Recommended for readers of XXx-level MSC-LEVEL | ECONOMETRICS | SPECIALTY | NL Prices versus Quantities in a Mixed Oligopoly BY: A mixed oligopoly is an oligopoly where firms face different objective functions. Current literature on mixed oligopolies, originating from Singh and Vives (1984), suggests that quantities as strategic variable dominates prices as strategic variable when goods are substitutes (and vice versa for complements). However, the dynamics of pure quantity-setting competition (Cournot competition) can become unstable as the number of firms passes a certain threshold. This research investigates a mixed oligopoly with one part of the firms using prices as strategic variables, and the other part using quantities. The current literature on oligopolies revolves mainly around two names: Cournot (1838) and Bertrand (1883). These two forms of competition are both commonly acknowledged, but lead to substantially different predictions. Cournot competition assumes firms use quantities as strategic variable, whereas Bertrand competition assumes firms use prices as strategic variable. Even when considering the exact same market structure, these types of competition lead to different outcomes. An important question therefore is which type of competition gives a better description of market outcomes. Singh and Vives (1984) investigate a differentiated duopoly where firms can choose whether to use price or quantity as strategic variable. They find that competing in quantities (prices) is a dominant strategy when goods are substitutes (complements). These findings are backed by other research as well, for example Häckner (2000). Other research finds that dominance is strongly dependent on the degree of substitutability between goods (Sakai et al. (1995)), as well as qualitative differences (Matsumoto and Szidarovsky (2010)). In all this research a two-stage model is considered, where in the first stage firms choose their strategy, and in the second stage they determine the value of their strategic variable. At first we assume that in the second stage the Nash equilibrium of the corresponding subgame is played, but later on we relax this assumption. ROB GOEDHART An important finding by Theocharis (1960) is that Cournot-Nash equilibrium becomes unstable under best-response dynamics as the number of firms gets larger. This instability will typically decrease profits and may make the quantity-setting strategy less profitable than the price-setting strategy in the same circumstances. It is unclear whether this instability is indeed large enough to prevent the price-setting strategy from being completely driven out by the quantity-setting strategy. To study this we consider a mixed oligopoly with an arbitrary number of firms. At first we consider two model setups, which differ in the information given to the firm about the other firms. The first model assumes that firms enter the same market each period, so that each firm is aware of the number of price-setters and quantity-setters in their market in every period. However, since we are looking to introduce an evolutionary model that allows firms to switch between strategies over time, we consider a second model. In the second model we assume that each period new markets are drawn from a very large population. A certain fraction of this population uses prices as strategic variable, and the other part uses quantities. This structure is similar to Hommes et al. (2013). The first model is used as a comparison for the second model. Since the results of the models are similar, this first model will not be discussed much further. 32 Lemma 1. The model in terms of prices of the pricesetters, and quantities of the quantity-setters can be written as many markets of n firms are randomly drawn from this population. Each period new random draws will be made to match firms in markets of size n, and thus the firms are not aware of the exact composition of their market when it comes to number of quantity- or pricesetters. However, they are aware of the fraction of the population of which the draws are made, and thus they do know the probabilities of each possible sample of n firms. We assume that firms make a decision based on naive best-responses, which means that each firm bestresponds to the quantities and prices set by the other firms. Since the decision problem for each quantitysetter (price-setter) is the same by symmetry, we assume that every quantity-setter (price-setter) sets the same quantity (price). Such a system is called a quasi-symmetric system. However, since in this model the firms are not aware of the exact distribution of their market, they use the binomial probabilities to calculate the expected profits. This is because the profits that come with a specific choice of quantity or price may vary between different markets. We therefore give an indication of the average equilibrium profits by calculating the expected equilibrium profits. The equilibrium price (quantity) of the price-setters (quantity-setters) and their expected profits can be written as a function of n, and . Lemma 2. The Nash equilibrium prices and quantities of price-setters and quantity-setters respectively, as function of n, and equal where with Here is a k×n-k matrix with each element equal to , and a k×k matrix with all diagonal elements equal to one, and all the off-diagonal elements equal to . Firms Know the Population Parameter ρ This model uses a structure similar to Hommes et al. (2013).We assume that there is a very large population of firms where a fraction of the firms uses prices as strategic variables, and a fraction 1- uses quantities. We consider a two-shot game where in the first period 33 MSC-LEVEL | ECONOMETRICS | SPECIALTY | NL Analogously to earlier research (e.g. Singh and Vives (1984)) we consider the market game as a two-stage game, where in the first stage each firm chooses its strategy (price-setting or quantity-setting) and in the second stage they choose the actual value of their strategic variable. We assume that in the second stage the Nash-equilibrium is played. In order to determine this equilibrium we start with defining our market demand structure. We use a demand structure based on the one used by Häckner (2000) and Matsumoto and Szidarovszky (2010). We consider an oligopoly that consists of n firms. In our model represents the degree of substitutability between the goods. We consider the case where goods are imperfect substitutes, 0< <1. Suppose that there are k price-setters in the market, we then denote the vectors of prices and quantities of price-setters by and respectively, and the vectors of prices and quantities of quantitysetters by and respectively. Without loss of generality, assume from now on that the first k firms are price-setters and the last n-k firms are quantity-setters. This means that the first k firms will determine their prices ( ) while the last n-k firms will determine their quantities ( ). We have derived that the model in terms of the strategic variables can be written as Recommended for readers of XXx-level The Market Recommended for readers of XXx-level MSC-LEVEL | ECONOMETRICS | SPECIALTY | NL and the resulting expected equilibrium profits equal Assuming a sufficiently large population means that one firm switching from one strategy to another does not significantly influence the population parameter . Therefore we compare ( ) and ( ). Numerical simulations show that the equilibrium profits of the quantity-setters are higher than these of the price-setters. Thus our simulations suggest that for this model we have Conjecture 1. The quantity-setting strategy yields higher profits than the price-setting strategy for all n, and . This represents the conjecture that in the two-stage game as described in this section, it is always more profitable to choose the quantity-setting strategy in the first stage. This result has important implications for the twostage game that we consider. It implies that in the first stage each firm will choose the quantity-setting strategy. In that case, in the second stage the Cournot equilibrium will be the outcome. Note that this result is not dependent on the actual degree of substitutability between the goods. The conjecture is in agreement with earlier research (e.g. Singh and Vives (1984), Tanaka (2001)) that suggest that the quantity-setting strategy is more profitable than the price-setting strategy. Note that this result holds only under the rational assumption that in the second stage always the Nash equilibrium is played. However, the assumption of perfectly rational agents has been put into question by recent literature (originating from Simon (1957)). In earlier research on oligopolies we have also seen that Cournot competition (competition where all firms are quantity-setters) under naive expectations can become unstable as the number of firms n in the market becomes larger ((Theocaris (1960)). This makes it interesting to investigate the stability of the system. Stability: Firms Know the Population Parameter ρ The first step to investigate the stability of the system is to write the system as a quasi-symmetric dynamic system, which is necessary to reduce it to a 2D system. Remember that quasi-symmetric means that each quantity-setter sets the same quantity, and each pricesetter sets the same price. In each period the firms best respond (based on naive expectations) to the quantities and prices in the previous period. This leads us to the following dynamics of the price set by price-setters (1) and the quantity set by the quantity-setters (2) (1) 34 (7) (2) Since the system is linear, we can use the Jacobian of the system to determine the stability. A straightforward computation shows that this Jacobian equals (3) Recommended for readers of XXx-level MSC-LEVEL | ECONOMETRICS | SPECIALTY | NL where Since eigenvalues of this Jacobian are complicated functions of n, and , we will discuss stability of the extreme cases of Cournot competition ( = 0) and Bertrand competition ( = 1). In the case of Cournot competition there are no price-setters, so that the first equation is not relevant. We then only have to consider the bottom right element of (3) (the derivative of with respect to ), which for = 0 simplifies to (4) 35 Indications of the stability of the system for =(blue) 0.25,(green) 0.5 and (red) 0.75, with n on the horizontal axis and on the vertical axis.The area above the graph is the stable area, whereas the area below the graph is the unstable area. so that we find that Cournot competition is unstable for (5) Evolutionary Population Parameter In the case of perfect substitutes ( =1) this means that the system is unstable for n > 3. This corresponds to the special case investigated by Theocharis (1960). As can be seen from (5), for every value of with 0< <1 there exists a value n* such that the best-response dynamics in Cournot competition are unstable if n>n*. Similarly for Bertrand competition we consider only the top left element of (3) (the derivative of with respect to ), which is for = 1 simplifies to Recommended for readers of XXx-level MSC-LEVEL | ECONOMETRICS | SPECIALTY | NL (6) which is always smaller than 1/2 for 0< <1.This means that Bertrand competition is always stable for these values of , independent of n. The eigenvalues of the Jacobian (3) under mixed competition are complicated functions of n, and . In order to investigate stability we therefore use a numerical approach. Figure 1 illustrates for each n the highest value of for which the dynamics are unstable. When dynamics are always stable for a particular value of n, the critical value of is shown as zero. It can be seen that as n grows larger, the value of for which the dynamics become stable tends towards a value slightly below 0.25, with only a small difference based on . Differences between the three different degrees of substitutability are more noteworthy in the movement for low values of n.The higher the degree of substitutability, the higher needs to be to have a stable system for low values of n.This result is intuitively clear, since a higher degree of substitutability means that firms are more dependent on each other, and thus their best-responses will be more narrowly connected. In the extreme case of =0, changes in other firms’ strategies would have no impact on the decision of the firm. Previously we have assumed that the population parameter remained unchanged over time. However, as performance of the two different strategies is known to the firms, this assumption does not seem very realistic. The next step is thus to let this parameter depend on the past performance of the strategies, so that better performing strategies are more likely to be used. To implement an evolutionary population parameter we consider two dynamics. The first dynamic is based on Anufriev and Hommes (2012), and the second dynamic builds on replicator dynamics (Droste et al. (2002)). We start by defining a performance measure for time t, denoted by . This performance measure represents the performance of the price-setting strategy compared to the quantity-setting strategy, and it is based on past profits of both strategies. We use a performance measure of the following form (7) where represents the memory of the performance rule. Next, we calculate the population parameter on time t, , which for the first model is done according to the following formula (8) where represents the idea that not all firms directly update their strategy. The parameter represents the intensity of choice, as a higher value of this parameter causes firms to switch to the best performing strategy more rapidly. However, under the evolution described in (8) will always converge to = 0.5 when each of the strategies is equally profitable. Since we would like to capture the idea that will remain constant in that case (people will not change their strategy if both strategies are equally profitable), instead of converging to 0.5, we adapted the evolution of to follow the exponential replicator dynamics, similar to Droste et al. (2002). Using the same performance measure as before, we now determine by (9) This function has the property that as long as the price-setting strategy is more (less) profitable than the quantity-setting strategy, the value of will slowly move up (down). Note however that the extreme cases of =0 and =1 are steady states. This can lead to hghghbhhg Figure 1. Stability of the System with Population Parameter 36 problems in the initial state of the process if is high. In order to deal with this we consider two possible solutions. The first solution is to lower the value of , and increase the amount of time periods. This can be done without loss of generality, since the time is not an absolute measure here. The general behavior of the dynamics thus remains representative. The second solution introduces a small noise, such that steady states are not available. Evolution according to the dynamics with noise is described by (10) Findings using the exponential replicator dynamics for the population parameter are in agreement with the findings using (8) for both of the mentioned solutions. Starting in a stable state, with sufficiently large, the quantity-setting strategy will be most profitable and thus lead to a decrease of . However, as the value of becomes too low the instability of the system causes the price-setting strategy to be most profitable. As a result the value of drastically increases again to re-stabilize the system, and the cycle repeats. Although we do notice that the price-setting strategy will not be completely driven out, instability alone is not sufficient to make this strategy most profitable. Good examples of these dynamics are shown in Figure 2. We added the critical value for instability (given the parameter values) in the figure and observe that it is higher than the turning point for the values considered. This is the case for other combinations of parameter values as well. As it seems, even when the system is unstable it is still possible that the quantity-setting strategy is most profitable. However we always observe a turning point where this is no longer the case, and the system moves back to a stable state to restart the cycle. In the model with noise, as described in (10), a low value of will lead to a fraction of price-setters slightly below 0.5. This is similar to the findings when using the model described in (8), and it happens because the noise is in that case too large compared to the change caused by the difference in profits, so that the fraction of price-setter will remain somewhere slightly below 0.5. This can again be limited by reducing the value of the shock further (e.g. to a uniform(0, 0.01) distributed variable), but the general idea remains the same. An example is shown in Figure 3. Figure 2b. Exponential Replicator Dynamics Typical dynamics using the exponential replicator function (a) (9) or (b) (10) for. The vertical axis shows , and the horizontal axes shows the period t. The red line indicates the critical value of for the given parameter values, as calculated in section 4. Parameter values are n = 10, =0.75, =0.5, =0.4 and =(a) 50 or (b) 43000. Initial values are =0.4, = =0.01, and = =0. Figure 3 - Exponential Replicator Dynamics with noise Figure 2a. Exponential Replicator Dynamics The evolution of the fraction of price-setters under dynamics described in (10), when the value of is ‘too low’.The vertical axis shows , and the horizontal axes shows the period t. The red line indicates critical value 37 MSC-LEVEL | ECONOMETRICS | SPECIALTY | NL to be uniformly distributed between 0 and 0.05 to represent a small but sufficient shock. Recommended for readers of XXx-level where we choose of for the given parameter values, as calculated in section 4. Parameter values are n=10, =0.75, = =0.5 and =4000. Initial values are =0.4, = =0.01, and = =0. Sakai, Y., S. Aguchi, H. Ishigaki (1995). “Price and Quantity Competition: Do Mixed Oligopolies Constitute an Equilibrium?” Keio economic studies 32, 15-25. Conclusion Simon, H. A. (1957). “A Behavior Model of Rational Choice,” Models of Man: Social and Rational, 196-279. In this research we establish that instability caused by competition in quantities is sufficiently large to allow the pricesetting strategy to remain in a mixed oligopoly. We observe a cyclical switching between both strategies in most cases, but the quantity-setting strategy is never able to drive out the price-setting strategy. Thus, despite the fact that the quantitysetting strategy is always most profitable inside equilibrium, the price-setting strategy can still be a more profitable option when perfect rational play is no longer assumed. Literature Recommended for readers of XXx-level MSC-LEVEL | ECONOMETRICS | SPECIALTY | NL Anufriev, M. and C. Hommes (2012). Evolution of Market Heuristics. Knowledge Engineering Review, 27, 255 - 271. Singh, N. and X.Vives (1984). “Price and Quantity Competition in a Differentiated Duopoly,” RAND Journal of Economics 15, 546-554. Tanaka, Y. (2001). “Profitability of Price and Quantity Strategies in an Oligopoly,” Journal of Mathematical Economics 35, 409-418. Theocharis, R. D. (1960). “On the Stability of the Cournot Solution on the Oligopoly Problem,” Review of Economic Studies, 27, 133-134. Anufriev, M., D. Kopányi, and J.Tuinstra (2013).“Learning Cycles in Bertrand Competition with Differentiated Commodities and Competing Learning Rules,” Journal of Economic Dynamics and Control 37, 2562-2581. Bertrand, J (1883). “Revue de la Théorie Mathématique de la Richesse Sociale et des Recherches sur les Principles Mathématiques de la Théorie des Richesses.” Journal des Savants, 499-508. Cournot, A. (1838). “Researches into the Mathematical Principles of the Theory of Wealth.” Dixit, A. (1979). “A Model of Duopoly Suggesting a Theory of Entry Barriers,” Bell Journal of Economics 10, 20-32. Droste, E., C.H. Hommes, and J. Tuinstra (2002). “Endogenous Fluctuations under Evolutionary Pressure in Cournot Competition,” Games and Economic Behaviour 40, 232-269. Hackner, J. (2000).“A Note on Price and Quantity Competition in Differentiated Oligopolies,” Journal of Economic Theory 93, 233-239. Hommes, C., J.Tuinstra and M. Ochea (2013). “On the Stability of the Cournot Equilibrium under Competing Learning Rules,” Working Paper. Matsumoto, A. and F. Szidarovszky. “Mixed Cournot-Bertrand Competition in N-Firm Differentiated Oligopolies.” Working Paper [1]. Matsumoto, A. and F. Szidarovszky (2010) “Price and Quantity Competition in Differentiated Oligopoly Revisited,” Working Paper [2]. 38 Rob Goedhart Rob will start his PhD at IBIS UvA. He obtained his Masters degree Mathematical Economics at the University of Amsterdam. This article summarizes part of his master thesis, which he wrote under supervision of Prof. dr. J. Tuinstra.
© Copyright 2026 Paperzz