Internet Channel Entry: Retail Coverage and Entry Cost Advantage

Transcription

1 Internet Chnnel Entry: Retil Coverge nd Entry Cost Advntge June Cheng Fculty of Business The Hong Kong Polytechnic University Hung Hom, Kowloon, Hong Kong Brrie R. Nult Hskyne School of Business University of Clgry Clgry, Alert, Cnd Jnury 16, 2006 We thnk the editors nd the reviewers of this ITM Specil Issue, the prticipnts of the INFORMS My 2004 Conference, of the T.J. Burns Colloquium t the Ohio Stte University, nd of the INFORMS CIST 2004 for their helpful comments. We lso thnk the support from the Deprtment of Accounting nd MIS in the Fisher College t The Ohio Stte University, the Dvid B. Roson Professorship fund nd the Informtics Reserch Centre t Hskyne School of Business t the University of Clgry, nd the Socil Sciences nd Humnities Reserch Council of Cnd. Copyright c 2006 y June Cheng nd Brrie R. Nult. All rights reserved.

2 Internet Chnnel Entry: Retil Coverge nd Entry Cost Advntge Astrct In this reserch we study how existing mrket coverge ffects the outcome of the Internet chnnel entry gme etween n existing retiler nd new entrnt. A mrket is not covered when some consumers with low reservtion prices re priced out y existing retilers nd do not purchse. In model with multiple existing retilers nd potentil new entrnt, we demonstrte tht when costs re equl, one of the existing retilers enters the Internet chnnel first. However, if the mrket is covered y existing retilers efore entry, then ecuse of the thret of Internet chnnel entry y the potentil new entrnt, retiler entry cnnilizes existing retil profits cnnilizing t loss. In ddition, if potentil new entrnt hs slight dvntge in Internet chnnel entry costs nd the mrket is not covered y existing retilers, then the new entrnt enters the Internet chnnel first. If the mrket is covered y existing retilers, then the new entrnt must hve lrger Internet chnnel entry cost dvntge to e first to enter the Internet chnnel. Keywords: retil pricing, B2C electronic commerce, mrket entry, stnd-lone incentive, preemption incentive, timing gme, cost dvntge.

3 1 Introduction Direct mrketing hs long een n importnt usiness prctice. According to survey conducted y the Direct Mrketing Assocition, direct mrketing generted n estimted $2.17 trillion in 2003 sles $1.17 trillion in consumer nd $998.4 illion in the usiness-tousiness mrket ( The trditionl direct chnnel includes ctlog miling nd TV dvertising. Incresingly firms re using the Internet s direct mrket chnnel to rech consumers. Prior to the dvent of the Internet some of these firms lredy hd n existing rick-nd-mortr storefront (e.g., Brnes & Nole), while others did not (e.g., Amzon). A puzzling question is why in some industries existing firms with well-known retil presence nd fcing significnt retil nd direct mrket competition were not firstmovers into the Internet chnnel. In this work we study how the existing mrket coverge ffects the outcome of the Internet chnnel entry gme etween n existing retiler nd new entrnt, where the existing retil mrket coverge is whether every consumer purchses from one of the existing retilers (covered) or some consumers with low reservtion prices re priced out nd do not purchse (uncovered). Our strting point is retil mrket where there re severl existing retilers (incuments) in loction nd price equilirium. 1 Then the technology for n Internet chnnel ecomes ville to them nd the new entrnt. As the entry cost into the Internet chnnel, mostly technology relted, declines over time the existing retilers nd the new entrnt strtegiclly decide when to lunch their Internet chnnel. We exmine the effect of the existing retil mrket coverge on the outcome of the entry gme - oth on the order of entry nd on the profitility of entry. Intuitively we expect tht in the uncovered mrket, oth the incument nd the new entrnt would try to move first to gr the uncovered portion of the mrket nd increse mrket shre nd profit, while, in the covered mrket, the incument might e reluctnt to lunch the Internet chnnel in fer of the possile chnnel conflict: conflict etween the Internet chnnel nd retil chnnel driving down prices nd profits. However, we find tht the results run somewht counter to tht intuition. We show tht n entry cost dvntge over the incument is necessry for the new entrnt to enter the Internet chnnel first. 1 Therefter we use existing retiler nd incument interchngely. 1

4 In ddition, the necessry entry cost dvntge is smller in the uncovered mrket, so n uncovered mrket does not fvor incument entry. Finlly, without sufficient entry cost dvntge for the entrnt, in the covered mrket the incument enters the Internet chnnel first ut loses money t the mrgin from entry ecuse of the resulting chnnel conflict. Thus, we find tht the existing retil mrket coverge hs significnt impct on the entry gme, especilly in view of the literture tht mostly ssumes covered mrket. Mrkets tht re uncovered hve importnt implictions for Internet chnnels oth ecuse of the potentil of the Internet to rech nywhere, nd the low fixed setup cost reltive to incuments tht locte in high-cost shopping res. An exmple of n uncovered mrket is the ethnic ook mrket in the United Sttes. Becuse the ethnic popultion is sufficiently scttered, demnd in vrious geogrphicl loctions does not justify the opening of physicl ookstore, nd in cses where there is sustntil locl popultion tht justifies retil store, the trnsporttion cost prohiits people further wy from physiclly visiting these stores. Therefore, the ethnic ook mrket in the United Sttes is n uncovered mrket. The rest of the pper is orgnized s follows. In the next section, we review the relted literture. Then we present our sic model setup. In the susequent two sections, we discuss seprtely the price setting gme when the mrket is covered nd when it is not, nd compre the incentives for the existing retiler nd the new entrnt in oth cses. Our nlysis of the timing gme, where the incument nd the new entrnt simultneously decide time to enter the Internet chnnel, follows from the comprison of the incentives. In the finl section we conclude y discussing the implictions nd limittions. 2 Relted Literture nd Model Formultion The fundmentl difference etween the existing retilers nd the Internet chnnel from the consumers stndpoint is tht for existing retilers consumers hve to physiclly trvel to the store in order to do the shopping, while in the Internet chnnel no physicl trnsporttion cost is incurred y the consumers. Rther, the cost of visiting Internet store is mostly fixed cost, consisting of the shipping cost, the delyed grtifiction, nd so on. We use the Slop s (1979) circle model to cpture this distnce-relted retil differentition. And 2

5 we plce the Internet chnnel in the center of the circle to cpture its nowhere-everywhere presence, following Blsurmnin (1998). We employ the circulr city model ecuse it cptures two fetures simultneously: the influence of incuments on ech other, nd the impct of the Internet chnnel on ech of the incuments. Assuming tht consumers hve perfect knowledge out the loctions of the firms nd the prices the firms chrge, consumers decide which firm to uy from sed on not only the reltive prices, ut lso the reltive distnces to ech firm. 2 In Blsurmnin (1998), the competition etween direct mrketers nd existing retilers is modeled y distriuting the retilers round circle nd putting third-prty direct mrketer in the middle of the circle. Thus, ech consumer is the sme distnce from the direct mrketer. The presence of the direct mrketer lters the mrket in such wy tht with profitle direct mrketer, ech retiler is forced to compete ginst the direct mrketer rther thn ginst neighoring retiler. We orrow the sic modeling ide of plcing the Internet chnnel in the middle of the circle. However, rther thn ssuming covered mrket, we lso consider the cse of n uncovered mrket. And we llow the entry y oth the new entrnt nd the existing retiler, while in Blsurmnin (1998) only the new entrnt cn enter the Internet chnnel. The previous literture on the timing of entry hs minly focused on the entry into new product mrkets. Lilien nd Yoon (1990) rgue tht the decision to enter the mrket should e timed to lnce the risks of premture entry ginst the missed opportunity of lte entry, nd empiriclly test set of propositions out the reltionship etween the mrketentry time nd the likelihood of success for new industril products. Using dt from French firms, they find tht firms re more successful when the new product is lunched during the introduction or growth stge of the product life cycle. Mitchell (1991) lso empiriclly tests the reltionship etween entry time nd the post-entry performnce. His rgument is tht incuments re likely to possess strong sets of ssets required for the commerciliztion of goods in new technicl sufield, nd s consequence, the effects of eing erly or lte vry with the type of entrnt. Our pper, however, investigtes the entry into new mrketing 2 In the sptil differentition models, the distnce etween the consumer nd the firm cn e interpreted oth s physicl distnce nd s the degree of the lck of fit etween the consumer s idel product nd the firm s ctul product offer. In our setting, we tke the first interprettion to highlight the importnce of the physicl presence of the existing retilers. 3

6 chnnel the Internet chnnel. In the Internet chnnel incuments usul entry dvntges re mitigted y expertise required to pply new technology - expertise they my not possess. Schoenecker nd Cooper (1998) study the role of firm resources nd orgniztionl ttriutes in determining entry timing nd find tht two ctegories of resources, technologicl nd mrketing, re ssocited with erly entry. This is consistent with our results, s we show tht n entry cost dvntge, which cn e the consequence of resources or cpilities, is necessry for the new entrnt to enter successfully. In ddition, our results re consistent with Nult nd Vndenosch (1996, 2000), who show tht the incument my lunch premturely, nd sometimes lose t the mrgin, in order to preempt the entrnt, nd the incument my e preempted y n entrnt with cpilities dvntge in the next genertion product. Our pper is lso relted to the multi-chnnel literture. Zettelmeyer (2000) discusses how, s the Internet expnds, firms should price their products nd whether they should fcilitte consumer serch in the retil chnnel nd in the Internet chnnel. This mtters ecuse y vrying the mount of product informtion provided, firms cn chieve finer consumer segmenttion nd increse their mrket power. In similr vein, Riggins (2002) showed how the digitl divide, where high type consumers dominte the Internet chnnel nd low type consumers dominte the retil chnnel, rtificilly segments the mrketplce therey mitigting the clssic cnniliztion prolem. Their focus is on the post-internet-chnnelentry decisions regrding to prices, communiction strtegies, or qulity differentition. Our pper, however, is on the trnsition from pre-entry to post-entry, focusing on the entry decision. 3 3 Bsic Setup We dopt the circulr sptil competition model introduced y Slop (1979). Consumers re ssumed to e distriuted uniformly on the edge of circle of unit circumference. Ech consumer is in the mrket for either zero or one unit of homogeneous good. Consumers hve the sme reservtion price, denoted y R. 3 There re other strems of reserch tht use the circle model for different purposes. For exmple, Bkos (1997) studies the role of uyer serch costs for differentited products in n electronic mrketplce, nd Dewn et l. (2000) study how the distriution of specil commodity informtion goods should e orgnized through proprietry networks nd the Internet. 4

7 The pre-entry mrket is in equilirium. We ssume tht existing retilers re locted t equl distnces from ech other on the circumference, which implies tht they re t the mximl distnce from ech other. This is consistent with the principle of mximl differentition (Tirole 2000, p.286), nd Kts (1995) who shows tht the equl distnce is n equilirium in the circle model. Equilirium lso mens tht there is no further retil entry. Blsurmnin (1998) shows tht even though the mrket is closed to further retil entry, it might still e open to direct mrket entry s long s the fixed setup cost is low compred to tht of the retil entry. In our model we ssume equilirium in the pre-entry mrket, so the fixed setup cost of retil entry is not relevnt, while the Internet chnnel entry cost is. Our interest is in n Internet chnnel entry either y existing retilers or y new entrnt. For simplicity, we ssume there re two incuments (denoted y A nd B) in the retil mrket, nd potentil new entrnt (denoted y E) not in the retil mrket. In order to focus our ttention on the entry gme etween the incuments nd the new entrnt, we ssume tht the entry cost for one of the incuments (sy A) is lower thn tht for the other. And we ssume tht the entry cost difference for the incuments is sufficiently lrge so tht we cn ignore the other incument in our timing gme lter. 4 In setting where there re more thn two retilers, this is the sme s picking the one with the lowest entry cost, or the lrgest cpilities dvntge (Nult nd Vndenosch 2000). Consumers incur trvel costs t (liner) rte t per unit distnce when visiting retiler long the circle. Following Blsurmnin (1998), ll consumers who uy from the Internet chnnel incur fixed shipping nd disutility cost of µ. Exmples of the disutility re delyed grtifiction nd lck of n opportunity for physicl inspection. To concentrte on the impct of mrket coverge, we ssume µ is the sme regrdless of which firms re in the Internet chnnel, so the Internet chnnel is undifferentited. Mrginl cost of procurement nd distriution of the good is equl for the incuments nd entrnt, nd is normlized to zero. We discuss implictions of this in our Conclusion. We tke t, µ, nd R to e in R +. The prices chrged y the two incuments A nd B for the retil goods re p nd p, respectively. The price chrged for the good in the Internet chnnel is denoted y p d. We 4 The purpose of our pper is to investigte whether the incument nd the new entrnt enters the online chnnel. This ssumption llows us to focus on the gme etween the incument nd the new entrnt nd simplifies our nlysis. Without this ssumption we would hve to consider the competition etween the incuments, which is not our min theme. 5

8 llow price discrimintion y incument if it opertes oth in the retil chnnel nd in the Internet chnnel. Profits re π nd π for the two incuments, nd π e for the entrnt. Prior to Internet chnnel entry, the two retilers mrket shre cn e determined using Figure 1 s 2x nd 2y (x or y on ech side of the firm) (Figure 1). ***Insert Figure 1 out here*** We ssume incuments nd the entrnt hve full informtion concerning ech other s costs, prices nd loctions, s well s consumer s disutility of uying online, their distriution round the circle, their trvel costs nd reservtion prices. Price Setting Gme In mking the entry decision, the incument nd the new entrnt consider their pre-entry profits, post-entry profits, nd the entry cost. Our first step is to solve the price setting gme nd clculte the equilirium profit in the pre-entry stte where no firm is in the Internet chnnel (denoted s Stte 1 or s1), nd in the following three possile post-entry sttes: incument A in the Internet chnnel (denoted s Stte 2 or s2), new entrnt E in the Internet chnnel (denoted s Stte 3 or s3), nd oth incument A nd new entrnt E in the Internet chnnel (denoted s Stte 4 or s4). Stte 1 is Slop s (1979) model, nd when the retil mrket is covered Stte 3 is Blsurmnin s (1998) model. We follow the convention of using superscipts for sttes so tht, for exmple, profits of incument A in Stte 1 is denoted y π s1. We solve the price setting gme nd clculte the equilirium profits in Section 4 nd 5. Timing Gme In order to study whether the incument or the new entrnt enters the Internet chnnel first, we model the entry gme s timing gme with declining entry cost. The timing gme is gme in which plyers decide on the optiml time to enter. Internet chnnel entry cost is mostly technology-relted cost driven y computing nd telecommuniction devices, nd softwre development. We ssume the cost of Internet chnnel entry is declining ecuse of dvnces in technology nd lerning from other pplictions. Declining technology doption cost ssumptions re common in the literture (e.g., Reingnum 1981, Fudenerg nd Tirole 1984, Ktz nd Shpiro 1987, Nult nd Vndenosch 1996, 6

9 2000). Let K i (T ) e the present vlue of the entry cost t time T, so tht the current cost is K i (T )e rt, i =,, e, where r > 0 is the interest rte. Declining entry cost mens tht d(k i (T )e rt )/dt < 0. We lso ssume tht entry cost flls t decresing rte, tht is, d 2 (K i (T )e rt )/dt 2 < 0. To mke the cse interesting, we ssume tht initilly entry is too costly so tht no firm enters t time zero. We consider the cse where the entry cost is the sme for incument A nd the new entrnt (equl entry costs), nd the cse where the entry cost is not the sme (unequl entry costs). The timing gme is presented in Section 6. Mrket Coverge Conditions The mrket eing covered mens tht the reservtion price R is high reltive to the trnsporttion cost t. In our setup the mrket is covered when R t/2 nd is not covered when R < t/2. To see this, suppose the mrket is not covered. Using incument A, the mrket shre of 2x cn e derived from the limiting eqution for the indifferent consumer, p + tx = R. Incument A mximizes profit, π = 2p x = 2p ( R p ), t y choosing p. First order condition yields p = R/2 nd x = R/2t. For the mrket to e uncovered ech incument s mrket shre must e less thn 1/2, therefore x < 1/4 mening R < t/2. If R t/2, then the mrket is covered. 5 R < t/2 s the mrket coverge conditions. We refer to the inequlities R t/2 nd Mrket Comprison Condition For the Internet chnnel to hve positive mrket shre the reservtion price must e greter thn the disutility of purchsing from the Internet chnnel, R > µ. If the mrket is not covered, then R > µ is sufficient for the Internet chnnel to hve positive mrket shre. Comining R > µ with the mrket coverge condition when the mrket is not covered gives t/2 > R > µ, which simplifies to µ/t < 1/2. To enle us to compre the covered mrket nd the uncovered mrket, we restrict our nlysis to µ/t < 1/2, nd refer to this s the mrket comprison condition. Essentilly this mens tht the disutility of purchsing from the Internet chnnel is smll reltive to the trvel cost to existing retilers. 5 See Lemm 1 for prices nd profits in this cse. 7

10 Tle 1: Conditions nd constrints Mrket Coverge Conditions R t/2 (covered mrket) R < t/2 (uncovered mrket) Mrket Comprison Condition µ/t < 1/2 Internet chnnel Prticiption Constrint C1 p d + µ R Internet chnnel Prticiption Constrint C2 x + y 1/2 Internet Chnnel Prticiption Constrints The reservtion price R plys centrl role in the price setting gme. Ignoring the incuments for the moment, for the Internet chnnel to hve positive sles the Internet chnnel price plus the disutility cost is constrined to e no greter thn the consumer s reservtion price. We denote this constrint, p d +µ R, y C1. Reintroducing competition from the incuments, we must lso constrin the Internet chnnel to nonnegtive mrket shre. We denote this constrint, x + y 1/2, y C2. Although C1 nd C2 re oth relted to Internet chnnel prticiption, the former is price constrint nd the ltter is mrket shre constrint. They re oth required s they ind under different conditions. The conditions nd constrints re summrized in Tle 1. We first solve the price setting gme in the covered mrket nd uncovered mrket, find the equilirium profits, nd compre the incentives to enter the Internet chnnel for the incument nd the new entrnt. Then, we nlyze the timing gme where the incument nd the new entrnt decide on time to enter the Internet chnnel. 4 Price Setting Gme nd Comprison of Incentives in the Covered Mrket 4.1 Equilirium in the price setting gme When the retil mrket is covered the mrket coverge condition is R t/2. In order to compre this cse to the cse where the retil mrket is not covered we dd the mrket comprison condition, µ/t < 1/2. In the following nlysis these two conditions re implicitly imposed. 8

11 Stte 1: No Internet Chnnel Slop s (1979) Model Stte 1 provides the seline cse of strictly retil competition, nd the equilirium depends on the mgnitude of the reservtion price reltive to the trnsporttion cost. 6 Incument A s profit mximiztion prolem is to choose p s1, nd incument B s profit mximiztion prolem is to choose p s1 : mx π s1 p s1 = mx 2p s1 p s1 x p s1 + tx = p s1 mx π p s1 s1 + ty R, x + y 1/2, (p s1 = mx 2p s1 p s1 y + tx R)(x + y 1/2) = 0, p s1, p s1 0. Lemm 1 If there is no firm in the Internet chnnel, then Nsh equilirium profits re s follows: If t/2 R 3t/4, then π s1 = π s1 = R t/4. If R > 3t/4, then π s1 2 = π s1 = t/4. Although R t/2 ssures the mrket is covered, R = 3t/4 is the rek point for competition etween the two incuments. Tht is, when R 3t/4, the reservtion price is sufficiently low tht the incuments no longer compete with ech other. The equiliri when R < t/2, when t/2 R 3t/4, nd when R > 3t/4, correspond to the monopoly equilirium, the kinked equilirium, nd the competitive equilirium, respectively, in Slop (1979). Our covered mrket comines the kinked equilirium nd the competitive equilirium, nd our uncovered mrket corresponds to the monopoly equilirium. Stte 2: Incument A Alone in the Internet Chnnel Referring to Figure 1, incument A s retil mrket shre is 2x, where x represents the loction where the consumer is indifferent etween purchsing from incument A nd from the Internet chnnel, i.e., p s2 +tx = p s2 +µ. Similrly, incument B s retil mrket shre is 2y, where y cn e derived from p s2 d +ty = ps2 +µ. Consequently, incument A s Internet chnnel shre is (1 2x 2y). d 6 Unless stted elow, our proofs re in n ppendix tht ccompnies this mnuscript. 9

12 Moreover, since incument A lwys hs the option not to sell in the Internet chnnel, i.e., chrge higher thn the reservtion price, the equilirium would e the sme s tht in Stte 1 if incument A finds tht opening n Internet chnnel is not profitle. Incument A s profit mximiztion prolem is to choose p s2 C1 nd C2: mx p s2,ps2 d π s2 = mx p s2,ps2 d { 2p s2 ( ps2 d ps2 + µ t ) + 2p s2 d ( 1 2 ps2 d ps2 + µ ps2 d t nd p s2 d, under constrints } + µ ) t ps2 (1) p s2 d + µ R (C1), x + y 1/2 (C2), p s2, p s2 d 0. Incument B s profit mximiztion prolem is to choose p s2 mx π p s2 s2 { = mx 2p s2 p s2 ( ps2 d } + µ ) t ps2 And incument A lwys hs the option to chrge p s2 d : p s2 0. (2) + µ > R nd not sell in the Internet chnnel if the mximizing profit is less thn tht in Stte 1. In this cse, the profit mximiztion prolem is the sme s tht in Stte 1. Lemm 2 descries the incuments equilirium profits in Stte 2. Lemm 2 If only incument A is in the Internet chnnel, then Nsh equilirium profits re s follows: If µ < 1 nd t R < 52µ2 32µt+25t 2, or 1 µ < nd t+2µ R < 52µ2 32µt+25t 2, then t t 4 t t π s2 = 13µ2 8µt+4t 2 nd π s2 18t = (2µ+t)2. 18t If 1 µ < nd t R < t+2µ, or µ < 1 nd t R < ( 3+1)t 2( 3 1)µ, then 4 t t π s2 = µ2 +2µR 2R 2 + R µ nd π s2 2t = R2. 2t If µ t < 1 4 nd 52µ2 32µt+25t µ t < 1 2 nd ( R < 3t, or 1 µ < t 36t 3+1)t 2( 3 1)µ 4 R < 3t 4, then πs1 nd 52µ2 32µt+25t t = π s1 = R t/4 2. R < 3t 4, or If R > 3t/4, then π s1 = π s1 = t/4. The first equilirium is n interior solution to (1) nd (2). In the second equilirium C1 is inding, which mens tht incument A chrges Internet chnnel price s high s 10

13 the consumer reservtion price less the disutility of uying online. In the third nd fourth equiliri incument A finds selling in the Internet chnnel is not profitle, nd chrges p s2 d + µ > R. In this cse the equilirium is the sme s tht in Stte 1. From Lemm 2, we cn see tht in some cses when the mrket is covered, ecuse of chnnel conflict, incument A chooses not to sell in the Internet chnnel. Stte 3: New Entrnt E Alone in the Internet Chnnel Blsurmnin s (1998) Model In Stte 3, price competition is etween new entrnt in the Internet chnnel nd the two incuments in the retil mrket. When the mrket is covered this is the sme nlysis s in Blsurmnin (1998), resulting in Figure 1 where the top nd ottom segments of the circle represent the entrnt s Internet chnnel shre. As we will see lter on, Blsurmnin s results re sensitive to the ssumption of mrket coverge. Lemm 3 If only the entrnt is in the Internet chnnel, then Nsh equilirium profits re π s3 = π s3 = (t+4µ)2 72t nd π s3 e = (t 2µ)2 9t. The new entrnt drws some consumers wy from the incuments. As result, the incuments profit is less thn the profit efore entry. Our formultion of the entrnt s profit mximiztion prolem differs from Blsurmnin (1998) ecuse we explicitly include C1 nd C2. However the results re the sme ecuse n interior solution otins. Stte 4: Both Incument A nd New Entrnt E in the Internet Chnnel With oth A nd E in the Internet chnnel, ecuse the Internet chnnel is undifferentited, Bertrnd competition cuses the Internet chnnel price p s4 d to equl the mrginl cost.7 The profit for the new entrnt is zero. For the incuments, C1 is utomticlly stisfied y the mrket comprison condition. With the Internet chnnel price set, the incuments choose p s4 nd p s4 to mximize their profits. Agin using Figure 1 to illustrte, x is given y 7 In Bertrnd competition, firms compete in prices rther thn quntities. The lterntive is the Cournot model where firms choose quntities first nd the mrket price is set t level such tht demnd equls the totl quntity. The Cournot model is etter model if output is difficult to djust (e.g., hotel rooms) (Tirole 2000). However, in the online retil industry, output cn e esily chnged, nd Bertrnd model is etter model. In Bertrnd competition, prices ove mrginl cost cn only e chieved through product differentition on dimensions like rnd, wesite design nd customer service. 11

14 p s4 + tx = 0 + µ, nd y is given y p s4 + ty = 0 + µ. Ech incument mximizes its profit suject to C2, where using A s the exmple, { mx π s4 p s4 = mx 2p s4 p s4 ( µ } ps4 ) + 0 t Lemm 4 descries equilirium profits in Stte 4. x + y 1/2 (C2), p s4 0. (3) Lemm 4 If oth the incument nd the new entrnt re in the Internet chnnel, then Bertrnd competition cuses the Internet chnnel price to equl mrginl cost. Nsh equilirium profits re π s4 = π s4 = µ 2 /2t nd π s4 e = 0. Competition in the Internet chnnel not only drives the Internet chnnel price to mrginl cost, ut ecuse the Internet chnnel is sustitute for the retil mrket, this competition lso lowers the retil price. Tht is, profits re strictly less in Stte 4 thn other sttes ecuse of the dditionl competition in the Internet chnnel. We stte this in the following corollry. Corollry 1 Incument profits in Stte 1, 2 nd 3 dominte incument profits in Stte Incentives for Internet Chnnel Entry Hving worked out the profits for the incuments nd the new entrnt in ech of the sttes, we re redy to compre their incentives for Internet chnnel entry. Both stnd-lone incentives nd preemption incentives re relevnt in the timing gme which we discuss in Section 6. Our definition of stnd-lone incentives nd preemption incentives follows from Ktz nd Shpiro (1987). A firm s stnd-lone incentive is the difference etween its post-entry profit nd its profit when no entry hs occurred (seline profit). If firm elieves tht its rivls will not enter, the incument or the new entrnt ses its timing of entry on its stnd-lone incentive. However, if firm elieves tht its rivl will enter the Internet chnnel if it does not, then it compres its profit s the winner in the entry gme nd its profit s the loser where it is preempted y rivl. This difference is clled its preemption incentive. For the incument, 12

15 its stnd-lone incentive is π s2 π s1, nd its preemption incentive is π s2 π s3. For the new entrnt, ecuse it hs nothing to lose eing preempted, the stnd-lone incentive nd preemption incentive re the sme: π s3 e 0 = π s3 e. We compre their stnd-lone incentives nd preemption incentives in the following theorems which we use in our nlysis of the timing gme. The prmeter rnges specified in Theorem 1 only represent very smll proportion of the possile rnge. See Figure 2 for n illustrtion. ***Insert Figure 2 out here*** Theorem 1 (Comprison of stnd-lone incentives) The new entrnt hs greter stndlone incentives, tht is, (π s µ 6 11 nd t R 20µ2 +17t 2, t t or 6 11 µ < 1 nd t R 3(t+2µ)+ 11(t 2µ) 10 t 2 2 π s1 ) < π s3 e, except for the following prmeter rnges: 12. Theorem 2 (Comprison of preemption incentives) The incument hs greter preemption incentives, tht is, (π s2 π s3 ) > π s3 e. 5 Price Setting Gme nd Comprison of Incentives in the Uncovered Mrket 5.1 Equilirium in the price setting gme The retil mrket is not covered when R < t/2. This mrket coverge condition comined with reservtion price higher thn the disutility of purchsing from the Internet chnnel, R > µ, yields the mrket comprison condition, µ/t < 1/2. Any entry into the Internet chnnel leds to covered mrket ecuse the cost of uying from the Internet chnnel, µ, is the sme for ll the consumers so tht if one consumer finds it ffordle to uy then ll consumers do. Stte 1: No Internet Chnnel Slop s (1979) Model As efore, Stte 1 provides the seline cse of strictly retil competition. Lemm 5 is proven s prt of Lemm 1. 13

16 Lemm 5 If there is no firm in the Internet chnnel (Stte 1), then Nsh equilirium profits re π s1 = π s1 = R 2 /2t. When the mrket is not covered, the Slop (1979) model hs incuments pricing sed on the consumers trvel costs rther thn on competition. Stte 2: Incument A Alone in the Internet Chnnel When only incument A is in the Internet chnnel, incument A chooses p s2 in (1). Incument B chooses p s2 nd p s2 d nd its profit mximiztion is s to mximize its profit s in (2). As in the cse when the mrket is covered, incument A lwys hs the option not to sell in the Internet chnnel, i.e., chrge higher thn the reservtion price. If incument A finds tht opening n Internet chnnel is not profitle, then the profit mximiztion prolem would e the sme s tht in Stte 1. Lemm 6 nd its proof re similr to the proof of Lemm 2, except tht now the mrket coverge condition is R < t/2 nd tht ffects the rnges over which the different equilirium prices hold. Lemm 6 If only incument A is in the Internet chnnel, then Nsh equilirium profits re s follows: If µ t < 1 4 t+2µ nd < R < t, then πs2 3 2 = 1 18t (13µ2 8µt + 4t 2 ), π s2 = 1 (2µ + 18t t)2. If µ t < 1 4 π s2 t+2µ nd R, or 1 µ < 1 nd R < t, then πs2 3 4 t 2 2 = 1 2t (µ2 + 2µR 2R 2 ) + R µ, = R2. 2t Unlike when the mrket is covered, when the mrket is not covered the incument lwys chooses to sell in the Internet chnnel, i.e., chrge p s2 d R µ. The intuition is tht ecuse the mrket is not covered prior to the Internet entry, serving the uncovered mrket niche is lwys profitle. In ddition, in some of the prmeter rnge, constrint C1 is inding (the second equilirium in Lemm 6) where incument A chrges R µ in Internet chnnel nd just serves the 14

17 consumers tht were not reched efore it entered Internet chnnel, nd does not compete with incument B for mrket shre. In this cse, incument B s profit remins the sme s efore incument A enters the Internet chnnel (i.e., π s2 the notion tht firms lwys try to void competition. = π s1 ). This is consistent with Only when the disutility for the consumers of uying from the Internet chnnel is smll compred to the trnsporttion cost (i.e., µ/t < 1/4), nd the reservtion price is high (i.e., R > (t + 2µ)/3), will incument A find it profitle to use the Internet chnnel to compete with incument B, ecuse the Internet chnnel is more likely to ttrct consumers under these conditions. In this cse, C1 no longer inds nd incument B erns less profit thn it does efore incument A enters the Internet chnnel (i.e., π s2 < π s1 ). When the mrket is not covered, C2 never inds ecuse the Internet chnnel cptures the uncovered mrket nd gets positive mrket shre. Stte 3: New Entrnt E Alone in the Internet Chnnel In this cse, when new entrnt is in the Internet chnnel nd the mrket is not covered, Blsurmnin s (1998) results no longer hold. The entrnt chooses p s3 d profit mximiztion prolem: mx π p s3 e s3 d = mx p s3 d p s3 d { 2p s3 d ( 1 2 ps3 d to mximize its profit yielding the following ps3 + µ ps3 d t } + µ ) t ps3 + µ R (C1), x + y 1/2 (C2), p s3 d 0. Incument A s profit mximiztion prolem choosing p s3 is { mx π = mx 2p s3 p s3 p s3 ( ps3 d } ps3 + µ ) t Incument B s profit mximiztion prolem choosing p s3 is { mx π = mx 2p s3 p s3 p s3 ( ps3 d } ps3 + µ ) t The resulting equilirium profits re given in the lemm elow. (4) p 0. (5) p 0. (6) Lemm 7 If only the entrnt is in the Internet chnnel, then Nsh equilirium profits re s follows: If µ t < 1 2 If µ t < 1 2 t+4µ nd < R < t, then πs3 6 2 = π s3 = (t+4µ)2 72t t+4µ nd R, then π s3 6 = π s3 = R2 2t nd π s3 e nd π s3 e = (t 2µ)2 9t. = (R µ)(t 2R) t. 15

18 When the reservtion price is smll (i.e., R < (t+4µ)/6), C1 inds, which mens tht the new entrnt only serves consumers who were not reched efore nd does not compete with the incuments. This is ecuse, with low reservtion price, the mrket left uncovered y the incument is lredy enough for the new entrnt in the Internet chnnel to mximize its profit. Even though competing with the incuments would enlrge the new entrnt s mrket shre, it would lso drive the Internet chnnel price down. With these two competing forces, the Internet chnnel profit is less when the reservtion price is low. However, when the reservtion price is high (i.e., R > (t + 4µ)/6), the Internet chnnel finds it more profitle to compete with the incuments, nd C1 does not ind. Stte 4: Both Incument A nd New Entrnt E in the Internet Chnnel With oth the incument nd the new entrnt in the Internet chnnel, s efore Bertrnd price competition cuses the Internet chnnel price to equl mrginl cost, p s4 d constrint, incuments A nd B choose p s4 = 0. Within this nd p s4 to mximize their profit. Lemm 8 descries incuments profits in Stte 4, nd the proof is then sme s for Lemm 4. Lemm 8 If oth the incument nd the new entrnt in the Internet chnnel (Stte 4), then Bertrnd competition cuses the Internet chnnel price to equl mrginl cost. Nsh equilirium profits re π s4 = π s4 = µ 2 /2t nd π s4 e = 0. Compring the profits of oth incuments in Stte 4 versus other sttes, we hve similr corollry when the mrket is not covered s when the mrket is covered. And s with Corollry 1, from Corollry 2, Stte 4 is dominted y other sttes. Corollry 2 Incument profits in Stte 1, 2 nd 3 dominte incument profits in Stte Incentives for Internet Chnnel Entry We compre the stnd-lone incentives nd preemption incentives for oth the incuments nd the new entrnt when the mrket is not covered. As in Theorem 1, the prmeter rnge specified in Theorem 3 only represents smll portion of the possile rnge. See Figure 2 for n illustrtion. 16

19 Tle 2: Who hs igger incentives? Uncovered mrket Covered mrket Stnd-lone incentive The incument The new entrnt Preemption incentive The incument Theorem 3 (Comprison of stnd-lone incentives) The incument hs greter stndlone incentive, tht is, (π s2 0 < µ t 1/ 20 nd π s1 ) > π s3 e, except for the following prmeter rnge: (5µ 2 +2t 2 ) 3 R < t/2. Theorem 4 (Comprison of preemption incentives) The incument hs greter preemption incentive, tht is, (π s2 π s3 ) > π s3 e. Before we move on to the next section, we riefly compre the reltive incentive structure for Internet chnnel entry when the mrket is covered nd when it is not. The incument lwys hs stronger preemption incentive, regrdless of whether the mrket is covered ecuse, s the incument, the incument suffers more thn the new entrnt if it is preempted s it fces new competition. However, the stnd-lone incentives differ depending on mrket coverge. When the mrket is covered, the incument hs smller stnd-lone incentives except for the smll prmeter rnges specified in Theorem 1. This is ecuse when the mrket is covered the incuments compete with ech other. Opening profitle Internet chnnel cn only intensify competition, driving down prices nd the profits. When the mrket is not covered y incuments, the incuments hve greter stnd-lone incentives, except for the smll prmeter rnges specified in Theorem 3. The intuition is tht the incuments do not compete with ech other when the mrket is not covered, nd there is n unserved mrket niche. By opening n Internet chnnel, the incument cn ttrct the unserved consumers while leving the pre-entry monopolistic sitution intct. Tle 2 shows the reltive incentive structure, ignoring the smll prmeter rnges specified in Theorem 1 nd Theorem 3. With this reltive incentive structure, we next turn to our nlysis of the timing gme. 17

20 6 Timing Gme Consider when incument A enters the Internet chnnel first. Incument A s (leder) pyoff if it enters the Internet chnnel first t time T is: L (T ) = T 0 π s1 e rt dt + T π s2 e rt dt K (T ) = 1 e rt π s1 + e rt r r πs2 K (T ). The new entrnt s (follower) pyoff if incument A enters the Internet chnnel first t time T is F e (T ) = 0. 8 Consider when the new entrnt enters the Internet chnnel first. The new entrnt s pyoff if it enters the Internet chnnel first t time T is: L e (T ) = e rt r πs3 e K e (T ). Incument A s pyoff is: F (T ) = 1 e rt r π s1 + e rt r πs3. Consider now when oth incument A nd the new entrnt enter the Internet chnnel simultneously. Incument A s pyoff if oth enter t time T is: M (T ) = 1 e rt r π s1 + e rt r πs4 K (T ). The new entrnt s pyoff is M e (T ) = K e (T ). From Corollry 1 nd 2, it cn e esily shown tht L i (T ) > M i (T ) nd F i (T ) > M i (T ), i =, e. Therefore, we drop simultneous entry from further nlysis s neither firm would choose it in equilirium. Define T i y L i (T ) = F i (T ), i =, e. For the incument nd the entrnt T i defines the time when the pyoff of leding is equivlent to the pyoff of following. Equl Entry Costs We first consider the cse where entry cost is the sme for incument A nd the new entrnt. From the comprison of the preemption incentive, it cn e esily shown tht T occurs prior to T e : the preemption incentive is greter for the incument nd 8 To e consistent with the literture (e.g., Fudenerg nd Tirole 1984, Nult nd Vndenosch 1996), we use the nottion of leder nd follower. It is worth noting tht in our cse follower never follows, s shown in Theorem 5. 18

21 entry costs re the sme. Consequently, from the definition of T i we get tht K ( T )e r T = 1 r (πs2 π s3 ) nd K e ( T e )e r T e > 1 r πs3 e. Using Theorem 2 nd 4, we know tht K ( T )e r T > K e ( T e )e r T e, which implies tht T < T e y the ssumption of declining entry cost. Define ˆT s ˆT = rgmx L (T ). This is the time when entry is most profitle. If there is no entry thret from the entrnt, the incument will enter t this time. However, s shown in the following theorem, the incument enters premturely in order to preempt the entrnt. The proof follows from Theorem 1 from Nult nd Vndenosch (1996) Theorem 5 If K (T ) = K e (T ), then in oth the covered mrket nd the uncovered mrket, the incument enters the Internet chnnel t time T e, nd the new entrnt never enters. The equilirium in the competition etween incument A nd the entrnt is isomorphic to the incument nd entrnt in Nult nd Vndeosch s (1996) Theorem 1. Nult nd Vndenosch s (1996) Assumption 1 sys tht the incument lwys hs greter preemption incentive thn the entrnt. In our setting, the incument hs greter preemption incentive thn the new entrnt oth when the mrket is covered nd when it is not. See Figure 3 for n illustrtion. The intuition is tht idelly, the incument would like to enter when it is most profitle: t ˆT. But the new entrnt mkes positive profit from entry ny time from T e onwrd, nd therefore the incument must preempt the new entrnt t time T e. Even though the new entrnt never enters, it still plys role in the timing gme: it lters the time when the incument enters the Internet chnnel. ***Insert Figure 3 out here*** In ddition, if the entry cost is the sme for the incument nd the new entrnt, whether the mrket is covered determines if the incument mkes profit t the mrgin from entry. When the mrket is covered the incument mkes negtive profit t the mrgin from entry in the covered mrket, ut not in the uncovered mrket. The following theorem sttes this result. Theorem 6 Assume tht K (T ) = K e (T ). If the mrket is covered, then the incument mkes negtive profit t the mrgin from entering the Internet chnnel except for the prmeter rnge stted in Theorem 1. If the mrket is not covered, then the incument mkes 19

22 positive profit t the mrgin from entering the Internet chnnel, except for the prmeter rnge stted in Theorem 3. Proof: The incument s incrementl profit from entering t T e is: e r T e r (π s2 π s1 ) K ( T e ). T e is defined y L e (T ) = F e (T ), which is (e r T e /r)π s3 e = K e ( T e ). With K ( T e ) = K e ( T e ), nd sustituting K e ( T e ) into the incrementl profit for the incument, we get: e r T e r [(π s2 π s1 ) π s3 e ]. The term in the squre rcket is the comprison of the stnd-lone incentives of the incument nd the new entrnt. Therefore the comprison of the stnd-lone incentives determines the sign of the incrementl profit for the incument. In the covered mrket, the incument hs smller stnd-lone incentive thn the new entrnt except for the prmeter rnge stted in Theorem 1, so it mkes negtive profit t the mrgin. In the uncovered mrket, the incument hs greter stnd-lone incentive thn the new entrnt except for the prmeter rnge stted in Theorem 3, so it mkes positive profit t the mrgin. Q.E.D. Whether the profit mrgin for the incument is negtive or positive is determined y the reltive stnd-lone incentive. Mking negtive profit t the mrgin is known s cnnilizing t loss in Nult nd Vndeosch (1996). Cnnilizing t loss hppens in the covered mrket, ut not in the uncovered mrket (except for the prmeter rnge stted in Theorem 3) ecuse in the covered mrket the incuments lredy compete with ech other nd entry into the Internet chnnel only intensifies the competition. However, the rtionle for the incument to cnnilize t loss is strtegic: if the incument wits, then it would incur even greter profit loss should the new entrnt e first to enter. Unequl Entry Costs If the new entrnt hs lower entry cost, it is possile tht it my enter first. The reson is s follows. For the new entrnt to enter requires tht t some T, L e (T ) F e (T ) > L (T ) F (T ), from which we get K (T ) K e (T ) > e rt r [(π s2 π s3 ) πe s3 ]. Since the incument lwys hs greter preemption incentive, tht is possile only if K e (T ) < K (T ). Therefore, necessry condition for the new entrnt to enter first is tht the new 20

23 entrnt hs lower entry cost thn the incuments. This result is similr to Corollry 1 in Nult nd Vndenosch (1996). Lower entry cost is defined s cpilities dvntge y Nult nd Vndenosch (2000), cpilities dvntge tht results from vriety of resources, processes, nd situtions. In the cse of Internet chnnel entry, cpilities dvntge might e the result of more effective we development technology, etter consumer dtse mngement, or more flexile orgniztionl structure. A perfect exmple here is the Amzon s technologicl ledership over Brnes & Nole. Since its founding, Amzon hs crried strong reputtion for its leding-edge usiness intelligence, nlytics, nd dtse opertions. The compny hs mde huge investment in uilding out nd integrting the technology tht rn its We site, customer service unit, pyment processing systems, nd wrehouse opertions (Leschly et l. 2003). Well-known exmples of the technologicl innovtions mde y Amzon include its one-click uying, which it hd ptented, We site personliztion, nd online recommendtions. This technologicl ledership is criticl for Amzon s successful entry into the Internet retiling usiness. The following theorem descries the condition under which the new entrnt enters nd the timing of entry. Theorem 7 If the difference in the entry costs stisfies K (T ) K e (T ) > e rt r [(π s2 π s3 ) π s3 e ], then the new entrnt enters the Internet chnnel t time T, nd the incument never enters. Proof: It is esy to show tht T e < T under the condition stted in the theorem. For the rest of the proof, see Theorem 1 from Nult nd Vndenosch (1996). Q.E.D. The condition in Theorem 7 sys tht the minimum entry cost dvntge required for the new entrnt to enter is proportionl to the difference in the preemption incentives n entry cost dvntge is needed for the new entrnt to overcome the incument s stronger preemption incentive. Figure 4 shows the curve of the minimum entry cost dvntge when µ/t [1/4, ). 9 We cn see tht when the new entrnt cost dvntge is ove the 9 The curve hs similr shpe when µ/t is in other rnges. 21

24 curve (i.e., sufficient cost dvntge), the new entrnt enters, s stted in Theorem 7. When the new entrnt cost dvntge is elow the curve (i.e., insufficient cost dvntge), the incument enters. ***insert Figure 4 out here*** From the non-decresing pttern of the curve, it is cler tht the cost dvntge is required to e smller in the uncovered mrket thn in the covered mrket. Corollry 3 The necessry entry cost dvntge for Internet chnnel entry y the new entrnt is smller in the uncovered mrket thn in the covered mrket. Agin we see the impct of mrket coverge. This result essentilly sys tht it is esier for the new entrnt to enter the Internet chnnel when the mrket is not covered. It might seem counterintuitive, s one might think tht esier entry would occur in the covered mrket since the entrnt could tke dvntge of the incument s reluctnce to enter the Internet chnnel resulted from the negtive profit mrgin. But this is not the cse. In the covered mrket, the entrnt mkes lrger profit compred to the uncovered mrket, ecuse of lrger consumer reservtion price R. However, the incument is le to mke even lrger profit compred to the uncovered mrket. Therefore, the needed entry cost dvntge is lrger in the covered mrket to overcome the even stronger preemption incentive of the incument. This result helps to explin the success of Amzon. It is estimted tht the numer of ook titles ville t Amzon.com is more thn 23 times lrger thn the numer of ooks on the shelves of typicl Brnes & Nole superstore (Brynjolfsson et l. 2003). By offering oscure ook titles (e.g., ethnic ooks) tht re not ville t physicl ookstores, tht is, serving the uncovered mrket, Amzon.com ws successful t preempting Brnes & Nole s entry into the online ook retil industry. A contrsting cse is the filure of Wevn, which ws n online grocery store. The grocery mrket in generl is covered mrket, with physicl grocery stores offering wide vriety of selections in close proximity to most people s homes. When the mrket is covered, it tkes lrger entry cost dvntge for the new entrnt to enter profitly. Unless Wevn possessed this cost dvntge over existing retilers, such s Sfewy, Wevn ws doomed to fil. 22

25 7 Conclusion In this reserch, we study the Internet chnnel entry gme etween n existing rick-ndmortr retiler nd new entrnt nd nlyze how the existing mrket coverge ffects the outcome of the entry gme. We use Slop s (1979) circle model to cpture the distncerelted differentition of existing retilers. We plce the Internet chnnel in the center of the circle to cpture the nowhere-everywhere presence (Blsurmnin 1998). We first compre the stnd-lone nd preemption incentives of the existing retilers nd the new entrnt for the Internet chnnel entry nd find tht mrket coverge is n importnt fctor determining the reltive incentives. In prticulr, when the mrket is covered, the new entrnt hs stronger stnd-lone incentives except for the smll prmeter rnges specified in Theorem 1. When the mrket is not covered, the existing retiler hs stronger stndlone incentives except for the smll prmeter rnges specified in Theorem 3. On the other hnd, in oth covered nd uncovered mrkets, the existing retiler hs stronger preemption incentive ecuse, s the incument, it hs more to lose if it is preempted. With this reltive incentive structure, we then nlyzed the timing gme where the existing retiler with the lowest entry cost nd the new entrnt strtegiclly decide when to lunch the Internet chnnel s the entry cost mostly technology-relted declines over time. We find tht, if the entry costs re equl, then the existing retiler enters the Internet chnnel first. However, when the mrket is covered, the incument mkes negtive profit t the mrgin from entry. For the new entrnt to enter the Internet chnnel, n entry cost dvntge is necessry, nd the needed entry cost dvntge for the entrnt smller when the mrket is not covered. Our results hve the following implictions. First, when n existing retiler chooses when to lunch n Internet chnnel, it must consider mrket coverge nd the possiility of entry from new entrnt. The existing retiler must e prepred to preempt s the mrket coverge determines whether entry results in mrginl profit loss. Even though entry might entil loss t the mrgin in the covered mrket, it is justifile in order to preempt the new entrnt nd void even greter profit loss. When there is n unfilled mrket niche, ech of the incuments would like to enter the Internet chnnel to gr the unfilled mrket niche. 23

26 The firm with the lowest entry cost the gretest cpilities dvntge enters first nd enjoys the first-mover dvntge. This demonstrtes tht striving to develop the lrgest cpilities dvntge in order to e the first mover mong the incuments is crucil to rep the profit gin derived from the incresed mrket shre. Second, from the stndpoint of the new entrnt, since the entry cost dvntge needed to overcome the existing retiler s stronger preemption incentive is smller in the uncovered mrket thn the covered mrket, it is importnt for the new entrnt to choose the right spot sed on its cpilities dvntge reltive to the existing retilers. The new entrnt should focus more on situtions where there is uncovered mrket niche, s less dvntge is required in this kind of mrket. Third, our reserch sheds light on the reltionship etween the Internet chnnel nd the retil chnnel. The Internet chnnel my compete ginst or complement the firm s own retil outlets, depending on mrket coverge. When the mrket is covered, to gin positive shre, the Internet chnnel hs to compete ginst incuments for consumers, which would cuse profit loss for incuments. The incuments my choose not sell in the Internet chnnel in order to void the chnnel conflict. In the uncovered mrket, it is possile for the Internet chnnel to ttrct the un-served consumers while leving the pre-entry monopolistic sitution intct. Socil welfre is incresed, oth ecuse of higher profit, nd more consumers eing reched. We relize tht our model is highly stylized. First, we ssume undifferentited competition in the Internet chnnel, which results in Bertrnd competition with more thn one firm in the Internet chnnel. In ctulity Internet chnnel memers cn differentite themselves long vriety of dimensions, for exmple, rnd, wesite design, nd suggestion tools. 10 Under product differentition, there might e second entry into the online chnnel. However, in the Internet world, there is lwys first-mover dvntge in uilding up the customer se. Therefore, who enters first is the most importnt question. Second, we ssume tht the firms mximize profit t ech time period. If, however, we ssume sticky demnd (for exmple, through rnd loylty), the firms cn ctully mximize mrket shre during the erly periods to lock in customers nd rep greter profit from them lter on. If the de- 10 For review on price nd product differentition in the Internet chnnel, see Smith et l. (1999). 24