ ..
- -
56
-- - -- -- - - , -- o _
. . ...... " N .... r- I~ 'N"
r--' ... .... -" .- ... 1 - .... ... I- UN r-'
• N ... - ,- - - 1-- - -- I-- - ,- - --I- -- - 1--
65 ~ 66 6 6 :) 6' 6'6 77 5 5 74 6
Fig. 46. P252 Value Queue
Quantity Traded
Price
PM 1.00
.15
.00
I ~ 3
0
- --. . .... --
- -
5 5 5
'I '5 II!
I- 1-- f--I
.... . 1'-" ....
- - -
5 6 6
7 g q 10 11 1;1. 13 1'1 15 110 3.~l
r1 I~'
- - I- - 1-- - .- - .- f-- - 1--
. .... • . . . ... f-~ .... .. ... "'M .... ... ...
io •• : ...
~.
- - - -- -f- -- - I- - -
5 5 6 6 6 6 6 6 6 4 6 6
Fig. 47. P242 Random Queue
,q
-
. ~ . ---7
(}O
-
. _ . -
6
.. \
Trading Period
106
Quantity Traded
Price Trading I ;;t 3 J.j 5 0 7 \? q 10 II IJ.. 13 '" IS lIP 17 18 Iq c)D
.15
.00
-
.. ...
1--
6
- -
.. -.... i-- -
7 7
. - -
...... ", ..
--I--
7 5
• l'o c
- ... -.. .. .. j-. - 1-- - - I- - - - l- . - --....
10'" . .
•• .. • = ", .. ... ....
.~ - .M ..:0 .... 1-' .- ". " ... - --- - - - - -1-- 1-- - I- -I- -
4 5 6 6 6 5 7 7 4 6 5 6 5 7
Fig. 48. P245 Random Queue
-_ .. -
. . .... -
6
Period
Quantity Traded
Pdce
1.00
.15
.00
I .;l 3
-- I-
l-!-. ....
0- -1--
s s~
'i S
-I-
joo •• i.M"
'--
6 6
10 7 K 9 /0 " 1:1. 13 If IS I/' 17 II 19 ",0
- - - - - - - 1--I- --.. .. . .
r" foo .. " joooo' .... !-. .... _ . .. I- - ... .... ....
1--00 '_. 'M'
-I- -I- - -- - t-- - -- -- -I- - -'s '5 6 s 5 6 6 6 5" 6 6 6
1-,
6 66
Fig. 49. P247 Random Queue
Trading Period
Quantity Traded
107
I J. 3 '{ 5 r, 7 g 9 10 II I~ 1,3 i1 15 It" 17 18 I~ l-w Price Trading
1.00
. 15
.00
.
1-"
---s
-
.. ... --
7
. - -
... ... N' -- -
7 ~
(
. . - -- - --- r- - - - - I- - I-..
.. .. .... 00 .. ... '" ... .. I ... oo '00-". - 1-' ... 1-' ... ... I- - - - -- -- - - - - -
-~
.. .. - - .. 4 7 7 6 7 6 7 6 7 S 7 7
Fig. 50. P248 Random Queue
-- -
....... .... - -,- .. -7 7
-
... .. . -
7
Period
Quantity Traded
108
D '71'; ,., ..
Trading Period
,1., .2 J AI S 6 1 8 9 10 " 12 13 /'1 /5 I~ 17 /I IV ':0 :11 ~ 01' ~ .:15 III E E E E
Price 0
1.00 _1_ - - - -'- - -- -f- - - --- - - -- - - -- -- -
.. .. • . . . , ~"
.;. l-I-" .. , ~
1-0- ~ I- !-oM I-,"- ~ I-~ r-" -:..... .
~ ... r-- ", .. ... H' i-': LJ ... I-" I-"
.15
.00 - - -- -I- .- - -I- -f- -I- - f-- -- - - - - - - - --- -,- - -'-- - l- I- - I- -i- I-- - - - -1-- - - -
14 5 5 6 6 6 5 6 6 6 6 6 6 7 6 7 7 7 6 7 6 7 7 7 7
Quantity Traded
Fig. 51. P276 Random Queue, 3 Sellers, Alternate Market
Trading Period 1 J. 3 ., S " 1 8 9 10
£ £ Price
. . I..., -l- I- - -I- ' -1.00
.. ~ ~ I-1-, 1-' ~~ r-- l- ...
.15
.00 f- .- -- --I-
- - .- - .- -f-- -I-4 5 5 5 5 5 5 5 5 5
Quantity Traded
1/ 12. 13 ,., £ £
-- -
r-- 0
"- ,
I-
,- --f-
- f-- - f--5 5 5 6
I~ II, n 18 19 ~o <11 .:101. :!3 ;,'/ ~s
£ E f£ 300.
-- -- - f-- f- 1-- --I-- f-- 1--,
M
f-• ... "- .. _. -'- - .-~-:.. 10-;-- ,;,.- f;..--- - - - - -1-- - -- -- -6 6 7 7 8 4 8 7 8 8 8
Fig. 52. P275 Random Queue, 3 Sellers, Alternate Market
Tradin
Price
1.00
.15
.00
Penod J .:l .3
~
I- ---
. . M
l- I- -'" -- --
I- - -6 6 7
'f
-
..... -
1--7
Quantity Traded
5
1--
~;
-7
(, 7 i 9 10 1/ E £ E EE
--I-- -1---- -. .
.o. io-H.
I-~. _0- ... f.o;"--~ -- - - -- ~- --8 8 6 8 7
IA 13 1'1 /5 I~ J7 IS 19 AO ~ I~A ;/3 ~'1
EEt. E
. - - . .- -1--1-- 1--1-- -- - - -
, . ... . M • ...I -I-~ ~ .. "'" r,..-F-- --I-- - -- .- -- -I-- -1-- - --
8 8 8 4 5 7 7 8 8 7 8 8 8
Fig. 53. P280 Random Queue, 3 Sellers, Alternate Market
Tra
Price
1.00
• IS .00
dlng Period I .:I. 3 'I
E
- - - -,
",.J ... " , ..
1-'"
-- .- -- - - 1--5 6 6 7
Quantlty Traded
S fD
- 1--
M, --1--
1-- -7 7
7 8 9 10 111.3. £ £ E£ E
- - 1-- - l- f-
... M .. ......
i-l r:."" l-- '-- - - t- ·5-7 7 8 8 4
13 /'/ IS 16 17 II " .:!O .;)./ dol ~3 :J.'I E E £ £ E E E. £ E E.
-- - - 1-- - I- ,.... 1-- - - - 1--
~
... . - .. 'M'" I- f=". ~ ...... .- ~~ ...
'.-~ I->- .;;-.~ .or-
- -I- - I- -- I-- - - - -7 7 7 7 7 7 6 7 7 8 8 7
Fig. 54. P282 Random Queue, 3 Sellers, Alternate Market
109
"'5 I:
--
--8
~ E
-
....~
-7
110
Trading Period J :J.. 3 <I 5 " 7
Price q{() E 71XJ 11.00
1.00 - 1-8 - - !- :- -
.15
.00
,.. .. ....
.- --, 44
~ ....
--5
__ .i .... .... ...
-- ,- -- - - l-5 5 5 5
Quantity Traded
.
8 '1 10\ ~ 1:2. 13 1M E
- l- i- - - -. . . ....
1--" .. .... ...... : .... .... -t-- -- --
I- i- - - - -5 5 6 6 5 5
1'/ 15 It" 17 I~ 1'1 :;0 ~I ~ 013 ",1 015 ££ II/)() E 1100 r;. IU~ E. lEE IE '1.0/) 5.S()
IE E Il6b '/.0£ (',D~
l- i- t-- - - - .- --i- -- -I .... . -. ..oo .... .... " . .. .. " i""
.. .. ~!... "" -- - ..... ... - -
l- e...._ - -- -- - -- -- --4 6 6 7 4 7 4 7 4 0 4 5
Fig. 55. P284 Value Queue, 3 Sellers, Alternate Market
Trading Period
1 .:t .3 "I 5 G. 7 Price
1.00 -- - .- - --
.-. ~.~
. -" .... ... " H." .. ~.
.15
.00 i--
I-5
- -,- -5 6
- ----- -l- I-6 6 6 6
Quantlty Traded
8 9 10 /I
'-- - - -
~.- ...... --- -- ~- ........ -- I-- f-- -
7 7 7 8
I;), 13 1'/ /s I" 17 ~l 19 dO :)1 ClO! :3 EIE ~ £ £ E
o
-- -- - -- -- - -- -- - -c
-"' - .. _ .... "".J.
.w ~~ .- 0;.. - -- ,..;.'! . - .;;; I--~- .-: .... "
'- - - - - --I- -- - - f-- --4 5 6 7 7 7 4 '/ 6 8 7 8
Fig. 56. P286 Value Queue, 3 Sellers, Alternate Market
~'I
c
--
~ ... 1---
8
~S E-
--
;;;.----8
111
Trading Peciod
~;". -:; 3 '( 5 " 7 8 '1 10 II 1:1. 13 /'1 15 '" 17 It 19 ;0 <3.1 -'0:2. ;23 ~¥ .25 E E ~ £ E E E £ E E IE e
price
1.00 I- - -- - -I- -- I-- - - - - i- - --- -- -- -- - - - --
__ I- I- : 1--I-~ 1-. 1--1-.. .. - ~-:: I=-~ ~~ ~ .... - ~. 0- _ -- - 1---~- ~"- f~ I"""' w .15
.00 - - 1-- r- 1-- -- - i-- - 1--1-- - 1-- -- -- -1---
5 5 6 6 7 7 7 7 7 6 7 7 8 8 8 7 8 8 8 8 8 8 8 7 Quantlty Traded
E denotes exit by a seller to the alternate market
Fi 57 P278 Value Queue, 3 Sellers, Altarnate Market g. .
Trading Period
Price
1.00
.15
.00
I .:t 3 '/ £
-- 0- _ ,
. I-
,
I- -1:=
_o~
--1-- -- -6 6 7 7
~
-
1-:::.'
1--
7
ID 7 £ E
-
~-= I--I-7 5
Quantlty Traded
8 9 10 E
- --
.. r-:-I-- i--
-- I--7 7 8
/I I;J. IJ 1"1 15 Ie 17 18 1'1 £ E E E E E
-- .- - 1-- I- ---
I-=- ~ 1=.' r- ~~ ~., ~ l-I'-O
- l- .- - -- - - 1--6 6 4 6 7 8 7 7 8
E denotes exit by a seller to the alternate market
~o CJ.I ~ .l3 J"I :/; E £ £ £ r £
-- - - 1-- -- -
'"' F.- f:;;.- I--,.!!' 1--- - - - 1-- --
8 8 8 8 7 8
Fig. 58. P279 Value Queue, 3 Sellers, Alternate Market
8
112
Trading Period 1 2 3 " 5 b 7 8 9 10 /I Ii!. IJ 11 15
0 '" 17 18 19 20 Price;
. • . 0
~ ..:. -.! . ... -~ ... - ... I- -'" -.... ~ ... i- -! fooo -N' --'" i-......... - ... ... -... --. I- ...
r--Pc
- -- - -- -- -- -- -- --I-- - -- - -- -- - -- -- -- ---7 8 7 8 8 9 9 10 10 8 8 8 9 9 8 8 8 7 8 8
Quantity Traded
Fig. 59. P322 Random Queue, 3 Sellers
Trading Period
Price I, 2 3 'i 5 6 7 F 9 10 /I 12 13 1'1 15 Ib 17 Ii 19 20
, .. ... _. .... ... 0 -- ..: ....... ....... - ... , .... -.. -'- .. - ... . ...... ....... ... ....... .. ._--~ ..... ...
. _-- - - - -- - -- - -- -- - -- -- -- - - .- - -- --5 8 8 8 7 9 6 8 8 8 8 8 8 8 9 8 9 8 8 8
Quantity Traded
Fig. 60. P326 Random Queue, 3 Sellers
LIST OF REFERENCES
Allen, B. And M. Hellwig, "Price-Setting in Firms and the Oligopolistic Foundations of Perfect Competition", American Economic Review Proceedings. May 1986, 76, 387-392.
Bain, J. S., Barriers to New Competition, Cambridge: Harvard University Press, 1956.
Baumo1, W. J., "Contestable Markets: An Uprising in the Theory of Industr.y Structure." American Economic Review March 1982, 72, 1-15.
Baumo1, William J., E. E. Bailey, and R. D. Willig, "Weak Invisible Hand Theorems on the Sustainability of Multiproduct Natural Monopoly," American Economic Review, June 1977, 67, 350-365.
Baumol, William J., Panzar, John C., and Willig, Robert D., Contestable Markets and the Theory of Industrial Structure, New York, Harcourt, Brace, Jovanovich, Inc., 1982.
Beckman, M. J. (with the assistance of Dieter Hochstadter), "Edgeworth-Bertrand Duopoly Revisited," Operations ResearchVerfahren. III, edited by Rudolf Henn, Sonderdruck, Verlag, Anton Hain, Meisenheim, 1965, 55-68.
Bertrand, J., "Review of the Theorie Mathematique de 1a Richesse and Recherches sur 1e Principes Mathematiques de la theorie des Rechesses," Journal des Savants, 1883, 499-508.
Chou, Ya-lu, Statistical Analysis with Business and Economic Applications, New York, Holt, Rinehart and Winston, Inc., 1969.
Cournot, A. A., Researches into the Mathematical Principles of the Theory ofVea1th, New York, Macmillan, 1897.
Coursey, D., M. Isaac, M. Luke andV. L. Smith, "Market Contestability in the Presence of Sunk Entry Costs," The Rand Journal of Economics Spring 1984, 15, 69-84.
Coursey, D., M. Isaac, and V. L. Smith, "Natural Monopoly and Contested Markets: Some Experimental Results," Journal of Law and Economics. April 1984, 27, 91-113.
114
115
LIST OF REFERENCES--Continued
Davidson, C., and R. Deneckere, "Long-Run Competition in Capacity, Short-Run Competition in Price, and the Cournot Model," The Rand Journal of Econohlics, Autumn 1986, 17, 404-415.
Dasgupta, P. and E. Haskin, "The Existence of Equilibrium in Discontinuous Economic Games, 1: Theory," Review of Economic Studies, 1986a, 53.
Dasgupta,P. and E. Maskin, "The Existence of Equilibrium in Discontinuous Economic Games, 2: Applications," Review of Economic Studies, 1986b, 53.
Demsetz, H., "Why Regulate Utilities?," Journal of Law and Economics, April 1968, II, 55-65 .
. Edgeworth, F. Y., Papers relating to Political Economy I, London, Macmillan, 1925.
Harrison, Glenn W., "Experimental Evalutation of the Contestable Market Hypothesis," Working Paper 1984.
Harrison, Glenn W. and Michael McKee, "Monopoly Behavior, Decentralized Regulation, and Contestable Markets: An Experimental Evaluation," The Rand Journal of Economics, Spring 1985.
Ketchum, J., V. L. Smith and A. W. Williams, "A Comparison of Posted-Offer and Double-Auction Pricing Institutions," Review of Economic Studies,1984, LI, 595-614.
Kreps, D. and Scheinkman, J., "Cournot Precommitment and Bertrand Competition Yield Cournot Outcomes," Bell Journal of Economics, 1983, 14, 326-337.
Kruse, Jamie L., "Experimenta1 Tests of the Contestable Market Hypothesis in a Multiple Market Setting: A Research Proposal," 1L!1publ ished Manuscri.pt, Depar tment of Economics, Univexsity of Arizona, December 1985.
Kruse, J., S. Rassenti, S. S. Reynolds, and V.I. Smi th, "BertrandEdgeworth Competition in Experimental Markets," Discussion Paper No. 87-12, Department of Economics, University of Arizona, November 1987.
116
LIST OF REFERENCES--Continued
Levitan, R. and M. Shubik, "Price Duopoly and Capacity Constraints," International Economic Review, 1972, 13, 111-121.
Maskin, E., "The Existence of Equilibrium with Pric-Setting Firms," American Economic Review Proceedings. May 1986, 76, 382-386.
Osborne, M. J. and C. Pitchik, "Price Competition in a Capacity Constrained Duopoly", Journal of Economic Theory, 1986, 38, 238-260. .
Panzar, J. C. and R. D. Willig, "Free Entry and the Sustainability of Natural Monopoly," Bell Journal of Economics, Spring 1977, 8, 1-22.
Plott, C. R. and V. L. Smith, "An Experimental Examination of Two Exchange Institutions," Revie'tl1 of Economic Studies, February 1978, 45, 133-153.
Scherer, F. M., Industrial Market Structure and Economic Performance. Second Edition, Boston: Houghton Mifflin Company, 1980.
Schwartz, Marius and Robert Reynolds, "Contestable Markets: An Uprising in the Theory of Industry Structure: Comment," America .. Economic Review, September 1984, 74, 572-587.
Sharkey, W. W., The Theory of Natural Monopoly, Cambridge: Cambridge University Press, 1982.
Shepherd, William G., "Contestabi1ity vs. Competition," American Economic Review, June 1983, 73, 488-490.
Shubik, M., Strategy and Market Structure, New York: John Wiley and Sons, 1959.
Smith, V. L., "Induced Value Theory," American Economic Review, 1976, 66, 274-279.
Smith, V. L., "An Empirical Study of Decentralized Institutions of Monopoly Restraint," in G. Horwood and J. Qurik(eds.), Essays in Contemporary Fields of Economics, West Lafayette: Purdue University Press, 1981.
, LIST OF REFERENCES--Continued
Sp&nce, A. M., "Entry, Capacity, Investment and Oligopolistic Pricing," Bell Journal of Economics, 1977, 8, 534-544.
Sylos-Lambini, P., Oligopoly and Technical Progress, Cambridge, Mass.: Harvard U.P., 1962.
117
Vives, X., "Rationing Rules and Bertrand-Edgeworth Equilibria in Large Markets," Economic Letters 1986, 21, 113-116.
Waterson, M., Economic Theory of the Industry, Cabridge: Cambridge University Press, 1984.
Weitzman, Martin L., "Contestable Markets: An Uprising in the Theory of Industry Structure: Comment," American Economic Review, June 1983, 73, 486-487.