This action might not be possible to undo. Are you sure you want to continue?

OF ECONOMIC

THEORY

38, 238-260

(1986)

**Price Competition a Capacity-Constrained
**

MARTIN J. OSBORNE

Department of Economics, Columbia and Department of Economics, Hamilton, Canada AND University, MeMaster L8S 4M4

in Duopoly*

New York University.

10017,

CAROLYN PITCHIK

Department

**of Economics, Universily q/ Toronto. Toronto. Canada MSS IA1
**

June 15, 1984; revised July 25. 1985

Received

This paper characterizes the set of Nash equilibria in a model of price-setting duopoly in which each tirm has limited capacity, and demand is continuous and decreasing. In general there is a unique equilibrium, so that comparative static exercises are meaningful. The properties of the equilibrium conform with a number of stylized facts. The equilibrium prices are lower, the smaller is demand relative to capacity. When demand is in an intermediate range, the firms use mixed are chosen simultaneously, strategies-they randomly hold “sales.” If capacities before prices, the set of equilibrium capacities coincides with the set of Cournot quantities. Journal of Economic Literature Classification Numbers: 022, 611.

I? 1986 Academic Press. Inc.

1. INTRODUCTION A model of a price-setting duopoly is a natural starting point for a theory of the behavior of oligopolists. However, such a model has been completely solved only under a particularly unrealistic assumptionnamely, that each firm can produce an unlimited quantity (or, at least as much as is demanded at the breakeven price) at constant unit cost. The

* We thank Jean-Pierre Benoit, Avinash Dixit, Vijay Krishna, Martin Shubik, and Aloysius Siow for useful comments. Osborne’s research was partially supported by a grant from the Columbia University Council for Research in the Social Sciences, and by National Science Foundation Grants SES-8318978 and SES-8510800. Pitchik’s research was partially supported by National Science Foundation Grant SES-8207765 and Office of Naval Research Contract NOO14-78-C-0598 at New York University.

238

0022-0531/86

Copyright All rights

$3.00

0 1986 by Academic Press, Inc. of reproduction in any form reserved.

PRICE

COMPETITION

IN A DUOPOLY

239

unique equilibrium in this case involves each firm setting the breakeven price, so that (unless capacity can be bought and sold at will) at most half of the available capacity is used in equilibrium. Given that the firms must at some point choose their capacities, it is difficult to see how such a situation could come about. We study the case where the capacity of each firm is limited, and cannot be instantly changed. We describe, for each pair of capacities, the set of Nash equilibria, which is in general a singleton. We assume that the unit cost is the same for each firm, and is constant up to capacity, and demand is continuous and decreasing. Previous analyses of the model are of two types. Either strong assumptions (e.g., linearity) are imposed on demand, or relatively weak results are proved about the character of equilibria. (For a discussion see Sect. 6.) In particular, little attention has been devoted to the uniqueness of equilibrium (except when demand is linear). This causes problems for economic applications, since without a uniqueness result we do not know if the characteristics of a particular equilibrium are shared by others, and we cannot legitimately perform comparative static exercises. Some consequences of our characterization are as follows (more details and discussion are contained in Sect. 4). The larger is capacity relative to demand, the lower are the equilibrium prices. This is, of course, intuitively plausible. It is noteworthy because it emerges, even though, up to capacity, the technologies of both firms are the same. One can obtain a similar prediction in a competitive model only by assuming, for example, that there are more and less efficient firms; when demand is low, the latter are forced out of business and the price falls. In our model the result emerges from the fact that there is, in a precise sense, more competition when there is excess capacity. In contrast to the competitive outcome, profit is at the monopoly level if capacities are small, even though there is more than one firm. When industry capacity is in an intermediate range, the equilibrium strategies involve randomization. The nature of the equilibrium distributions depends on the shape of the market demand function. Under some conditions, the large firm is most likely to charge either a high or a relatively low price, while the small firm is most likely to charge a low price. (For details, see Sect. 4.) Thus in this case, the large firm randomly holds “sales,” as in Varian [ 18 ]. As the size of the large firm increases, its equilibrium profit increases; as the size of the small firm increases, the greater competition induced may offset the direct effect on its profit from larger sales, and the net effect is uncertain. The force at work here is closely related to that studied by Gelman and Salop [9]; even if capacity is free, a small firm may not find it advantageous to expand indefinitely. As the size of the small firm converges to zero, the equilibrium approaches the monopoly outcome.

this is not enough to determine which scheme is chosen. We assume that if pi < pi then consumers first try to buy from firm i.e. Finally. 3 of Osborne and Pitchik [ 141). we make an assumption which appears to be consonant with the .) p < pO. 2 k. since i’s profit is independent of the scheme used (only j’s profit is affected). p2). somewhat loosely.1) We now wish to define the profits of the firms at each pair of prices (pi. The aggregate demand at the high price pj then depends on the way the limited supply k. in the Appendix we provide proofs.c. If demand varies cyclically. d(p) is the aggregate demand for the good when its price exceeds the unit cost by p. is allocated among the consumers. each with preferences which have no “income effect” (for details. pO). they turn to firm j. > 0. and the firms choose constant capacities appropriate for some intermediate level of demand.) There is a large number of identical consumers. while in slumps the firms will hold random “sales. when its supply (k. However. see Sect. 2. and d is continuous There exists p0 > 0 such that d(p) = 0 if p 2 p0 and d(p) > 0 if and decreasing on ( -c.240 OSBORNE AND PITCHIK In the game in which capacities are chosen simultaneously before prices. and constant. (Possible relaxations are discussed in Sect. cc ). we assume throughout that k. the set of pure capacity pairs chosen in Nash equilibria coincides with the set of (pure) Cournot equilibria (i.. Each firm can produce the same good at the same. and our results. the aggregate demand for the output of the firms as a function of price is D:R+ -+R+. there may be (depending on the shape of the demand function) Cournot equilibria which are not associated with subgame perfect equilibria of the two stage game. to an element of S as a “price. then in booms prices will be high. and in Section 6 we discuss the related literature.) is exhausted. Given the prices of all other goods. 5. (Whenever i and j appear in the same expression. Given that this is so. we mean that i is not equal to j. a slightly weaker version of the main result of Kreps and Scheinkman [ 1l] holds under our more general assumptions on demand). Firm i has capacity ki. However.” In the next two sections we describe the model. THE MODEL There are two firms. in Section 5 we consider possible extensions. We refer. Let p denote the excess of price over unit cost and let S = [ .” Define d: S + R + by d(p) = D(p + c). In Section 4 we discuss economic implications of our characterization. We make the following assumption on the demand function (which implies that profit p d(p) attains a maximum on 5’). constant unit cost c>O up to its capacity. It is natural to assume that the rationing scheme is chosen by firm i. (2.

d. 4~~)) P mWj..k.) Under these assumptions. each of the (identical) consumers is allowed to purchase the same fraction of ki.) P(k$ p-- pd(p) / / / - ’ .) = pi min(k. d(p. those with the highest reservation prices are served first by the low price frm.) Levitan and Shubik [12] and Kreps and Scheinkman [ 1l] assume that the residual demand has this form.)) -c P(k) / Y Plk.2) 4i(p) z if pi =pj =p. pj) = = pi Minsk. for i= 1. =p. respectively. _ . d(p.) is the profit of firm i when it sets p. the demand faced by firm j when p.... < pj and d(pi) > k. The functions L.) Under these assumptions.2. and p.. #.) . In this scheme. the profit of firm i when it sets the price p. it is one of two possibilities studied by Gelman and Salop [9]. E S and firm j sets pj E S is UP. \ FIG.) are its prolits when p. ki d(p)/k) if Pi < Pj3 (2. (Note that this means that the shape of the residual demand function does not depend on the value of p. L.(p. sp. > pi. (There is another assumption on the microeconomic structure of demand which generates the same result: there is a large number of consumers.PRICE COMPETITION IN A DUOPOLY 241 competitive nature of the model: firm i chooses the scheme which minimizes the profit of firm j.(p. is just max(O. and M.. < p. the price 1.(p..e. M..kj).) hitpi. . (I. max(O. each with a reservation price.(p. a rule like “limit two per customer” is imposed. rather than allowing some fraction of the customers to buy as much as they want.) and M. if p.) .

k.) = FEa. kz) in the strategy set of each firm is Y and the payoff function of firm i is 1. we can show that no firm ever charges a price above P(kz). Examples of the functions Li. 2) then (0. 1. . cumulative probability distribution functions on S). 3.e. each pair (k. These equilibria can be given clear interpretations.. Note that if p < P(k) then M. No firm will charge a price below P(k) (since it could then raise its price and still sell all its output). 2). where prices are driven down to unit costs. P(k)) is a pure strategy equilibrium (region III). Further.+ is the maximal profit of firm i from charging a price in excessof that charged by firm j. (3. At the other extreme.) when this does not cause confusion. Let 9’ be the set of mixed strategies (i. so the capacity limits are irrelevant. at p..3) a.).47 rather than M. and that if M. k.e.) It is also easy to check that if k.f+(ki. and hence MT a k. k.F. k.(P. kz). are shown in Fig.(P)(Fj(P)-a. M.) let WYk.(p) = pk.(p))+~i(P)crj(P) + L.. E 9 we have h. k. nor below 0 (since 0 guarantees a zero profit.1) We frequently write A..242 OSBORNE AND PITCHIK where k = k.O) is a pure strategy equilibrium. and M.)=M.. we study the price-setting game H(k. an(O) (region I of Fig. DESCRIPTION OF THE NASH EQUILIBRIA OF H(k. (see Fig.(P).(PH 1 -F.(P))... and of all subsequent arguments are given in the Appendix. and we are back to the standard Bertrand model. The domain of hi can be extended to 9 x Y in the natural way. In particular. di. if F. a price below unit cost) yields at least one of the firms a negative profit). so that the maximizer of Mi is at least equal to P(k). It is easy to put some restrictions on the range within which equilibrium prices must lie. and we assume that if pi = pl then demand is allocated in proportion to capacities. In region I each firm has more than enough capacity to meet the demand even at the breakeven price (p =0 under our normalization). l).) depend on the value of (k.(p) is the size of the atom (if any) in F.x M. while a negative price (i.P(k). For each (k. k. is maximized at P(k) (in which case the same is true for M2) then (P(k).) The qualitative characteristics of the Nash equilibria of H(k. . + k2. (The details of these. where For which hi (i= (2. in .

since L.) are of three types. P(k)) (region II). we can show that the support of every equilibrium strategy is a subset of g= [X(k). F. as follows: Region Type of (k.)... Using (2. P( k )) (and. in degenerate cases.(p) then F. <d(O) and MT > M. >. where k.3). The basic idea behind the construction of an equilibrium is very simple.(E. E.(P(k)) ( = k. where X(z) = max(O.. equilibria) Pure strategy equilibrium: (P(k). The equilibria of the of Equilibrium I II 111 k.k?) as a function game H(k. possibly also mixed strategy equilibria) region III there is undercapacity in the industry. Type of equilibrium of H(k. the only price not covered by this .) < Ei for all prices p. = h.PRICE COMPETITION IN A DUOPOLY 243 FE 2. In the remaining case.). where As argued above. this implies that if L. a unique mixed strategy equilibrium. In this case there is no incentive for “competition”: each firm is producing at capacity..(p) > Mi(p) if X(k) < p< X(k>).0) Mixed strategy equilibrium (or.(p) 2 Qi(p. F?)E Y x Y be an equilibrium and let E. no pure strategy equilibrium exists. P(z)).. k. in degenerate cases. and so cannot benelit from undercutting its rival. Let (F.(P(k)) My = M. 40) and M: > M.(P(k)) k. We show that in this case there is. F. Then h..(p) > M.(p.k?). <d(O) Pure strategy equilibrium: (0. X(k2)]. in general. the equilibrium profit of i.

Ei)) for i= 1. nondecreasing cover of Qi. since there is a number of possible degeneracies which must be taken into account. Then. Now let Gj(p) = and Up)>Mi(P) max(O.) = MT. in order for (F. while if it charges the same or higher the profit of firm i exceeds MT. =S\{X(k)} =(X(k). is not a strategy. E. E. our argument has taken the equilibrium profits as given. Now.(p) 2 Q. Now let hi be the highest price in the support of the equilibrium strategy of firm i. F2) to be an equilibrium we also need hi(p.) for all p. then (G. that is. However.) for all p. F.. firm i must get its equilibrium profit when it charges bi. Suppose firm i charges a price which maximizes M. If each G.9Qi be the nonnegative.(p. and thus determine the equilibrium profit of firm 2. (If some set of prices with positive F.. we certainly need F. must be nonnegative and nondecreasing. the function Q. 2. the profit of firm i is MF. remains the basis for the construction of equilibria. it can profitably be eliminated from the support. Q.(p. since bi $ J(F. Let 3.)].)). this means that if p E (supp F.. G.) is an equilibrium: each G. and Fj cannot both have an atom at the same price (since one firm could then always do better by charging a slightly lower price). and hence restrict the nature of the equilibrium strategies of firm 1. It turns out that this must be firm 1. we can considerably restrict the nature of the equilibrium strategy of firm 2 (by arguments like those in the paragraphs above).. so that the equilibrium profit of firm i is Mr.)\Zi. is continuous in p. h. The equilibrium strategy Fz of firm 2 is nonatomic. In the nondegenerate case there is a unique equilibrium and its structure is quite simple. in particular it is not nondecreasing..)\Zi is not an atom of Fj then F. since we need F. X(li.-measure yields less.(p) = 9Q2(p. under our assumptions.(p) 3 YQ. the profit of firm i is at least MT. and bi $J(I. Thus whatever firm j does. and supp G.(p..). G.244 OSBORNE AND PITCHIK argument is X(k). So far. and less outside. Now. in this case F.(F. But this profit is just Mi(bi) (firm j charges a lower price with probability one. Then the support of each equilibrium strategy is the same subset of 2. is a strategy. F. That is. so the argument establishes that the equilibrium profit of one of the firms is equal to MF.(p.) = Ei for all p E (supp F. Hence i’s equilibrium profit is at least MT.(p)= Q. where supp Fi is the support of Fi and Zi is a set of Fi-measure zero. That is.) Using (2. with F. In general. is nonatomic. = supp G. almost all prices in the support must yield the equilibrium profit. Ei = hi(Fi. and hence each firm’s profit is constant on the support of its strategy.2. Then if hi 2 b. MT) for all p E $. F.) for i= 1.) (the set of atoms of F.3) again. To describe .. If firm j charges a lower price. Ei). A precise description of all possible equilibria is complex. since each Q. Given this. and F. Let .

are increasing. In this case there is always a pure equilibrium (P(k). Theorem 1 also allows for a degeneracy which may occur when M. QS (i= strategy equilibrium 1. is not uniquely maximized at P(k)) then (i= FE. then Q. or if the set of prices such that Qz(p. (If the lowest price in the common support of F. let p be the highest price in the support of F2 which is at most equal to P(k. k. which account for the nature of the conditions on F.) = 1 . F ./k.2).) (with p = P(k)) if no such price exists).). M...F. E.. M. If p < p then F. so that the support of each equilibrium strategy is the same interval.(p)/k.PRICE COMPETITION IN A DUOPOLY 245 the equilibrium strategy F.).C) = F?(p) =O is not a singleton. = P(k. 3. P(k)). is less than P(k. PI..(p) = 9Q.2) and (3. 3). if PE [P(k). Figure 4 illustrates these possibilities.(p. defined in (3.) Degeneracies occur if Q2(p.(p) = Q.2)) we have F. MT) over some range. and the only atom in F. of firm 1.(p. But if there exists p> P(k) such that MT = M. is maximized at P(k) (so that MT = k.) then E. a. P(k)) (as discussed above). while on the rest of each interval F.E2/k2p (refer to Fig. An example of a nondegenerate I. occurs at the highest price in the common support.(p.YQ. E. 2) is the function of H(k.e.)). in part (b)(ii) of Theorem I below. 2) is the equilibrium strategy of lirm mixed i. MT) is constant and equal to . and F.(p) (i. so that (using (2.(p)=k. and Q. = k*E. which is included in part (b)(i) of Theorem I (u= 1. has an atom at the end of each interval in the support of F. if p 3 p then F.) (If the profit function p d(p) is concave.

(b) Ifk. F. M:) supp F. 4’1 II II II II II I FIG. a0 <p < P(k. . An example of degeneracies.F. p4] and [pS. c A. there may be other. The example Q*(p) = 0 and Qz(x) < 0 for constant. is of A (see (b) of the theorem) is pr. &+. III III .)c~ and only if it satisfies the following conditions. 2) is defined in (3. (3. where A = {p E & : F2( p) = UP. 4.4) r ~Qz(P. mixed equilibria..)EYxYisanequilibriumofH(k.. while the smallest member that there are many choices for a (see (b)(ii) of the theorem). P(k) (region III) then supp F. so a mixed strategy equilibrium of H(k. 2) is the function 1. < P(k. II . equal to (0.2). ph]. n 3. 3 d(0) (region I) the unique equilibrium of H(k. . degenerate since ( 1) there are many prices p for which I < p. (a) If k. c 3 for i= 1. THEOREM 1.2 and for some0 d r < 1 and P(k) Q a.3) (3.k.)I. Ql (i= 1. (i) Zf MT = k.246 OSBORNE AND PITCHIK /j’ 0 //I 11 I. II II II . III .. (i= 1. kz) exhibiting some the equilibrium strategy of tirm i.) we have Fz(P)= I if if X(k) d p < a.. The equilibrium profits are (0. Q..O).<d(O)then(F.O). k2) is pure. This accounts for the complexity in part (b)(i) of Theorem 1.). and (2) over the intervals [p3. The smallest member of the support of Fz is pz.

This means that if the duopoly lasts for more than one period. b). The nature of the distributions of prices charged by the firms depends on the shape of the demand function. the equilibrium strategies are concave separately on[a. his interpretation is that the firms sometimes hold “sales. )) and (P(k. if F. kJ.“(k.) is outside this interval.(P) 2 yQI(p. while that of the large firm is atomless and concave on [a. except that prices just above P(k. b]. F. MT) for all PE so. Fig. ). min supp F2].for i= I. there are “holes” in the supports of the strategies over those intervals of prices where demand decreases most rapidly.4) and (3. 2.(p. 3 and 4. If demand is such that the profit function p d(p) is concave then the support of the equilibrium strategy of each firm is an interval [a. it does not depend on imperfect information. APPLICATIONS Characteristics of Equilibrium In region II all Nash equilibria involve mixed strategies.a) right-increasing at p.a). while the small firm tends to charge low prices most of the time. k. b). b]. In his model.(p) = Ql(p. If the profit function is not concave then the equilibrium strategies can be quite irregular.) is between a and 6. In the case that P(k. as shown in Figs.PRICE COMPETITION IN A DUOPOLY 247 and for a = P(k) we have F. In both cases(i) and (ii) the equilibrium profits are (M. but are not concave on [a.(p) = 4Q.. 4. P(k. k. is (3. Varian [ 181. and f’. in a model in which some consumers are imperfectly informed. c $. and has an atom at b. the model predicts variations in prices between periods (as the firms’ random devices generate different realizations).5) (ii) If MT > k. the randomization by firms emerges from the process of competition itself. the equilibrium density of prices is U-shaped. also finds that firms randomize in equilibrium.5) for some aE [min A. P(k) (region II) then supp F.a) if PE so. Comparative Statics The comparative statics of some features of the equilibria can be studied with the aid of Fig. and F.” In our model. 5 (cf. satisfies (3. k. Thus in this case the large firm either charges a high price or is likely to offer a substantial discount (as in Varian’s model). 2 of Kreps and Scheinkman [ 111). b). If P(k. then the equilibrium strategy of the small firm is atomless and concave on [a. Thus the pattern of prices charged is similar to the one in the previous case. This .) are now relatively likely to be offered by both firms.

. k. functions of 4 (where ph FIG. j = 2.248 OSBORNE AND PITCHIK FIG. Consider first the effect of an increase in the size k. while that of the small firm is nonincreasing. The only functions in Fig. as shown. + k.( . If the new value of P(k) is between pz and p3 then the equilibrium is pure (region III). and kza. 5. 5). MT) is zero.) =m. . = k...). equilibrium An example of the construction of an equilibrium. dependence of the equilibrium strategies and profits shown in Fig. The paths these variables take for the example of Fig.. and pr are respectively the lowest and highest prices in the supports of the equilibrium strategies. the equilibrium profits are E. 5. = MT(k. Thus a. 5 are shown in Fig. . . in which there is a single price (ai) at which Q. if it is below p3 then it is mixed again (region II).+ on (k. and E. of the diagram shows a nondegenerate case for region II. of the small firm.) of the equilibrium strategies (a and b) are shown as capacity x. . 6. for a tixed value of k... and a single maximizer (p. of the firms.a.. kj > k... is such that P(x. the equilibrium profits The limits of the supports are m.. strategies are a. the equilibrium profits (E. The equilibrium profit (MT) of the large firm is nondecreasing in k.) = ph for h = I. of the large firm. the linear parts of which rotate counterclockwise about the origin. holding fixed the size k. 5 affected by this increase are L1 and M. . the endpoints of the supports of the equilibrium strategies are also nonincreasing.) of M.. An example or the For the profit function p d(p) and the limits of the supports k. ” kj XI ~2 ~3 ~4 k.. and p. The is given in Fig. kZ). and E. 6 (i = 1.

The impact on the equilibrium strategies and profits for the example of Fig. is very small.(x) 6 MT if x < p.p rotates clockwise about the origin. Any increase in demand raises (or leaves constant) M.). pp. then simultaneously choose prices. slightly lower prices.(p) at each price p. decreases: the fact that the firm is smaller reduces the competition with the large firm and raises equilibrium prices. when k. They argue that even if capacity is free. roughly. the equilibrium approaches the monopoly outcome. j = 1. we can show that the set of capacity pairs chosen in Nash ’ There is at least one Cournot equilibrium. . converges to zero. given the size k. of the large firm. until region III is reached. the firms begin to use mixed strategies and. The equilibrium profit of the large firm and the endpoints of the supports of the equilibrium strategies increases as k. the large firm almost always charges a price close to the monopoly price. under the assumptions that the inverse demand function is concave and the cost of capacity is increasing and convex. The equilibrium profit of the small firm may rise or fall as k.) As k.PRICE COMPETITION IN A DUOPOLY 249 If the size k. decreases. prices are equal to unit costs (region I). When demand is very small (relative to industry capacity). by a result of McManus 1133. The limits of this range of payoffs are nonincreasing in the size of the large firm. while the small firm charges slightly variable. and the function k. and prices are high. then there are many equilibria. (The arguments of Gelman and Salop [9] are closely related to this point. the curvilinear part of M1 increases. prices increase.) If there is more than one price p for which L. They show. where the equilibrium strategy is again pure. because of the lower equilibrium prices associated with larger industry capacity. ki <k. The comparative statics of a shift in demand are more difficult to determine. the effects on the equilibrium profit of the small firm and on the range of prices charged in equilibrium depend on the precise nature of the increase in demand. Under our weaker assumptions on demand. of the small firm decreases. 5 is again shown in Fig. 497-81 suggests is reasonable. possibly offsetting the direct effect on firm 2’s profits of its smaller size. so that the equilibrium profit MT of the large firm does not decrease.(p) = MT and L. there may be many Cournot equilibria’. yielding different payoffs to the small firm. Capacity Choice Kreps and Scheinkman [ 111 study the game G in which the firms first simultaneously choose capacities. As demand increases. that in the unique Nash equilibrium the capacities chosen are the (unique) Cournot quantities. However. 6 (i = 2. a firm may not want to expand indefinitely. (The solution thus possesses a characteristic that Shitovitz [ 17.

where the equilibrium profits are equal to the Cournot levels.C) . The price strategies used in the second stage of this equilibrium must be an equilibrium of H(k:.) for all k. so the profit of firm i at this equilibrium is E. in region III. k. +kT)-u(k.(kT.*.*. However. and increasing. (Under the assumptions of Kreps and Scheinkman [ 1 I]. k:). kf) is the nondecreasing cover of k.u(k.*). at least equal to k. so that there can be Cournot equilibrium quantities which are not associated with subgame perfect equilibria of G. Thus it pays firm i to decrease its capacity from the Cournot level x. However. as argued in [ll. We can complete the argument that (k:.) That is. then by Theorem 1. + kf).u(k.*) be the capacity pair chosen in the lirst stage of a Nash equilibrium of G. i’s equilibrium profit in the subgame following (ki. Since i can guarantee a profit of kjP(ki + k:) .C(k. E. > x then firm i’s equilibrium profit in the subgame following (ki. since if (x.*. An example (the details of which we omit) shows that a Cournot equilibrium may not be associated with a subgame perfect equilibrium of G. x) with k. x) exceeds its Cournot profit only when the former is decreasing (this follows from the properties of MT mentioned in the previous paragraph). let (k:. The example works as follows.*) is an equilibrium profit of i given in Theorem 1 and u is the cost function of capacity. P(k. M:(k. and setting the price P(k. P(k.. If k..) Thus (x. x) in region II. in regions I and II.*) for i= 1. and hence E. assume that k: 2 k.*) is a Cournot equilibrium by showing that E.2 (see (b)(i) of Theorem I). we know that EJk.f+.250 OSBORNE AND PITCHIK equilibria of G for which the capacity choices are pure coincides with the set of (pure) Cournot quantity pairs. + k. +kT) (as a function of k. and the price 0 if j chose k. k:).f+). + k:).(k. for some values of ki with kj <x and (k. Now let (k:.+ P(kT + k. under our assumptions the set of subgame perfect equilibria of G can be a proper subset of the set of Nash equilibria. X) exceeds its Cournot profit even when the former is increasing. First. X) is in region III then any pair (ki. k. <x is also in region III. are such that (k:. Thus those values of k: for which M. k.*) > kiP(ki + k:) . k:)-u(k:)>k. P(k. this cannot happen. k:) is constant in k.) in G by choosing k. x) is not associated with a subgame perfect equilibrium.*.. k:) is in region III. Proofs of these results can be outlined as follows.* in the first stage.).?) = k. the result of Kreps and Scheinkman generalizes naturally.(k. k. and then sets the price P(kT + k. where Ei(k. so that i cannot benefit from increasing capacity. NOW.?) = k. There is a symmetric Cournot equilibrium (x. kz)=M:(kf. k:) be a (pure) Cournot equilibrium. These strategies clearly constitute a Nash equilibrium of G. Suppose that in G each firm i first chooses k..) if j chose k. Without loss of generality. equal to k.) for all k. P(k. # k:. x) in region III.u(k. (We have not investigated mixedstrategy Cournot equilibria. (In our example there is another Cournot .? P(kT + kf ). k. k.u(k.*(kT.(k:. M..?. kl) . 3271. p.

RELATION TO THE LITERATURE The game we analyze is a nonzerosum “noisy” game of timing (i. There are two generalizations which appear to require more significant changes in our arguments: the existence of more than two firms. 2. it is natural to assume that the low-price firm chooses the method of rationing. the outcome is indeterminate (since the payoff of the low-price firm is independent of the rule used). but if it is chosen at the same time as price. GENERALIZATIONS AND EXTENSIONS Our assumptions can be relaxed in a number of ways. preserving the continuity of CaQ. so long as it is leftcontinuous (since Qi can then jump only down. (In the latter case each firm should announce both a price and the maximum amount it is willing to sell at that price.PRICE COMPETITION IN A DUOPOLY 251 of G. and our assumption of the absence of an income effect is quite restrictive. and within each region of continuity each player’s payoff depends only on his own action). though this would require an analysis of the price-setting game for each choice of rationing rules.). As argued previously. except possibly when both players use the same strategy. not on the price of the other firm. Even given the rationing rule we assume. variations in the assumptions which preserve their character can be made. and costs which are not linear up to capacity. for i = 1. Since the functions of central importance in the construction of the equilibria are 9Q. For example. convex cost function. That is.e. Dixon [7] shows. It is possible that the rationing rule could be chosen first. Almost all our arguments apply also when the unit costs of the firms differ and when demand may increase over some range.) Perhaps the most significant way in which our arguments are limited concerns the form of the residual demand. the nature of the residual demand depends on consumers’ preferences. 6. that an equilibrium exists in the case that both firms have the same. An important feature is that the demand faced by a high-price firm depends only on its price. using the results of Dasgupta and Maskin [4]. equilibrium which is associated with a subgame perfect equilibrium we do not know if there always exists such an equilibrium. though the description of the equilibria is then somewhat more complex. M. and p. Previous work on . demand may have discontinuities.. though the outcome may then be sensitive to the form of the dependence. is a function of p1 alone.) 5. a game in which the payoff functions are continuous. Many of our arguments can probably be extended to the case where Mi depends on both p..

Dasgupta and Maskin [S] show that if d satisfies (2. 293-2951 gives a uniqueness proof which relies heavily on the fact that the equilibrium payoffs in such a game are unique. Edgeworth [8] argues that. He considers uniqueness. It is easy to argue that if this second rationing scheme is used. we establish that there are no other equilibria. though they do not give a complete characterization and do not study the uniqueness of the equilibrium strategies.. Allen and Hellwig prove more detailed results on the nature of these supports when the demand function is not necessarily decreasing and there are two (see [2]) or more (see Cl]) firms. Shapley [ 161 reports (in an abstract of a paper which seems to be unobtainable) a characterization of the equilibrium of a price-setting duopoly game. Several authors have worked with the rationing scheme in which the low-price firm allows some fraction of the customers to buy all they wish (rather than allowing all customers to buy some fraction of their demand)*. For the games we consider. but his argument is flawed3. 3 For example. and the the argument . In the economics literature. Davidson and Deneckere [6] verify this in an example with linear demand. a substantially more involved argument appears to be necessary. Kreps and Scheinkman [ 1 l] assume that inverse demand is decreasing and concave. As a result. there is a large number of customers. * Equivalently. = k2 then the supports of the equilibrium strategies satisfy a certain property. Karlin [ 10. and if k. then the result of Kreps and Scheinkman [l l] on the coincidence of the Nash equilibrium capacity pair in the two-stage game and the Cournot quantity pair no longer holds in general. = d(0). For a class of zerosum games of timing. Levitan and Shubik [ 121 describe the Nash equilibrium of H(k. each with a reservation low-price firm serves a random sample of them. Finally. k2) in the case where d is linear and either k. Then. and uniqueness has not previously been examined. they do not study uniqueness.1) then an equilibrium exists. in a series of results. which follows (IS) does not establish uniqueness of the equilibrium payoffs. APPENDIX: PROOFS Here we prove Theorem 1 of Section 3. it is not clear precisely what his model or assumptions were. they establish a result concerning the equilibrium payoffs.. = k. there is no equilibrium in pure strategies. Beckmann [3] assumes that d is linear and k. for a range of capacity pairs.252 OSBORNE AND PITCHIK such games is limited. = k2 or k. price. pp. First we check that any pair of strategies satisfying the stated conditions is an equilibrium (Proposition 1). the inequality in his (15) should be reversed.

) = W(P)(F. or Fi is left-increasing at p (i. E supp F. and qn < p for all n. then either in which case F.(p)) hi(Fi. 2 we have hi(p. F2) is an equilibrium then in the first case fact (A) implies that hi(p. k(Fi.h +4.(p)) + L. Fj) = MiP)(f”(p)k =hi(P.) for all PE (supp F. F1) is an equilibrium p E J(F.. Fj) < hi( F. jumps or atoms) of F. 211 of [lS])..(P)) @j(P).PRICE COMPETITION IN A DUOPOLY 253 We begin by stating some basic properties of equilibrium strategies which we shall use repeatedly. there is a decreasing sequence with similar properties). and p E supp F. F. Fj) = M.)\Z. F. . in which case -F. -F. F. is left-increasing at p.(F. or (c) Fi is right-increasing at p. F. Fj) = h.. and .e. p.(P) + ‘i(P)(l -q(p). F.).(p)(@j. (A. in the second case.). we mean that i is not equal to i. where Z..4i(p)) xj(p).. in which case a.) for all p E S. or Fi is right-increasing at p (i. there is a sequence (p. let supp F be the support of F.) = hi(F. and h..e..) = h.) for all n.) E Y x 9’ such that for i = 1.(F.. F.(P) F.11 (Recall that whenever we use the indices i and j in an expresssion. If p E supp F. 1) that (A) (F.. ) with pn E supp F. F2) is an equilibrium if and only if for i = 1.e.(p) + 1 -F. qn T p.D~t p).) For FE 9.. 2.. F. there exists a sequence ( qn > with q.(P) a.(P) + L(pNl =hi(P. and hi(q. F.. in the third case there is a decreasing sequence with similar properties (see p. kZ) is a pair (F. F. is a set of Fi-measure zero. For FEY. < p for all n.).(p.. let J(F) be the set of points of discontinuity (i.. F/j+ (M..(P)) hi(F. If (F. A (Nash) equilibrium of H(k.(p) . hi(F3 Fj) G hi(Fi.(P) or (b) F.) for all FE Y.) + (L. then either p E J(F. F..(P)-di(P)) . It follows from (A. By taking limits in the second and third cases we have the following: (B) (a) If (F.) = h.

Recall that we write X(z) = max(O.2 or bi > 0 and h. > 0. We now turn to the specific features of our game. using or (c) of fact (B) again using (A. -cdp.)].) > 0.ai( (A.3 (A. Fj)>O. then h.) = 0.). ai 3 0. If ai EJ(F. 2 X(k) for i = 1.(p). and either (a) holds. F.) = 0.)=k.(p)F.) 2 0 for i= 1. = 0 and h..(p)(l -F. P(z)).. L.) = 0 f or i = 1.2 by fact (A).(Fi. If F. 2 fact (B). F.)=O. for i= 1. we need and hence.2 (A. If a. F. .(F..)=M. (e) if PE supp F.4 if if P(k)<p<O X(k)<p<p.254 In particular: OSBORNE AND PITCHIK (d) h. Then hi(p.) is a weighted average of M. E 9 we write ai = min supp F... F. using (a) of 0 6 h. Zf (F.p (A.(P) G 0 oG”i(P)cdi(P)<Li(P) (A.2) we have hi(O.(p..(Fi. F2) is an equilibrium and p EJ( F. <a. Zf (F.) then F. We can now prove Theorem 1. LEMMA ProoJ: (i) aj 2 P(k).(a. strategies LEMMA 2. k.(Fi.(p) = M. F.) > 0 if 0 < p < min(b. so that we need 0 6 hj(Fi..(p)for i= 1.<P(k) if p=Oorp>p. (iii) Suppose b.(p)+L. and LAP). Fj) = L.4)..(a. Fj) > 0 for i= 1. Let a. so hi(F.(p)) if PE tsuPP Fi)\J(Fj).2) and Fig.3). which shows that we can work with the restricted game Z?(k. (see (2. F. and hi = max supp F.(F. I$~..(p).(O. F.) = hj(ai.f’j)=hi(p. Fi) = dj(ai) u.) + Li(ai)( 1 .. F2) is an equilibrium then (a) a. 2.(F. F. By (A. we have bi > 0 and hence h.2 and (b) either b.) then. and hence.2. by fact (A). 1) are easy to establish. so that h. F. The following properties of the functions Li.) in which the (pure) strategy set of each firm is 3 = [X(k). 1 This allows us to restrict the prices at which both equilibrium can have atoms. t.3) ai 20. Also since ai = bj =0 if bj =0 (by (ii)). since this is increasing in p. 4. we have a. and M.(P) < 4i(P) < M.2).(F. $J(F. X(k. and h.. . Our first few arguments lead to Lemma 4.) n J(F2) then Li(p) = b. 1.(a.. If p6 P(k) then h. pO) by (A. 2 P(k) by (ii) ai > 0.

and hence M. m .X(k)>O.(p)-d. By fact (A) we need hi(p. LEMMA 4. F..2) and (A.. and SO h. 2.(p)=O if p > X(k. F. . F. Taking the limit of the left-hand side gives (L. and (A.. (F..) or h $ J(F. F.(p))aj(p)<O. Then h...(p.) E Y x Y is an equilibrium it is an equilibrium of fi(k. 2. so by fact (A) we have hi(Fj. F.. If 6. But by (A.(p)>~$jp) (by 1 Lemma 1. so that h. if bj > b. since s = { 0) in this case. F. and M.. = b then there exists i such that either b E J(F.) = Mj(b) = M. or bi = b. so by Lemma l(b) assume bj >O for i= 1. We need to show that F.* 7 p. If bi > b. by (A.) 6 hi(Fi.(X(k). PROPOSITION 1.?. k.?.4)) we have Li(p) = hi(p) = M.(F.(b. equal to (0.). 3 d(0) then the unique equilibrium pure.).2) and (A. We need to show that hi(p.) if p > 0 (the first two inequalities by (A.(h)=M. If b. and in both cases h.i) for all n.. F. so we need to show that bi < X(k. Proof We have 0 < M.) = Mj(bi) = M. 2. F.(b.)n J(Fz) then by Lemma 2 we have L. = 15.2.* fbr i= 1.)=M..* for i= 1.*.).(F.)=M. kz). qf H(k.. Fi) = hi(p. then hi(F.. LEMMA 3.. Hence hi(F. Now suppose that (F.PRICE COMPETITION IN A DUOPOLY 255 Proof: Consider a sequence (p.)>O for i= 1.) is an equilibrium of fi(k. It is easy to check that (0. then by (a) or (b) of fact (B) we have hi(Fi. k2). k2) if and on/v $‘ Proof: First suppose that (F. F. 2.O).(F. we restrict the strategy space of each player to 9). E . 2 if p 3 P(k.)=h. F’) < hi(F. .(F.) whenever p < X(k) or p > X(kz). > with p. F.. We can immediately prove part (a) of Theorem 1. which pins down the equilibrium payoff of one of the firms. Fz) is an equilibrium of H(k. < P(k. so (F.). F.(p) < hi(p. F. By Lemma 1 we have ai > X(k). F?) is an equilibrium of H(k. = h.)= M:.) d hi(F.O) is an equilibrium. Now. If h.). Now. From now on we use Lemma4 to restrict attention to the game f?(k.2) and (e) of fact (B) we have h.)a k. it follows from Lemma 3 that b.4)). strategy pair is by Lemma 4 Proof.(p).X(k) if p<X(k) and h. .) 2 M. . Fj) = Mi(bi).)bk. F. Let (F.. kz). E’. If k. and bi $ J(F. = b2 = 0 this is certainly true. F2) he an equilibrium.. there can be no other. F)= kiX(k) for any FEY. We can now establish the following useful result..(b). I. F. 2..(b)=4i(b)= Mi(b) for 1 i= 1. So sincepEJ(Fj) and L. k2) (that is. and hence M.(p.).P for i = 1.). (A.(p) 60 if p ~0. Since Mi(p)=O 1 for i= 1. Fi) > M. = b and bE J(F.3) we have h.2).

(X(k).) = k. But then from (2. n S.(p)3SQz(p. LEMMA 6.(p. E. the set 2.a>k.) for all p~$. 1 PROPOSITION 2.) = k. we have h. a.(p.(P(k).2)). F. =a2 =a then and . F?) = k.4)).). =a.3) and (3.(p)>. so that M.5) and the continuity of Q. NMI. Note that if k.(p.(a. and 5. = {p E 3. if a. F.2) we have hi(p. for all p E 3. since F.(P. Hence h.5) and Lemma 5. since F. We need to check that the conditions of fact (A) are satisfied.(F. ProoJ If F.. and if p Esupp F. F?)=L.a for i= 1. h.a for all pus by (3..) (by (3. <d(O) then P(k. k. Also h. F2) always denotesan equilibrium.).) (see (2. From now on we assumethat k2 < d(O).) 6 E. is nonatomic on S. F.. This is done in Lemma 7.)< k.(P(k). and h. The following preliminary result allows us to prove.(p)= Q.(p)>. Also h. k.: F2 is not right-increasing at p} has F.X(k) d Ei (see (A. Qi(p.2)).a for all p E (supp F2)\Z2.(p. Zf Ei >k.(p) = Qz(p.(p. F.(a) for i= 1.P(k) we have h. Fz) = MT. da. First we pin down the equilibrium payoff of firm 1.)=L. Hence h. if a2 $ P(k..) h. F2) be a pair of strategies satisfying the conditions of part (b) of Theorem 1. in Proposition 2. P(k). Any pair of strategies satisS?)ing conditions in part the (b) qf Theorem 1 is an equilibrium sf H(k. Since MT > k. after a preliminary result (Lemma 6) which gives a relation between the equilibrium payoffs and the smallest prices in the supports of any equilibrium strategies.. Ej) for all p E 3. and the conditions of fact (A) are satisfied for i= 1. FJd MTfor all pus by Lemma 5. 1 It remains to show that there is no other equilibrium. for all PE so.) = k. we have h. =a then a. So the conditions of fact (A) are satisfied for i = 2..)..X(k) h. Now consider the payoff of firm 2. F.)= k2 P(k). and hence M: > 0. Recall that &.(p.(p. Since k.(p) >O if 0~ p< P(k. F. MT) (see (3.) >O. Fj) < E.) = MT for all p~supp F. . First consider the payoff of firm 1. so that.256 OSBORNE AND PITCHIK We now address part (b) of Theorem 1. and (F.} with pn 1 p). If Fz is right-increasing at p then F. M:) for all PE&. that any pair of strategies satisfying the conditions of Theorem 1 is an equilibrium. 2. Finally.(p) > Q. Ei) + C+(P) for all p E So (consider a sequence {p. = &\{‘w)} = (X(k).5) so p$J(F. then in fact F.(F. then LEMMA 5. P(k) and F. Now. is nonatomic on so.(p.2. F.a) by (3. Proof: Let (F. Q. F.-measure zero. hi(Fi. then F.

=a. If p 3 P(k. (b)(ii) and (iii) do not depend on our previous results) gives some relations between the equilibrium strategies and equilibrium payoffs.YQ.. Proof.(d(b.(d(&) .(d(b.(a) < L. so since h.(a) Fj(u) + L.. F. so that h.b. 2 by (A.(F. . and d(b. F2).F. .(bi) = MF for some i.k.2).)-k..(b) = b(d(b2) -k. OY(iii) p Esupp F. which is increasing in p. F.).b.) >O (by Lemma l(b) and the fact that MT >O). =a2 =a and k. or (ii) Fj is right-increasing at p. If a. Let E.. with X(k) < x < p then F. Now we show that if a. so that hi(Fi. h. F.a=k. < P(k. so aj < ai by fact (A). If p <a.b..(a) = 0. By Lemma 3 we have hi(Fj. <a2 then means that a.PRICE COMPETITION IN A DUOPOLY 257 Proof. this then a. = hj(Fj. so that h. and hence M. and thus if azdP(k. =k.) = M. 6 P(k.)=M:. = a2 = a > X(k) then a & J(F. da.) by Lemma 4.2 (by (d) of fact (B)). F2) = M.) . t.). Fj)=Li(a) for i= 1. Then F.).)<k we have Mf>k.(p.hz(d(bz)-k.)= a = b. or a # .)/k. F.). Now. or d(b.)=k. E. and P 4 J(FJ. Suppose h2(F2.) = M.(b) = MT.(d(bz)-k. F2)= t. (c) if F.(F?. E. F. F.(p.). F.(b?) and hence (by the first part of Lemma 3). (a) (b) tf (i) p E J(F. we have b. < P(ki) then h..(p) so from fact (A) and the fact that M.. E.)-(k.(F.4)) we have F.(bJ = k.(a) for i= 1..). So if k. =a2 =a= X(k) then Iz. E. I The following general result (parts (a). LEMMA 8. By Lemma 2 we have a #J(F.). But if p <a we have hi(p.) Iz. We know that k. This means that there exists a sequence as in (c) of fact (B).M.. Proof. Since a. F.) -k2)b?(d(b’)-k).I( F.)/kz =h. We now need to deal with the case k.) for some i.(p) > YQ. if k.(F.(F.) then M.cj(a)). contradicting the first part of Lemma 3.(a)( 1 .(F.).(a.)=k (so that b2 = P(k)) this implies that h.F..)= Li(p)=kip by (2...(p.)= L. F2) = Mr.(F.)/k. =k. > kz and d(bz) <k. >li. h.2) and (e) of fact (B). and the fact that F.a= h.(P) 2 ~QAP. (A. F2) = M.(F.) = L.) by (d) of fact (B). If a.(p) =O. (a) follows immediately from fact (A). Then by Lemma 6 we have a.(F.(F.) = M2(b2) = MI.) and there exists x E supp F.(p) = .) and p E 3.M: > k. then Fip) = QAp..(a) (see (A. 1 LEMMA 7.(F. Hence we must have h.k.4).

E.d(p)) if X(k)<p<P(k. E.) is constant on [P(k. contradicting Fi(p) > YQi(p./k. MF) is right-increasing at p then so is SQ. this follows from (b)(i). the argument above.).)/k./kzp .( ’ . so that p > a.a.)} for i= 1. E.) is right-increasing on [p. Lemma 8(b) implies that 0 d F..xi) = Q..) d dQi(p.).). E. p.)].) then a.2) and (3. P . > P(k.(x.E.E. so Fi(xj) = Qi(xj.(P(k. MT) = 0 if p d EJk. If ?cj <xi then xi y!J(F. E2) = I.).]. is maximized at P(k).lk. E. Q./k.)=(pd(p)-E.): Q. E. so that if EJk.Q.)).) is constant on [P(k.. the result follows in this case.Wkdlpk. since F. since . But then xi # J(F.2)).(p.) is rightincreasing on this set (see (A. E. so that X. and let xk = max(x E supp Fk : x < p ) for k= 1.(p. (A. .(x. . (see Lemma 6). Ei) so that Fj(P)=yQj(P./kz. 2. suppose that t’. Ej) 2 0).(P(k.E.Wp(k 1 . Hence xi 2 xj. Let Pi = min(p3 P(k. E. Ej)=YQi(P(k. P(k. = a. EJkz < p 6 P(k..E.E. (using Lemma 8(a)). E.(p . we have pI < pz. Ei) = 9Q. Hence . If a.)lpk. 2 P(k. so that F.6)). .) d P G P(k. (b).) 3 Qj(xj.(p. E. < P(k.) = Qj(xi.(xi. El) = 1 .(p. Finally consider the case P(k 1) < P d Pfk.(p. Since either a2 EJ(F~) or F2 is rightincreasing at a. the result follows.(p.p (see (AS)).) (see (2. To prove (c).(x.(p) > YQ. E.(p) = Fi(xj) = Q. 1 .(a. E. and (c) of fact (B) and Lemma 2. E.). Hence a. G by k.(p. p2].) < YQjp. &I = if P(k.(p) = F.).(p.PQ. so by (b)(iii) we have Fi(. Since .).) is./k.) is right-increasing whenever YQ. First note that (k.2. E2) = k..( . If9Q2( Proof: 3) to deal with the case . E.(p) =0 if p < E.SQ./k.). LEMMA 9.51 Q. so that YQi(p. (b) follows from (a).(p. and hence YQ. > X(k) (since aj 2X(k)). -Q/k2 if W.) = k.)/k. If p > p1 then YQ.)... Ej).). = k.p Now suppose that a.) = Q. if X(k) -Cp 6 P(k.(a. E. so that Ql(p.) (if xi = x. I The next two results allow us (in Proposition where M.) = F. so that SQ1( . > E.(p) > SQ. P(k.(xj. MT) is right-increasing on a subset of [E2/k2. = a and E.). Since YQz(p. E.258 OSBORNE AND PITCHIK must be nonnegative and nondecreasing.).) < p 6 P(k. E. Ej) by (b)(iii) and hence F.).( . Ej).)]. and Q2(p. .) and hence F. (A. = k.) (using (a)).). E2) = Q.) then YQ. E. (since YQj(p.YQJ./k. and hence in this interval YQi( . E.(P.E.61 1 .9Q.) = I 0 if if X(k)<p<E.

(a. so since F. so that a.E. = P(k) and h.)). MT) is right-increasing that .. F2) is an equilibrium only if it satisfies the conditions in (b)(i) of Theorem 1.. Proofi If F..5) follows from Lemmas 11(c) and S(b)(ii).)= Q..(p. so a. ~1 and . M:) if P E &.. . I PROPOSITION 4.PRICE COMPETITION IN A DUOPOLY LEMMA 259 10. = a. Then since F2 is right-increasing at a2 (by (b) and the continuity of YQ2( . (c) If a. So exists x~suppF* with X(k)<x<p then F. and (3.2.) = k. Proof: The conditions on F2 follow from (a) and (b) of Lemma 11. P(k) then (a) supp F. satisfies (3. > Ez. P(k.YQ. = a then h. so h. MT) then there exists y > p such that YQ2(z. Zf k.(P) = yQz(p.. EJ(F]) or F. Lemma 10 implies that F. Proof: (a) We have X(k) 4 supp F. MT)>0 if PEG. E. Zf k. by Lemma 6. MT) then supp F2 c {X(k)} u F?(z) > (P. 1 .P(k) by Lemma 6. by (e) of fact (B) and (A.5) from Lemma 8(a) and (b)(ii). we need the following: and LEMMA 11.(P(k))= MT we have QX(p.3) and hence is nonatomic on so. <d(O) and M: = k.) -E [p. so if there for all y. <d(O) and MT > k. (b) F. con- PROPOSITION 3.( .v if it satisfies the conditions in (b)(ii) of Theorem 1.(y.(F. .4) follows from Lemmas 1 l(a) and S(b)(iii). P(k) then (F. < u2.(a.(F. and (c) h2(F1.a.a.)>. 1 To complete the proof of Theorem 1.(a. (if pn Ta.. > X(k) and hence a2 > X(k) by Lemma 6. Now.)= k. a?].M~) by Lemma 8(b).4) follows from Lemma S(b)(iii) (3. and F.)=h2(p. F. Proof First note that since MT > 0 we have P(k) > 0 and hence X(k) = P(k). Zf MT > k.)2k2 min A.(p) > 4Q. is right-increasing at a.(a.(z. we have F.) = k. F.(z)=-aQ. F.. P(k) then (F. M:) for all z E [p.)= k.. Since either a.a for some a~ [min A. 1 at y.(q) = k. M:)) we have F.) is also right-increasing at y (by Lemma 9)..O we have &(a.(F. MT)). (b) This is immediate from (a) and Lemma 10. F. EA.YQ. Suppose that a.2).(a. But then F.)d&).) = L. GE.Y&(p. = P(k) by Lemma S(a).( . we need b(p. The condition (3. so that a. and hence a. EJ by Lemma S(b)(ii). Since M. Ifp E 3. F.a. JJ] by Lemma 8(c). c so for i= 1.)=Q. Thus (3. But now L.(p) > . F2) is an equilibrium 0nI. is right-increasing at kadicting Lemma S(b)(ii) (since F2(~) > $Q.(a2.

KARLIN. Rev.” III (R. J. 16. 6. GELMAN AND S. BECKMANN (with D.” Discussion Paper 82/54. “Bertrand-Edgeworth Oligopoly in Large Markets. 1959. J. DASGUPTA AND E. R. Equilibrium. 10. Ed. Amer. HELLWIG. VARIAN. ALLEN AND M. S. Meisenheim am Glan. ALLEN AND M. C. Henn. 1. D.” Discussion Paper 186. “Cartels. June 1982. Columbia University. J. Universitiit Bonn. Center for Analytical Research in the Social Sciences. 326337.” Discussion Paper 82/55. EDGEWORTH. London.) 9. 8. Programming. SHUBIK. La teoria pura del monopolio. hf. Profits. MASKIN. 14 (1983). B. Edgeworth. Econ. numbers and size in Cournot oligopoly. 205-212. SHITOVITZ. Econ. F. M. A model of sales. Edgeworth-Bertrand duopoly revisited. 13-31. 5. A. OSBORNE AND C. P. B. pp. 11. Bell J. P. S. “Mathematical Methods and Theory in Games. Bell J. PITCHIK. Reu.. R. DAVIDSON AND R. Vol. SW. II. 12. Macmillan. and the Cournot model. H. Y. 15. 307-320. 1925. and Excess Capacity. Addison-Wesley. Ser. March 1982. “The Existence of Equilibrium in Discontinuous Economic Games. B. London School of Economics. short run competition in price. 17. 14. 3. 405414. 315-325. The existence of mixed-strategy equilibria in a price-setting oligopoly with convex costs. 1965. Mass. Yorkshire Bull. SCHEINKMAN.” Discussion Paper 141. London School of Economics.260 OSBORNE AND PITCHIK REFERENCES I. C. Ed. Quantity precommitment and Bertrand competition yield Cournot outcomes. M.). Hain. International Centre for Economics and Related Disciplines. Price duopoly and capacity constraints. August 1984. Econ. 651-659. 2077221. KREPS AND J. DASGUPTA AND E. PITCHIK. Econ. in “Papers Relating to Political Economy” (F. Department of Economics. L. Judo economics: Capacity limitation and coupon competition. Y. DIXON. 13 (1972). 1. Lett. Giorn. 7. 4. Econ. “The Existence of Equilibrium in Discontinuous Economic Games. M. “The Range of Equilibrium Prices in Bertrand-Edgeworth Duopoly.). MASKIN. DENECKERE. Int. . 14 (1983). transl. 2: Applications. 70 (19801. Reading. Hochstadter). pp. 55-68.” Working Paper 83-27. Econometrica 41 (1973). Game Theory 10 (1982). Long run competition in capacity. M.: The pure theory of monopoly. Abstract of “A duopoly model with price competition. December 1983. 111-122. Econ. J. International Centre for Economics and Related Disciplines. Theory. 467-501. Res. and Economics. Equilibria of a two-person non-zerosum noisy game of timing. 18. LEVITAN AND M. MCMANUS. draft. lnstitut Fiir Gesellschaftsund Wirtschaftswissenschaften. 16 (1984). 2. 13. SHAPLEY. Econ. 354355. University of Pennsylvania. SALOP. 16 (1964). 1983. H. 2 15 (1897). 1983 (revised March 1984). in “Operations Research-Verfahren.” Vol.” Econome/rica 25 (1957). 68-75. (Engl. 111-142. HELLWIG. Oligopoly in markets with a continuum of traders.

- Math Review
- 6400LectureSpreadsheets.xlsx
- Law of one price.docx
- 2_Binomial_Monte_Carlo.doc
- Expected_Utility.pptx
- ECON 350 ST Lecture 2.pptx
- ECON 350 ST Lecture 1.pptx
- ECON 130 Lecture 20(2).pptx
- ECON 130 Lecture 19(2).pptx
- BondIndexData(1).xls
- Tri2scheduleTODATE(29-06).xlsx
- 4340LectureSpreadsheets.xls
- Econometrics-I-12.pptx
- Equilibrium.xls
- WEOApr2013all.xls
- PA.NUS.PPPC.RF_Indicator_MetaData_en_EXCEL.xls
- FINA301-Tut 7.doc
- FINA301-Tut 5.doc
- UseR-.pdf
- LU SAH - Law
- Albano2006-Japanese Eng Ascending Auction B JOURNAL
- Hazlett Munoz2009-Welfare Analysis of Spectrum Alloc
- A Swift Tour of Auction Theory
- Dana

Sign up to vote on this title

UsefulNot usefulauciton

auciton

- Markets for ownership
- Huanxing Yang - Understanding Asymmetric Price Adjustments
- 2006 Final With Key
- Theory of Price and Output Determination
- Macro Oo
- DEM&SUP
- Aggregate Demand and Aggregate Supply
- qjae2_2_1
- Shone - An Introduction to Economic Dynamic
- Economic Analysis
- Unit 3 Final
- exam tips
- Macroeconomics-1 by Prof. N.C.nayak
- Chatterjee - The Equilibrium Theory of Arms Races - Some Extensions
- Shi und Xu
- Stable Instability
- marketstructure.ppt
- 09 Ndp_throughsvd (Tristan2001)
- Figueroa sigma society
- Exchange rate
- Rub in Stein - Edgar Allan Poe's Riddle - Games & Economic Behavior 2011
- Model on Attention Economy
- 14 Monopoly
- Lecture 1 (Maneco2)
- Rv Adv. Micro V10
- Chap11 Auctions
- 10e 12 Chap Student Workbook
- 522 Economic Analysis b
- Article
- Analysis of Competitive Markets Lecture_Price Ceilings
- Osborne

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.