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

102, No. 2 (May, 1987), pp. 375-394 Published by: Oxford University Press Stable URL: http://www.jstor.org/stable/1885068 . Accessed: 13/12/2013 16:54

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

.

Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to The Quarterly Journal of Economics.

http://www.jstor.org

This content downloaded from 181.177.239.29 on Fri, 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions

**MARKETS WITH CONSUMER SWITCHING COSTS*
**

PAUL KLEMPERER Ex ante homogeneousproductsmay, after the purchaseof one of them, be ex post differentiatedby switchingcosts includinglearningcosts, transactioncosts, or "artificial"costs imposed by firms, such as repeat-purchasediscounts. The noncooperative equilibriumin an oligopolywith switching costs may be the same as the collusive outcome in an otherwise identical market without switching costs. However,the prospectof futurecollusiveprofitsleads to vigorouscompetitionfor market share in the early stages of a market'sdevelopment.The model thus explains the emphasisplaced on marketshare as a goal of corporatestrategy. I. INTRODUCTION

In many markets consumersface substantial costs of switching between brands of products that are ex ante undifferentiated. There are at least three types of switching costs: transaction costs, learningcosts, and artificialor contractualcosts. There may be transaction costs in switching between completely identical brands. Two banks may offer identical checking accounts, but there are high transaction costs in closing an account with one bank and opening another with a competitor. Similarly, it is costly to change one's long distance telephone service,or to return rented equipment to one firm and rent identical equipment from an alternative supplier. The learningrequiredto use one brandmay not be transferable to other brands of the same product, even though all brands are functionally identical. A number of computer manufacturers,for example, may make machines that are functionally identical, but if a consumer has learned to use one firm's product line and has invested in the appropriatesoftware, he has a strong incentive to continue both to buy machines from the same firm and to buy software compatible with them. Similarly, when choosing a cake mix, it is easiest for a consumer to buy the brand that he already

*This paperis a shortenedversionof a chapterof my Ph.D. thesis. Researchfor this paper was begun at Stanford University in the summer of 1983, and was supportedthere by the Stanford GraduateSchool of Business and the Center for EconomicPolicy Research.I am gratefulto Doug Bernheim,Adam Brandenburger, Tim Bresnahan,Jeremy Bulow, David Kreps, Meg Meyer, James Mirrlees,Barry Nalebuff,John Roberts,LarrySummers,and the refereesfor help and advice.

? 1987 by the President and Fellows of HarvardCollege and the MassachusettsInstitute of Technology. The Quarterly Journal of Economics,May 1987

This content downloaded from 181.177.239.29 on Fri, 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions

after the purchase of one of them. Corporateplanning models often rest heavily on the assumption that the future demandfor a productwill be positively correlatedwith its presentsales.even if he knows that all brands are of identical quality. this is exactly equivalent to paying s for a discount coupon of value s.There is.the consumerfaces a loss of utility fromswitchingfrom a brandthat he has tried and liked to an untested rival. S & H Green Stamps.376 Y JOURNALOFECONOMICS QUARTERL knows exactly how to prepare. and 1. 2.as when. etc. consumers who switch between different companies are penalized relative to those who remain with a single firm. Spence [1981] and Clarkeet al.) In these examples. for example).239. (If a customer signs a contractcommittinghimself to either buy from a firm or pay damages s.' These two types of switching costs reflect real social costs of switching between brands.Airlinesenroll passengers programsthat rewardthem for repeated travel in "frequent-flyer" on the same carrier. even though their size can be influenced by firms (by their product-design choices.and many groceryproducts are sold with a discount coupon valid for the next purchase of the same item. ex post heterogeneous. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . Similar switching costs can be created by contracts. This content downloaded from 181. some retailers offer trading stamps (Greenshield stamps. fromsimple for no obviouseconomically rationalreason. There may also be psychologicalcosts of switchingbetweenbrands. arises entirely at firms' discretion.consumersrepeat-purchase habit or loyalty.) that customers can exchange for prizes after sufficiently many have been accumulated from repeated purchases.177. In all these markets rational consumers display brand loyalty when faced with a choice between functionally identical products.2 Products that are ex ante homogeneousbecome. an importantdifferencebetween a model of switchingcosts and models of learningon learningraises the demand side in which advertisingor other nonconsumer-specific all consumers'willingnessesto pay for a particularbrand (see. This correlationof demands across time is an immediate implication of a model with switchingcosts. in expectation. 3. If a consumermust purchasea brandto discoverwhetheror not it is suitable for him (that is. each brand is an experience good in the terminology of Nelson [1970]). A third type of switching cost. artificial or contractual switching costs. Ourmodel can apply to marketsin whichconsumers'preferencesare shaped by knowledgeacquiredfromexposureto other consumers'purchases. [1982]):with switchingcosts the distinction between those consumerswho have and who have not previouslybought or been otherwiseexposed to a particularbrandis critical. however. and due to the existence of groupsof consumerswho tried a brand and did not like it.then.3 In this paper we take switching costs as exogenous.29 on Fri.as brand-specific well as to markets in which consumers repeat-purchase. and is distinguished by the absence of social costs of brandswitching. however.and so the concept of a customerbase is important.there are additionalcomplicationsdue to the possibility of prices being used as signals of quality. for example. In this case.

Hence the assumption that firms charge the same price in every period leads. and also assumesthat consumers'tastes for the underlying product characteristicsmay change. switching costs make each individual firm's demand more inelastic and so reduce rivalry. to markets that are more competitive with switching costs than without switchingcosts. however. Switching costs segment the market into submarkets. Von Weizsdcker[1984] has built a simple model that incorporates switching costs. and so abstracts from the issues we shall consider.Prices in the second period are always higher both than in the first period and than if there were no switchingcosts. in contrast to ours. In his model. in this model. Since with switching costs a consumer'schoice today is also influenced by the future when his tastes may be different. but he assumes that firms commit to charging the same price in every period (including the introductoryperiod).239.Each submarket contains consumerswho have previously bought from a particularfirm and may in effect be monopolized by that firm. The resulting (noncooperative) equilibrium may be the same as the collusive solution in an otherwise identical market with no switching costs. switching costs do not necessarily make firms better off.and obtains results closerto those of our paper.yields no benefits to set against the cost of restricted output. argue that both the assumption that firms can credibly commit to charging the same 4. This content downloaded from 181. In a standard differen- tiated products model the social cost of firms' increased monopoly power is mitigated by the benefit of increased consumer choice. Differentiating functionally identical products through switching costs. We thus provide an explanation for the importance that many companies attach to building market share and for the emphasis that is commonlyplaced on market share as a measureof corporate success. the central result is that switching costs make markets more competitive. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . However. Klemperer[1987b] considers a model based on von Weizsicker'sbut without the assumptionthat firmsmust choose the same price in everyperiod.177. however.29 on Fri. Von Weizsicker considers products that are differentiatedfunctionally as well as differentiatedby switchingcosts. Second.4 We. The ferociouscompetition to attract new customersso as to be able to fleece them (and other consumerswhom they teach or influence) after they have "bought in" to a particular brand may more than dissipate firms' extra monopolistic returns and leave them worse off than in a standard oligopoly. We make two main points.MARKETS WITHSWITCHINGCOSTS 377 examine their implications for the competitiveness of markets.today'stastes becomeless importantrelativeto any pricedifferencethat is expected to last. First. the monopoly power that firms gain over their respective market segments leads to vigorouscompetition for market share before consumershave attached themselves to suppliers.

We conclude in Section V. Note that a monopolist (or collusive oligopoly) would choose quantity. In labor markets there may be switching costs of all three kinds. p = 55. and rules that employees' pensions are vested only after a certain number of years are "artificial"costs of switching. and the assumption that they will necessarily want to. read workersselling labor for low (high) wages. flying a route in each of two periods. Summers [1985]. Job-specific training leads to "learning"costs.) that would allow agents to make credible commitments about the future. and price.5 Section II uses an exampleto illustrate the point that switching costs can lead to a noncooperativeequilibriumthat is the same as the collusive solution. Industry demand is q = 100 .which assumptionsimply trainingor movingcosts withoutgivingcurrent that a firmcannotpay a newworker's employees a bonus of equal financial value.6Then Section IV considers the first period.Thus. and shows how the second-period monopoly profits lead to competition for market share in the first period. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . q = 1/2(100 .10) = 900 for each firm. including related models in which a consumer is reluctantto buy an untestedbrandbecauseof uncertaintyaboutproductquality (see note 1). but in Cournot competition (imagine each airline choosinga numberof flights to fill in each period) each firm chooses a quantity of 30. This period is formallyquite similar to a single-periodmodel of functional productdifferentiationin which heterogeneityin consumers'tastes (as in standard models a la Hotelling [1929]) or firm-specific elements in consumers' utility functions (see.239. This content downloaded from 181. each discount of 10 to firm announcesthat it will offer a "frequent-flyer" anyone who flew on its flight duringperiod 1. yielding an industry profit of 2025.de Palma et al. This is the second period of a two-period model. are unjustifiedin many markets. Are these discounts a sign of ferocioussecond-periodcompetition? 5. Consider two airlines.378 JOURNALOF ECONOMICS QUARTERLY price in every period. and Wernerfelt[1985] have independently developed models of switchingcosts that contain some of the results of our paper. for example.29 on Fri. For discussion of other models. Now imagine that after Cournotcompetition in period 1. physical relocation costs. We assume that consumers (workers) bear the switchingcosts and that firms cannot price discriminate. We also ignore uncertainty and risk to productmarkets. etc. II. NUMERICALEXAMPLE A simple example introduces the main idea.) However. and profits are (30)(40 .p. A and B. aversion. before consumers have attached themselves to suppliers. the market price p = 40. and each firm'sconstant marginalcost is 10. Farrell [1985]. (For consumerspurchasingproducts at high (low) prices. see Klemperer[1986]. Section III analyzes the collusive nature of equilibrium in a mature market in which switching costs have already been built up. and other hiringand-firing costs are "transaction"costs. [1985])gives firmssome local monopoly powerover consumers. reputation.177. Our model can be interpreted as representing an oligopsonistic labor market. the model is more appropriate 6.we are assuming away the possibility of mechanisms (long-term contracts.10) = 45.

B P2 65 and P2 = (100 . but a discount of at least 45 is then required.TWA has offereddiscountsto AT&T long distancephoneserviceusers. Neither firm can benefit by deviating from the monopolyprice (55 + s) to its old customers. the maximumprice differencebetween the firms is 10 (otherwisethe higher price firm marketclearingrequiresP2 = P2 + 10.10) = 10121/2 for each firm. and so each firm sells a quantity of 221/2. and if q2 ? will sell nothing).3 is sufficientfor the period-2equilibriumto be qA = q2 =221/2. let q2 = 22'/2. see note 19. market clearingrequires that pB = pA .See Klemperer[1984]for detailed computations.and manyfirmsmail discountcouponsto potential purchasersof their products. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . 7I A B we have E [171/2. This content downloaded from 181. the same result can be obtained.MARKETS WITHSWITCHINGCOSTS 379 On the contrary:in the second-period noncooperativequantiso the ty-setting equilibrium.if each firm must first announcewhether or not it will honor its rival's discount couponsbefore each firm chooses its quantity or price. precommitmentto not honoringthe rival's couponsis each firm'soptimal (in fact dominant)strategyand so arises naturally.and acts as a monopolist against them. each firm finds that it is too expensive to increase output far enough to take any consumersthat it did not serve in the previous period.29 on Fri.So each firm serves only its old customers. the market price p = 65.Any discount in excess of 9. For s . The discounts have implications for the first-period equilibrium also. and profits are (221/2)(65. We do not allowa firmthe strategyof honoringits rival'sdiscountcoupons(that is. Not only do the "discounts" allow the airlines to achieve the collusive level of output.it may be possibleto 8. It is then easy to solve for the 171/2.177. attracting the rival's customersrequirespricing at least s below the rival. straightforwardto check. discountcoupons.each firm chooses a quantity of 221/2. and so gives up at least as much profit on these customers as it can gain by stealing its rival'scustomers. For example. total quantity q = 45.With a discount of s 2 45 each firm sets a price of 55 + s in the noncooperative equilibrium. hence the effective net-of-discount price to "old" (previous-period)customers This equilibriumis is 55.10.10 . However.8 However. Of course this is only a partial analysis. see the Interpretationsubsection of Section III. and to check maximizes A's profits. For further intuition. However. The optimality of qB given qA follows by that qA = 221/2 symmetry. offeringdiscountsto consumerswho flewwith the rival in the first period). (unique) price p2 such that the total purchasedis q2 + q2 (noting that consumers with reservationprices below 40 can get discounts from neither firm).239.so if q4 2 271/2.unless it sells to any of its rival's customers. Whenswitchingcosts are "artificial.2q2) + 10 (since the discount has value 10). Given the discounts. but the customers do not benefit in any real way from them-the quoted prices simply end up higher than the collusive price by the amount of the discounts. the ~ 27/2].7 If firms engage in price competition rather than quantity competition.45 this cuts its effective price to old customersto its marginalcost or lower.A tacit agreementnot to do this may be easy to monitorand to enforce(a best responseto a rival honoringyour discountsis to honor its). check the eriod-2 equilibrium. Then if q2 C To 7." avoid any effects on the first-periodequilibriumby distributingthem in other ways than throughfirst-periodsales. Furthermore.

and not as a complete explanationfor "frequent-flyer" part of their explanation is probablythat they exploit a principal-agentproblem within the purchasingfirms. taking account of how first-period incentives are affected by the dependence of secondperiod profits on first-periodsales. (Any "start-up" transaction or learningcosts are assumed to be the same whichever firm a consumer buys from. Our numericalexample is intended as an illustration of our more general programs. f (q) would be inverse demand if there were no switching costs. second-period switching costs are created by first-period sales.9 III. However. A and B.g. It finds the symmetric equilibrium in a generalmodel of consumerswith different switching costs. while the complementaryfraction aB(= 1 _ UA) must pay a switching cost to purchase A's. In the first period consumers have no ties to any particular firm. We consider two firms. These fractions are most naturallythe firms'respective shares of the previous period's sales. as for computers). There is also the advantagethat the employees are receivinga tax-freebenefit. A fraction of the market CA.and it is typically not worth companies'paying the monitoring costs to avoid this.A large analysis. lead to the collusive solution even with noncooperative behavior-is the focus of our paper. and emphasizes that noncooperative behavior in the presence of switching costs leads to outcomes that look collusive. Thus. We assume that q consumers have reservationprices greaterthan or equal to f (q) for the product that they previously bought or were otherwise exposed to (e.) However.177.29 on Fri. that is. the "mature market" after consumers' switching costs have been built up. In Section IV we shall return to analyze the initial period.239.380 QUARTERLY JOURNALOFECONOMICS basic point-that repeat-purchase discounts or. This content downloaded from 181. by observationof others' purchasesor through specific training."must pay a switching cost to buy B's product. consumers face a "switching cost" of purchasingthe productthat they were less exposed to.. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . consumers' costs of switching between competing brands. but employeesget the free vacations. switching costs and firms' previous-periodmarket shares are given. In period 2. and this will be assumed when we examine the full two-period 9. THE SECOND PERIOD: BEHAVIORLOOKS COLLUSIVE NONCOOPERATIVE The next two sections look at the development over two periods of a market with switching costs. producing functionally identical products. more generally.Companiespay higherticket prices. This section analyzes the second period. "A's consumers.

r (o) = 0. as is the case when switchingcosts are real learningor transactioncosts.177. that is. (We discuss below the case in which r (o) > 0. pB) and reservationprice less switching cost greater than or equal to pA (the third term). pA Without loss of generality. to those of B's consumerswith reservationprices greaterthan or equal to pB and switching costs less than or equal to pB _ pA (the second term). This content downloaded from 181.and no consumerspayingthe switchingcost to switch betweenfirms. switchingcosts. Here we are assuming that all a firm'scustomerspay it the same price.). This model differs slightly from that of Section II.239.as is the general case when switchingcosts are artificial. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . at least in the cases we consider. but for a given size of switching costs a market with artificialswitching costs will be somewhatthe more competitivesince new customersare relativelymorevaluableand so the incentiveto cut price or increasequantity is greater. Most naturally.MARKETS WITHSWITCHINGCOSTS 381 model in Section IV. but they could also depend on other factors. Firm A. With real switchingcosts of s the fully collusiveoutcomeinvolvespricesequalto the monopoly price. the quoted price is higherthan the effective (net-of-discount) price by an amount equal to the switching cost. there is an atom of consumers without switching costs.see notes 14 and 18. With artificialswitchingcosts of s the fully collusive outcome involves prices s above the monopoly price and all purchasing consumers receiving discount s. customersstolen from a competitor pay a higher price than the net-of-discount price paid by the firm'sold customers. Market equilibrium Ah (pA) + (TBr (pB - pA)h(pB) + JB fpB r(r - pA)[-dh(r)] and (lb) qB jB (1 - r(pB _ PA))h(pB)- Firm B sells only to its own consumers with reservation prices greaterthan or equal to pB and switchingcosts greaterthan or equal to pB _ pA. Thus. and also to B's consumerswith reservationprices in the range (pA. y(w) = ar (w)/aw : 0 is the density function of the switching costs. let requiresthat (la) qA = '.10 10. (We assume that each firm has the same proportion of consumers with each reservationprice. on the other hand.) We let h(*) = f1 (.) The switching costs need not be the same for each consumer. which would arise naturally.29 on Fri.but not real. We let r(w) be the proportionof a firm's consumers whose cost of switching to the other firm's product is less than or equal to w. In the exampleof Section II. Anotherdistinction is that with artificial. sells to all its own consumers with reservationprices greaterthan or equal to PA (the first term of (la)).This distinction is of no importanceto the qualitative results. Klemperer[1984] generalizesthe model to allow for different distributions of switching costs for different firms and at different reservation prices.

depend on the economics(see Singh and Vives [1984]and Klempererand Meyer [1986]). This is cost-minimizing if cA" = cB" 0.177. If y (0) = 0.239. This is just the first-order condition for a monopolist (or collusive oligopoly) in a market without switching costs. pA with CA = =-B 1/2 (3) 2 [h(p) + (p (h'(p) - = y(0)h(p))] O. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . (We assume that the monopolymust operate both firms'plants equally. and where we have assumed that cA(.) cB (. Substituting from (la) yields (2) 0 = oAh(p A) + qB]U(pB pA )h(p [p B) + -qAJ[j B fPB r=pA (r pA) dh(r)] + h (p A) aB (pB _ppA)h (pB) ?OB + PB _-y (rpnA) [-dh(r)y] = AB So at a symmetric equilibrium (in pure strategies).as it is easier to derive.382 QUARTERLY JOURNALOF ECONOMICS Noncooperative Equilibrium We can use (la) and (lb) to solve for either a price-setting or a quantity-setting equilibrium. This content downloaded from 181. where q = 2qA = h(p).) It is easy to use (2) to check that this result-that the first-ordercondition for a symmetric pure-strategy equilibrium is the same as that for a monopolist in a market without switching 11.). and an oligopoly might anyway be constrained to do this by an inability to make sidepayments. of course.' We begin with the price-competition equilibrium.but in the absence of any fully satisfactorytheory to choose between them we consider both kinds of competition. The choice between the two equilibria should. then we can rewrite (3) as (4) h (p) + (p - I(CAt() + CB'(q))) h (p) =0.29 on Fri. Firm A's first-ordercondition is Or apA q A + HA acA A aq aqA [P dpA =? where 7rA are A's profits and cA is A's total cost.

. where R = maxF=A. on the other hand. More generally. Thus. marginalcosts. but not necessarilyequal marketshares. where qmis the monopoly 12.14 (We shall see that with quantity competition it is much easier for the joint-profit-maximizing outcome to be a noncooperative equilibrium.R).ooin the symmetric equilibrium. With y(0) between these extreme cases the equilibrium is between the competitive and collusive equilibria.MARKETS WITHSWITCHINGCOSTS 383 costs if y (0) = 0-also holds when firms have unequal market shares (aA / GB). the rest of the distribution is crucial in determining whether the prices satisfying the first-orderconditions are global best responses for the firms. which more fully exploits its own customers without losing any to its competitor. each firm has an incentive to slightly increase its price. S > (( /5 1)/4)(a -C) is sufficient for the joint-profit-maximizingoutcome to be a noncooperativeequilibrium. marginalcosts. y (0)-these are the marginal consumers who are sensitive to a small deviation in one firm's price from its competitor's price.0. This assumesa monopolist'sprofitfunctionwouldbe quasi concaveif there were no switching costs. see note 10. equal. and its profits are irA = oA4rm. the market price approaches the competitive price (firms' marginal cost) as we approachthe case of no switching costs. Thus. S > ((a . 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions .177. equal. However.C)/2) is sufficient.. and B for CA 1/2.)For s large enough that pA B=p = pm in equilibrium. f (q) = a ."3 At any lower common price. 13.(acA/aqA)) -.fq.B ((1 - oF)/V) is a measure of the relative market shares of the firms. Other consumers'switching costs are also more important in asymmetric equilibria.12 Consider the important special case in which all consumers have a switching cost of at least s > 0 and firms have constant. if they have constant.that is.239.) We can check that for linear demand and costs. its market share of the monopoly output. firm A's sales are Aq. When switching costs are artificial. 14. This content downloaded from 181.. s 2 (a . cA(q) = cB(q) = Cq.C)/4)( R2 + 4R . As y (0) . a price that is locally optimal for a monopolistcould be an equilibrium.C)/2 is sufficient for the outcometo be a noncooperativeequilibriumif aA = a joint-profit-maximizing 1/2. That is. (3) implies that (p . the first-order conditions do in fact define an equilibrium-neither firm has an incentive to make a large deviation-if s > ((a . for all o-Aand oB.Here 'y(0) = 0.29 on Fri. in a symmetricpure-strategyequilibriumthe only information about the distribution of switching costs that matters is the density of consumerswith zero switching costs. so the only possible symmetric pure-strategyequilibrium is for each firm to choose the monopoly price Pm for an otherwise identical market without switching costs.

if there is an atom of consumerswithout switching costs (for example. the only possible symmetric pure-strategy equilibrium has price equal to marginal cost. With smaller switching costs mixed-strategyequilibriaarise but these are difficultto calculate. equal linear costs. (That is. (At any greater common price.fq CAc(q) = CB(q) = Cq. = 1/2. k]. equal market shares. In that model pure-strategyequilibriain pricesare obtainedeven when a fractionof the second-periodconsumersare "new"(uncommitted)consumers.15 At the other extreme.16 To obtain pure-strategy equilibria with prices between the competitive and monopoly prices. 17.177. y(w) = 1/k for w < k f(q) = a . Klemperer[1987b]avoids this problemby consideringa model in which productsare functionallydifferentiatedas well as differentiatedby switchingcosts. for switchingcosts less than those supportingthe joint-profit-maximizing outcome. So qA = 7Ah(pA) + UB h(pB) A + UB (h(pA) . 16.and we confirm the intuition that the second-periodprice is decreasingin the proportionof new consumers. oo).17 The equilibriumis a- 15. Shilony [1977]in anothercontext solves for the mixed-strategyequilibriaof a price-competitionmodel that is mathematicallyequivalent to the special case of ours in which all consumershave the same reservationprice as well as the same switching cost and all firms have the same market share and constant identical marginalcosts. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . each firm has an incentive to slightly lower its price and capture the entire atom of consumers without switching costs. a proportion of consumers is new in the second period and uncommitted to either firm).h(pB)) XA and qB = B h(pB) _ Bh(pB) A. To confirmthat mixed-strategyequilibriado in generalexist. the expected market price and firms'expectedprofitsincreasecontinuouslyand monotonicallyas the switching costs increase. an equilibrium with constant marginal costs if any consumers have positive switching costs-so mixed-strategyequilibriaarise in this case. To see that this is true for any demand (and also that if rA + JB the firm chooses the higher price in any asymmetricequilibrium).384 QUARTERLY JOURNALOFECONOMICS output and rm the monopoly profits in the otherwise identical market without switching costs. The technical problemhere is that there is a discontinuity in firms' profit functions. and switching costs uniformly distributed on the interval [0. we consider the intermediate case in which there is a positive density of consumerswith zero switching costs but no atom at zero switching costs. His solutionconfirmsthe intuition that. use Theorem5 of Dasguptaand Maskin [1986].fromthe competitiveequilibriumwhen switchingcosts are zero up to the collusiveoutcome. therefore. for example.239. Consider linear demand. This content downloaded from 181. No asymmetricpure-strategyequilibriaexist with aA = a' = 1/2 and equal linear costs.) Then the second-order conditions are globally satisfied and a unique equilibriumin pure strategies exists and is describedby (3) for all k G [0.) Clearly this is not an equilibrium in general-it is not. CA = and y (w) = 0 for w > k.29 on Fri. higher-market-share assumethat pA < pB in equilibriumand write r (pB _ pA) = A.

Since we found earlierthat (PA . tive equilibrium. For B not to preferchargingPA we requirethat (pB - C)qB> (pA - C) UBh(pA) (pB - C)h(pB) 2 (pA - C)(h(pA)) + (pB _ C)Ah(pB).239. the market price and industry profits increase monotonically from the competitive equilibrium.even whenfirmshave costs.C).29 on Fri. linear demands and costs.x)h(pB))) (UB/OrA) ((pA (pB - C)h(pB) (aB/ A)(pA - C)h(pA) > (pB C)h(pB). As k increases. it must be that aB < orA.and is presented in the Appendix. for the joint-profit-maximizing see note 10. and equal market shares.x)h(pB))) 2 + (.7 percent of the monopoly outcometo be a noncooperaprofitmarkupper unit.h(pB))i] E (0. This content downloaded from 181. For A not to preferchargingpB we requirethat (pA - C)qA 2 (pB - C)UAh(pB) ((PA (PA - C)h(pA) + (1 . per unit). if lower-reservation-price equal previous-period ing costs. as k oo. p = C at k = 0. 18.TB/vA) C)A(xh(pA) (pB - C)h(pB).C) (that is.the equilibriummoves from the joint-profit-maximizing quantity-setting the to is. 17. As the density of consumerswith zero switching costs increases from 0 to equilib0o. that is. 20. which condition seems quite plausible.Scotchmer[1986]showsthat there are in generalno symmetricpure-strategyequilibriain this model if there are morethan two firms.177. F) Pm.C) (xh(pA) + (1 . to the fully collusive outcome.2 percent of firms' monopoly profit markup.C)h(pB). the joint-profit-maximizingoutcome is a noncooperative equilibrium if all consumers have a switching cost of at least (3/2 (a .C. 1). (that equilibrium Cournot rium to the equilibrium in a market without switching costs). The analysis of the noncooperative quantity-setting equilibrium is closely analogous. Togetherthese inequalitiesrequirethat .18 Interpretation The intuition for these results is that firms' demands are less elastic than if there were no switchingcosts-if either firmincreases where x= [IP (r _ PA) [-dh(r)I/(h(pA) .see Klemperer[1986]. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . but there may exist asymmetricpure-strategy consumershave lowerswitchsales. When switchingcosts are artificial.1)/4) (a .C)h(pA) < (pB .MARKETS WITHSWITCHINGCOSTS 385 (5) pA = PB k +( ) k ( 2C)}. p = (a + C)/2. The analysis is robust to the addition of an atom of consumerswithout switching equilibria. With quantity competition.it is sufficientthat all consumershave a switchingcost of at least (( 1 .

However. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . The higher are switching costs. the difficulty with colluding on output levels is that each firm has incentive to increase its quantity slightly (see. that is yy(0)= 0. A firmmust drive down the price at which it sells to its own consumersby s before it takes any consumers away from its competitors.386 QUARTERLY JOURNALOF ECONOMICS output. we have shown that switching costs mean that competitive behaviormay look collusive.So with switchingcosts a firmhas no incentive to increase its output only slightly. Each firm therefore chose an output exactly one-half of the originalmonopoly output. Consider for simplicity the special case in which all consumers have a switching cost of at least s > 0. Collusive Behavior So far. Absent switching costs. its price falls faster than its opponent's. where p' was the price net of the discount. even when they are not large enough for the collusive outcome to be a noncooperativeequilibrium. In the general model if there are some consumerswith switching costs arbitrarily close to zero. The discounts divided the customers into two identical groups:one that had bought from A in period 1. Stigler [1964] and Green and Porter [1984]). This is precisely what happened in the example of Section II.p'). When the density of consumerswithout switching costs is zero.so that only large changes in output can help a firm.239. Recall that the collusive output is always a local optimum for each firm.this incentive is reducedrelative to the case This content downloaded from 181. the fewer consumers are attracted by a price cut. then locally each firm's demand is as if it were a monopolist in its share (fraction a) of the total market demand.177. and anotherthat had bought fromB in period 1. and with imperfect monitoring. Switching costs reduce or remove this incentive.and thereforefaster than without switching costs (in which case both prices would fall together).29 on Fri.they may also facilitate collusive behavior. each firm locally faced a demand q = 1/2 (100 . then firms still have some incentive to chisel on the collusive agreementby slightly increasing their outputs. for example. even if 'the switching cost is not large enough for it to be a global optimum. However. hence the smaller the incentive to cut price (or increase output). Each firm therefore acts as a monopolist in its share of the market (provided that the first-order conditions define an equilibrium). so p = p' + 10 was the quoted price. Provided that the firms' strategies were not too different. and the closer the equilibrium price is to the collusive price. Such large changes are far easier to monitor and so (tacitly or otherwise) agree to eschew.

It is convenient to assume that a highervalue of vArepresents more aggressiveplay (so for quantity competition we write VA = qA and for price competition we write vA = 1/pA) so that ao-A/aVA > 0. and the forces for collusion are correspondingly strengthened. firm A chooses its first-periodstrategic variablevj to maximizeits total discounted future profits (6) 7AA i(VA.MARKETS WITHSWITCHINGCOSTS 387 of no switching costs. THE FIRST PERIOD: COMPETITION FOR MARKET SHARE The previous section described the second period of a twoperiod market in which second-period switching costs are created by first-period sales.whereaswithout switchingcosts it is necessaryto collude on output levels that may be very hard to observe. and are discounted by a factor X in first-period terms." IV. This content downloaded from 181. that a Therefore. With switchingcosts it is sufficientto agree(tacitly or otherwise)not to price discriminate in favor of competitors'consumers(or honor competitors'discount coupons)which agreementmay be easy to monitorand enforce. Thus. 0 drA dA= aXA A A a-A aA aVA > 0. In period 1.29 on Fri. provided that a7rA/aoA higher market share makes the firm better off in the second period. but by collectivelyacting in this way. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . chooses its first-period strategic variable at a level higher than that which maximizes first-period profits given the opponent's behavior: with switching costs firms compete more aggressivelyin the first period than they otherwise 19. We now consider the first period.in Section II. and so providingnatural"focal-points" for tacit collusivedivision of the market. in which consumersare not attached to any particularfirm. Here iA are the firm's first-period profits. firms will compete more aggressivelyin the first period. the role of switching costs can be thought of as making collusion easier. air/3vA < 0. even in our main model and numericalexample.see note 7. and firm B also. Thus.Each firm typically has an incentive to do this. we did not allowfirmsthe formallysimilarstrategyof honoringtheir competitors' discountcoupons. In noncooperativeequilibrium. VB) + XX(jA(v14 Vt)) taking B's first-periodstrategic variablevB as given. that is.239. With switching costs in the second period.we assumed that firms are unableto price discriminatebetweentheir old customersand consumersthat prefer their competitor'sproduct. In our analysis of the noncooperativeequilibrium. because increased sales increase market share and so increase second-periodprofits. firmswould reducetheir profits. and irA are the firm's second-period profits which can be written as a function of the firm's first-period market share orA. Switchingcosts may also facilitate collusion by breakingup a market into well-defined submarketsof groups of customerswho bought from different firms.177.Similarly. firm A.

239. (See Klemperer [1987b]. K. (See Farrell [1985]. and subsequently impose high bank charges to "milk" their customers after they graduate (the "second period").Two years later the averagerate paid was one-half percent below that of money market funds (see the Wall Street Journal. Two caveats should be noted to the result that firms will compete more aggressively in the first period than if there were no future switching costs. This content downloaded from 181. U. An examplefromthe U. each firm balances the firstperiod profit it gives up by more aggressive play with its secondperiod gains at the margin (2a/OVA = -X ( Ia&/aVA)). On average.) If either caveat applies. in order to gain market share that will be valuable to them in the second. educational institutions. for example. the result is that the firms dissipate some of their second-period gains. book-tokens. banks give college students money.) Second. switching costs are unambiguously beneficial to firms. if higher-market-share firms charge higher prices in the second period. so switching costs can either help or hurt firms overall.177. of course. it is possible for a higher market share to hurt a firm (Oir-/OA/ < 0) by making its competitor more aggressive. Summers [1985].29 on Fri. Thus. to avoid gaining market share. 1984). The total first-period profit given up can be in any relationship to the total second-period gains from switching costs. First. firms might compete less aggressively in the first period than without switching costs. rational consumers recognize that a lower price in the first period increases a firm's market share and so foretells a higher price in the second period. Specifically. for example. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . after industry deregulation. S. and Klemperer [1987a]. In this case consumers are less tempted by a price cut.388 JOURNALOFECONOMICS QUARTERLY would. Firms still compete more aggressively in the first period than they would if they were maximizing short-run (first-period) profits given first-period demand. and free banking services to induce them to open accounts (the "first period"). As the oligopoly would anyway produce more than the joint-profit-maximizing output in the first period. November21. firms end up with no more market share as a result of this fiercer competition.20Firms compete most aggressively for consumers who influence many others. Thus. but they may compete less aggressively in the first period than in the otherwise identical market without second-period switching costs. This motive explains the fierce competition for the university and high school markets that we observe in the computer industry. so first-period demand is less elastic than in an otherwise identical market without switching costs in the future period.with a promotionalfrenzyof high interest rates (morethan 10 percent abovethe rates of moneymarketfunds) and cash bonusesfor openingaccounts. 20. bankingindustryis providedby the introduction of "NOW" money market checking accounts in December 1982.

on the other hand. If demand is sufficiently concave below the duopoly equilibrium point when there are no switching costs.since marginal first-periodconsumersget no utility frombuyingin the second periodso first-period demandis unalteredaroundits equilibriumprice. equal. firms with constant. since a switching costs. 22. then switchingcosts will generallyimply higherfirst-periodstart-up highersecond-period costs and so lower first-perioddemand.177. It is easy to alter the demand curve slightly either so that firms prefer large switching costs or so that firms prefer no switching costs.This has no effect on the equilibrium. firms prefer switching costs. but the details of the comparisonare slightly different. We assume that consumers'reservationprices are identically orderedin the two periods. With linear demand.C. It follows that if demand has this form in both periods. even in our example in which second-periodprices are independent of first-periodmarket shares.22 With price competition. (7) 7rA = qA( fl(qA + q ) _ C) X(( + 4 ) Each firm therefore acts in period 1 as if an additional segment of demand of constant elasticity -1 had been added to the market. is that all consumershave switchingcosts of at least f (qm) . With quantity competition. marginal costs C per unit. If switching costs are transaction costs. and consumers have switching costs that are large enough that ir = oAlrm.MARKETS WITHSWITCHINGCOSTS 389 Considerthe case analyzed in Section III in which firms have constant. equal marginal costs are exactly as well off with large switching costs in the second conditionsto defineequilibriumfor 21. so if f1(*) is first-period inverse demand (taking account of any first-period learning or transaction start-up costs).21 firm's second-period profits are always increasing in its market share and its second-periodprice is unaffected by its market share. This content downloaded from 181. irA(qA. where -xm is the monopoly profit in an otherwise identical market without Then neither of the caveats above applies.29 on Fri.so that the consumerswho have the qmhighest reservationprices in the second periodall purchasedin the first period.Note that we have ignoredthe fact that. In these exampleswe are thinking of adding second-periodswitchingcosts that do not affect customers'first-periodconsumptionutilities. where qmis the output either firm would produceas a monopolistin the second period if there were no switching costs. or purely psychologicalcosts. firms always make less profits with large switching costs than with no switching costs. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions .where qA and qB are firms' first-periodquantities.the second-periodswitchingcosts affect firms'commonfirst-perioddemand fl(q) by affecting consumers'decisions about whether or not to buy in the first period(by buyingin the first period.consumersare also buyingthe rightto buy more cheaply in the second period).qB) = (qAI(qA + qB))irm. It then remainstrue that firms may either prefer large switching costs or prefer no switching costs. A sufficientconditionfor the first-order the two-period problem. This is natural if switching costs are artificial costs. for either price or quantity competition.firmsmake the same total profits over the two periods with infinite switching costs as with no switching costs. learning costs (the compatibility of firms' products may affect second-period switching costs but not first-period start-up costs).239.

in order to attract high-reservation-price consumerswho will become valuable repeat customers. 25.with switching costs.25 23. The extra units sold in the first period are less socially valuable (their social value is closer to their marginalcost) than the units of output that are lost in the second period when output is contracted to the monopoly level. including linear demand and costs and demand of constant elasticity -1. The example with price competition most clearly illustrates a welfare cost of switching costs. then some of these consumerspay the switchingcost to switchbetweenfirmsin the secondperiod. switching costs cause an allocative inefficiency even though no consumers actually pay the switchingcost. even though they also attract low-reservation-price consumerswho will never purchase again. not because they want to attract the lower-reservation-priceconsumers but because they want to win the largest possible share of the higher-reservation-price consumers. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions .and book clubs give away books as introductory offers.Thus.23 Although it is possible to find demand curves for which switching costs increase total welfare. The result is similar with quantity competition. the competition for market share is sufficiently fierce that first-period prices are below firms' costs. industry output is socially optimal in both periods if there are no switching costs. In it. A more competitive first period typically increaseswelfare.29 on Fri. banks give gifts and cash to undergraduatesopening their first bank-accounts. In a model in which products are functionally differentiated as well as differentiatedby switchingcosts (see Klemperer[1987b]). This intuition is exact for the case of demandof constant elasticity -1. Consider a demand curve that is sufficiently concave above the noswitching-costsduopolyequilibriumpoint. Thus.switchingcosts may also causea welfareloss akin to that in a standardmodel in whichthere is a cost increase. In all these examples. 24. However.390 QUARTERLY JOURNALOFECONOMICS period as they are without switching costs in the second period. fewerthan the no-switching-costs these numberswill not be exactly equal. and even though no consumerswould want to switch between the products if switching costs were zero. Their first-periodprices are low. but that is linear or convexbelow it. Also.of course.but in general. while second-periodprices are at the monopoly level. changeover If any consumers' preferencesfor the underlyingproductcharacteristics time. for which the numberof extra units sold in the first period equals the numberof units output sold in the second.since output is less than socially optimal in Cournot competition without switching costs.24simple examples.while others buy the good that is not the one they wouldchoose if there were no switching costs. in a model in which switching costs are This content downloaded from 181. but firms produce excessive output in the first period and too little in the second if there are switching costs.177. but this welfare gain is generallyoutweighedby the second-periodwelfare loss.239. firms sell to more consumersin the first period than in the second period. suggest that switching costs generallyreduce welfare.

APPENDIX: NONCOOPERATIVE QUANTITY-SETTINGEQUILIBRIUM This Appendix analyzes the noncooperative quantity-setting equilibriumin the second-period of a market with switching costs. this simplifies to the first-ordercondition for the If ~y(O) endogenous. and should also examine the role that switchingcosts may play in determiningindustry structure. switching costs do not necessarily make firms better off. This work has relied on a two-period model. this paper has taken switching costs as an exogenous feature of markets. 26.firms may choose socially inefficienttechnologies in order to build up switchingcosts becauseof the implicationsfor marketequilibriumdiscussed in this paper. standards. contracts. Aghion and Bolton [1985]showhow switchingcosts createdby contractscan be used to deter new entry. Future researchshould also look at markets in which in each period a proportionof consumersleaves the market In such markets the distinction and is replacedby new consumers.1986a.1986]and Katz and Shapiro[1985.. but that these rents induce greater competition in the early stages of the market's development. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . policies that encouragestandardization. This content downloaded from 181. Using (lb) to find apliapA Iqiconst and then (la) for aqA/OpA IqsBconst the first-order condition can be simplified to show that. Future research should directly examine firms' incentives to increase or reduce switching costs through product-design choices. 27. symmetricequilibriumwith a = (Al) f (2qA) IfqA dcA f (2qA) q q I 4 Af'(2 A fio(2q 4q A)Y(O)] A)y(o =? f = 0. and Klemperer [1987a] analyzes the effect of real switching costs on entry.26 between the first and subsequent periods would be of less consequence than in our model. Finally. at any = 1/2.emphasizingnetwork-externalityeffects.29 on Fri. Klemperer[1987b]examinesa two-periodmodelwith new consumersin the second period. so the model provides some support for regulatorypolicies that lower switching costs: for example.177. and it would be useful to determine the extent to which the results carryover into a multiperiod setting. Thus.etc.MARKETS WITHSWITCHINGCOSTS V.239. CONCLUSION 391 This paper shows that switching costs in a mature market lead to monopoly rents. Farrelland Saloner [1985. and it would be interesting to establish the properties of steady-state equilibria.1986b] discuss incentives to standardize. The presumption is that welfare is reduced.

0 symmetricCournotequilibriumwithout switching costs: f(2qA) qA + qA f (2qA) As with price competition. the only possible symmetric pure-strategyequilibrium has each firm producingits Cournotoutput qA = qB = (a . if the first-order conditions define the equilibrium. equal marginal costs.177. the noncooperative equilibrium is for each firm to choose q = aF(a . this result also holds for unequal market shares. Thus. in an otherwise identical market without switching costs. we need to check whether the quantities given by the first-order conditions actually form an equilibrium. respectively. for large enough s each firm chooses "its share" of the monopoly output. and 0A = a = 1/2.C)/30.k +k2(_ .29 on Fri. a6A :& CyB.Vk2 2+ (a -C)2k(a a-) hence (A3) pA=pB=1/3{k?2aC -C)} This content downloaded from 181. we can verify the followingresults. (this is (4) multiplied by f '(q). or.239. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . and this is an equilibrium if the proportion of consumers without switching costs is at least one-third. cA(q) = C (q) = Cq. As y(O) -P cc. which they do if s ?a -C) 1 IF=AB. the first-order condition approaches that for the . and the market price and total profits are the monopoly price and profits. as before q = 2qA). and oa = 1/2.fq. a unique symmetricpure-strategyequilibriumexists and has (A2) q = q =(1/60) {(a-C) . If a proportionof consumershave no switching costs.392 QUARTERLY JOURNALOFECONOMICS monopoly or collusive oligopoly equilibrium f (2qA) - dqA + 2qA f (2qA) =0. f (q) = a . If all consumers have a switching cost of at least s > 0. With any distribution of switching costs with y(O)= 1/k that B A ' = satisfies the second-order conditions. f (q) - 1/2(CA (q/2) + cB (q/2)) + qf '(q) = 0.C)/23. If firms have constant.For linear demands and costs.

Katz."Stability in Competition. Suzanne.) ST.177. LIII (July 1985)."Journal of Business."EconomicJournal. my. and Carl Shapiro."Reviewof EconomicStudies." Stanford GraduateSchool of Business WorkingPaper 786. XVII (Winter 1986).29 on Fri. Customers. XXXIX (March 1929). This content downloaded from 181. December1985. Joseph von R. "Entry Deterrence in Markets with ConsumerSwitching Costs. V." Hotelling." . LV (October1982). J.Compatibility. forthcoming. XXXVIII (November 1986a). . Thesis.forthcoming.ed. Ph. XCIV (August1986b)."OxfordEconomic Papers. and J. "The Existence of Equilibriumin Discontinuous EconomicGames. Heineke.424-40." Clarke. Porter. and Robert H. Miller.. XCVII (1987a).70-83. Rand Journal of .767-81.-F. "The Competitivenessof Markets with Switching Costs." Nelson. Quantity Competition:The Role of Uncertainty."HarvardUniversityWorkingPaper."Econometrica. "Network Externalities. 517-30..LII (January1984). Paul D. Trading Stamps. Analogously to price competition. Harold. and Eric Maskin. 87-100. LXXVIII (March/April Scotchmer." ." North-Holland. "OptimalPricingin the Presence of ExperienceEffects.. "The Principleof Y.forthcoming. September1985.D. LIII (January 1986). "Price Competitionvs. CATHERINE'S COLLEGEAND INSTITUTE OF ECONOMICSAND STATISTICS. forthcoming. Michael L. in Telecommunica. November1985a. September1986..and Economics."EconomicIssues in Standardization. Klemperer. and AmericanEconomicReview. Compatibility."Standardization.822-41."Technology Adoption in the Presence of Network Externalities. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . Y. Thisse. 311-29. De Palma.XVIII (Spring1987b). and Technology Choices Aid Collusion.FrankH." Rand Journal of Economics. Journcl of Political Econoand ConsumerBehavior." Rand Journal of Economics. Phillip." GTE LaboratoriesWorking Paper." Economic Journal. Competition. Partha.A. . Philippe.MARKETS WITHSWITCHINGCOSTS 393 for all k v [0. tions and Equity. Papageorgiou. "NoncooperativeCollusionunder ImperEconometrica. "Entry Prevention through Contractswith HarvardUniversityWorkingPaper. "Market Share Inertia with More Than Two Firms: An Existence Problem.41-57. "Information 1970). oo). p = (a + C)/2.EdwardJ. 1-26." . p = (a + 2C)/3. and MargaretA. May 1984. MinimumDifferentiationHolds Under Sufficient Heterogeneity. Meyer.LXXV (June 1985).239. as h -k o. "Markets with Consumer Switching Costs. the price and profits are higher than with price competition for any given level of switching costs k.XVI (Spring1985).1986).and Journal of Political Economy. Dasgupta. "Product Compatibility Choice in a Market with Technological and Progress. "Collusionvia Switching Costs: How 'Frequent-Flyer'Programs.and John M. and GarthSaloner. to the fully collusive outcome.I: Theory. see Klemperer [1984]. "Competitionwith Lock-in. (Amsterdam: Green. MasakoN. . OXFORD REFERENCES Aghion. and Patrick Bolton. (Compare (A3) with (5): the intuition is the same as that for a market without switching costs. However. Ginsburgh.. at k = 0. Darrough." Stanford University.. fect Price Information.and Innovation. the market price and industry profits rise monotonically from the quantitysetting equilibrium without switching costs. Farrell.

Birger. Yuval. Spence. XIV (April 1977). "Price and Quantity Competitionin a Differentiated Duopoly." EconomicsLetters.and Xavier Vives.XII (Spring1981).Pricing. "A Theory of Oligopoly." Bell Journal of Economics. October 1985. LXXII (February1964). LII (September1984). 1085-1116. von Weizsdcker. Christian. "NontatonnementEconomies: Market Organization. XXI (May 1986). "The Costs of Substitution.XV (Winter1984).. Wernerfelt. Michael."Rand Journal of Economics. Nirvikar." Econometrica."Journal of Political Economy. 546-54.177.239.394 QUARTERLY JOURNALOFECONOMICS Shilony. GeorgeJ.373-88.C. and GeneralEquilibrium. This content downloaded from 181. "The Learning Curve and Competition. Stigler. Singh. "Mixed Pricing in Oligopoly."Journal of Economic Theory.LawrenceH.29 on Fri.. 13 Dec 2013 16:54:22 PM All use subject to JSTOR Terms and Conditions . 77-79.49-70. "OnFrequent-FlyerProgramsand other Loyalty-Inducing EconomicArrangements. A.44-61."NorthwesternUniversity WorkingPaper. Summers.

- stataguide.pdf
- 2008 - Apuntes de Álgebre Lineal - José Cánovas Peña
- Modelos Ecuaciones Diferenciales
- Saavedra
- Guia Analisis Concevf
- Abuso de Posición de Dominio en La Modalidad de Abuso de Procesos Legales
- La Nueva Microeconomia - Comentarios
- Presentación Precios de Transferencia
- 2010 - Macroeconomic Theory - A Dynamic General Equilibrium Approach - Exercises and Solutions - University of York - Mike Wickens
- Sesión 2 - Spss Básico
- Sesión 1 - Spss Básico
- 2002 - Universidad Carlos III Madrid - Apuntes de Teoría de La Decisión y de Los Juegos - Juegos Estáticos
- Primer Set de Ejercicios Calificados de Microeconomía I
- 09 Evaluacion de Politicas Publicas Con Microsimulaciones
- 2009 - Universidad Nacional de Colombia - Dynare
- Modelos de Primera Practica Calificada _ Calculo i
- Apuntes de Macroeconomía Avanzada - Torres Chacon
- Vampiro s
- B Regresion
- Deloitte Peru - Curso NIC 12
- Anuario Cuanto 2013
- ME1-2014-I
- Matemática Básica
- Club Teoria
- Modelo de 1era Practica de Mate Basica

Sign up to vote on this title

UsefulNot useful- dp796
- [Oz Shy] the Economics of Network Industries
- frbrich_wp05-6.pdf
- Synopsis on Cellular Network
- Sheffi Urban Trans Networks
- Solucionario Shy.pdf
- HW 4
- Chap007 - Copy
- Managerial Economics
- The+Balance+of+Payments Unit 3
- chapter 6
- Monopoly
- Crotty Uncertainty
- 2006
- new.dvi
- VV-P2
- heckscher ohlen
- [Israel M Kirzner] the Driving Force of the Marke(BookZZ.org)
- 1. Introducion of Micro Economics -.pdf
- Exchange (Chapter 17)
- Economics.doc
- Managenent&Marketing
- Copy of Economics F.Y.B.com
- Economics F.Y.B.com
- internationalbusinessenvironment-fullsyllabus-090816082904-phpapp02
- hamza new.docx
- Group 10 Switching Cost
- Financial Market Simulation Using Multi Agent Approach (Cut)
- Book Chapter 11 Submission
- Supply
- Markets With Switching Cost