On the Competitive E ects of Mobile Virtual Network Operators

Transcription

1 On the Compettve E ects of Moble Vrtual Network Operators Phlp Kalmus, Lars Wethaus September 2006 Prelmnary - please don t quote wthout permsson Abstract Moble vrtual network operators (MVNOs) o er moble telecoms servces to customers by purchasng capacty from moble network operators (MNOs) whlst competng wth them at the retal level. Ths paper explores to what extent MVNOs exert a compettve constrant on network operators. We employ a two stage model. In the rst stage MNOs determne capacty both to wholesale and to retal. In the second stage MNOs and MVNOs compete n the retal market. We nd that MNOs host MVNOs only f the latter do not exert a compettve constrant on MNOs retal busnesses. Thus, absent access regulaton, MVNO entry may happen but s unlklely to reduce consumer prces. JEL Class caton: L13, L42, L96 Keywords: Competton, Moble Telecommuncaton, MVNO, merger regulaton. Correspondng author, LECG Ltd., 5 Southampton Street, London WC2E 7HA, UK, lwethaus@lecg.com. The paper bene ted substantally from dscussons wth Matthew Bennett. Moreover we thank Nuno Ruz and Len Waverman for useful suggestons. Fnancal support from Portugal Telecom s gratefully acknowledged. 1

2 1 Introducton Moble Vrtual Network Operators (MVNOs) o er moble telecoms servces to customers by purchasng capacty or wholesale mnutes from a Moble Network Operator (MNO) and sellng t under ther own brand to end-customers. In ths way, MVNOs avod the need to own and operate ther own end-to-end moble rado access networks. The number of MVNOs n Europe has ncreased substantally over the last three years, wth seven new operators launchng n the UK alone over ths perod. Many well known consumer brands have launched servces, or have publcly announced ther ntent to enter the market. However, the degree to whch these MVNOs have successfully attracted customers (ether new moble customer, or customers from other operators), and the busness models adopted, vary consderably from case to case. (Ofcom 2006) Despte ther obvous presence MVNOs have receved lttle attenton n the academc economc lterature. 1 In partcular the compettve e ects of MVNOs entry n moble communcatons markets s unclear. On the one hand, one mght conjecture that an ncreased number of retal operators ncreases the ntensty of competton n the retal market and therefore reduces prces for end-customers. On the other hand, network operators mght refuse to supply capacty to MVNOs for ths reason. Yet, as ponted out by Brto and Perera (2005), the threat that one s rval mght o er capacty rather than oneself could potentally create a prsoners dlemma of hostng MVNOs and nduce competton n the retal market respectvely. A proper understandng of these e ects seems mportant. Frst t may help both network operators and MVNOs to formulate busness agreements. Secondly, regulators mght be nterested n whether MVNOs can negotate compettve access condtons absent regulaton. Thrdly, anttrust authortes may consder MVNO entry as a remedy for moble mergers. However, 1 We dscuss the study by Brto and Perera (2005) below. 2

3 such a remedy would only be sutable f MVNOs provded an as strong compettve constrant to the market than a network operator. To analyse these questons we construct a two-stage Cournot model. In the rst stage network operators decde how much capacty to buld up ) for ther own retal busness and ) for ther wholesale supply. In the second stage MVNOs decde how much capacty to purchase from network operators and both MVNOs and network operators compete n the retal market. We account for product d erentaton through a parameter for substtuton between MVNOs and MNOs and demonstrate how network operators decsons to host an MVNO are n uenced by the value of ths parameter. Our ndngs suggest that the mere ntroducton of MVNOs does not ncrease the ntensty of competton n the retal market. End-costumer prces do not decrease, rrespectve of the number of MVNOs and the extent to whch MVNOs servces consttute a substtute to the network operators servces. The reason s that network operators choose wholesale capactes su cently low, such that MVNOs retal supples do not compromse the network operators own retal busness. In the extreme, f MVNOs servces are a perfect substtute to the network operators servces, the latter o er no wholesale capacty at all. If MVNOs servces, are d erentated from those of network operators, network operators wll o er a lmted amount of wholesale capacty to MVNOs. Yet n ths case servce d erentaton secures stable retal prces for network operators. Brto and Perera (2005) have ndependently analysed the compettve e ects from MVNO entry. In contrast to the present work, the authors employ a two-stage Bertrand model n whch network capacty s gven. In the rst stage three network operators make a take-t-or-leave-t o er to a sngle MVNO. The MVNO accepts the best o er and n the second stage rms compete on Salop s crcle. Brto and Perera establsh cases n whch the ncumbent network operators face a prsoner s dlemma wth respect to hostng the MVNO. In contrast to our ndngs, then, network operators grant favourable access condtons to the MVNO and retal prces may decrease. Whle theoretcal and emprcal nvestgatons regardng the compettve e ects of MVNOs are scarce some studes on energy markets exst. Boots et al. (2004) apply a successve olgopoly approach to the European natural gas 3

4 market. In ths market, (upstream) producers supply gas to (downstream) traders who compete n the market for end-customers. Ths structure s smlar to network operators supplyng capacty to MVNOs who sell mnutes n the retal market. Ther model yelds results that are consstent wth ours. Our model d ers, however, as not only non-ntegrated downstream supplers (MVNOs) compete n the retal market but also the ntegrated upstream suppler themselves (network operators). Recently Hastngs and Glbert (2005) carred out an emprcal study of wholesale prcng of vertcally ntegrated gasolne re ners n the U.S. They found hgher wholesale prces for ndependent gasolne retalers n regons close to re ner-retalers. Appled to the present context, ths would mean that network operators ncreased ther wholesale prce the less d erentated MVNOs servces would be to the network operators 2. Ther result s smlar to ours. Due to our model set-up, less servce d erentaton between the network operator and the MVNO results n less (equlbrum) wholesale supply rather than a hgher (equlbrum) wholesale prce. Our theoretcal model s based on standard formulatons n the lterature. We employ a two-stage olgopoly set-up (Greenhut and Ohta 1979) as has been used n a number of prevous studes. Salnger (1988), Ordover, Saloner and Salop (1990), McAfee (1999), and, more recently, Economdes (2005) are several papers that study the ncentves for vertcal ntegraton wthn ths type of model. The paper s organzed as follows. Secton 2 presents the model and results. Secton 3 provdes a non-techncal nterpretaton of our man results. Secton 5 concludes and provdes suggestons for future research. 2 The model We consder a market for moble telecommuncaton servces wth two network operators (MNOs), = 1; 2, and n moble vrtual network operators (MVNOs), vo = 1; 2; :::n. Network operators and MVNOs compete for end- 2 Product or servce d erentaton may ether take the form of geographcal d erentaton (as n the gasolne retal market) or d erentaton regardng the product or servce attrbutes tself (lke n the case of moble communcaton servces). 4

5 costumers n quanttes, q and q vo. Quanttes q and q vo are the amount of voce mnutes suppled to end-customers 3. Pror to quantty competton at the retal stage network operators expand nfrastructure capacty alongsde two dmensons. Frst they buld up capacty, k, to satsfy ther ntended own retal supply, q. Secondly network operators expand nfrastructure capacty, k vo to compete wth each other n the supply of wholesale mnutes for MVNOs. We assume that there exsts a drect and trval relatonshp between nfrastructure capacty, k mnutes), q and q vo, of the form k q and k vo of nfrastructure capacty relates to one unt of output quantty. and k vo, and output quanttes (voce q vo. That s one unt That s we consder capacty n terms of base statons, glass bre cables, mcrowave transmtters and addtonal transmsson equpment rather than n terms of rado spectrum. Whle the amount of the latter mght be xed and excessve even n the longer run, the former has to be adjusted regardng actual moble tra c. The structure of ths ndustry s vsualzed n Fgure 1. It s helpul to thnk of competton takng place n two stages. In the rst stage, network operators decde how much capacty they buld up. In partcular each network operator, ; determnes the amount of capacty k whch t bulds up for ts own supply to end-customers and the amount of capacty whch t supples to the MVNO, k vo. As stated above, we assume that the amount of nstalled capacty translates drectly nto the amount of servces whch can be suppled to end-consumers,.e. mnutes of moble communcaton. Costs of capacty take the followng lnear form: C (k ; k vo ) = c(k +k vo ). The demand for capacty by MVNOs s derved from the MVNO s maxmzaton problem n the second stage. In the second stage, gven capacty nstalled n the rst stage, the network operators and the MVNO supply q and q vo mnutes of moble communcaton 3 Ths paper does not go nto the detals of access and termnaton charges. Rather we consder Cournot competton as an approxmaton for mperfect competton n the market for moble call orgnaton and abstract from nput procurement n terms of voce termnaton. Parker and Röller (1997) con rm the approprateness of ths smpl caton emprcally. In any case, the abstracton from termnaton charges s not crtcal to our results snce the larger sze of MNOs networks would tend to ncrease MVNOs e ectve end-costumer prces relatve to MNOs. That s n ths paper we assume condtons for the MVNOs that are more favourable than those n realty. 5

6 Network Operator 1 Retal Wholesale Capacty Capacty Network Operator 2 Wholesale Retal Capacty Capacty k 1 k 1 vo k 2 vo k 2 Network Operator 1 Retal Supply q 1 MVNOs 1,2,...n Retal Supply q vo Network Operator 2 Retal Supply q 2 End-customers Fgure 1: Moble communcaton ndustry structure wth network operators and MVNOs. servces to end-customers. For smplcty, we abstract from operatng costs at the retal level. In the second stage, only the MVNO bears costs n terms of purchasng capacty from the network operators at the wholesale prce w. 4 At the retal level, we assume that there s a lnear nverse demand functon facng the th network operator. P = a q q j X vo q vo ; = 1; 2; 6= j; (1) where a > c s the prohbtve prce and ; ; measures the degree of servce substtutablty between the th network operator s servces and the MVNO s servce from the end-customers pont of vew. In partcular a low value for means that the MVNO s servce does not consttute a close 4 As Ofcom (2004) states t s essentally ths cost-asymmetry that dstngushes a MVNO from a network operator: "MVNOs can avod many of the xed costs ncurred by MNOs such as spectrum lcenses, network nfrastructure, and platforms for new servces. [...] However, snce MVNOs pay out a large proporton of ther revenues n xed-agreement wholesale fees to network operators, ther operatng margns are far lower than those of MNOs." 6

7 substtute to s servces whereas a hgh value refers to a hgh degree of substtutablty. D erent degrees of substtutablty may re ect, for nstance, customers preferences for certan types of prcng or brand mage. Also a low value for could be nterpreted that the MVNO targets customer segments through spec c dstrbuton channels whch are a) not avalable to the network operators and b) attractve to the MVNO s customer segment. To some extent supermarket MVNOs attract unque customer segments. In order to focus on the dependence of the hostng decson on the substtuton parameter, we assume that end-customers consder network operators servces as perfect substtutes (.e. a substtutablty parameter of value one). MNOs, n contrast, serve the (same) mass market. Therefore we do not assume servce d erentaton among MNOs and, furthermore, that both MNOs exhbt the same degree of servce d erentaton wth respect to MVNOs. The correspondng nverse demand functon facng the MVNO can be wrtten as X P vo = a (q + q j ) q vo ; vo = 1; :::n: (2) The network operators pro t functons take the form vo = P q + w X vo k vo c(k + X vo k vo ) ; = 1; 2; (3) Retal Revenues Wholesale Revenues Costs of Capacty where P s gven by (1). As dsplayed by (3) the network operators profts consst of revenues through ther retal actvty as well as through ther wholesale actvty. In order to generate revenues n the retal market and n the wholesale market, costly buld-up of capacty n terms of spectrum and nfrastructure s requred. The MVNO s pro t functon s vo = P vo q vo wq vo ; vo = 1; :::n; (4) Retal Revenues Costs of wholesale mnutes whch P vo s gven by (2). A comparson of (4) wth (3) hghlghts the man d erences between the MVNO and MNO busness models. Frst, network 7

8 operators generate revenues through both the retal and the wholesale markets whereas the MVNO only operates at the retal level. Secondly network operators have sgn cantly lower margnal costs at the retal level (namely the margnal cost of purchasng wholesale mnutes, w) whereas MVNOs don t have to ncur any costs of capacty (nfrastructure) nstalment. We now analyse the rms decsons wthn the above outlned two stage competton process. In dong so we proceed backwards and derve rst the rms equlbrum output quanttes at the retal level. Wth these results at hand, secondly, we derve the network operators equlbrum capacty nstalment. 2.1 Competton at the retal level (2nd stage analyss) At the retal level the network operators,, = 1; 2; and the MVNOs, vo, vo = 1; :::n, smultaneously compete n quanttes. As wll be clear below, however, the two stage game n whch network operators rst set capactes and then engage n Cournot competton at the retal level reduces to one n whch network operators set capactes and output quanttes n the rst stage (as Stackelberg leaders) whereas the MVNOs set output quanttes n the second stage (as Stackelberg followers). The reason s the asymmetrc cost structure of network operators and MVNOs where network operators ncur costs of capacty n the rst stage but have no such costs at the second stage (retal level). In contrast, MVNOs have operatng costs n the second stage equal to the wholesale prce set by the network operators. Let us start by assumng that, n the second stage, there are no capacty constrants at all. Then network operators and MVNOs would maxmze (3) and (4) wth respect to q and q vo leadng to, q = a(1 + n(1 )) + nw ; = 1; 2 (5) 3 + n(3 2 2 ) and q vo = a(3 2) 3w ; vo = 1; :::n (6) 3 + n(3 2 2 ) where q and qvo denote the (hypothetcal) 2nd stage Nash equlbrum n the absence of capacty constrants. 8

9 Next consder capacty constrants and dstngush the followng cases: I : q < k ; = 1; 2 II : q > k ; = 1; 2: In case I; capacty constrants are non-bndng and network operators set quanttes accordng to (5). In case II, capacty constrants are bndng and network operators have to supply less than q. In fact t s optmal for them to supply k n ths case 5. In case I; the MVNO sets ts output quantty accordng to (6). In case II, the MVNO knows that the network operators set k < q. Gven ths t s optmal for the MVNOs to maxmze 0 vo = (a (k + k j ) wth respect to q vo leadng to X q vo )q vo wq vo ; vo = 1; :::n (7) vo qvo r = a (k + k j ) w ; vo = 1; :::n (8) (n + 1) where q r vo denotes the MVNOs optmal response as a functon of the network operators capactes, k ; = 1; 2. To determne the second stage Nash equlbrum t thus crucal to know whether case I or case II holds n stage 2. If t s case I then rms play (q ; q j ; q vo) n the second stage whereas n case II rms play (k ; k j ; q r vo). Yet t s easy to see that case I would not consttute a subgame perfect Nash equlbrum of the rst stage because ths would leave the th rm wth unused but costly capacty (see Dxt 1982). Rather the th rm could pro tably reduce ts capacty whch s true as long as case I holds. Thus we are secured that case II must hold n stage Note that pro t functons are strctly concave n 2 =@q 2 = c < 0. Hence for any q < q the rst-order-condton to the network operators second stage maxmzaton problem are postve. 6 The remanng queston s just whether there exsts a closed form soluton to the network operators rst stage maxmzaton problem, k, such that k 6 q or, else, that the closed loop soluton mples k > q. In the former case the closed form soluton would consttute the subgame perfect Nash equlbrum whereas n the latter case, network 9

10 Gven case II we can derve the MVNO s demand curve for wholesale mnutes usng (8). Keepng n mnd the assumpton that one output unt q requres one unt of capacty k and hence that the supply of q r vo requres that q r vo > k vo + k vo j ; we get w = a (k + k j ) (n + 1)(k vo + k vo j ): (9) where w s the wholesale prce that clears the wholesale market, gven the network operators supply of n(k vo + k vo j ). 2.2 Capacty competton (1st stage analyss) Upon substtuton of k for q and (k vo + kj vo ) for q vo n (3) and (1) we can wrte the th network operators rst stage pro t functon as = P k + wk vo c(k + nk vo ); = 1; 2; 6= j; (10) where P = a k k j n(k vo + k vo j ); = 1; 2 6= j: (11) The network operators maxmze ther objectve functon wth respect to k and k vo. The rst-order-condtons = a 2k k j n(2k vo + kj vo ) c = 0; = 1; 2; 6= j = n(a (n+1)(2k vo +kj vo ) (2k +k j ) c) = 0; = 1; 2; 6= j: (13) Equlbrum capactes Solvng (12) and (13) smultaneously for k and k vo ; = 1; 2; yelds the (unque) subgame perfect Nash equlbrum n the network operators capacty strateges, k = (a c)(1 + (1 )n) ; = 1; 2 (14) 3 + 3(1 2 )n operators would just nstall capactes up to k = q, as a corner soluton. The latter soluton s unlkely to emerge because capacty s costly n the rst stage whereas output quanttes are not n the second stage. We shall see below (see footnote 10) that our closed form soluton to the rst stage maxmzaton problem, k, ndeed sats es k 6 q. 10

11 and k vo = An mmedate mplcaton from (15) s the followng (a c)(1 ) ; = 1; 2: (15) 3 + 3(1 2 )n Clam 1 If MVNOs servces consttute a perfect substtute to network operators servces,.e. = 1, the network operators do not o er wholesale mnutes to the MVNO, k vo = 0; = 1; 2: Ths s the rst of our man results: network operators have no ncentve to o er wholesale capacty to MVNOs as long as MVNOs o er servces whch are und erentated from those of the network operators. An alternatve nterpretaton of ths result s that network operators would set ther wholesale prce prohbtvely hgh such that no wholesale mnutes would be demanded. Accordngly we would not expect MVNOs to o er a perfect substtute to the network operators servces. In ths way, MVNOs do not exert compettve pressure on MNOs and the ntroducton of MVNOs cannot compensate for the loss of MNOs. To understand the ntuton behnd Clam 1 t s useful to examne closer the network operators rst-order condtons. Snce the Nash equlbrum s symmetrc we can re-wrte (12) and (13) slghtly, omttng Retal capacty = a 3k n3k vo c = 0;(16) =j Wholesale capacty = a vo Retal Prce E ect (n + 1)3k vo Wholesale Prce E ect WholesalePrce E ect 3k Retal Prce E ect c = (17) 0: Equatons (16) and (17) dsplay the margnal e ects of an ncrease n a network operators capacty and, respectvely, supply n the retal and the wholesale market. The margnal e ect of capacty/supply ncreases s ambguous. On the one hand t may ncrease revenues as a consequence of the larger volume sold. On the other hand t may decrease revenues because ncreased output volumes decrease market prces. By (16) and (17) network operators have to take four prce e ects nto account when makng ther capacty/supply decsons: 11

12 Retal Prce E ect of retal capacty: If a network operator ncreases ts retal capacty/supply n the retal market, ths decreases the prce n the retal market (drectly). Wholesale Prce E ect of retal capacty: If a network operator ncreases ts retal capacty/supply, ths decreases ts wholesale prce ndrectly because ts ncreased retal supply has a downward pressure on the retal prce whch reduces the MVNOs margnal pro tablty of retal o ers and hence the MVNOs demand for wholesale mnutes. Wholesale Prce E ect of wholesale capacty: If a network operator ncreases ts wholesale capacty/supply ths decreases the market prce (drectly). Retal Prce E ect of wholesale capacty: If a network operator ncreases ts wholesale capacty/supply ths decreases the the retal prce ndrectly because the ncreased wholesale o er ncreases the MVNO s supply n the retal market. If servces are perfect substtutes ( = 1), t s apparent by (16) and (17) that an ncrease n wholesale capacty has as strong a negatve retal prce e ect (.e 3k) as an ncrease of retal capacty (.e. 3k). Clearly, snce servces are perfect substtutes, a supply ncrease of ether a network operator or the MVNO cause the same downward pressure on market prces. At the same tme an ncrease n retal capacty has a weaker negatve e ect on the network operator s wholesale prce (.e. n3k vo ) as compared to a drect ncrease n wholesale capacty (.e. (n + 1)3k vo ). An ncrease of wholesale capacty has a double mpact: the addtonal capacty lowers prces n the retal market whch reduces demand for wholesale mnutes and, secondly, ncreased wholesale capacty reduces the wholesale prce drectly due pro t maxmsaton of MVNOs. Ths s a representaton of the double margnalsaton e ect. Intutvely network operators prefer to earn the MVNOs margn themselves and hence supply retal mnutes rather than wholesale mnutes (see Salnger (1988) for smlar arguments). As can also be observed from (16) and (17) that wth an ncreasng number of MVNOs the d erence between these two e ects s dmnshng. Due to ther larger number, MVNOs 12

13 lose barganng power relatve to the network operator, hence MVNOs margns decrease and accordngly the e ect of double margnalsaton s reduced. However, f the degree of servce substtutablty between the network operator and the MVNO decreases ( < 1), (16) and (17) reveal that both the (negatve) wholesale prce e ect of ncreased retal capacty and the (negatve) retal prce e ect of ncreased wholesale capacty decrease. Ths means that ncreases n retal capacty tend to have a stronger e ect on the retal prce than ncreases n wholesale capacty. Ths dvergence may, to some extent, counterbalance the loss of double margnalsaton and n turn lead to postve wholesale capacty. In partcular, let us consder how network operators supply capactes change n the degree of servce substtutablty. The dervaton of (14) and (15) wth respect to = (a c)n(1 + n(1 )2 2) 3( 1 + ( )n) 2 7 0; = 1; 2; = (a c)(1 + n(1 )2 ) < 0; = 1; 2: (19) 3( 1 + ( )n) 2 Equatons (18) and (19) lead to Clam 2 (a) A hgher degree of servce substtutablty,, decreases (ncreases) network operators supply n the retal market, k, f the product substtutablty s low (hgh). (b) A hgher degree of servce substtutablty,, decreases network operators supply n the wholesale market, k vo. Proof. Frst clam: (18) s negatve 1 + n(1 ) 2 2 > 0. Second clam: obvous by (19). Regardng the network operators ncentves to supply wholesale capacty, Proposton 2 con rms the ntuton, developed n the prevous secton: network operators o er more capacty n the wholesale market when MVNO s servces are d erentated at the retal level. Next consder the change n the network operators capactes due to a change n the number of MVNOs. The dervaton of (14) and (15) wth 13

14 respect to n = (a c)(1 ) 6 0; = 1; 2; (20) 3( 1 + ( )n) = Equatons (20) and (21) mply the followng: (a c)(1 ) > 0; = 1; 2: (21) 3( 1 + ( )n) 2 Clam 3 If the number of MVNOs ncreases, network operators reduce ther own retal supply and ncrease ther wholesale supply. The ntuton behnd Clam 3 s that the e ect of double margnalsaton decreases f the number of MVNOs ncreases. MVNOs become more compettve, lower ther retal prce to the compettve level (.e. at the level of the wholesale prce), and hence, ceters parbus, wholesale o ers become more attractve for the network operators as they can capture (more of) the complete margn themselves or, respectvely, can ncrease output wth less dstortons of the retal prce 7. Equlbrum retal prces obtan the network operators retal prces Upon substtuton of (14) and (15) n (11) we P = a (k + k j ) n(k vo = 1 (a + 2c); 3 + k vo j ) (22) and substtuton of (14) and (15) for q ; q j and q vo ; where q vo = k v (4) gves the MVNO s retal prces, where P vo = a (k + k j ) n(k vo + kj v, n + k vo j ) (23) = 2c( + n( )) + a( n(1 2 )) ; 3 + 3( )n lm P vo = 1 (a + 2c): n!1 3 Equatons (22) and (23) mply the followng 7 As dscussed for (16) and (17) above the gap between the retal prce e ect of wholesale and retal capacty decreases. 14

15 Clam 4 (a) The network operators equlbrum retal prces are constant n the degree of servce substtutablty and the number of MVNOs. (b) The MVNOs retal prce s decreasng n the degree of servce substtutablty and the number of MVNOs n the market. Proof. (a) s obvous. For (b), note that the dervatve of (23) wth vo =@ = (a c)( )=( ) 2 < 0 vo =@n = (a c)(1 =(3( 1 + ( )n) 2 > 0. The fact that the MVNO s retal prce s decreasng n the degree of servce substtutablty to the network operators s not surprsng. The less the MVNO manages to target dstnct customer segments, say due to spec c tar plans or unque dstrbuton channels, the more t competes drectly wth the servce operators. Therefore the more ntense s competton and the lower ts retal prces. Also not surprsng, the MVNOs retal prce decreases n the number of MVNOs and approaches the wholesale prce level (24) f the MVNO segment becomes perfectly compettve (.e. n! 1). In contrast, network operators retal prces reman constant n the degree of servce substtutablty. Ths s our second mportant result of the paper. Even f network operators suppled capacty to a (d erentated) MVNO, prces for the network operators end-customers reman as hgh as f there would be no MVNO at all. In other words, network operators would support MVNOs to the extent that these do not put compettve pressure on ther own exstng customer base. Ths, however, contradcts the noton of MVNOs as a remedy to an mpedment of e ectve competton n the market for moble telecommuncaton servces (arsng from ncreased concentraton). To understand the reason for ths result, recall that all rms compete n the same retal market 8 whereas (22) represents the retal prce that network operators would attan f they competed n the retal market wth margnal costs of output, c, but absent MVNOs. Even n the presence of MVNOs, network operators essentally replcate the (Cournot) outcome n the retal market of moble communcaton servces 9. From a network operators pont 8 Wth respect to competton between network operators and MVNOs ths s true, of course, to the extent that servces consttute substtutes to each other (measured by ). 9 It s easy check that, f q vo = 0, and margnal costs of producton of c, then maxmsaton of (3) wth respect to q ; = 1; 2, yelds equlbrum output quanttes, q whch 15

16 of vew, MVNOs are just another means of dstrbutng moble communcaton capacty. Then, gven that there exsts a Cournot Nash equlbrum n output quanttes wth margnal costs of capacty, c, network operators should have no ncentve to adjust capactes. An addtonal unt of capacty, produced ether for wholesale or retal supply, would ncur larger margnal costs than margnal revenues. Due to ths ratonale, network operators wll not nd themselves caught n a prsoners dlemma of wholesale capacty supply to MVNOs. Each network operator s aware that ts counterpart could not pro tably supply too much wholesale capacty to the MVNO as ths would cannablse ts own pro ts n the retal market. Equlbrum wholesale prces Substtuton of (14) and (15) nto (9) yelds the wholesale prce determned by the network operators equlbrum retal and wholesale supply, k and k v, w(k ; k v ) = a (k + kj ) (n + 1)(k v + kj v ) (24) = 1 (a + 2c): 3 Clam 5 The equlbrum wholesale prce, w(k ; k v ), s equal to the network operators retal prces and hence constant n the number of MVNOs, n, and the degree of servce substtutablty,. The ntuton behnd Clam 5 s that network operators would strctly prefer to make costly capacty nstallatons for that busness, wholesale or retal, that yelds hgher returns n terms of hgher prces. Hence as long as w > P ; network operators would only nstall wholesale capacty, whereas, f w < P, network operators would only nstall retal capacty. In any equlbrum wth postve wholesale and retal capactes the wholesale and the retal prce must be equal, whereby the retal prce, as explaned above, s smply the market prce one would derve from standard Cournot analyss. 10 mply (22). 10 It remans to be shown that q > k. By (5), (14) and (24) ths amounts to showng that a(1 + n(1 )) + nw 3 + n(3 2 2 ) > (a c)(1 + (1 )n) 3 + 3n(1 2 ; ) 16

17 3 Non-techncal summary We have presented a smple formal model to analyse the compettve e ects of MVNOs. Our analyss provdes the followng answers: Network operators retal prces are not a ected by the mere exstence of MVNOs. Fgures 1 a) and b) dsplay retal and wholesale prce levels n a market for moble communcaton servces. In partcular Fgure 1 a) consders prce levels as a functon of servce substtutablty between network operators and MVNOs. A hgher (lower) degree of servce substtutablty means that end-customers regard the network operators and MVNOs retal o ers as more (less) smlar servces to each other. Fgure 1 b) consders prce levels as a functon of the number of MVNOs. It can be seen from a) and b) that the network operators (MNOs ) retal prces are nether a ected through the degree of servce substtutablty between MNOs and MVNOs nor through the number of MVNOs. Network operators equalse the prce for hostng MVNOs wth ther retal prce. Fgures 1 a) and b) also show that network operators set ther wholesale prce equal to ther retal prce. Keepng n mnd that the necessary nfrastructure s costly both for wholesale and retal o ers, t s clear that network operators prefer to nvest n a busness that has the hgher pro ts. Ths ratonale causes wholesale and retal prces to equalze n equlbrum: as long as the retal prce s hgher than the wholesale prce, network operators ncrease ther retal supply and decrease ther wholesale supply whch causes the retal prce to decrease and the wholesale prce to ncrease (and vce versa). where w = 1 (a + 2c): 3 Substtuton for w and rearrangng terms gves (1 + n)[a( 1 + )n + c( 3 + ( )n] 3( 1 + ( )n)( 3 + ( )n) whch s true snce the denomnator s strctly negatve and the numerator s non-postve. 17 > 0

18 a) Servce Substtutablty and Prces b) Number of MVNOs and Prces P w MNO Retal Prces MVNO Retal Prces P w MNO Retal Prces MVNO Retal Prces Wholesale Prces Wholesale Prces 0 1 (Independent (Perfect Servces) Substtutablty) 1 n 20 (No Competton (Intense Competton among MVNOs) among MVNOs) nk v c) Servce Substtutablty and Wholesale Supply nk v d) Number of MVNOs and Wholesale Supply Wholesale Capacty/supply Wholesale Capacty/supply (Independent Servces) (Perfect Substtutablty) 1 n 20 (No Competton (Intense Competton among MVNOs) among MVNOs) Fgure 2: Equlbrum Results 18

19 MVNOs set hgher retal prces, the more d erentated ther servce s to the network operators servces. Fgure 1 a) ndcates that MVNOs retal prces fall f ther servces are more smlar to those of network operators. The ntuton here s that a hgher degree of d erentaton ncreases MVNOs market power and hence ther ablty to ncrease prces. In contrast, f MVNOs o er a perfect substtute to the network operators servces, ther retal prce falls to the network operators retal and wholesale prce respectvely. In turn there remans no margn for the MVNO. MVNOs set lower retal prces, the hgher the number of MVNOs. Fgure 1 b) vsualzes the e ect of an ncrease n the number of MVNOs on MVNOs ablty to set retal prces above the wholesale prce. As one would expect, f the number of MVNOs ncreases ths causes ther retal prce to fall to the compettve level. Network operators decrease ther wholesale supply f MVNOs servces consttute closer substtutes to ther own servces. Fgure 1 c) dsplays the network operators wholesale supply as a functon of the substtutablty between the network operators servces and the MVNOs servces. In partcular, there exsts no wholesale o er at all, f servces are perfect substtutes. Ths s consstent to Fgure 1 a), where, n the case of perfect substtutablty, MVNOs would not be able to earn a postve margn. Network operators ncrease ther wholesale supply f the ntensty of competton between MVNOs ncreases. Fgure 1 d) vsualzes ths e ect. If competton between MVNOs ncreases, ther retal prce approaches the compettve level (.e. the wholesale prce, see Fgure 1 b)). Ths means that network operators no longer have to share the margns wth MVNOs and hence makes wholesale o ers more attractve. Havng sad that, Fgure 1 d) also ndcates that overall wholesale supply (asymptotcally) approaches the maxmum level, regardless of addtonal MVNOs. Clearly the overall wholesale supply s stll lmted by the extent that MVNOs wll not 19

20 pose a compettve constrant on the MNOs retal supply. 4 Concluson Our paper clar es several e ects related to the ntroducton of MVNOs. In our model, absent regulaton, MVNOs do not provde a compettve constrant on MNOs. MVNOs that have smlar busness plans to MNOs are rejected by them. MVNOs wth d erentated busness plans do not exert compettve pressure on them. MNOs make sure that they balance ther o er of supply nely wth the degree of substtutablty of the MVNOs busness model. 20

21 References [1] Boots, G.,M., Rjkers, F.A.M, Hobbs, B.F. (2004), Tradng n the Downstream European Gas Market: A Successve Olgopoly Approach, Energy Journal, 25, [2] Brto, D., Perera, P. (2005), Moble Vrtual Network Operators: A Vrtual Prsoner s Dlemma?, Workng Paper. [3] Dxt, A. (1980), The role of nvestment n entry-deterrence, Economc Journal, 90, [4] Economdes, N. (2005), The Incentves for Vertcal Integraton, NET Insttute Workng paper # [5] Greenhut, M.L., Ohta, H. (1979), Vertcal Integraton and Successve Olgopolsts, Amercan Economc Revew, 69, [6] Hastng, J.S., Glbert, R.J. (2005), Market Power, Vertcal Integraton and the Wholesale Prce of Gasolne, Journal of Industral Economcs, LIII, [7] McAfee, R.P. (1999), The E ects of Vertcal Integraton on Competng Input Supplers, Federal Reserve Bank Cleveland Economc Revew, 35, 2-8. [8] Parker, P.M., Röller, L.-H., Collusve conduct n duopoles: multmarket contact and cross-ownershp n the moble telephone ndustry, Rand Journal of Economcs, 28, [9] Ofcom (2006), The Communcatons Market, Interm report, February 2006, [10] Ordover, J.A., Saloner, G., Salop, S.C. (1990), Equlbrum Vertcal Foreclosure, Amercan Economc Revew, 80, [11] Salnger, M.A. (2005), Vertcal Mergers and Market Foreclosure, Quarterly Journal of Economcs, 103,