Posted prices versus bargaining in markets_7 pdf

CHAPTER 9 The Role of the Trading Procedure 9.1 Introduction In this chapter we focus on the role of the trading procedure in determining the outcome of trade. The models of markets in the previous three chapters have in common the following three features. 1. The bargaining is always bilateral. All negotiations take place be- tween two agents. In particular, an agent is not allowed to make offers s imultaneously to more than one other agent. 2. The termination of an unsuccessful match is exogenous. No agent has the option of deciding to stop the negotiations. 3. An agreement is restricted to be a price at which the good is ex- changed. Other agreements are not allowed: a pair of agents cannot agree that one of them will pay the other to leave the market, or that they will execute a trade only under certain conditions. The strategic approach has the advantage that it allows us to construct models in which we can explore the role of these three features. 173 174 Chapter 9. The Role of the Trading Procedure As in other parts of the book, we aim to exhibit only the main ideas in the field. To do so we study several models, in all of which we make the following assumptions. Goods A single indivisible good is traded for some quantity of a divisible good (“money”). Time Time is discrete and is indexed by the nonnegative integers. Economic Agents In period 0 a single seller, whom we refer to as S, and two buyers, whom we refer to as B H and B L , enter the market. The seller owns one unit of the indivisible goo d. The two buyers have reservation values for the good of v H and v L , respectively, where v H ≥ v L > 0. No more agents enter the market at any later date (cf. Model B in Chapter 6). All three agents have time preferences with a constant discount factor of 0 < δ < 1. An agreement on the price p in period t yields a payoff of δ t p for the se ller and of δ t (v −p) for a buyer with reservation value v. If an agent does not trade then his payoff is zero. When uncertainty is involved we assume that the agents maximize their expected utilities. Information All agents have full information about the history of the mar- ket at all times: the seller always knows the buyer with whom she is matched, and every agent learns about, and remembers, all events that occur in the market, including the events in matches in which he does not take part. In a market containing only S and B H , the price at which the good is sold in the unique s ubgame perfect equilibrium of the bargaining game of alternating offers in which S makes the first offer is v H /(1 +δ). We denote this price by p ∗ H . When bargaining with B H , the seller can threaten to trade with B L , so that it appears that the presence of B L enhances her bargaining position. However, the threat to trade with B L may not be credible, since the surplus available to S and B L is lower than that available to S and B H . Thus the extent to which the seller can profit from the existence of B L is not clear; it depends on the exact trading procedure. We start, in Section 9.2, with a model in which the three features men- tioned at the beginning of this section are retained. As in the previous three chapters we assume that the matching process is random and is given ex- ogenously. A buyer who rejects an offer runs the risk of losing the seller and having to wait to be matched anew. We show that if v H = v L then this fact improves the seller’s bargaining pos ition: the price at which the good is sold exceeds p ∗ H . 9.2 Random Matching 175 Next, in Section 9.3, we study a model in which the seller can make an offer that is heard simultaneously by the two buyers. We find that if v H is not too large and δ is close to 1, then once again the presence of B L increases the equilibrium price above p ∗ H . In Section 9.4 we assume that in each period the seller can choos e the buyer with whom to negotiate. The results in this case depend on the times at which the seller can switch to a new buyer. If she can switch only after she rejects an offer, then the equilibrium price is precisely p ∗ H : in this case a threat by S to abandon B H is not credible. If the seller can switch only after the buyer rejects an offer, then there are many subgame perfect equilibria. In some of these, the equilibrium price exceeds p ∗ H . Finally, in Section 9.5 we allow B H to make a payment to B L in ex- change for which B L leaves the market, and we allow the seller to make a payment to B L in exchange for which B L is committed to buying the good at the price v L in the event that S doe s not reach agreement with B H . The equilibrium payoffs in this model coincide with those predicted by the Shapley value; the equilibrium payoff of the seller exceeds p ∗ H . We see that the results we obtain are sensitive to the precise character- istics of the trading procedure. One general conclusion is that only when the proce dure allows the seller to effectively commit to trade with B L in the event she does not reach agreement with B H does she obtain a price that exceeds p ∗ H . 9.2 Random Matching At the beginning of each period the seller is randomly matched with one of the two buyers, and one of the matched parties is selected randomly to make a proposal. Each random event occurs with probability 1/2, independent of all past events. The other party can either accept or reject the proposal. In the event of acceptance, the parties trade, and the game ends. In the event of rejection, the match dissolves, and the seller is (randomly) matched anew in the next period. Note that the game between the seller and the buyer with whom she is matched is similar to the model of alternating offers with breakdown that we studied in Section 4.2 (with a probability of breakdown of 1/2). The main difference is that the payoffs of the agents in the event of breakdown are determined endogenously rather than being fixed. 9.2.1 The Case v H = v L Without loss of generality we let v H = v L = 1. The game has a unique subgame perfect equilibrium, in which the good is sold to the first buyer to be matched at a price close to the competitive price of 1. 176 Chapter 9. The Role of the Trading Procedure Proposition 9.1 If v H = v L = 1 then the game has a unique subgame perfect equilibrium, in which the good is sold immediately at the price p s = (2 − δ) 2 /(4 − 3δ) if the seller is selected to make the first offer, and at the price p b = δ(2 − δ)/(4 − 3δ) if the matched buyer is selected to make the first offer. These prices converge to 1 as δ converges to 1. Proof. Define M s and m s to be the supremum and the infimum of the seller’s payoff over all subgame perfect equilibria of the game. Similarly, define M b and m b to be the corresponding values for either of the buyers in the same game. Four equally probable events may occur at the beginning of each period. Denoting by i/j the event that i is s elec ted to make an offer to j, these events are S/B H , B H /S, S/B L , and B L /S. Step 1. M s ≥ (2(1 −δm b ) + 2δM s ) /4 and m b ≤ (1 −δM s + δm b )/4. Proof. For every subgame perfect equilibrium that gives j a payoff of v we can construct a subgame perfect equilibrium for the subgame starting with the event i/j such that agreement is reached immediately, j’s payoff is δv and i’s payoff is 1 −δv. The inequalities follow from the fact that there exists a subgame perfect equilibrium such that after each of the events S/B I the good is sold at a price arbitrarily close to 1 −δm b , and after each of the events B I /S the good is sold at a price arbitrarily close to δM s . Step 2. m b = (1 −δ)/(4 − 3δ) and M s = (2 −δ)/(4 − 3δ). Proof. The seller obtains no more than δM s when she has to respond, and no more than 1−δm b when she is the proposer. Hence M s ≤ (2δM s +2(1− δm b ))/4. Combined with Step 1 we obtain M s = (2δM s + 2(1 −δm b )) /4. Similarly, a buyer obtains at least 1−δM s when he is matched and is chosen to be the proposer, and at least δm b when he is matched and is chosen to respond. Therefore m b ≥ (1−δM s +δm b )/4, which, combined with Step 1, means that m b = (1 − δM s + δm b )/4. The two equalities imply the result. Step 3. M b ≤ 1 −m b − m s . Proof. This follows from the fact that the most that a buyer gets in equilibrium does not exceed the surplus minus the sum of the minima of the two other agents’ payoffs. Step 4. M s = m s = (2 −δ)/(4 − 3δ) and M b = m b = (1 −δ)/(4 − 3δ). Proof. If the seller is the responder then she obtains at least δm s , and if she is the proposer then she obtains at least 1−δM b.ByStep 3wehave 1−δM b ≥ 1 −δ(1 −m b −m s ), so that m s ≥ [2δm s + 2(1 −δ(1 −m b −m s ))]/4, which implies that m s ≥ 1/2+δm b /[2(1−δ)] = 1/2+δ/[2(4−3δ)] = M s . Finally, we have M b ≤ 1 −m b − m s = (1 −δ)/(4 − 3δ) = m b . 9.2 Random Matching 177 By the same argument as in the proof of Theorem 3.4 it follows that there is a unique subgame perfect equilibrium in which the seller always prop os es the price 1 − δM b = p s , and each buyer always offe rs the price δM s = p b .  Note that the technique used in the proof of Step 1 is different from that used in the proofs of Steps 1 and 2 of Theorem 3.4. Given a collection of subgame perfect equilibria in the subgames starting in the second period we construct a subgame perfect equilibrium for the game starting in the first period. This line of argument is useful in other models that are similar to the one here. So far we have assumed that a match may be broken after any offer is rejected. If instead a match may be broken only after the seller rejects an offer, then the unique subgame perfect equilibrium coincides with that in the game in which the seller faces a single buyer (and the proposer is chosen randomly at the start of each period). The prices the agents propose thus converge to 1/2 as δ converges to 1. On the other hand, if a match may be broken only after a buyer rejects an offer, then there is a unique subgame perfect equilibrium, which coincides with the one given in Proposition 9.1. This leads us to a conclusion about how to model competitive forces. If we want to capture the pressure on the price caused by the presence of more than one buyer, we must include in the model the risk that a match may be broken after the buyer rejects an offer; it is not enough that there be this risk only after the seller rejects an offer. We now consider briefly the case in which the probability that a match terminates after an offer is rejected is one, rather than 1/2: that is, the case in which the seller is matched in alternate periods with B H and B L . Retaining the assumption that the proposer is selected randomly, the game has a unique subgame perfect equilibrium, in which the seller always pro- poses the price 1, and each buyer always proposes the price p b = δ/(2 −δ). (The equation that determines p b is p b = δ(1/2 + p b /2).) A buyer accepts the price 1, since if he does not then the good will be sold to the other buyer. When a buyer is selected to make a proposal he is able to extract some surplus from the seller since she is uncertain whether she will be the prop os er or the responder in the next match. If we assume that the matches and the selection of proposer are both deterministic, then the subgame perfect equilibrium depends on the order in which the agents are m atched and chosen to propose. If the order is S/B I , B I /S, S/B J , B J /S (for {I, J} = {L, H}), then the unique subgame perfect equilibrium is essentially the same as if there were only one buyer: the seller always proposes the price 1/(1 + δ), while each buyer always proposes δ/(1 + δ). If the order is B I /S, S/B I , B J /S, S/B J then in the unique 178 Chapter 9. The Role of the Trading Procedure subgame perfect equilibrium the seller always proposes the price 1, while each buyer always proposes the price δ. The comparison between these two protocols demonstrates again that in order to model the competition between the two buyers we need to construct a model in which a match is broken after a buyer, rather than a seller, rejects an offer. 9.2.2 The Case v H > v L We now turn to the case in which the buyers have different reservation values, with v H > v L . We return to our initial assumptions in this section that each match is terminated with probability 1/2 after a rejection, and that the probability that each of the parties is chosen to be the proposer is also 1/2. If v H /2 > v L and δ is close enough to 1, then there is a unique subgame perfect equilibrium in which the good is sold to B H at a price close to v H /2. The intuition is that the seller prefers to sell the good to B H at the price that would prevail were B L absent from the market, so that both the seller and B H consider the termination of their match to be equally appalling. We now consider the case v H /2 < v L . (This is the case we considered in Section 6.5.) In this case, the game does not have a stationary subgame perfect equilibrium if δ is close to 1. The intuition is as follows. Assume that there is a stationary subgame perfect equilibrium in which the seller trades with B L when she is matched with him, for at least one of the two choices of proposer. The interaction between S and B H is then the same as in a bilateral bargaining game in which with probability at least 1/4 the match does not continue: negotiations between S and B H break down, and an agreement is reached between S and B L . This breakdown is exogenous from the point of view of the interaction between S and B H . The payoff of B H of such a breakdown is zero, and some number u ≤ 3v H /4+v L /4 < v H for the seller. The equilibrium price in the bargaining between S and B H is therefore approximately (u + v H )/2 when δ is close to 1. Since (u + v H )/2 > u, it is thus better for the seller to wait for an opportunity to trade with B H than to trade with B L . Thus in no stationary equilibrium does the seller trade with B L . Now consider a stationary subgame perfect equilibrium in which the seller trades only with B H . If δ is close to 1, the surplus v H is split more or less equally between the seller and B H . However, given the assump- tion that v L > v H /2, buyer B L should agree to a price between v L and v H /2, and the seller is better off waiting until she is matched with B L and has the opportunity to make him such an offer. Therefore there is no stationary equilibrium in which with probability 1 the unit is sold to B H . 9.2 Random Matching 179 T H T HL prop os es to B H p ∗ p ∗ S proposes to B L p ∗ v L accepts from B H p ≥ v L p ≥ v L accepts from B L p > v L p ≥ v L B H proposes v L v L accepts p ≤ p ∗ p ≤ p ∗ B L proposes v L v L accepts p ≤ v L p ≤ v L Transitions Go to T HL if B H re- jects a price p ≤ p ∗ . Go to T H after any re- jection except a rejec- tion of p ≤ p ∗ by B H . Table 9.1 A nonstationary subgame perfect equilibrium for the model of Section 9.2.2, under the assumption that v L < v H < 2v L . The price p ∗ is equal to (4 − 3δ)v L /δ (> v L ). We now describe a nonstationary subgame perfect equilibrium. There are two states, T H (“trade only with B H ”) and T HL (“trade with both B H and B L ”), and p ∗ = (4 − 3δ)v L /δ > v L . The initial state is T H . The strategies are given in Table 9.1. We now check that this strategy profile is a subgame perfect equilibrium for δ close enough to 1. The price p ∗ is chosen so that in each state the expected utility of the seller before being matched is v L /δ. (In state T H this utility is the number V that satisfies V = (v L + p ∗ )/4 + δV/2; in state T HL it is p ∗ /4 + 3v L /4.) Therefore in each state the seller is indifferent between selling the good at the price v L and taking an action that delays agreement. Hence her strategy is optimal. Now consider the strategy of B H . It is optimal for him to accept p ∗ in state T H since if he rejects it then the state changes to T HL , in which he obtains the good only with probability 1/2. More precisely, if he ac- cepts p ∗ he obtains v H − p ∗ , while if he rejects it he obtains δ[(1/2) · 0 + (1/4) · (v H − p ∗ ) + (1/4) · (v H − v L )] < v H − p ∗ if δ is close enough to 1. For a similar reason, B H cannot benefit by proposing a price less than v L in either state. It is optimal for him to reject p > p ∗ in both states since if he accepts it he obtains v H − p, while if he rejects it, the state either remains or becomes T H , and he obtains close to the average of v H − p ∗ 180 Chapter 9. The Role of the Trading Procedure and v H − v L if δ is close to 1. Precisely, his expected utility before being matched in state T H is v H /(2 − δ) − v L /δ (the number V that satisfies V = (1/2)(v H − (v L + p ∗ )/2) + (1/2)δV ), which exceeds v H − p if δ is close e nough to 1 and p > p ∗ . Finally, B L ’s strategy is optimal since his expected utility is zero in both states. This equilibrium is efficient, since the good is sold to B H at the first opportunity. However, the argument shows that there is another subgame perfect equilibrium, in which the initial state is T HL rather than T H , which is inefficient. In this equilibrium the good is sold to B L with probability 1/2. We know of no characterization of the set of all subgame perfect equilibria. 9.3 A Model of Public Price Announcements In this section we relax the assumption that bargaining is bilateral. The seller starts the game by announcing a price, which both buyers hear. Then B H responds to the offer. If he accepts the offer then he trades with the seller, and the game ends. If he rejects it, then B L responds to the offer. If both buyers reject the offer, then play passes into the next period, in which both buyers simultaneously make counteroffers. The seller may accept one of these, or neither of them. In the latter case, play passes to the next period, in which it is once again the seller’s turn to announce a price. Recall that p ∗ H = v H /(1 + δ), the unique subgame perfect equilibrium price in the bargaining game of alternating offers between the seller and B H in which the seller makes the first offer. Proposition 9.2 If δp ∗ H < v L , then the model of public price announce- ments has a subgame perfect equilibrium, and in all subgame perfect equi- libria the good is sold (to B H if v H > v L ) at the price p ∗ = δv L +(1−δ)v H . If δp ∗ H > v L then the game has a unique subgame perfect equilibrium. In this equilibrium the good is sold to B H at the price p ∗ H . Thus if the value to the seller of receiving p ∗ H with one period of delay is less than v L then the seller gains from the existence of B L : p ∗ > p ∗ H . The price p ∗ lies between v L and v H ; it exceeds v L if v H > v L , and converges to v L as δ converges to 1. By contrast, if the value to the seller of receiving p ∗ H with one period of delay exceeds v L , then the existence of B L does not improve the seller’s position. This part of the result is similar to the first part of Proposition 3.5, which shows that the fact that a player has an outside option with a payoff lower than the equilibrium payoff in bilateral bargaining does not affect the bargaining outcome. Proof of P roposition 9.2. If δp ∗ H > v L then there is a subgame perfect equilibrium in which S and B H behave as they do in the unique subgame 9.3 A Model of Public Price Announcements 181 perfect equilibrium of the bargaining game of alternating offers between themselves. The argument for the uniqueness of the equilibrium outcome is similar to that in the proof of the first part of Proposition 3.5. Now consider the case δp ∗ H < v L . The game has a stationary subgame perfect equilibrium in which the seller always proposes the price p ∗ , and accepts the highest proposed price when that price is at least v L , trading with B H if the proposed prices are equal. Both buyers propose the price v L ; B H accepts any price at most equal to p ∗ , and B L accepts any price less than v L . Notice that the seller is better off accepting the price v L than waiting to get the price p ∗ since δp ∗ H < v L implies that δp ∗ = δ 2 v L + δ(1 − δ)v H <v L , and B H is indifferent between accepting p ∗ and waiting to get the price v L , since v H − p ∗ = δ(v H − v L ). We now show that in all s ubgame perfect equilibria the good is sold (to B H if v H > v L ) at the price p ∗ . Let M s and m s be the supremum and infimum, respectively, of the seller’s payoff over all subgame perfect equilibria of the game in which the seller makes the first offer, and let M I and m I (I = H, L) be the supremum and infimum, respectively, of B I ’s payoff over all subgame perfect equilibria of the game in which the buyers make the first offers. Step 1. m H ≥ v H − max{v L , δM s }. Proof. This follows from the facts that the seller must acc ept any price in excess of δM s , and B L never proposes a price in excess of v L . Step 2. M s ≤ p ∗ (= δv L + (1 −δ)v H ). Proof. We have M s ≤ v H −δm H by the argument in the proof of Step 2 of Theorem 3.4, and thus by Step 1 we have M s ≤ v H −δ(v H −max{v L , δM s }). If δM s ≤ v L the result follows. If δM s > v L then the result follows from the assumption that δp ∗ H < v L . Step 3. M H ≤ v H − v L . Proof. From Step 2 and δp ∗ H < v L we have δM s < v L , so the seller must accept any price slightly less than v L . If there is an equilibrium of the game in which the buyers make the first offers for which B H ’s payoff exceeds v H − v L then in this equilibrium B L ’s payoff is 0, and hence B L can profitably deviate by proposing a price close to v L , which the seller accepts. Step 4. m s ≥ p ∗ . Proof. Since B H must accept any price p for which v H − p > δM H , we have m s ≥ v H − δM H ≥ p ∗ (using Step 3). 182 Chapter 9. The Role of the Trading Procedure We have now shown that M s = m s = p ∗ and M H = m H = v H −v L . Since p ∗ ≥ v L , the sum of the payoffs of S and B H is at least p ∗ +δ(v H −v L ) = v H , so that the game must end with immediate agreement on the price p ∗ . If v H > v L then p ∗ > v L , so that it is B H who accepts the first offer of the seller.  Note that if δ = 1 in this model then immediate agreement on any price between v L and v H is a subgame perfect equilibrium outcome. Note also that if B L responds to an offer of the seller before rather than after B H , or if the responses are simultaneous, then the result is the same. 9.4 Models with Choice of Partner Here we study two models in which the seller chooses the buyer with whom to bargain. The models are related to those in Section 3.12; choosing to abandon one’s current partner is akin to “opting out”. In Section 3.12, the payoff to opting out is exogenous. Here, the corresponding payoff is determined by the outcome of the negotiations with the new buyer, which in turn is affected by the possibility that the seller can move back to the first buyer. In both models, the seller and a buyer alternate offers until either one of them accepts an offer, or the seller abandons the buyer. In the latter case, the seller starts negotiating with the other buyer, until an offer is ac ce pted or the seller returns to the first buyer. The m ain difference between the models lies in the times at which the seller may replace her partner. In the first model, the seller is the first to make an offer in any partnership, and can switch to the other buyer only at the beginning of a period in which she has to make an offer (cf. the model in Section 3.12.1). In the second model, it is the buyer who makes the first offer in any partnership, and the seller can switch to another buyer only at the beginning of a period in which the buyer has to make an offer (cf. the model in Section 3.12.2). By comparison with the model of Section 9.3, the seller has an extra tool: she can threaten to terminate her negotiations with one of the buyers if he does not accept her demand. On the other hand, when matched with the seller a buyer is in a less competitive situation than in the model of public price announcements since he is the only buyer conversing with the seller. 9.4.1 The Case in Which the Seller Can Switch Partners Only Before Making an Offer This model predicts a price equal to the equilibrium price in bilateral bar- gaining between the seller and B H . The fact that the seller confronts more [...]... or may retain only partial information about his experience He may obtain information about the histories of other agents or may have no information at all about the events in bargaining sessions in which he did not take part In this chapter we focus on an assumption made in Chapter 8 that agents cannot condition their behavior in a bargaining encounter on their experience in previous encounters, or... of models in which, as in Section 9.3, sellers announce prices, which all buyers hear (ex ante pricing), and the equilibria of models in which (as in Section 9.2, for example) prices are determined by bargaining after a match is made (ex post pricing) Peters (1991) considers a model of a large market; when the agents’ common discount factor is close to 1 the equilibrium sequence of ex ante prices as... that in a model in which the seller chooses whether or not to terminate bargaining with one buyer and move to the other, the results do not capture our intuition about competition between the buyers In the first model, the presence of BL is irrelevant for the equilibrium The reason is clear, in light of the analysis in Section 3.12.1 The seller can never obtain more by moving to BL than by staying with... a simpler case in which each agent does possess information about other matches Information In period t each agent has perfect information about all the events that occurred through period t − 1, including the events in matches in which he did not participate When taking an action in period t, however, each agent does not have any information about the other matches that are formed in that period or... We conclude by investigating a model in which the range of contracts available to the agents is greater than it is in the models of the previous sections The buyers are allowed to agree to a contract according to which BH pays BL a sum of money and in exchange BL leaves the market This contract leaves BH alone in the market and thus presumably puts him in a better bargaining position In addition, the... monetary holdings of her existing partner relative to that of a fresh buyer Peters (1988) studies a model that contains elements from the models of Sections 9.3 and 9.4 Sellers post prices, but a buyer who is matched with a seller has the option of making a counteroffer; the seller can accept this offer, reject it and continue bargaining, or terminate the match When excess demand is small, posted prices are... previous sections and, as in Chapter 6, use the Nash solution to model bargaining; we restrict attention to the case vH > vL and assume that δ = 1 The trading procedure is the following In each period, two agents are matched and reach the agreement given by the Nash solution of the appropriate bargaining problem If no agreement has been reached, then all three possible matches (including that between BH... equally probable, and the matching is independent across periods Bargaining After a buyer and a seller have been matched they engage in a short bargaining process First, one of the matched agents is selected randomly (with probability 1/2) to propose a price between 0 and 1 Then the other agent responds by accepting the proposed price or rejecting it Rejection dissolves the match, in which case the agents... to vH /2, since δ = 1): the seller gains from the existence of BL We have already mentioned that one of the attractions of models of matching and bargaining is that they enable us to interpret and better understand solution concepts from cooperative game theory The model of this section illustrates this point Notes The random matching model of Section 9.2 is based on Rubinstein and Wolinsky (1990);... utility is zero In most of the chapter, we consider the case δ = 1 Matching In each period any remaining sellers and buyers are matched pairwise The matching technology is such that each seller meets exactly one buyer and no buyer meets more than one seller in any period Since there are fewer sellers than buyers, B − S buyers are thus left unmatched in each period The matching process is random: in each period . Trading Procedure 9.1 Introduction In this chapter we focus on the role of the trading procedure in determining the outcome of trade. The models of markets in the previous three chapters have in. events that occur in the market, including the events in matches in which he does not take part. In a market containing only S and B H , the price at which the good is sold in the unique s ubgame. in bilateral bargaining does not affect the bargaining outcome. Proof of P roposition 9.2. If δp ∗ H > v L then there is a subgame perfect equilibrium in which S and B H behave as they do in