Competitive and Cooperative Inventory Policies in a Two-Stage Supply Chain

Gerard P. Cachon • Paul H. Zipkin
The Fuqua School of Business, Duke University, Durham, North Carolina 27708 gpc@mail.duke.edu zipkin@mail.duke.edu

W

e investigate a two-stage serial supply chain with stationary stochastic demand and fixed transportation times. Inventory holding costs are charged at each stage, and each stage may incur a consumer backorder penalty cost, e.g. the upper stage (the supplier) may dislike backorders at the lower stage (the retailer). We consider two games. In both, the stages independently choose base stock policies to minimize their costs. The games differ in how the firms track their inventory levels (in one, the firms are committed to tracking echelon inventory; in the other they track local inventory). We compare the policies chosen under this competitive regime to those selected to minimize total supply chain costs, i.e., the optimal solution. We show that the games (nearly always) have a unique Nash equilibrium, and it differs from the optimal solution. Hence, competition reduces efficiency. Furthermore, the two games' equilibria are different, so the tracking method infiuences strategic behavior. We show that the system optimal solution can be achieved as a Nash equilibrium using simple linear transfer payments. The value of cooperation is context specific: In some settings competition increases total cost by only a fraction of a percent, whereas in other settings the cost increase is enormous. We also discuss Stackelberg equilibria. (Supply Chain; Game Theory; Multiechelon Inventory; Incentive Contracts)

1. Introduction
How should a supply chain manage inventory? If the members care only about overall system performance, they should choose policies to minimize total costs, i.e., the optimal solution. While this approach is appealing, it harbors an important weakness. Each member may incur only a portion of the supply chain's costs, so the optimal solution may not minimize each member's own costs. For example, a supplier may care more than a retailer about consumer backorders for the supplier's product, or the retailer's cost to hold inventory may be higher than the supplier's. While the firms may agree in principal to cooperate, each may face a temptation to deviate from any agreement, to reduce its own costs. Supposing each firm can anticipate these temptations, how will the firms behave?

Furthermore, to what extent will these temptations lead to supply chain inefficiency? This paper studies the difference between global/ cooperative and independent/competitive optimization in a serial supply chain with one supplier and one retailer. (We assume there are two independent firms, but the model also applies to independent agents within the same firm.) Consumer demand is stochastic, but independent and stationary across periods. There are inventory holding costs and consumer backorder penalty costs, but no ordering costs. There is a constant transportation time between stages, and the supplier's source has infinite capacity. Inventory is tracked using either echelon inventory or local inventory (A firm's local inventory is its on-hand inventory, and its echelon inventory is its local inventory plus all
0025-1909/99/4507/0936$05.00
Copyright © 1999, Institute for OpcraHoas Research and the Management Sciences

936

MANAGEMENT SCIENCE/VOI. 45, No. 7,

July 1999

and Porteus (1997). Section 6 compares the games' equilibria with the optimal solution. Three exceptions are Lee and Whang (1996).) In the optimal solution. We also study Stackelberg versions of the games. (A pair of strategies is a Nash equilibrium. Implementation of the cooperative solution requires that the firms eliminate the incentives to deviate. so competitive decision making degrades supply chain efficiency. they should modify their costs so that the optimal solution becomes a Nash equilibrium. They develop a nonlinear transfer payment contract that induces each firm to choose the system optimal base stock policies. This is an important assumption. the firms choose base stock policies. We distinguish between echelon inventory and local inventory and investigate how these different methods for tracking inventory infiuence strategic behavior. (Section 3 discusses this modeling issue. in which one dominant player chooses its strategy before the other. We compare the games' equilibria to each other and to the optimal solution. Our model differs from theirs on several dimensions. In these games neither player dominates the other. and therefore neither firm has an incentive to deviate unilaterally from the equilibrium. To model independent decision making we consider two games. and the retailer pays holding costs on units it possesses. but he proposes a different coordination scheme.. The optimal solution is typically not a Nash equilibrium. we develop linear transfer payment contracts. We assume the upper stage (the supplier) may care about consumer backorders. Hence. They consider a setup cost at the upper echelon. and §5 analyzes the two games. Therefore. and §10 concludes. In Lee and Whang (1996). • Chen (1997) studies a game similar to the popular Beer Game (Sterman 1989). the firms use echelon stock policies and all backorder penalties are charged to the lowest stage. and it cannot be modified once it is announced. each firm makes an optimal decision given the behavior of the other firm. July 1999 . This goal can be achieved by a contract that specifies linear transfer payments based on easily verifiable performance measures like inventory and backorders. Chen (1997).CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies inventory held lower in the supply chain. the performance of the system.) Thus. Both firms are concerned about consumer backorders. described in §3. called responsibility tokens. except that the demands in 937 MANAGEMENT SCIENCE/VOI. the competitive decisions are nontrivial. We find that in each game there is (usually) a unique Nash equilibrium. No. We develop a set of linear contracts that meet this objective. and briefiy discuss other techniques for aligning incentives. The upper stage incurs holding costs only. and §3 formulates the model. 7. if each firm minimizes its own cost assuming the other player chooses its equilibrium strategy. because it allows us to study how the firms' relative preferences infiuence their strategic behavior and. Section 8 discusses the numerical study. so it may carry inventory even when inventory policies are chosen competitively. Section 4 describes the system optimal solution. In both games the firms simultaneously choose their base stock levels. the Echelon Inventory (EI) game and the Local Inventory (LI) game. We evaluate the magnitude of this effect with an extensive numerical study. the supplier pays a consumer backorder penalty as does the retailer. The next section reviews the related literature. 2. while we do not. in turn. These policies can be implemented by tracking either echelon inventory or local inventory. whereas in the LI game both firms track local inventory. with competitive selection of policies. 45. Literature Review The literature on supply chain inventory management mostly assumes policies are set by a central decision maker to optimize total supply chain performance.) The EI and LI games differ in only one way: In the EI game both firms are committed to tracking echelon inventory. i. and the firms simultaneously choose their strategies. Porteus (1997) studies a model similar to Lee and Whang's model. Section 9 analyzes the Stackelberg games. The firms in each game play a Nash equilibrium. thereby minimizing its own cost.e. Section 7 describes contracts that make the system optimal solution a Nash equilibrium. the upper stage carries no inventory. This is their only strategic decision. Finally. The supplier pays holding costs for inventory in its possession or in-transit to the retailer.

to model lost sales directly. The parameter p is the total system backorder cost..>:: • ? •.. This scheme is more complex than ours in some ways. The supplier is charged holding cost h2 per period for each unit in its stock or enroute to the retailer. Time is divided into an infinite number of discrete periods. The parameter a is exogenous. Model Description Consider a one-product inventory system with one supplier and one retailer..) Alternatively. = IL^.. local inventory level. Finally. Each firm may order any nonnegative amount in each period. : ? * ILj. yet their models are significantly different. Kouvelis and Lariviere 1996). Each firm uses a base stock policy. where a consumer may switch among firms to find available inventory. No. analogous to the treatment of inventory financing costs. There is a lead time for shipments from the source to the supplier. 0 ^ a ^ 1.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies different periods are independent random variables with a common distribution that is known to all players. Chen et al.. as explained in §7. is all inventory at stage / or lower in the system minus consumer backorders.a)p. The supplier is Stage 2 and the retailer is Stage 1. Such costs need not affect the firms equally. These backorder costs have several possible interpretations. IP.) In period t before demand define the following for stage /: in-transit inventory.. Muckstadt and Thomas (1980). Ha 1996. but that introduces considerable analytical difficulties. (3) consumer demand occurs. Unmet demands are backlogged. They may represent the costs of financing receivables. Many papers investigate how a supplier can induce a retailer to behave in a manner that is more favorable to the supplier (e. per period for each unit in its stock.. echelon inventory position. Tsay 1996. Assume /ij > 0 and /z. and from the supplier to the retailer. they provide a crude approximation to lost sales. if customers pay only upon the fulfillment of demands. Lai and Staelin 1984.. they have no competing interests. and all backorders are ultimately filled. though simpler in others. He outlines an accounting scheme that allows each player to optimize its own costs and yet choose the system optimal solution. A firm's local MANAGEMENT SCIENCE/VOI. (This requires a discounted-cost model to represent exactly. = IL.. Both the retailer and the supplier may incur costs when demand is backordered. 7. The behavior of a central planner has also been investigated in settings with moral hazard (e. ••" ='ii . and Axsater (1996) investigate a centralized control system that allows each firm to optimize its own costs and still choose an outcome desirable to the central planner. Each firm pays a constant price per unit ordered.g. IT„. which in tum lead to long-run declines in demand. Donohue 1996. 45. IP. In a multiechelon model with multiple retailers. is inventory at stage i minus backorders at stage / (the supplier's backorders are unfilled retailer orders). IP. which is why we allow the fiexibility to choose a G [0. Consumer demand at the retailer is stochastic. whereas we only assume rational players.. they may be proxies for losses in customer good-will.. IP.. and a specifies how this cost is divided among the firms. Using an echelon base stock level./?T v. + IT. Several other papers address related issues. 3. ILj. of course. Pasternack 1985).g. Porteus and Whang 1991. each period the firm orders a sufficient amount to raise its echelon inventory position plus outstanding orders to that level. Lippman and McCardle (1997) study competition between two or more firms in a one-period setting. all standard.. Narayanan and Raman 1996.. Unlike our players. but the approximation here is standard in the average-cost context. (It would be better. The retailer is charged ap for each backorder. There is no fixed cost for placing or processing an order. L. (2) orders are submitted and shipments are released. (4) holding and backorder penalty costs are charged. The retailer's holding cost is /JJ + /i. ^-i: . He also studies the behavior of boundedly rational players. echelon inventory level. + IT. July 1999 . > 0. 1]. Even the optimal policy is unknown. Moses and Seshadri 1996. and the supplier (1 .. and local inventory position. L. his share the objective of minimizing total system costs. (1997) study competitive selection of inventory policies in a multiechelon model with deterministic demand. so there are no quantity discounts. independent across periods and stationary. The following is the sequence of events during a period: (1) shipments arrive at each stage. Parlar (1988) and Li (1992) also stiidy the role of inventory in the competition among retailers... Hausman and Erkip (1994).

where Gl(y) = E[Gl(y . Again. MANAGEMENT SCIENCE/VOI. Let D ^ denote random total demand over T periods.). [x]~ = max(O. Sj + s.. Echelon and Local Inventory Games In the Echelon Inventory (EI) game.(s.(Sj. Let </)' and <> be the density and distribution functions of 1^ demand over r periods. equals player i's echelon base stock level. the players simultaneously choose their strategies. A prime denotes the derivative of a fimction of one variable. (However. horizon. S].D^n •• The supplier's optimal echelon base stock level. S2) = Hi(s. After their choices. Furthermore. 7. the two stages are independent firms or players. define G°(IP. except the local inventory position replaces the echelon inventory position.. so positive demand occurs in each period.. Math notation follows: [x]* = max{0. each firm chooses a policy that minimizes its ers are committed to their strategies over an infinite cost function.D') equal the retailer's charge in The players know which game they are playing. the period t. x}. In period t charge the supplier G°2(IP2f). all model parameters are common knowlgruen and Zipkin (1984). it may not be possible to run the system optimally with local base stock policies. Sj) in the sense that H. (S is sufficiently large that it never constrains the players. where pair (S]. and all parameters are common knowledge. there is no need to define distinct cost functions with local arguments. In a nonstationary and define demand environment.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies base stock level is similar. respectively. Since any echelon base stock pair can be converted into an equivalent local Define sJ as the value of y that minimizes Gl(y): pair. System Optimal Solution is a pair (s. where choice between the EI and LI games is not one of their decisions. strategies are chosen simultaneously. so the same is true of <i>\ T > 0. increasing. We assume tl>'(x) is continuous. Define s^ as stage f's echelon base stock level and s.. $'(0) = 0. the playThen. where s. G a. S2) is equivalent to (s. the players The system optimal solution minimizes the total averimplement their policies over an infinite horizon. and /i' denote mean total demand over T periods. and S is a very large constant. and Chen and Zheng (1994) edge. When S2~S2 + s. and s. —x]. July 1999 939 . In the game's only move. [a. Define H. demonstrate that an echelon base stock policy is In the Local Inventory (LI) game the supplier and optimal in this setting.) A joint strategy s 4. minimizes 5.. s. This section briefiy outlines this method. Also in period t.. No. Although there is little operational distinction between echelon and local This is the retailer's optimal base stock level.. G (T = [0.. Let G''^(IL^. Clark and Scarf (1960). = s. we later show that they differ the induced penalty function. the operational equivalence of echelon and local base stocks does depend on the assumption of stationary demand. . as its local base stock level. Sj). In age cost per period. Federaddition. The optimal solution is found the retailer choose local base stock levels. We will frequently switch a pair of base stock levels from one tracking method to (1) another to facilitate comparisons. and E[x] is the expected value of x. s. Define base stock policies. b] is the closed interval from a to b. s^) as player i's expected per-period cost when players use echelon base stock levels (s. and differentiable for X s 0.) as the retailer's Sj). by allocating costs to the firms in a particular way. s. 45. the local base stock expected charge in period t + L^.) . a is player i's strategy space. strategically.

. We use a standard derivation. the retailer is charged /z. = S2 . Define Gi(/Li. Hi(s. such that each player chooses a best reply to the other player's equilibrium base stock level: The retailer's true expected cost depends on both its own base stock as well as the supplier's base stock. S2) = min H2(Si. in the LI game.) 5.. (s\. • and G2(/P]. assuming retailer orders are shipped immediately. Likewise. - i. but before period t demand. s'j). the supplier can completely fill the retailer's period t order. and 7P. x)}.(s2 . No. is strictly convex.x)dx.S . ' .(y). the supplier's echelon inventory level equals S2 . When S2 . Define (xDefine sJ as the value that minimizes this function. S2) = min . with a subset of a according to the following rules: . When S2 .) as the supplier's expected period t + L^ backorder cost.. We generally find a unique pure strategy equilibrium. the supplier cannot fill all of the retailer's order. j i" i.. in the EI game. E[s2 . S 2 . After inventory arrives in period t. • Sl = arg min G. that is. . the supplier's echelon inventory position equals S2.. S2)} er's costs. ap "1 + "2 + «P si) = {S2 E a I H2(Si. . or local base stock levels.. x + Si)}.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies For the EI game the best reply mapping for firm i is a set-valued relationship associating each strategy S:.D'" > s.. Sj + x)} xScr ''2(si) = {§2 G o-1 H2(si.D^^] is the supply chain's expected inventory level (average supply chain inventory minus average backorders). 1 Define 62(^^1/ t backorder cost. (s[. 45. After the firms place their orders in period t . A pure strategy Nash equilibrium is a pair of echelon base stock levels. .D^\ Hence. S2) = E[Gi(min{s2 - (We do not consider mixed strategies. 0}). MANAGEMENT SCIENCE/VOI.D') as the sum of these costs in period t. sJ). the best reply mappings are rSi) = {Si e o-1 Hi(si. so s. . Differentiation verifies that G. Actual Cost Functions In each period. 7. S2 + Sl) = min H2(si.D''\ Hence.D*"^ < Sj. the supplier's actual period Define Gi(/P]. S2 + Sl) = min H^(x.) as the retailer's expected cost in period t + L. is determined by G'j(sJ) = 0. the base stock level that minimizes the retailso i. so IP^. + /I2 per unit held in inventory and ap per unit backordered. for the LI game. = S]. July 1999 .1.L^.. x) = m\n{x.

(We subsequently consider the extreme cases a = 0 and a = 1. the second term is the expected cost for inventory held at the supplier and the final two terms are the expected backorder cost charged to the supplier. The second term. ... LEMMA 1. dH2 . When s. However.e. < s^ and constant for s. from the implicit function theorem. because then ^''(S2 .. This condition cannot hold at S2 < s..e.X)dx. > s^. . and strictly convex when Sj ^ 0 and Sj > 0. s^ > 0. we assume that each firm incurs some backorder cost. the retailer's Sj only influences the allocation of inventory between the supplier and the retailer. JEchelon Inventory Game Equilibria with Shared Backorder Costs When Sj < s°. so T-jCSi) is a function (i. increasing for s" < Sj :S Sj..))]. Hj is decreasing for s. s^) is strictly convex in s^. i. bfta Both terms are convex. 0 < a < 1. Hence. and constant for Sj > Sj.. the supplier's inventory declines as s. However. r^Csi) is a function. -I. but when s^ stays constant. and strictly convex 941 MANAGEMENT SCIENCB/VOI. Consider H. (2) where s. When s^ > s". Assuming a < 1. the total system inventory depends on Sj only. H. = r^{s.. holding Sj constant.s. Consider the following function of Sji -. and S2 = s. From Lemma 1. - + ^ sJ).) < 1..CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies for S2 s: s. 45. The first term above is the expected holding cost for the units in-transit to the retailer (from Little's Law). is convex. a The following lemma characterizes the supplier's best reply mapping. increases. is decreasing for s.. the retailer can increase total system inventory by raising s. H.Sj.X)dx = 0. PROOF. H2(s. Hj has a unique minimum) and is determined by the first-order condition PROOF. Hj is strictly convex in Sj. H^{s. is convex.2.) > s. > Sj.]. 7.. is quasiconvex in s. holding Sj fixed. < Sj and differentiate Hj.. Fix D'-' and s. We mentioned above the operational equivalence of local and echelon base stock policies when s. H.V-y]. Assuming a < 1. and H^is^. Given s^ > Sj. the change in player i's cost due to a shift in player j's strategy depends on the inventory tracking method.. No. the supplier's expected on-hand inventory is independent of s. Sj) is strictly convex in Sj s: 0. LEMMA 2. h^E[{s^ . while the second term is strictly convex in the interval Sj £ [D^\ D'-' + s. = s.. In this section. and 0 < r^(s. Hence. For example. Assume s. is constant with respect to s. July 1999 .D^\ s.Si) = 0 and G'^i}/) < 0.) We begin with some preliminary results on the players' cost functioris and best reply mappings. Furthermore. 5. Therefore. r-^is^) > s. < s".. So holding s^ fixed. Now take the expectation over D^\ The first term. s^) is quasiconvex in s. s^. E[G2(min|s2 .

80 3. + S2) is strictly convex in S2. {s{. 0 < r'^is^) < 1. H^is.00 3.20 240 2. 3. When Sj < sJ.60 r. . this implies sJ = sJ. ( S .s. • :•:-.70 3.(s. S2) is strictly convex in S2. Differentiation of H2(s. sQ..00 2. = 0.: MANAGEMENT SCIENCE/VOI. Local Inventory Game Equilibria with Shared Backorder Costs The analysis of the LI game also begins by characterizing the cost functions and the best reply mappings.50 3. H.^2 DEI equilibrium AU equilibrium From Theorem 1.:. . so H2(s. = /ij = 0. s. so = [s^. S] S 2 < S r Recall that G. there is only one Sj = r^is").90 2.20 LEMMA 5. + S2) reveals that S2) i. s. PROOF.(52) + .) > 0 for s^ > s.00 2.(i/) is strictly convex and minimized by y = sJ.30. 2. n Although in the El game the supplier does not always fill the retailer's orders immediately. 7. > S2 minimizes G. p = 5. No.. LEMMA 3. n The retailer's best reply is not necessarily unique. iO 3. By the assumptions and Lemma 1.60 3. the equilibrium is unique. Since r2 is a function.. Assuming 0 < a < 1. Let x = D^\ When s.3. compact convex subset of a Euclidean space. . When S2 > s". the supplier's echelon base stock level has little influence over the retailer's strategy.80 2. If s'2 ^ si.^. only s. From Lemma 1.20 2.(nain{s2 PROOF. For the El game. PROOF.90 2 80 1.80 3. s. and (2) player i's cost function is continuous in s and quasiconvex in s.(s2) = s". a pure strategy Nash equilibrium exists if (1) each player's strategy space is a nonempty. in the El game (s{ = s". SetS2 = s.20 Figure 2 3.00 2t 2.40 Si 2. /?.. Hence Sj > s. From Lemma 2 in any equilibrium. S2 = ?'2(Si)) is the unique Nash equilibrium.x > s°. + S2) is strictly convex in s. a contradiction.50 3. but from Lemma 3.)). these conditions are met. 4. a = 0..x < s".10 3.5. S].. all X.70 3. Figure 1 ReactJon Functions. p = 5.30 O Optimal DEI equilibrium ALI equilibrium 3. so r.00 2. _ ]\A. When s. Differentiate H . Sj .. D Figures 1 and 2 plot the firms' reaction functions and the resulting equilibrium for two examples. the retailer's best reply 3. THEOREM OOpiimal /^ 3. . ft. = s" minimizes Gi(minis2 .00 2. .20 / 3. . July 1999 . s.20 3.80 1. = s.2 in Fudenberg and Tirole (1991). s. + S2) is strictly convex in I2 and.90. .)).60 3. S2) (3) 5. so there is at least one equilibrium.80 3.(min{s2 — x. + S2 and s. H2(Si. a = 0.10 / mapping is V '•l(*2) 2. s.D'-' < s j .40 Reaction Functions.40 3.5. ::• f . but there is only one Nash equilibrium.80 3. for i = Si minimizes G.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies The cross partial of H j is negative because G'2 < 0 < h^ and <t>'-'{s2 . Therefore..x. Lemma 3 implies s\ s s^. Si + S2). Since G ^ > 0. 45.60 2. s^ = r^is^) > s\. s.

The existence proof in Theorem 4 applies even when a = 1. s*). (s'. Suppose (s*. When a = 1. From Lemma 3.) . r. $'^'(s2)G'.. PROOF.(-) is strictly convex. the Nash equilibrium in the El game is no longer unique..) > 0. 5^H. Hence ri(s2) > sj. see (4).CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies . PROOF. in the LI game {s[ = ri(O). .(s. S] + S2) When s. S]. (s. Now. ri(s2) = s". s. ^. so s* > sJ cannot be an equilibrium. D LEMMA 7. Hence. > 0. No. From the implicit function theorem 5. + §2 = Sj.s. Lemma 5 confirms the required conditions for the existence of an equilibrium (in the proof of Theorem 4). The retailer's best reply is determined by the first order condition aH. Equilibria Under Extreme Backorder Cost Allocations ^ Suppose the retailer is charged all of the backorder costs. s*) is an equilibrium.. When §2 > 0.) = rj(s.]. Assuming a > 0. From > . there S2) S2) .) + s. + S2)/5s. July 1999 MANAGEMENT SCIENCE/VOI. so s* > S2 implies that s* > s'. r^is^) > s. + Sj = PROOF. there is a unique equilibrium.4. i.(Si) = 0 and therefore r'^is^) = .(S.) > 0.(s..1 < r'. Since r. and therefore dH^(s. 45. so a pure strategy equilibrium exists.. which implies that r^is^) > 0. From Lemma 7.. In this situation. s. assume §2 < s*.e. Without loss of generality. G.1 < f'^is. where s* > s'. + s. < 0. the supplier incurs no backorder costs. n Figures 1 and 2 also display the reaction functions in the LI game as well as the Nash equilibrium.1 < f'^is. S2) and (s*. so s* = s^ + s* > s\ the same lemma. .. When a = 1 the supplier chooses §2 = 0. a = 1. Assuming a < I. in the El game the Nash equilibria are (s\ e [Sj. THEOREM 9.) . S]. Sj) is the unique Nash equilibrium. When s. the supplier picks S2 < s. but an equilibrium only occurs when s* < s*. PROOF.) < 0. = r^{s^). r2(s. 7. First. LEMMA 6.) = r'^{s. s. r. = 0. only holding costs. s.)/as.) < 1.h •. From the same lemma. 0 < r2(s. and . Also r[(s. is strictly convex in s. Assuming 0 < a < 1.) whenever S2 = s. there is a unique Nash equilibrium in the LI game.I. a • • .1 . (s[. . From Lemma 2. S]. D The next two lemmas characterize the best reply mappings. Sj e [0.) = [0. a In the LI game there is a unique equilibrium even when the retailer incurs all of the backorder cost. suppose there are two equilibria. + Sj and s.(s2) < 0. Hence.x)dx ' Assume S2 > 0. = Sj.~:-^5- When 0 < a < 1. so S2 > 0. from Lemma 6. Since ^''(Sj)G'[{s.e. because H2(S|.. this implies that s* < s'.(0) is unique.1 .(s.. a contradiction. which means that H. r'jCsj) = .. so . §2 = 0) is the unique Nash equilibrium.. = 0. For the supplier ^^(s. r2 is increasing.1 . n mim When the supplier incurs all backorder costs. But from Lemma 2. . Assuming a . {s* E [s*. so for any s* ^ s". r. For a = 1.(s2) is a function.) < 0. s.1 < f\(S2) < 0. s. Suppose s* ^ s'.) < 0. ri(s2) = [s*. the numerator above is positive and the numerator equals the second term in the denominator. i. + > 0.a. s*) is an equilibrium. s"])PROOF. + S2) = H2(Si.1. < s". THEOREM 10. 943 . From Lemma 3.(s2) > sj. THEOREM 8. and when S2 = 0. S2) + (4) Since G.

a In the LI game either sj > sJ or sJ < Sj is possible. Since s[ > s[. In an El game equilibrium. . THEOREM 14. i. Assuming 0 < a < 1. J)).s.'(y) < G'. (In Figure 1 the retailer chooses s[ < sJ. s^). but in Figure 2 s'l > s°. No.s'l) is the unique Nash equilibrium of the LI game. (s\. The supplier's best reply mapping is a function in either game. Sj). From Lemma 2. s' = since dHjis^. is the unique Nash equilibrium in the El game. so Sj = '"2(0). in Si. The retailer always chooses a lower base stock in the El game than it does in the LI game. dx x) = . assuming the supplier chooses S2 = r2(Si): ' ' -' PROOF. §2 = S2 . and {s[ = s'. Competitive Equilibria and the Optimal Solution In the El game the retailer's base stock level is lower than in the optimal solution. the base stock levels for both firms are higher in the LI game equilibrium than in the El game equilibrium. r2(s.(s2) > s. Since the retailer incurs no backorder cost sJ = sJ = 0. {s\ = 0..2.) However. Competitive Equilibria The firms choose higher base stock levels in the LI game than in the El game. 7. a Why does the supplier prefer the LI game equilibrium? The supplier always prefers the retailer to increase its base stock. when backorder costs are charged to the supplier. In the LI game the supplier chooses r2(S]) . (The numerical study confirms this.Furthermore. so it is sufficient to show that sj > s'2. hence the supplier is always better off in the LI game. convert the LI game equilibrium..) > s. 6. From Lemma 7. the supplier's cost declines as Sj increases. S2 = The equilibrium in the El game is {s[ = s". Hence. G. Assuming a < 1.1. 6. where sj = s[ and Sj = s\ + S-. into the equivalent pair of echelon base stock levels. the supplier's base stock level is lower than in the system optimal solution in both games. To facilitate these comparisons. For x < . Sj > S2 and s\ > s[. THEOREM dH-.a)p and MANAGEMENT SCIENCE/VOI. 5s. 11.) However. r^is^) is increasing PROOF.. PROOF. the supplier's base stock level in both the LI and the El equilibria is lower than in the system optimal solution. Comparing Equilibria This section compares the equilibria in the LI and El games to each other as well as to the optimal solution. which implies that s\ > sj = s°. s'l > s^ = s\. so S2 = r2(s[) > r2{s'^) = si a The retailer's cost in the LI game equilibrium can be more or less than in the El game equilibrium. the retailer's base stock level is lower than in the optimal solution. Since both Gr(y) and G'. the supplier has a definite preference for the LI game.e.s . and Gj < 0. Assuming a = 0. H2 is lower at s[. r.(y) are increasing in y. In the El game the supplier chooses r2(Si). r2{s^))/dS2 = 0.s^ as its local base stock level and the equivalent echelon base stock level is r2{s^). In any equilibrium Sj > Sj. July 1999 . < 0. T H E O R E M 12. Sj = si .( 1 . the supplier's cost in the Ll game equilibrium is lower than its cost in the El game equilibrium. so <I>'-'(S2 . Differentiate the supplier's cost function with respect to the retailer's base stock level. THEOREM 15.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies is a unique equilibrium in both games and they are identical.s' n PROOF. PROOF. so dH2/ds. 6.(y). 45. Thus. Note that Gi'ix) < G\ix). Assuming 0 < a < I. (s^. for all y.) > 0. for all x. From Lemma 2. thereby increasing the retailer's inventory and decreasing the supplier's backorder costs. THEOREM 13.

S o . so the proof of Theorem 15 demonstrates s^ ^ s^. (Nonlinear payment schedules. Hence. Sj > S2. assuming the supplier fills retailer orders im945 16. When the supplier incurs no backorder costs. For instance. in the LI game Sj = s[. 45. But. + ^2 + p) < {h^ + h2 + p). There are several schemes to achieve this goal: a per period fee for the supplier's backorder. Sj < s^. as in Lee and Whang (1996). There are several methods that enable the firms to minimize total costs and still remain confident that the other firm will not deviate from this agreement. or a subsidy for each unit of inventory in the system. THEOREM From Theorem 15. PROOF. the LI game Nash equilibrium is S2 = s[). the firms can lower total costs by acting cooperatively. In the El game Sj = sj < s" < s" < s". so the optimal solution is not a Nash equilibrium in either game. For instance. THEOREM 18. PROOF. July 1999 . e. the system optimal solution can be a Nash equilibrium under a very special condition.g. 7. Assuming a < 1. the firms could contract to choose (s°.. Cooperative Inventory Policies According to Theorem 17.s . Such stipulations are hard to enforce.) Transfer payments can also be imposed on the retailer. the system optimal solution is a Nash equilibrium in the LI game only when PROOF. the firms could write a contract that specifies transfer payments which eliminate incentives to deviate from the optimal solution. For x > 0. the system optimal solution is not a Nash equilibrium. a 17. Define accounting inventory and accounting backorders as the inventory and backorder levels. Sj = sJ = sJ = s^. with strict inequality for . Payments could also be based on the inventory and backorder levels that would have occurred had the supplier performed certain actions. For the supplier S2 = s'l when G2' = 0. a per unit fee for each unit the supplier does not ship immediately. a Recall that the supplier's echelon base stock determines the supply chain's average inventory level. n When the supplier incurs no backorder costs. Chen (1997) uses this approach. since each firm has an incentive to deviate from this contract (because it is not a Nash equilibrium). and si ^ s^. the contract must also specify a penalty for deviations. but these are necessarily more cumbersome. s^ ^ si < Sj. When a = 1. = MANAGEMENT SCIENCE/VOI. S o l v i n g f o r s[. the optimal solution is virtually never a Nash equilibrium. From Theorem 15 it follows that in either game's equilibrium the supply chain's average inventory level will be lower than in the optimal solution. Alternatively.' because (p'-^*'-'*'' stochastically dominates *'-'^' and p/(/i. could also be considered. suggesting that competition will also tend to lower the supply chain's average inventory. D In that case. GjX^) .s . S2) as their base stock levels. a per unit fee per consumer backorder. THEOREM 7. Therefore. Assuming a = 1. Assuming a = I.Sj < Sj. a subsidy on retailer inventories..CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies For . The numerical study confirms this observation. < X < 0 -(1 and again Gl'ix) = -p.dH2/dx.O] = h^ and It is possible that s[ = s'. with strict inequality for S2 > S). or a subsidy on backorders. the supplier's base stock level is no greater than in the system optimal solution. -a)p. the induced penalty function G"^ is nonlinear. When a = 1. This occurs precisely when with strict inequality for x < s. So in all cases G2'(^) < dH2/dx. hence. < x < sJ. No.

and the supplier charges the retailer ft. §2) is a Nash equilibrium for the cost functions H''. Suppose the firms track local inventory and they adopt a transfer payment contract with constant parameters ((. Then determine the contracts in which s] satisfies player i's first order condition. §2) as the expected per period transfer payment from the supplier to the retailer. + S2) = . MANAGEMENT SCIENCE/VOI. 45. there is no strategic interaction between the firms. the retailer will choose s°. (i. With these contracts the firms can choose {s\. is stage i's backorders. (We later impose some restrictions on the parameters. and s\ + Sj = Sj. given that player. . Let H'{s^. the supplier chooses S2. s. i. We study linear transfer payments based on actual inventory and backorder levels. but not the supplier's backorders.This contract specifies that the period t transfer payment from the supplier to the retailer is + 2. Define Ti(/P. + §2)..) We also assume the optimal solution is common knowledge. first assume that H] is strictly convex in s. this approach creates a challenging accounting problem. Sj + S2) be player i's costs after accounting for the transfer payment. + S2) = H. Since the supplier incurs all actual costs.. influences the retailer inventory and backorders. where sJ = s". Linear Contracts min{0. . (5) 052 Define T(s.. /32.(s...x)]dx. Note that s. With this scheme. While our approach avoids the problem of tracking accounting inventory and backorders. is the retailer's on-hand inventory.{s. The following are the first order conditions: . thereby minimizing total costs. his technique may be easier to implement in some situations. determine the subset of these contracts that also satisfy the original strict convexity assumption..). s. x)dx. July 1999 . 7.) as the expected transfer payment in period f + L. /3. e. H\(su s. Then. + where /. S2). To find the desired set of contracts. due to retailer inventory and backorders. and also be assured that no player has an incentive to deviate.1.g. Ours also requires a demand parameter.. s. all measured at the end of the period. i.. where Uy) = E[t. and B. No.r(Si.CACHON AND ZIPKIN Competitive and Cooperative Inventor\/ Policies mediately.^. ours in others. chooses s'^.D'^^T + We wish to determine the set of contracts. s. + S2) + r ( s . 7. Chen's method uses only cost parameters. such that {s\. s. > 0 represents a holding cost subsidy to the retailer and i. However. < 0 represents a holding fee. Sj). thereby minimizing player i's cost. Suppose the supplier pays all of the retailer's actual costs.„ ^2' /3i). Thus..[y . per unit of accounting inventory and /12 "•" P per accounting backorder. j ^ i.. the retailer's decision is independent of the supplier's. and the retailer chooses s°. There are no a priori sign restrictions on these parameters. S2). Finally. + S2) .

H. r /j. ^2+ /32+ (1 .CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies ••'• +T\{Si + S2-x)]dx. s . along with (8) and (9). hi + h2+ ap . .sj).a)p + ll +/3. and the third states that the supplier should not fully reimburse the retailer's backorder costs.: . When the firms choose (i. 7. To help interpret these results. Using (7).a)p = 0 J. No. essentially its fill rate. ^2/ Pi) ^° satisfy (8) and (9). (5). consider the three extreme contracts where one of the parameters is set to zero: =0 ii) I. assuming PROOF. (6) + S2 .a) (Of these three contracts. the second does not meet the conditions in Theorem 19. = .x))dx > 0. sufficient conditions are (1 . s . then the optimal policy {s\. Define 72 = ^'^'(s" . For the supplier.) Furthermore. and (6) yield the following two equations in three unknowns. July 1999 . < ap..(This is the supplier's in-stock probability. Substituting (8) yields i. The retail947 MANAGEMENT SCIENCB/VOI. n These are quite reasonable conditions: The first requires that the retailer's inventory subsidy not eliminate retailer holding costs. = When the following second order conditions are satisfied.a)p. The first inequality reduces to . because the supplier fully compensates the retailer for all of its costs.= ^ 72 (9) It remains to ensure that the costs functions are indeed strictly convex. 2: 0 and |3. = 0. the second stipulates that the supplier be penalized for its backorders.a)p. The second inequality.x)dx.( 1 .p + (p + X . > 0. < h. is stiictly convex in s. yields ^2 > 0.x))dx>0. /3.( 1 . . 45. and the retailer should not overcompensate the supplier's backorder costs. or. Combining the first inequality with (8) yields t.Li~ (7) Pi>0.a)p + /3. 19. and the following additional restrictions apply THEOREM (i) > ^ ^0 1 (ii) (iii) ap > /3. (8) h-. = (1 ..si) + (h. u l .( 1 . the supplier's first order condition in the optimal solution is 0 = . s[) is a Nash equilibrium. iii) t. + ^2 and /3.

this payment should be independent of all other costs and actions. + p + = 72- 1 - ft. even if there were additional Nash equilibria. but not fully (provided a > 0). one firm's cost may increase. However. once s. when firms use local base stock levels. one could seek a contract (i. only a supplier backorder penalty is required. With the third contract the supplier subsidizes the retailer's holding costs. 8. Clearly this increases the complexity of the contract. The second identity specifies a critical ratio for the supplier. with echelon stock base stock levels. the players can coordinate on this equilibrium.i-. This can create a perverse incentive. + ft: Consider the first identity./(ft. + ft2 . but less than in the first contract. this rule effectively reduces both stages' holding costs by the same fraction. > S2. the one correspond- ing to the optimal solution Pareto dominates any other. The i32^2t transfer payment could easily dominate the additional retailer inventory cost. does influence P2B2/./3i ft. this analysis assumes the firms use local base stock levels. These ratios must be identical to induce the retailer to minimize total system costs. Cooper et al.) There is a solution to this problem. Hence. Additional Contracting Issues Theorem 19 details the contracts that make the optimal solution a Nash equilibrium. Hence. (There is experimental evidence that players coordinate on a Pareto dominant equilibrium when they are able to converse before playing the game. (Furthermore. without increasing its inventory. (There is only a tiny probability of negative demand. above sJ'. + ft2)). 1989. including transfer payments. arbitrarily large (assuming S..i.e. we conducted a numerical study. In this context local policies have several advantages over echelon policies. is large too). the retailer would receive no additional benefit by raising s. is which is written more simply as ft. Numerical Study The system optimal solution is virtually never a Nash equilibrium. 7. taking the transfer payment into account.) The remaining 45. So the left-hand side is the retailer's critical ratio.. the supplier still carries inventory because it pays a penalty for its local backorders. For example. Thus..) With the first contract the retailer fully reimburses the supplier for the supplier's consumer backorder penalty. One period demand is normally distributed with mean 1 and standard deviation 1/4. (8) and (9) can be written ap . e. By increasing s. P2 is its (local) backorder cost. July 1999 948 MANAGEMENT SCIENCE/VOI. and its holding cost. The righthand side is the critical ratio for the total system costs controlled by the retailer.2. The transfer payment could assume that the retailer chooses s^ ^ s'^. 7.|3.. but how large is the difference between their costs? To answer this question. ^2' Pi) such that each firm's cost is no greater than in the original Nash equilibrium. The holding cost ft2 is multiplied by the factor (1 . Alternatively. The quantity ap .) Although total costs decline when the firms coordinate. Suppose P2 is large. s. the retailer can make B2. Cachon and Camerer 1996. P2 = 72''2/(l ~ 72)Incidentally. the limit on s. which is the fraction of actual holding costs paid by the retailer. > S2. However. . No. This firm will be unwilling to participate in the contract unless it receives an additional transfer payment. the supplier is penalized for its backorders. a = 1). is the retailer's backorder cost. To maintain the strategic balance of the contract. Finally. the retailer can increase s. but any s. the firms could transfer a fixed fee each period.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies er's incentive to choose the optimal policy is weak: s° is a Nash equilibrium strategy. Nevertheless. has no influence on ^2^2 supplier's backorder penalty. but this does not imply a unique equilibrium. Recall that s. In addition. Local measurements avoid this problem altogether.g. When the retailer incurs all backorder costs (i. holding S2 constant. is too.

and the maximum is 13% and 8%. 25} retailer incurs no backorder penalty (i. a declines).: . In the El game the median percentage is 84% when the a G {0.3. 0.939% 120% 27% 13% 19% 66% 34. In the competitive solution when the players care about backorder costs equally the supplier carries no inventory.493% 10. 8. = s.1 Echelon Inventory Game 0. 0. 7.e. respectively. 17% 34% 45% 116% MANAGEMENT SCIENCE/VOI.e. respeccosts as a function of . .7.edu/" gpc/.271% 5. we choose the one order cost (i.5 0. 0. a = 0.7 0.V 96% 19% 0% Maximum 10. Look at Figure 3.930% 97% 21% 11% 12% 40% 9..duke. system optimal solution: In the El and LI games the which displays the percentage increase in total system median competition penalty is 6% and 3%. G {1. The competitive outcome is poor when a = 0 because the retailer chooses s.930% . tively. July 1999 949 . the competition penalty is substantial the retailer holding cost is constant. as the retailer with the largest s. 5. However.5 0.: ''•• .5. a increases)..1. 0. No. The performance of the competitive solution www. 0.9} penalty (i.9 1 0 0. so all consumer demands are backordered. (2) the Nash equilibrium consumer backorders. Nevertheless. the Table 1 summarizes the percentage increase in cost maximum competition penalty increases rapidly.939% 119% 0 0. L2G{1.9.) (When a = 1 there are muldeteriorates rapidly as the retailer incurs lower backtiple Nash equilibria in the El game.) incurs higher backorder cost (i. because the supplier refuses to carry inventory. therefore.5). We call this percentage the competition pen..while the minimum and median penalties do not. a = 1). i . Consider the extreme case where all backorder costs alty. Table 1 The Competition Penalty Under Different Allocations of Backorder Costs Competition Penalty: Percentage Increase in Cost of the Nash Equilibrium Over the System Optimal Solution a Minimum 107% 5% 2% 1% 1% 1% 2% 107% 5% 2% 1% 0% 0% 0% 5th Percentile 117% Median 95th Percentile 5. = 0. 0.e. Several results are evident from the table. 1 % 4% 1* 1% . 1} p £ {1.. For each problem three solutions are evaluated: (1) thereby hampering the retailer's effort to mitigate the system optimal solution.3 0. a = 0) and 483% when the retailer incurs all of the backorder L.9 1 9% 4% 2% 1% 1% 15% 604% 41% 12% 8% 4% 8% 483% 804% Game 117% 8% 3% 1% 0% 0% 0% 37% ' -' • -'im : 26% "'[ • • jfifc•.e. When a = 1. 0.f t .3. of the Nash equilibrium over the system optimal solution.7. 0.2. 0. 16} fti e {0. First.1 Local Inventory 0.7 0..CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies parameters are chosen from the 2625 possible combinations of the following: The competition penalty increases as the backorder cost allocation becomes more asymmetric. and (3) the Nash equilibrium of the LI The LI game's response to changes in a is slightly game..5. 45. of the El game.1. (These data can be obtained from http:// different. 4. 2. 4.. the Nash equilibrium is close to the behavior is not always harmful. 16} ft2 = l . this (i.3 0. Note thatftJ +ft2= 1 in all of the problems. are allocated to the retailer.e. 8.

1. When ft2/(ft. and the retailer chooses s. the competition penalty is high when one of the firms has a substantial influence over a major portion of total system costs.. WalMart. Nevertheless.e. n In the LIS game the supplier anticipates that the retailer will choose r.:v. the supplier wishes to carry no inventory.According to the next theorem.. and then the other player chooses its base stock level. the first player cannot change its decision after observing the second player's. When a < 1. there exists st such that 45. When a < I. as large as possible. . so it chooses S2 G [0. leaving the supplier with some expected inventory.g. S]. (s. . .'). competition raises supply chain inventory by at most 4%. a player is committed to its choice. 0.. it would choose s. The supplier cannot choose S2 ^ s\.. One Dominant Player In the LI and El games the players choose their policies simultaneously. s°]. The Stackelberg version represents a situation where one player is the dominant member of the supply chain (e. In the Stackelberg version of either game one of the players chooses its base stock level first. announces its choice to the other player..) 9. S2 G [0. §2) Since this is continuous in §2. . the supplier's self serving behavior does little damage. so the supplier's decisions have a small impact. +ft2)is large. except in some cases when the retailer cares little about backorders (a is small). ^ S2.^:. ^ s\. .CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies Figure 3 Competition Penalty When the Retailer Incurs AJI Backorder .. there is little difference between the EIS and El games. i. As (10) increases. Supplier Stackelberg Games The Stackelberg game with the supplier leading is called either the Echelon Inventory Supplier game (EIS) or the Local Inventory Supplier game (LIS). .. the second player chooses an optimal response to the first player's strategy and the first player (correctly) anticipates this behavior.5 L . there is little benefit to holding inventory at the supplier rather than the retailer. We seek sub-game perfect equilibria. THEOREM 20. 9. July 1999 MANAGEMENT SCIENCE/VOI. because then the retailer chooses s.(s2).(s2) ^ s. = s°. No. so the supplier should anticipate s. Overall. • v .(s2) ('^. therefore the competitive solution's performance is nearly as good as the optimal solution's. Therefore.f.. Table 2 presents data on the percentage change in average supply chain inventory in the two equilibria relative to the optimal solution. . 7.5 1. As in the El and LI games.:. + L2 (10) What does (10) measure? When L^/{L^ + L2) is large.(S2)). ^.. sJ]} are the Stackelberg equilibria... Intel.^^. When a = 1. in the EIS game (s". . Average inventories in the competitive solutions are generally lower than in the optimal solution. = s°. In the EIS (LIS) game the supplier chooses S2 (S2) to minimize its cost. . Hence. depending on the inventory tracking method. the supplier's lead time is a small fraction of the total system lead time. the supplier's cost is PROOF. .. .. but little incentive to help manage that cost. G [S2. since then the retailer will choose s. r2(sj)) is the unique Stackelberg equilibrium. The proof of Theorem 13 shows that if the supplier could choose s. 140% Costs (a = 1) .e... : . . . the supplier chooses Sj = r2(s. When a = 1. r. given that it anticipates the retailer will choose r. i.

r'. PROOF. it is immaterial whether the firms use echelon or local inventory measurements. When a < 1. Differentiate the retailer's cost function: s.)r\{s. From Lemma 6. the retailer's cost for any base stock level is the same in the two games. ^urn MANAGEMENT SCIENCE/VOI. for §2 ^ ^(s..:-.e. 7.6 0.(s2) + S2 is decreasing in S2. Hence there exists a Stackelberg equilibrium.(s2) < 0.).9 1 0 0. PROOF.) when s.3 0..) > 0. i.).). and let {sf. Existence of an equilibrium is straightforward.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies Table 2 Change in Supply Chain Inventory Percentage Change in Average Supply Chain Inventory In the Nash Equilibrium Relative to the Optimal Solution Minimum 5th Percentile Median 95th Percentile ' '" Maximum 0 Echelon Inventory Game -18% -10% -12% 0. sl'l be the equivalent pair of echelon base stock levels. sf = sf and s^' = sf + s^ 21. ./ds. aH2/dSj s 0. July 1999 951 . THEOREM i.1 0. s^^ < f2{s[) = s^. G'2{y) < 0.s] ^2(82). differentiate the supplier's cost function with respect to §2 9. dH2/ds2 > 0.(s2).3 0.).7 0. When a < 1.5 0.. • tit f. When s^ s r2(s.7 0.••. Therefore. Since r. in the LIS game the supplier chooses a base stock level lower than in the LI game. Hence. No. Assuming s. Retailer Stackelberg Games In the Echelon Inventory Retailer (EIR) and Local Inventory Retailer (LIR) games the retailer anticipates the supplier will choose r2(Sj) and r2(s. Since r^isi) = -s.. = r. when the retailer is dominant.. respectively. Sj < §2 and S2 < Sj. . In the EIR and LIR games the retailer chooses a base stock level that is higher than in the El game (s{).2. = s.9 1 -8% -13% -22% -37% -73% -18% -10% -7% -S% -9% -15% -27% -56% -2% -3% -2% -3% -4% -7% -17% 2% 1% 0% -1% -1% -1% -3% 2% 1% 0% -1% -1% 0% 0% 4% 3% 0% 0% 0% -1% -2% 4% 3% 1% 0% 0% 0% 0% -12% Local Inventory Game -7% -10% -18% -34% -38% -7% -5% -7% -11% -15% -16% -2% -2% -2% -2% -2% -2% -1% H2(S2) = infs^eio. T H E O R E M 22. Let |sf. Hence.. but lower than in the LI game (s[)... + r2(s.e. sl'l be an equilibrium. The retailer anticipates that the supplier will choose r2(s. So {dH. 45.1 0.

When the players view consumer backorders as equally costly (i. there is a significant strategic difference. the above is negative for all s. 1995.. ' MANAGEMENT SCIENCE/VOI. These results highlight an important lesson for managers: While the lack of cooperation/coordination implies the system will not perform at its best efficiency. A framework for decentralized multi-echelon inventory control. Hence the retailer chooses a base stock level lower 10. at least one of them has a private incentive to deviate from the agreement. Working Paper. Lund. Econom. Camerer. R..CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies Since r'2(Si) > 0. Hence. Decentralized supply chains subject to information delays. Chen. 165-194. 1997. Nevertheless. the supply chain optimal solution is never a Nash equilibrium. the referees. the competition penalty can be huge.. July 1999 . competition generally lowers supply chain inventory relative to the optimal solution. C. is larger than s. and the associate editor are also graciously acknowledged. G. This is a surprising result. when the players have divergent backorder costs. 1996. 45.{s[. We draw this conclusion from a sample of 2625 problems. cooperation may lead to lower inventory. References Axsater. the University of Chicago. 'uS[ h) it holds that dH. Buzzell and Ortmeyer 1995.g. To appear in Management Sei. Kumar 1996). Ortmeyer. The supplier prefers local inventory.. so competitive selection of inventory policies decreases efficiency. we characterize a set of simple linear contracts which eliminate each player's incentive to deviate. Buzzell. However. 7. a = 0. and in the LI game equilibrium.5). Conclusion When both players care about consumer backorders. In other words. the University of Michigan. 1996. The helpful comments of Eric Anderson. = sJ. Sweden.e. G. No. since many authors suggest the opposite (e. while there is little operational distinction between tracking echelon inventory or local inventory (since we assume stationary demand). inventory remains a public good even here. we suspect there is always a strong tendency for comp)etitive firms to choose lower inventory than in the optimal solution. hence the optimal s. In other settings. The rationale is that inventory is a public good: Each firm benefits from more inventory. Although the players may agree to cooperate and choose supply chain optin\al policies. and the 1997 Multi-Echelon Inventory Conference at New York University. the median competition penalty in the El game is only 6% and in the LI game it is 3%. These contracts are based on actual inventories and backorders. Quart. and this might enable better policies than those available to competitive firms. the magnitude of the efficiency loss is context specific' ' The authors would like to thank the seminar participants at Rochester University.. S. cooperative firms could share sales information. if firms cooperate and choose the optimal solution. Furthermore. Sloan Management Rev. ]. Loss-avoidance and forward induction in experimental coordination games. ds. Marty Lariviere. the median competition penalty in the El game is 483%. It is well known that participants tend to underinvest in the provision of public goods (see Kreps 1990). Implementation of these contracts will not provide dramatic improvements when the players have similar preferences for reducing consumer backorders. when the supplier is indifferent to consumer backorders. F. and these equilibria differ. For instance. Chanel partnerships streamline distribution. but the retailer's preference depends on the parameters of the game. Eor each problem we measured the competition penalty. Since r'2 < 1. they will tend to increase inventory. In the games we study. the percentage increase in total cost of the Nash equilibrium over the optimal solution. there is a unique Nash equilibrium in either the El game or the LI game. Should the players wish to choose the optimal solution cooperatively. but each wants the other to invest in it. Cachon. Lund University. 36. For instance.

An approach for developing an optimal incentives. NC. A. 45 54-65. 1991. Working paper. 1980. Game Theory. Ross. R. Optimal pricing and return policies for perish32 818-836. Zheng. Decentralized multi-echelon supply chains: .. Seshadri. Are multi-echelon inventory Donohue. infinite horizon. Porteus. Columbia sity. 45. Forsythe. 1996. M. No.. able commodities. Management Sei. Nirmalya. S. coordination. V. Wharton School. Lower bounds for multi-echelon stochastic Li.. 6 475-490. M. Federgruen. J. Working paper. 26 483-494. 1997. 1994.-Dec. D. 1985. S. Cambridge.. McCardle. The power of trust in manufacturer-retailer ment Sei. 12 1524-1539. Game theoretic analysis of the substitutable product inventory control policies. multi-echelon inventory model. Computational issues in an 397-409. Management Sei. 1996. T. 1984. Management Sei. MA. D. Logist.CACHON AND ZIPKIN Competitive and Cooperative Inventory Policies Lee.. 1996. Duke University. Management Sei. Hausman. July 1999 . R. Decentralizing cross-functional ment. Stanford Univerdecentralized distribution systems. 1997. Narayanan. 1997..1 - MANAGEMENT SCIENCE/VOI. Kreps. The role of inventory in delivery time-competition. single-echelon Parlar. Naval Res. 1997. 1984. Durham. Yale University. New York University. Whang. A. W. Coordination mechanisms for Incentives and information. Policy mechanisms for supply chain the battle of the sexes games: Some experimental evidence. Accepte_d hy Luk Van Wassenhove. 1994. ManageKumar. Oper.. R?s. relationships. feedback in a dynamic decision making experiment.. Thomas. On manufacturing/marketing incentives. R. CA. Working paper. Working paper. Zheng. Harvard University. Muckstadt. PA. Fudenberg. D. 1960. K. Raman. Dejong. Nov. Philadelphia. Pastemack. 38 182-197.. New York. Zipkin. Working . 37 1166-1181. 1990. 35 321-339. decisions: Coordination through internal markets. Oper. . 1992. Erkip. 35 Federgruen. Princeton UniverSterman. 7. 40 1426-1443. A. N. CT. E. L. 1996. 1988. This paper has been zvith the authors 9 months for 2 revisions. Working paper. Tsay. Staelin. Management Sei. 1996. Modeling managerial behavior: Misperceptions of sity Press. University. Working paper. Supply contracts for fashion goods: Optimizing methods worth implementing in systems with low-demand charmel profits. 1989. Optimal policies for a multi-echelon Lippman. R. The competitive newsboy. 20 568-587. discount pricing policy. H. A Course in Microeconomic Theory. K. Contracting for inventory in a Ha. 1996. 1996. Clark. A. Cooper. 40 597-602. CA. rates? Management Sei. inventory problem. Cambridge. Supply contracts for a short-life-cycle product with distribution channel with stochastic demand and substitute demand uncertainty and asymmetric cost information. 1991. 1989. P. Whang. H. M. Rand J. Y. J. Princeton. L. New York. S. Scarf. paper. A. Workproducts. Multi-echelon vs. Stanford. A.. inventory problem with random demand. J. Management Sei. Responsibility tokens and supply chain manageKouvelis. Res. received March 7. To appear in Management Sei. 1996. Y. MIT Press. NJ. Rev.. ing paper. Stanford. Tirole. .. B. Harvard Bus. J. Lariviere. Stanford University. 4 166-176. Econom. J. inventory systems. Communication in Moses. P. The quantity flexibility contract and supplier-customer Lal. Marketing Sei..

Sign up to vote on this title
UsefulNot useful