Printer Friendly

The Role of Monitoring in Duopoly Market Outcomes.

JAMES H. HOLCOMB [*]

ABSTRACT: This research investigates monitoring and its effects on market outcomes. Monitoring is the ability of agents to observe each other's actions in the marketplace. We argue that monitoring has a strong influence on the behavior of economic agents and therefore can have strong policy implications. Previous theoretical research in oligopoly strongly suggests that monitoring, or the lack of it, will alter the behavior of economic agents and therefore affect market outcomes. Using experimental laboratory techniques, we demonstrate that monitoring does significantly affect market outcomes.

1. INTRODUCTION

Oligopolistic behavior is characterized by the firms' awareness of their interdependence. A firm's profitability depends not only on its own choices of market variables, but also on the choices made by its rivals. As a result, each firm is aware that any action it takes is likely to elicit a response from the other firms. Thus, a firm will develop, over time, some set of assumptions about the expected response of its rivals to stimuli. This behavioral expectation has, in differing circumstances, been called a reaction function or a conjectural variation.

Implicitly assumed in the development of a reaction function is that firms must be able to observe the actions of their rivals, that is, to react to a rival's change, a firm must be able to tell a change took place. If rivalistic firms are unable to monitor the other firms in their market, they will be unable to react to the strategic moves of their counterparts. While this seems a trivial assertion, a lack of monitoring is a key element in the story told about cartel instability.

Firms in a cartel have an incentive to cheat on their output quota, if they can do so without being detected. [1] Each firm may see a short-term, positive return to "chiseling" on the agreement. For chiseling to "work," it must be done in secret. If chiseling becomes apparent, the cheated firms can inflict some form of punishment, and the result is cartel instability. We argue that much of the conduct of economic agents is guided by the degree to which they can monitor the activities of other agents in the market. [2]

Monitoring is defined to be the observation of the actions of economic agents by rival agents, in any market setting where agents recognize their mutual interdependence. The issue of monitoring is tied to several other behavioral aspects of market conduct including signalling and punishment.

Generally, oligopolistic firms have an incentive to cooperate, or collude, acting as if they were a monopoly. However, firms are unable to convey explicitly their intentions and desires, particularly about collusive price-setting arrangements, to the other firms in the market. [3] Firms can only reach some form of tacit understanding about their desire to cooperate. [4] This indirect communication, called signalling, is accomplished by the firms' market actions, for example, setting prices. For instance, if a firm unilaterally reduces output (or raises price), this action could be interpreted by others as a desire to collude particularly if the firm places a significant amount of profit at risk.

Monitoring and signalling are linked closely. Monitoring is the ability of a firm to "see" a rival's signal. An examination of the process of signalling is an important issue and has an extensive literature of its own. However, this study focuses on the ability of firms to receive signals and does not consider, in any detail, a firm's ability to interpret signals correctly or of the process of sending a signal to convey intent. Before a firm can send a signal or interpret a signal, it must be shown that the firm and its rivals can receive (monitor) signals.

Monitoring is tied to punishment strategy. As mentioned, a collusive arrangement usually involves at least a tacit understanding to restrict output. Yet, if a firm unilaterally expands its output, it will benefit at the expense of its rivals. In order for a collusive agreement to remain viable, the colluders must exert some effort to enforce the agreement, that is, to detect and punish defectors. To punish defectors the nondefecting firms raise output until they reach a self-enforcing equilibrium where unilateral departures no longer increase profit (something analogous to the Cournot equilibrium). At this equilibrium, no firm wants to be the only one to change production; but, all firms earn reduced profits. Notice that any punishment strategy is predicated on the fact that firms could detect (monitor) a cheater.

Collusive agreements are not legally enforceable. The colluders must enforce their agreement on their own. Enforcement requires colluders be able to detect cheating, that is, to monitor. Monitoring is costly and difficult to carry out. Cartels have used several different methods to monitor more effectively, including: control over more than just price, market division, fixing market shares, most-favored-nation contracts, meeting-competition contracts, and trigger prices. Generally, a collusive agreement would be easier to enforce if the number of firms is small, demand is stable, pricing information is widely circulated and the firms produce homogeneous products.

To summarize, the problem that a lack of monitoring causes for a collusive consensus (cartel), can be expressed as a difference between group and individual incentives. There is an incentive for the group to cooperate. By tacitly cooperating, the oligopolistic firms can reach and maintain the shared-monopoly outcome. However, individual firms could have an incentive to cheat on the collusive agreement, if the chance of being detected and/or punished is sufficiently small. The amount of monitoring in these situations will influence the outcome of the market. The ability to monitor should lead to more stable collusive outcomes because monitoring raises the cost of defection. This implies that a lack of monitoring could lead to the opposite result. This is the issue raised by this research. Using experimental economic methodology, the effect of the inability to monitor accurately the actions of rival firms on an established collusive consensus is examined.

II. REVIEW OF SOME RELEVANT LITERATURE

In his classic article, Stigler (1964) suggested that the greatest obstacle to collusion, in the absence of entry, would be "secret price cutting." He viewed the cartel as a police officer who was required periodically to punish the "crimes" of individual cartel members. Stigler also introduced the idea that collusion could take place even with noncooperative (no explicit communication) firms, if collusion was the dominant strategy, that is, tacit collusion.

Cyert and DeGroot (1970) extended the oligopoly model to include uncertainty about demand. Using a simple linear demand function with an uncertain intercept, they reached the conclusion that without restrictive assumptions on behavior, a unique solution does not exist. By adopting different sets of behavioral assumptions, they could support a range of possible equilibrium outcomes from Cournot to collusive.

The uncertainty about demand reduces the ability of firms to monitor their rivals. A firm could not be certain if demand changed or a rival cheated on the agreement. Presumably, if demand were uncertain but there was perfect monitoring, collusive agreements would be no more difficult to enforce than if demand were known with certainty.

Osborne (1971, 1974) described firms as responding to changes in output by other firms to retain their own market share. If rival firms are aware of each other's reaction functions, each firm gradually realizes that departures from the collusive agreement are suboptimal, in either a simultaneous or a sequential choice market. However, if firms cannot directly observe general market conditions, their optimal response is unclear. This implies firms, if they are aware they are not being monitored, may find it profitable to cheat on the cartel agreement.

Friedman (1971) discussed the profitability of cheating on the cartel agreement in an uncertain monitoring environment. Firms will suspect cheating if the market price falls below a certain threshold, or trigger, price. The trigger price is the market price that would result if all firms produced the agreed-upon amounts. A firm's response, or punishment strategy, to suspected cheating is to produce the Cournot level output (generally to expand output) for the remainder of the planning horizon.

On the other hand, the shared-monopoly outcome can be the dominant strategy if the time rate of discount for the firms is sufficiently high. Moreover, a punishment reaction, by the cheated firm, may not always be an optimal response to suspected cheating. The cheated firm, desiring to punish the cheater, also punishes itself, that is, neither firm earns the collusive profits. Instead, it may make more sense for a firm to adopt a competitive strategy (increase output above the Cournot level), for a short time, to express its displeasure and then be willing revert to the collusive equilibrium to earn economic profits.

Green and Porter (1984) extended Friedman's model to include the possibility firms will revert to the collusive equilibrium after the punishment period. For example, assume firms have agreed upon a trigger price. If the market price drops below the trigger price, cheating is assumed to have taken place. The cheated firms adopt a competitive strategy (increase output or reduce price) for some length of time. Firms then gradually return to the collusive output level and remain there unless cheating is detected again. If cheating is again detected, the cycle repeats. Cheating on the output restriction (collusive agreement) increases profits to the cheater and the response by the cheated firms reduces profits to the cheater. A firm recognizes that increasing output above the collusive level will increase short-run profits at the expense of the increased probability the punishment response will follow. Additionally, each cheating episode increases the time necessary for all firms to return to the collusive level.

Radner (1986), Radner, Myerson and Maskin (1986) and Porter (1983a, 1983b) further developed the trigger strategy model considered by Friedman (1971) and Green and Porter (1984). After each market period, each firm tests the hypothesis that the other firms have been following the agreement. If a firm rejects the hypothesis, it reverts to a pre-determined punishment strategy, producing the Cournot-output level, for some number of periods. After a sufficient number of periods, the firms return to the output agreement and the testing process is continued.

Kreps and Spence (1985) suggested that market history may have an important impact on firms reaching and maintaining a collusive consensus. Firms in an industry with a collusive history tend to remain collusive even in the face of a change in the market environment. Conversely, industries with a history of competitive or rivalistic behavior may find collusion to be an elusive goal. [5]

Radner (1986) built on this discussion and depicted imperfect monitoring to be the situation when the information players possess does not include the complete history of previous observations and the moves of all players. Without a memory of the past, firms are unable to form a reaction function and therefore cannot interpret the signals of rival firms.

Abreu, Pearce, and Stacchetti (1987) generalized the Green and Porter (1984) analysis to address some problems discussed by Kreps and Spence (1985). They permitted the punishment strategy to depend upon the actual market price observed and allowed for the possibility that prior period prices affect current behavior. That is, they allowed for some quasi-reputations (market history) to be formed. This implies that a single low price is viewed differently in light of collusive prices in the past. Under ibis characterization, price wars are not begun as a response to cheating but due to low demand, since all firms have no incentive to increase output. This presents an interesting paradox. If no one has an incentive to cheat on the agreement, then there is nothing to be gained by having a trigger price (indeed profits will be lost due to false positives or interpreting cheating from a low revelation of demand). However, and here is the conundrum, without a punishment phase, no player has the incentive to remain c ollusive.

The effect of history on market performance was investigated experimentally by Phillips, Battalio, and Holcomb (1987). They looked at two basic situations: a quantity-setting duopoly and a price-setting duopoly. The markets were characterized by excess capacity, zero variable costs and homogeneous products. The quantity-setting environment was more conducive to collusive outcomes than was the price-setting environment. [6]

The impact of history was examined by placing subject pairs, from the quantity setting market, with a collusive history into a price-setting environment, where collusion should have been more difficult to maintain. Yet, the change in environment did not affect the collusive arrangements. However, when pairs from the price-setting markets were placed in the quantity-setting market, they were unable to cooperate. A history of cooperative or competitive behavior appeared to facilitate similar behavior in future periods, even under different conditions, lending support to Kreps and Spence (1985).

III. EXPERIMENTAL DESIGN

The role of monitoring in oligopolistic markets was examined by employing laboratory experimental methods. As a research technique, controlled experiments are well-suited to investigating the impact of institutional arrangements, such as the ability or inability to monitor, on market outcomes. Experiments allow for direct empirical tests of theoretical predictions while providing a high degree of control over information flows. The degree of monitoring may be controlled to any degree from perfect observation of rival firm's behavior to complete ignorance. Perhaps more significantly, the information flows can be altered during the operation of an experiment, providing for direct observation of the effect of monitoring on market outcomes.

How is a lack of monitoring characterized in terms of an oligopolistic market? In the real-world, firms generally cannot discuss explicitly their market intentions but can only signal their intentions through the choice of market variables, for example, price and output. Also, firms are not aware, with certainty, of market demand nor are they completely aware of the cost conditions of the rival firms. If one firm observes a change in market price or quantity, it cannot be sure if a rival changed its position or if market demand changed. This uncertainty with respect to the source of profit instability is the lack of monitoring.

The Phillips, Battalio, and Holcomb (1985), Green and Porter (1984) and Kreps and Spence (1985) results form the foundation for the experimental design. Subjects were placed in an environment designed to foster tacit collusion to build a collusive history. [7] The institutional structure was then altered (the ability to monitor was impaired) to observe the effect on the collusive consensus.

A) Complete Monitoring

In the experiments for this study, a total of 40 subjects were paired in duopoly markets. Subjects were paired at random and each kept the same counterpart for the duration of the experiment. Subjects were not told who their counterpart was before, during, or after the experiment. They were taken to a common room where instructions about the experiment were read to them and questions answered. A copy of the instructions is included in Appendix A. The subjects were then separated and one-half were led to another room.

The task for the subjects was to select a value for X from a payoff table for each period. Subjects wrote down their selections on slips of paper or "Data Entry Forms" that were collected and exchanged with the subjects' counterpart. The subjects then found their profit, their counterpart's profit and their new balance Subjects were paid their balance, in cash, at the conclusion of the experiment.

This experiment was a symmetric design, in that every subject was given the identical payment table. The table contained the possible payments that subjects could earn each period. The table listed the possible choices for each subjected and the payoff that would be earned based on their counterpart's selection of X. Embedded in a payment table was the profit function faced by these firms. The payment table is included in Appendix A.

The profit table used for the experiments was generated using the profit function given by (1).

[[pi].sub.i] = [[q.sub.i]{17 - (0.21([q.sub.i] + [q.sub.i]))} - 127]/100 (1)

where [q.sub.i], [q.sub.j] = 18, ... ,29.

The size of the table was limited purposefully so that subjects would have less information to digest. A value for [q.sub.i] of 18 corresponded to a value of X of 1 on the table and so on. The values in the table included both the joint profit maximum (X = 3) and the Cournot equilibrium (X = 10). The profit table employed in this experiment featured a cost function where marginal cost was zero.

The issue at question was whether imperfect monitoring could break down an established collusive consensus, that is, does imperfect monitoring lead to cartel instability. Consequently, a deliberate attempt was made to increase the percentage of collusive pairs. Each experiment began with five periods of explicit communication. Subjects were allowed to send written messages to their rival for the first four minutes each period to foster collusion. [8] After seeing the message, subjects chose their values. All messages were nonbinding in the sense that they were not enforceable by the experimenters.

After period 5, signals could not be sent in any fashion other than by the values for X that were selected by the participants. Monitoring remained perfect during this phase, in the sense that subjects knew after each period exactly what value their counterpart had selected in that period. [9]

The collusive outcome is the joint-profit maximizing level of output. The collusive level is a value for X of 3 chosen by both participants. The Cournot level is that output where firms would settle if they assumed their counterpart in the market would not change their output in response to a change in their value. The Cournot level of X is a value of 10 selected by both duopolists. [10] The movement toward values for X smaller than 10 was interpreted as becoming more collusive while a movement toward values larger than 3 as becoming more competitive.

B) Incomplete Monitoring

A minimum of 20 periods were conducted with perfect monitoring. Following the perfect monitoring condition, the experiment was halted and all subjects were given and read a new set of instructions (included in Appendix B). The subjects were told explicitly that they would now be correctly informed of their countepart's choice from the payment table only 50 percent of the time. [11] Otherwise they would be given a random value as their counterpart's selection.

During the imperfect monitoring condition, as before, each subject was asked to choose a value for X from the profit table. These choices were collected from the subjects and taken to a central location. A coin was flipped independently for each subject. If the coin was "heads," a participant's choice was reported accurately to their counterpart. If the coin was "tails," a value for X was drawn at random from the range of values on the table and that value was reported to their counterpart. A subject, receiving a value for X selected by their counterpart, did not know with certainty if the value they saw was what their counterpart selected or a randomly selected value. [12] Further, subjects were paid based on what they were told their counterpart selected--not on what was actually selected.

After a minimum of 20 periods with imperfect monitoring, subjects returned to the perfect monitoring environment. This A-B-A design was used to decide if the exposure to imperfect monitoring permanently changed subjects' behavior or whether they would revert to their previous strategies.

IV. RESULTS

A total of 20 duopoly markets were run. If the inability to monitor affected the ability to form and maintain a collusive consensus, we would expect to see either of two events. First, subjects who reached a collusive equilibrium during the perfect monitoring phase would be unable to maintain the collusive understanding during the imperfect monitoring phase. Second, for subjects who were not at a stable equilibrium during perfect monitoring, as well as those who were stable, we would expect to notice a significant increase in the mean of the total market output under the imperfect monitoring condition.

The effect of the imperfect monitoring condition may also implicitly change the payoff matrix. To put it simply, each subject faces a rival who, one-half of the time, apparently selects a random value between 1 and 12. Each subject, during the imperfect monitoring condition, may be choosing [X.sub.i] to maximize their expected payoff according to some function which describes their rival's choice of [X.sub.j]:

max 1/2[X.sub.j] + 1/2 [[[sigma].sup.12].sub.j = 1] 1/2[X.sub.j] (2)

If subjects believed that their counterpart intended to remain faithful to the collusive understanding during the imperfect monitoring phase, they would expect to see a value of 3 from their rival one-half of the time and, given the nature of the uncertain demand distribution, a value of 6 (or 7) the remainder of the time. It could be argued that the joint-payoff maximum position has changed from (3,3) in the certain monitoring condition to (7,7) in the uncertain monitoring condition. In which case, the above anticipated increase in market output would not represent a change in behavior but simply a continuation of cooperative behavior in the face of the new environment. In Phillips, Battalio, and Holcomb (1987) experimental subjects who had formed a collusive arrangement were able to maintain the arrangement even as the experimental environment changed.

Still, the actual payoff matrix has not been changed. The Cournot value of X, of 10 retains the same meaning in both conditions. If a subject moves to a value of 6, in response to the story above, approximately One-quarter of the time they will receive a value greater than 6 from their rival and therefore a decreased payment. Yet, by moving to the Cournot value, subjects insulate themselves from the impact of the random values. There is no advantage to selecting a value of 6 over the Cournot without the cooperation of the other subject. During the uncertain monitoring phase this cooperation is not as easy to achieve. The return to cheating on the agreement has increased because the likely of punishment (and the impact of any punishment) has decreased. Still, note the impact of imperfect monitoring remains the same, that is, market output rises, regardless of the effect on behavior.

The results of the experiments are summarized in Table 1. The data are broken down by perfect and imperfect monitoring. The column "Mean Output" gives the mean total market output actually selected by the two participants for the last ten periods of each condition. To provide a reference point, a value of 6 for mean output is the joint profit-maximizing level. A value of 20 for mean output is the Cournot equilibrium.

Looking over Table 1, two points stand out. First, in 19 of 20 markets mean output rose once imperfect monitoring was introduced. The exception is market 3, which remained unchanged from the initial period. The difference between the mean of the last ten periods in each condition was significant at the [alpha] = 0.05 level in 17 of the 20 markets that showed differences. [13] Second, 17 of the markets reached a stable collusive equilibrium prior to the last ten periods of the initial perfect monitoring condition.

It would appear that imperfect monitoring made it difficult for subjects to maintain a collusive consensus. We consider this an important result because evidence suggests that once a history of collusion is formed, the agreement usually continues, even in the face of unfavorable conditions (Phillips, Battalio, and Holcomb, 1987).

Moreover, the breakdown in collusion was peculiar to the imperfect monitoring regime. Following the imperfect monitoring periods, subjects returned to the no communication, perfect monitoring condition. All 16 pairs that had become noncollusive when faced with imperfect monitoring, returned to the collusive. output level.

V. CONCLUSIONS

The principal objective of this research was to investigate the role that monitoring plays in market outcomes. To date, a large and expanding theoretical literature exists on trigger price strategies and other methods to deal with imperfect monitoring. Porter (1983b) has found some historical evidence that trigger prices might have been used to foster collusion in the railroad industry in the. 19th century. However, beyond this one study there is little empirical basis for this body of theory. The lack of empirical validation is a major drawback especially, if at some point in the future, policy-makers attempt to make an industry or industries more competitive by outlawing market practices that serve to make monitoring less difficult.

There are several avenues of further research from this point. One would be to figure out if imperfect monitoring could be a reason that the number of firms in a market seems to affect industry performance. Of course, other environmental changes brought on by increasing the number of firms such as exponentially increasing negotiation and contracting costs, tend to make markets with more firms more competitive. Yet, monitoring apparently becomes more difficult as the number of firms increases. If weaker monitoring brought on by more firms is shown to inhibit collusion, this result enhances the "More firms mean more competitive" rule of thumb. More significantly, policy makers seeking to make industries more competitive could target practices and institutions which facilitate monitoring rather than breaking up existing firms.

Acknowledgments: We wish to acknowledge the financial assistance of the University Research Institutes of both The University of Texas at El Paso and of Western Illinois University.

(*.) Direct all correspondence to: James H. Holcomb, Department of Economics and Finance, The University of Texas at El Paso, El Paso, TX 79968.

NOTES

(1.) See among many others Axelrod (1984) and Holcomb and Nelson (1991).

(2.) A similar discussion has been developed with regard to public goods. There is some evidence that the level of cooperation seen in public good provision depends on whether the decisions of all participants are public knowledge.

(3.) Explicit cooperation is illegal, with some exceptions, under the Sherman Anti-trust Act of 1890 and others.

(4.) This cooperation, when and if achieved, often takes the form of output restrictions, or quotas. See Davidson and Martin (1985) for an illustration of how prices (or quotas) might be set.

(5.) The role of history in markets is also addressed in the rapidly developing literature of coordination games. See Van Huyck, Battalio, and Beil (1990) among others.

(6.) This representation is made due to the number of duopoly pairs reaching a collusive consensus under the quantity-setting paradigm relative to the price-setting model.

(7.) Subjects were undergraduate students recruited by means of campus-wide advertisements.

(8.) Subjects could write any message except those that involved threats of physical violence or post-experiment division of earnings. There have been a number of articles which have looked at the usefulness of communication to foster collusive arrangements. See Cooper et al. (1989) and Palfrey and Rosenthal (1991) among others. The design of this experiment did not intend to advance the literature in this area. Rather we intended to do everything possible (short of informing the subjects to pick the collusive outcome) to facilitate collusion in the initial phase. Our focus in this research is whether or not the inability to monitor with certainty causes an established collusive consensus to break down.

(9.) For the interested reader, the instructions provided in Appendix A give a more detailed description of the actual operation of the experiment.

(10.) Care was taken not put the collusive and Cournot choices on the corners of the table so as to try not influence subject behavior through location.

(11.) A market example of this treatment might be a commodity cartel. Each cartel member knows the price it charges to customers and can quickly get an accurate view of market price from the commodity exchange. If the firm observes a change in the future price, it can not be completely certain whether market demand changed or rival firms altered prices. The firm faces some uncertainty when it tries to determine the cause of the price change.

(12.) If the value was randomly chosen all values were equally likely to be selected. Of course there was a 1/12 chance that the randomly selected value was equal to the value a participant s counterpart actually chose.

(13.) The difference was significant at the [alpha] = 0.01 level in 16 of the 19 markets.

REFERENCES

Abreu, Dilip, Pearce, David, & Stacchetti, Ennio. (1987). Optimal cartel equilibria with imperfect monitoring. Journal of Economic Theory, 39, 251-269.

Axelrod, Robert. (1984). The evolution of cooperation. New York: Basic Books.

Cooper, Russell, DeJong, Douglas V., Forsythe, Robert, & Ross, Thomas W. (1989) Communication in the battle of the sexes game: some experimental results. Rand Journal of Economics, 20, 568-597.

Cyert, Richard M., & DeGroot, M. (1970). Multi-period decision models with alternating choice as a solution to the duopoly problem. Quarterly Journal of Economics, 84, 410-429.

Davidson, Carl, & Martin, Lawrence W. (1985). General equilibrium tax incidence under imperfect competition: a quantity-setting supergame analysis. Journal of Political Economy, 93, 1212-1223.

Friedman, James. (1971). A noncooperative equilibrium for supergames. Review of Economic Studies, 28, 1-12.

Green, Edward, & Porter, Robert H. (1984). Noncooperative collusion under imperfect price information. Econometrica, 52, 87-100.

Hay, George A., & Kelley, Daniel. (1974). An empirical survey of price fixing conspiracies. Journal of Law and Economics, 17, 13-38.

Holcomb, James H., & Nelson, Paul S. (1991). Cartel failure: a mistake or do they do it to each other on purpose? Journal of Socio-Economics, 20, 235-249.

Kreps, David, & Spence, A. Michael. (1985). Modelling the role of history in industrial organization and competition. In G.R. Feiwel (Ed.), Issues in contemporary microeconomics and welfare. Albany, NY: State University of New York Press.

Osborne, D.K. (1971). The duopoly game: output variations. American Economic Review, 61, 538-560.

Osborne, D.K. (1974). A duopoly price game. Economica, 41, 157-175.

Palfrey, Thomas R., & Rosenthal, Howard. (1991). Testing for the effects of cheap talk in a public goods game with private information. Games and Economic Behavior, 3, 183-220.

Phillips, Owen R., Battalio, Raymond C., & Holcomb, James H. (1987). Duopoly behavior with market history. Working Paper, University of Wyoming.

Porter, Robert. (1983a). Optimal cartel trigger price strategies. Journal of Economic Theory, 29, 313-338.

Porter, Robert. (1983b). A study of cartel stability: the joint executive committee, 1880-1886. Rand Journal of Economics 16, 41-50.

Radner, R. (1986). Repeated partnership games with imperfect monitoring and no discounting. Review of Economic Studies, 53, 43-57.

Radner, R., Myerson, R., & Maskin, E. (1986). An example of a repeated partnership game with discounting and with uniformly inefficient equilibria. Review of Economic Studies, 53, 59-69.

Stigler, George. (1964). A theory of oligopoly. Journal of Political Economy, 72, 44-61.

Von Huyck, John B., Battalio, Raymond C., & Beil, Richard O. (1990). Tacit coordination games, strategic uncertainty, and coordination failure. American Economnic Review, 80, 234-248.
Table 1.
Experimental Results *
        Perfect Monitoring                      Imperfect Monitoring
Market         Mean         Standard Deviation          Mean
   1          10.0                 4.65               18.3 ***
   2           6.0                 0.00               18.6 ***
   3           6.0                 0.00                6.0
   4           6.0                 0.00               10.1 **
   5           6.0                 0.00               20.9 ***
   6          12.4                 3.90               14.3
   7           6.0                 0.00               16.5 ***
   8           6.0                 0.00               15.9 ***
   9           6.0                 0.00               15.9 ***
  10           6.0                 0.00               19.0 ***
  11           6.0                 0.00               18.4 ***

  12           6.0                 0.00               19.1 ***
  13          12.9                 5.77               16.6
  14           6.0                 0.00               18.7 ***
  15           6.0                 0.00               16.6 ***
  16           6.0                 0.00               19.8 ***
  17           6.0                 0.00               18.3 ***
  18           6.0                 0.00               16.6 ***
  19           6.0                 0.00               17.9 ***
  20           6.0                 0.00               19.1 ***
Market  Standard Deviation
   1           4.60
   2           3.88
   3           0.00
   4           5.22
   5           0.54
   6           4.22
   7           3.20
   8           2.70
   9           2.77
  10           2.76
  11           5.31
  12           4.13
  13           4.48
  14           2.61
  15           3.88
  16           0.40
  17           2.45
  18           4.20
  19           1.37
  20           1.92
Notes: (*)Means and standard
deviations are calculated for the
last 10 periods of each condition.
(**)Significantly different from
mean of last ten periods of
perfect monitoriing at [alpha] = 0.05
(***)Significantly different from
mean of last en periods of perfect
monitoring at [alpha] = 0.01


APPENDIX A

Instructions (Monitoring Condition)

This is an experiment in the economics of market decision making. The University Research Council and other funding agencies have provided funds for the conduct of this research. The instructions are simple. If you follow them closely and make appropriate decisions, you may make an appreciable amount of money. These earnings will be paid to you, in cash, after the experiment.

You will be paired, at random, with another person, for the experiment. The identity of this other person will not be known to you. Also, the other participant will not be aware of your identity.

In this experiment there will be a sequence of market periods. Each period, you and the other participant, will pick a value for X from a table. The two values of X selected will be used to decide the payment you, and the other participant, will receive. Therefore, you will be required to select a value for X in each period of the experiment.

The payments that you and the other participant will decide from the values for X you select are yours to keep. The payment you receive each period will be added to your balance. You will be paid your balance and the participation fee of $4.00, in cash, at the conclusion of the experiment.

You are furnished with a table showing the various values of X you may select. The values you may select are written down the left-hand side of the table. The values the other participant may select are written across the top of the table. The payment you will receive is determined by the intersection of the value for X you select and the value selected by the other participant. For example, if you select a value for X of 5 and the other participant selects a value of 8, your payment will be the amount found at the intersection of the fifth row and the eighth column. Conversely, the other participant's payment would be found at the intersection of the eighth row and the fifth column. That is, you would receive a payment of $0.29 and the other participant would receive a payment of $0.51. The other participant is furnished with a table that is identical with your table.

At the start of each period you will may send written messages to the other participant. These messages can say anything you wish except that you may not make threats of physical harm and you may not suggest to split your or their earnings after the experiment. Otherwise you may send any message you wish. However, please work quickly since periods will be 4 minutes in length. Give completed messages to the runners who will relay them to the other participant.

Near the end of the period each participant will select a value for X, from the table, and write this value on the DATA ENTRY FORM (which you will find on the desk in front of you). When you have written down the value for X you wish to select on the DATA ENTRY FORM, give it to the runner who will come in to pick up your forms. The runner will take your form to another room where the other participant is located. After a moment the runner will return with the other participant's DATA ENTRY FORM. The form will now have the value for X selected by the other participant. The other participant will not see the value for X you selected before making his/her selection nor will you see the other participant's value for X before making your selection. Record the information on your RECORD SHEET and compute Your payment, your balance, and the other participant's payment.

To show how you can figure out your payment for the period we have included two examples.

Example 1

Your value for X = 6.

Value for X selected by the other participant = 4.

Your payment would be found at the intersection of the sixth row and the fourth column or $0.51. The other participant's payment would be found at the intersection of the fourth row and sixth column or $0.35.

Example 2

Your value for X = 5.

Value for X selected by the other participant = 7.

Your payment would be found at the intersection of the fifth row and the seventh column or $0.34. The other participant's payment would be found at the intersection of the seventh row and the fifth column or $0.49.

Your balance is calculated by adding your starting balance + your payment for the period. BE SURE TO RECORD YOUR BALANCE. Your balance is how you will be paid after the experiment.

To Sum Up

ENDING BALANCE = STARTING BALANCE + PAYMENT

Please do not speak to any other participant while the experiment is in progress. This is important to the validity of the study and no exceptions can be made. The experiment will consist of many periods.

If you have a question that you feel was not adequately answered in the instructions, please raise you hand and ask the monitor at this time. YOUR EARNINGS MAY SUFFER IF YOU PROCEED INTO THE EXPERIMENT WITHOUT UNDERSTANDING THE INSTRUCTIONS!!

Summary Instructions

1. Each period you must select a value for X from the table.

2. You may send messages to the other participant but you may not make threats or agree to split earnings

3. The payment you receive will depend on the value you choose and the value your counterpart chooses.

4. The payment you receive can be found on the interior of the table at the intersection of the value you chose and the value chosen by the other participant.

5. You will record the value for X you selected, the other participant's value, the other participant's payment, your payment, and your balance.

6. Please do not speak with anyone until the experiment is completed.

APPENDIX B

Instructions (Nonmonitoring Condition)

In the upcoming market periods a change will take place in the market. As before, you, and the other participant will enter your values for X on the DATA ENTRY FORM. However, after we leave the room with your data entry forms, we will flip a coin. If the coin is heads, the value you selected will be reported to the other participant. If the coin is tails, a value for X will be drawn randomly from one of the 12 possible values. The value drawn will then be reported to the other participant instead of the value that you selected. Also, when we leave, the other participant's room we will flip a coin; if the coin is heads, the value the other participant selected will be reported to you. If the coin is tails, then a value for X will be drawn randomly from one of the 12 possible values. The value drawn will then be reported to you instead of the value that the other participant selected.

Both you, and the other participant, will receive your payment based on the value for X you selected and the value for X reported to you whether that value was selected by the other participant or was selected randomly.

A different coin will be flipped for you and the other participant. You will not be able to know, for sure, if the value you selected was received by the other participant, nor will the other participant know, for sure, if you received the value for X he/she selected. And you will not be able to know, for sure, if the value for X reported to you was the value for X selected by the other participant or a value drawn randomly.

Therefore, 50 percent of the time, on average, the value for X reported to you will be the actual value selected by the other participant, and 50 percent of the time the value for X reported to you will be drawn randomly. If the value for X reported to you is drawn randomly, each of the values 1, 2, ... , 12 has an equal chance of being selected. Each value has a 1/12 chance of being selected. Notice it is possible that the value chosen by the other participant will be drawn randomly.

Summary

1. The payment table, the other participant, and your balance will continue as before.

2. The value that you see as the "Value Selected by the Other Participant" has a 50 percent chance of being the value he/she selected. Otherwise, one of the 12 possible values for X will be selected at random.

3. IF YOU HAVE ANY QUESTIONS, PLEASE BE SURE TO ASK THEM AT THIS TIME.
COPYRIGHT 1997 JAI Press, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1997 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:HOLCOMB, JAMES H.; NELSON, PAUL S.
Publication:The Journal of Socio-Economics
Article Type:Statistical Data Included
Geographic Code:1USA
Date:Jan 1, 1997
Words:6816
Previous Article:Pre-commitment vs. Flexibility: Uncertainty and Distribution Reform in P.R. China.
Next Article:The Socioeconomics of Drunk Driving.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters