Course Hero - We put you ahead of the curve!

You have requested the below document.

Coursehero >> Michigan >> Michigan >> LAW 625

Regulation Access and the Adoption of VoIP1 Paul W. J. de Bijl2 Tilburg University Martin Peitz3 International University in Germany and University of Mannheim 31 May 2006 1 Financial support from ENCORE (Economics Network for Competition and Regulation) is grate- fully acknowledged. We would like to thank Georg G tz; the Dutch telecommunications and postal authority OPTA, in particular, Jorn van Steenis and Elbert Jan van Veldhuizen; and seminar participants at OPTA, University of Antwerp, Tilburg University, and the 2006 meeting of the Industrial Economics Section of the Verein f r Socialpolitik for helpful comments and discussions. Only the authors are responsible for the contents of this paper and any errors it may contain. 2 TILEC (Tilburg Law and Economics Center) and CentER for Economic Research, Tilburg University, PO Box 90153, 5000 LE Tilburg, Netherlands, also a liated with the ENCORE Fellows Network; pdebijl@uvt.nl. 3 School of Business Administration, International University in Germany, Campus 3, 76646 Bruchsal, Germany, also a liated with ENCORE and CESifo; Martin.Peitz@i-u.de. Abstract The introduction of packet-switched telephony in the form of VoIP raises concerns about current regulatory practice. Access regulation has been designed for traditional telephony on PSTN networks. In this paper we analyze the e ect of access regulation of PSTN networks on the adoption of a new technology in the form of VoIP. In particular, we show that with endogenous consumer choice between PSTN and VoIP telephony, higher prices for terminating access to the PSTN network make VoIP less likely to succeed and lead to lower pro ts of operators that o er exclusively VoIP telephony. JEL-Classi cation: L96, L51, L13. Keywords: telecommunications, voice over broadband (VoB), voice over Internet protocol (VoIP), entry, access, regulation, imperfect competition. 1 Introduction With the emergence of voice telephony based on the Internet protocol, generally known as voice over IP or VoIP, the telecommunications landscape is rapidly changing. This new technology, which is fundamentally di erent from telephony over the PSTN (public switched telephone network), is providing a new impetus to local loop unbundling (LLU) and also stimulates entry into telephony markets by cable operators. Without even mentioning software applications for voice telephony that run completely over the Internet, it is clear that incumbent operators are facing a serious threat. In particular, they face the question of whether they should milk the PSTN as long as possible, or introduce VoIP quickly and at low prices, at the cost of cannibalizing PSTN revenues, in the hope of at least partially deterring entry of new operators. In this paper we explore a situation of imperfect competition between an incumbent and an entrant. While the incumbent, with a history in PSTN telephony, is assumed to have a complete local access network, the entrant is either a cable operator with a full-coverage broadband network or a newcomer who uses LLU to reach end-users. The incumbent o ers PSTN (public switched telephone network) voice telephony to one segment of customers, as well as VoIP services to another segment, while the entrant only o ers VoIP services in the latter segment. We distinguish two set-ups. In the rst one, the relative size of these segments is exogenously given, so that there is no migration from PSTN to VoIP services. In the second set-up, we allow for endogenous migration between the segments, so that consumers actually choose between staying with the PSTN network versus adopting VoIP services.1 In this set-up, we explore the nature of competition in a market for voice telephony between an incumbent trying to balance its tactics with regard to PSTN and VoIP telephony, and an entrant without ties to the past. In addition, we explicitly focus on the e ects of regulation on the market. Accordingly, there is a close link to regulatory practice. National regulatory authorities (NRAs) are currently struggling with the question whether they should restrict the incumbent s activities with regard to VoIP, or refrain from intervening so that the market will determine how this new technology will develop. Regulation may be necessary in order to prevent anticompetitive behavior, but on the other hand, intervening may easily distort the development and adoption processes of innovation. 1 In both cases, we assume that there is full coverage, that is, all consumers make a purchase. See De Bijl and Peitz (forthcoming) for an analysis of partial market coverage in a setting that focuses on unbundling rather than terminating access. 1 In this paper, we mainly focus on regulation of terminating access. We will also discuss the e ects that the retail price in the market for PSTN telephony may have on market outcomes in the VoIP segment. Accordingly, we will not so much be considering regulation of the VoIP market itself. Instead, we look at the broader regulatory picture, which may partly be motivated by considerations of a universal service obligation in the PSTN market, or by market power with regard to call termination on the PSTN. Within the European regulatory framework for communications markets, such considerations which may be legitimate in a relatively isolated context can easily trigger regulatory interventions, such as regulation of the incumbent s access price. However, because of network interconnection, regulators should be aware of the e ects that they may have on emerging markets. For a recent policy document on this, see OECD (2006).2 Our aim is to articulate some of the most salient side e ects of current regulatory interventions. The assumption of imperfect competition means that both operators have some market power; such a situation is more realistic than the more stylized set-up with a competitive fringe that needs to purchase the essential input from the incumbent. In practice, entry immediately tends to generate some discipline on incumbents. In such a situation, supposing that VoIP operators use bill-and-keep or some other predetermined scheme for call termination, we analyze the competitive e ects of terminating access at the PSTN network. In particular, we consider access that is not priced at its underlying marginal cost level. We will summarize our results in the concluding section of the paper. Literature review. Terminating access in telecommunications networks has been recently analyzed in situations in which operators need mutual access. This literature on two-way access includes the seminal papers by Armstrong (1998) and La ont, Rey and Tirole (1998); overviews are provided by La ont and Tirole (2000), Armstrong (2002), De Bijl and Peitz (2002), and Vogelsang (2003). Our paper builds on that literature by analyzing the emergence of VoIP networks in a PSTN environment, and in which the PSTN and VoIP networks are interconnected. In such an environment, metering of call tra c may still make sense on the PSTN, but this may be di erent for VoIP calls. Hence the natural starting point is to consider the case in which VoIP operators do not charge for call termination. Note that technically speaking, we look at a two-way access problem. However, when VoIP operators do not charge for access, our set-up boils down to a one-way access problem. Hence 2 We will see later that some of the views in OECD (2006), on the link between the PSTN access price and the intensity of competition in the VoIP market, may lead to the wrong conclusions. 2 our paper directly connects to the literature on one-way access, which has typically focused on access problems in which either all rms or at least the downstream entrants do not have market power in the retail segment. In the former case there is pure Bertrand competition, and in the latter case, downstream entrants form a (perfectly) competitive fringe. In such settings, when the incumbent s retail price is assumed to be xed, the e cient component pricing rule (ECPR) has been proposed as the socially optimal pricing method for setting the incumbent s access price (see Baumol, 1983, and Willig, 1979). The ECPR says that to obtain productive e ciency, the access price should be equal to the marginal cost of access plus the incumbent s opportunity cost of providing access (the incumbent s lost pro ts due to entry). If the incumbent and the entrant o er perfect substitutes, this rule reduces to the margin rule according to which the incumbent s lost pro t equals its lost retail revenues (Armstrong, 2002). Accordingly, the vertically integrated incumbent can increase its pro ts by granting access to a more e cient downstream entrant. Since foreclosure is unpro table if an entrant is more e cient, there is no need to regulate the incumbent s access price. However, the logic behind the ECPR has been challenged in several adaptations and extensions; see Armstrong (2002) for a thorough overview. Access pricing in a situation of imperfect competition (with regulated access prices) has been analyzed by La ont and Tirole (1994, 1996), Armstrong and Vickers (1998), Lewis and Sappington (1999) and De Bijl and Peitz (forthcoming), among others. La ont and Tirole (1994) analyze the implementation of Ramsey prices via a global price cap. Apparently, imperfect competition complicates the appropriate use of a global price cap, because supply-side and demand-side e ects have to be taken into account simultaneously. In a full information world, in which only the access price is regulated, the idea is to replace retail regulation by competition. Here, access prices can be used as a regulatory instrument to a ect retail price levels. If, in particular, the regulator can set di erent rates for a bottleneck owner and a nonintegrated competitor, the regulator may want to subsidize the competitor at the margin to increase competitive pressure (Ebrill and Slutsky, 1990; and Lewis and Sappington, 1999). In addition, in an asymmetric market the regulator may want to use the access price to favor the more e cient rm (Armstrong and Vickers, 1998; Lewis and Sappington, 1999; and De Bijl and Peitz, forthcoming). The reason behind such a policy is that the more e cient rm would otherwise obtain a market share that is less than socially optimal. Note that such a bias becomes less pronounced in a more competitive market (Lewis and Sappington, 1999). Foros (2004) analyzes a competitive situation which is somewhat related to ours. To consider the retail market for Internet access, he models a situation of a vertically integrated 3 rm controlling both local access and providing broadband access, and a downstream Internet retailer. The integrated rm can invest in the capacity of local connections, and given the outcome of that decision, the regulator chooses an access price. The rms compete in a Cournot fashion in the retail market. The focus of the paper is mainly on regulation as a way to induce the integrated rm to invest e ciently and to deter it from foreclosing the market. Hence, Foros results can be seen as complementary to ours. The economics literature has also looked at bypass possibilities (see e.g. Armstrong, Doyle and Vickers, 1996; La ont and Tirole, 1994, 1996). Note, however, that although VoIP can substitute PSTN telephony, it does not allow for full bypass as long as some consumers stay with the PSTN. In our model, with full penetration of VoIP the access problem has disappeared whereas at any intermediate situation, terminating access to PSTN remains essential and a bypass possibility is not available. With respect to our modeling framework, this paper also contributes to the literature on multiproduct rms in a multidimensional product space. This literature has been reviewed in Manez and Waterson (2001).3 In this paper we provide a tractable model of multiproduct competition which allows for endogenous formation of market segments and can be solved analytically. However, our model has the property that integration does not a ect the incentives with respect to prices (in the VoIP segment). Our framework may prove to be useful for other applications in industrial organization. The structure of this paper is as follows. Section 2 provides some illustrative background information on terminating access in relation to VoIP. In section 3, we explore a model in which the group of consumers purchasing VoIP services is exogenously given. Section 4 analyzes the case in which migration from PSTN to VoIP services is endogenously determined. Section 5 concludes the paper and summarizes the results. 2 Terminating access and VoIP in practice In this section, we provide some background on IP-based services, wholesale access to local access networks, and call termination. Also, based on this background information we clarify the focus of our analysis, as it is beyond the scope of this paper to provide a comprehensive analysis of VoIP. We abstract from LLU regulation in order to focus on regulation of 3 Contributions include Katz (1984), Lal and Matutes (1989) and Canoy and Peitz (1997). 4 terminating access.4 2.1 IP-based telephony The Internet Protocol (IP) is a data protocol, based on packet switching rather than circuit switching, that is used for routing and carriage of messages over the Internet. As any type of electronic information can be transported in packets, IP can also be applied to transport voice calls. The number and variety of IP-based telephony service propositions are large and increasing, and terms like VoIP, IP telephony and Internet telephony are often used interchangeably. In order to clarify the focus of our paper, it is useful to recapitulate the main types of IP-based telephony: 1. IP-based transport in the core of traditional networks: At the level of long-distance backbones, PSTN operators have been supplementing and replacing traditional circuitswitched technology with IP-based technology. To reach end-users, they may still use traditional (i.e., not upgraded to DSL) copper wires connections. 2. IP-based transport at the edges of traditional networks, allowing for IP-based o erings from traditional operators: PSTN operators may upgrade their local connections to digital subscriber lines (DSL), enabling broadband Internet access.5 Such connections may also be used to o er VoIP telephony. This type of VoIP is also know as voice over broadband (VoB) or voice over DSL (VoD). Operators may o er Internet access and VoIP services as a bundle. 3. IP-based o erings from cable operators: Cable operators may adapt their local lines so that they can carry high-speed two-way tra c, enabling broadband Internet access as well as VoIP telephony. Again, these o erings may be sold as a bundle. 4. IP-based o erings from entrants without local networks base on LLU: If the incumbent s local network is unbundled, entrants without their own local loops can lease unbundled local lines from the incumbent and o er broadband Internet access or VoIP services to 4 For an overview of the development of LLU throughout Europe and the European regulatory framework, see De Bijl and Peitz (2005). 5 DSL is a technique that increases the available frequency spectrum on copper wires, so that more data can be sent through a line. 5 end-users. This type of VoIP is also know as voice over broadband (VoB) or voice over DSL (VoD). 5. IP-based o erings from entrants without local networks based on bitstream access: Entrants without their own local loops can purchase bitstream access from the incumbent or an entrant using the incumbent s unbundled DSL connections, and o er broadband Internet access or VoIP services to end-users. 6. IP-based o erings through an ISP: End-users, subscribing to an ISP, may purchase VoIP from or via that ISP. 7. Fully IP-based next generation networks : Operators may roll out new networks or upgrade existing ones to create completely IP-based networks. An example is BT s 21st Century Network. 8. VoIP over the Internet: This is IP-based telephony that is purely Internet-based. Consumers with Internet access can download free, peer to-peer based, voice telephony software, enabling them to make free calls to consumers with the same software installed on their computers (computer-to-computer calls). A well-known example of this software is Skype. Calls to subscribers of other telecoms networks (computer-to-phone calls) are also possible, although they may be charged, as termination on other networks may be costly. 9. IP-based private branch exchanges (PBXs): Corporate customers may, for in-house telecommunications services on their local and wireless access networks (LAN and WAN), use IP-enabled switches. Traditionally, in-house switches were circuit-switched. Note that with an IP-PBX, calls to the outside world may be transformed into circuitswitched calls, depending on the nature of the network that a customer subscribes to. Experts expect that the current variety will remain for the foreseeable future, if only because of the wide diversity in the ways that end-users have access to services.6 In this paper, we restrict our focus to competition between an incumbent o ering IPbased services (type 2 or 7) and an entrant with or without a local network (type 3, 4 or 5). Note that for type 7 (the incumbent upgrading to a next generation network), we assume in our analysis that it is still in the process of upgrading its network from a PSTN to an all-IP 6 Ofcom (2006, p. 8-9). 6 network, so that during this transition, it o ers both PSTN and IP-based telephony. In case of a LLU-based entrant (type 4), we abstract from problems associated with the setting of the wholesale lease price of local loops. In our analysis, we assume that if a customer switches to a LLU-based entrant, he or she completely substitutes the PSTN service with the entrant s VoIP service. Hence our analysis captures naked DSL (also known as standalone DSL), a service proposition in which an entrant provides only a broadband Internet connection based on DSL (typically priced at a at rate) by leasing only the broadband part of the frequency spectrum of the copper wire. Accordingly, the narrowband part of the line is no longer used. Finally note that in our models, we implicitly allow for IP-based backbones (type 1), as we do not specify the nature of long-distance backbones. 2.2 Terminating access Public telecoms networks, whether PSTN or IP-based, must interconnect with one another, so that users can be reached irrespective of the network that they subscribe to. The process that makes this possible, network interconnection, consists of the mutual provision of terminating access. Traditionally, operators charge each other for call termination. Charging for call termination is typically done on a per-minute basis. An alternative to charging for access is bill-and-keep (or reciprocal settlement-free termination ), a system in which calls are terminated without access payments between operators. The emergence of VoIP may radically change operators wholesale deals on call termination.7 With IP-based telephony, the rationale behind termination charges is undermined, as the marginal cost of call termination is drastically reduced, and VoIP calls are often not metered anymore. Nevertheless, calls from an entrant s VoIP network to the incumbent s PSTN network are delivered at a traditional circuit-switched interconnection point or through a gateway , which allows for straightforward identi cation of incoming calls, and, hence, for termination charges. Accordingly, for calls from an IP-network to a PSTN, a VoIP provider may have to pay for call termination. Such charges create a perceived marginal cost for the VoIP provider, which possibly translates into a strictly positive per-minute price for this type of call. In the case of calls from one IP-based network to another, operators may nd it more e cient to implement bill-and-keep, in line with the packet-based nature of VoIP that, to a certain extent, eliminates the logic of metering incoming calls. In the basis case of our analysis, we suppose that the incumbent charges for call termina7 See Analysys (2004) for an overview of possible business models for call termination. 7 tion on its PSTN, and that no termination charges are used for other types of calls. This is in line with the observation that the marginal cost for termination at the PSTN is typically seen as being strictly positive, whereas call termination on IP-based networks comes virtually without a cost. It is straightforward to consider di erent wholesale pricing schemes in our models, though.8 3 3.1 Exogenously given consumer segments The model There are two rms, an incumbent (operator 1) and an entrant (operator 2). The incumbent is assumed to have a complete local access network. The incumbent s network can be used for PSTN-based telephony as well as IP-based telephony (VoIP). For instance, its local connections have been upgraded to allow for Digital Subscriber Line (DSL) technology, and its (long-distance) backbone to an IP-based network. The entrant uses only IP-based technology to o er voice services. The entrant may be a cable operator with a full-coverage broadband network. Alternatively, it may be using LLU to reach end-users, that is, it leases unbundled local connections from the incumbent. In the latter case, we assume that the line rental of the local loop is regulated at a cost-based level, so that the entrant is on an equal footing as the incumbent. This assumption allows us to abstract from regulatory issues that stem from LLU, and focus solely on terminating access. Consumers are heterogeneous with respect to their reluctance to use a new rather than an established technology (with some abuse of language, we will sometimes refer to consumer groups as old and new segments). The incumbent o ers PSTN-based voice telephony to customers with little technological savvy (the old segment), as well as VoIP to the new segment which is open to a new technology, while the entrant only aims at the latter segment by o ering VoIP. The group of old consumers is of size 0 and the other consumer group of size . The total size of the market is normalized to 1, so that 0 + = 1. More precisely, there is a continuum of consumers with mass 1. A possible interpretation is that consumers in the old segment are narrowband users, whereas consumers in the new segment are broadband users. 8 Such a structure of access prices may be a good approximation of the outcome of negotiations between VoIP and PSTN operators. As stated in an OECD report (OECD, 2006, p. 26), ...it seems likely in reality that VoIP operators might not charge PSTN operators for IP termination while PSTN operators would still charge VoIP operators for the same call in the opposite direction, due to the VoIP providers weaker negotiation power. 8 Throughout this section, we assume that consumers cannot migrate from one segment to the other, while the segment sizes are exogenously given (this assumption will be dropped in the subsequent section). All networks are interconnected, so that any consumer can make calls to any other consumer. To allow for a call from one operator s network to the other s network, the rst operator must purchase a wholesale service called terminating access from the second one. We assume that the marginal cost of call termination on a VoIP network is 0,9 and that operators do not charge for call termination to a customer subscribing to a VoIP service. This is in line with the tendency of VoIP providers to use bill-and-keep arrangements for call termination, and with the fact that interconnection typically has already been settled at the underlying level of Internet service providers. The marginal cost of call termination on the PSTN network is c > 0, and the incumbent charges a termination charge a for call termination to its PSTN customers. To keep the number of parameters small without loss of generality, we set all other costs equal to zero. Access price a is set by a regulator.10 Since we do not explicitly model the regulator as a player, access price a is an exogenous parameter in our model. By supposing that all consumers have an identical, inelastic demand to make calls once they have a subscription, each consumer will make a given number of calls. Without loss of generality, we normalize this number to 1.11 The retail price in the old segment is assumed to be given by p0 . For instance, it is set by the regulator or it may be determined by the presence of a competitive fringe in PSTN telephony (e.g. carrier-select based competitors competing on price).12 Thus we can treat p0 as a parameter. In the new segment, the operators compete by setting at fees. Operator i s retail price for VoIP telephony is denoted by pi , i = 1, 2. Note that implicitly, all per-minute 9 The true marginal costs of electronic communications are virtually zero. Nevertheless, in practice, operators allocate xed costs to tra c, and hence may partly treat these costs as marginal costs when setting prices. Thus, what we call the marginal cost of call termination is in fact the tra c-dependent cost of call termination. These costs are substantially lower for IP networks than for PSTN networks, and therefore we set them at 0 for VoIP calls. 10 For instance, the regulator has determined that the incumbent has signi cant market power (SMP) in the wholesale access market, and because of that, and in line with the regulatory framework that is in place, applies price controls (this illustration corresponds to the situation in EU member states). 11 See De Bijl and Peitz (2002) for a more elaborate speci cation. 12 Even if no actual retail regulation is in place the threat of regulatory intervention may e ectively be re ected by an upper bound on the PSTN retail price. When this upper bound is binding our actual analysis applies. 9 prices are 0. In a more elaborate model, one could incorporate that consumers have elastic demand to make calls, or to have access to the Internet, in addition to the demand for a subscription. Such extensions lead to additional interactions between the operators, for instance because there is call tra c between the networks see De Bijl and Peitz (2002) for an inclusion of call tra c. Nevertheless, the present model is rich enough to capture some crucial elements of the strategic interaction between PSTN and VoIP providers.13 When a consumer makes a call, the receiving consumer may be any other consumer with equal probability, independent of the network they are subscribed to. This implies that calling patterns are balanced, that is, the volumes of on-net and o -net calls are proportionate to market shares. This assumption, which is common in the literature on competition in telecommunications markets, simpli es the analysis and should be seen as the natural benchmark. Market shares in the segment of the new technology depend on the retail prices, and are denoted by si (p1 , p2 ), i = 1, 2. We assume that an operator s market share is decreasing in its own price and increasing in the price of its rival. Furthermore, we assume that market preferences when consumers have identical demand functions. With full participation, total model. The property that market share changes continuously with price implies that rms have market power. Consumers do not consider the services provided by the two rms as perfect substitutes and therefore do not necessarily go for the lowest price. In reality, imperfect substitutes seem to be common in telecommunications (as well as other services markets), for instance due to heterogeneity in brand recognition, corporate images, and consumer switching costs. Also, services o ered by operators are o ered in di erent bundles with other services: if the bundles are not the same, they will be considered as imperfect substitutes. 13 shares only depend on the price di erence p2 p1 . This assumption is satis ed for quasi-linear market demand is xed. For an example see below. Figure 1 illustrates the set-up of the the present model may be seen as an approximation of a model with positive usage charges but rather inelastic demand. The demand assumptions simplify the analysis considerably and allows us to focus on participation decisions, abstracting from usage intensity. In the light of the enormous variety in non-linear contracts, the case of at fees in a world of simple demand structures provides a natural benchmark. Traditionally, operators have set two-part tari s for PSTN telephony, while at present, operators seem to be inclined to set at fees for VoIP services (possibly combined with linear prices for calls terminating on the PSTN). Related to the emergence of IP-based telephony, there seems to be a trend towards at fees for all voice telephony services. 10 consumers with old technology subscription access (at price a) operator 1 operator 2 subscription 1 subscription 2 consumers with new technology (discrete choice) Figure 1: Illustration of the model. Pro t functions are as follows. Firm 1 s pro ts can be written as 1 (p1 , p2 ; a, p0 ) = 0 [p0 0 c] + [s1 (p1 , p2 )(p1 0 c) + s2 (p1 , p2 ) 0 (a c)], and rm 2 s pro ts as 2 (p1 , p2 ; a) = s2 (p1 , p2 )(p2 0 a). (2) (1) These pro t functions re ect the volumes of on-net and o -net tra c between operator 1 s PSTN network and both operators VoIP networks volumes that are proportionate to market shares as well as the wholesale payments for calls terminating on the PSTN network. A special case of our general model is obtained by assuming that the networks are horizontally di erentiated. Suppose, for instance, that consumers are uniformly distributed on the interval [0, 1]. Firm 1 is located at location y1 = 0 on the interval, and rm 2 at y2 = 1. A consumer located at z buying from rm i incurs a disutility |yi z|. Note that a higher value of parameter corresponds to more di erentiation between the networks. A consumer 11 at z buys from rm 1 if v1 (p1 , p2 ) z > v2 (p1 , p2 ) (1 z), where vi (p1 , p2 ) denotes the si (p1 , p2 ) = 1 2 conditional indirect utility of a network at the ideal location z. Market shares then satisfy cation which has also been widely used in models on two-way access (see e.g. La ont, Rey and Tirole, 1998; Armstrong, 1998; and the survey by Armstrong, 2002). Structure of the game and equilibrium: The structure of the model is as follows: t = 0: The regulator sets access price a and retail price p0 , or alternatively, the latter price + (vi (p1 , p2 ) vj (p1 , p2 ))/(2 ), where j 6= i. This is a simple Hotelling speci - is determined by a competitive fringe in the retail price for PSTN voice telephony. t = 1: Operators choose their prices for VoIP voice services in order to maximize pro ts. t = 2: Consumers observe retail prices and make purchasing decisions, based on utility maximization. Consequently, market shares and pro t levels are realized. We are interested in a Nash equilibrium in prices (p , p ), which is de ned in such a way 12 that given its rival price, neither rm has an incentive to change its own price. That is, each operator s price p maximizes pro ts i (p1 , p2 ; a, p0 ) when p is given, i 6= j. Accordingly, i j can be written as max 1 (p1 , p ; a, p0 ), 2 p1 given the equilibrium price of the competitor, the pro t maximization problem of operator 1 (3) while operator 2 maximizes max 2 (p , p2 ; a, p0 ). 1 p2 (4) 3.2 Analysis Suppose that there exists a unique pair (p , p ) which solves problems (3) and (4) simultane12 ously (hence it constitutes an equilibrium). We are then interested in which way a change in regulatory policy with regard to the access price a a ects market outcomes. Hence, consider the following increase of the termination charge: a0 = a + a, where a > 0. We can then show that this increase is passed through to consumers. Market shares in equilibrium, as well as the entrant s pro ts, are una ected. However, we will see that the incumbent bene ts in two ways: (i) it can charge a higher mark-up in the retail market, and (ii) it receives higher revenues from calls that terminate on its PSTN network. Consumers in the new segment are worse o , as the they face higher retail prices by both networks. We will now explore the underlying mechanism in detail. 12 Given the new access price a0 , we claim that equilibrium retail prices are p = p + 0 a 1 1 and p = p + 0 a. Our proof consists of establishing that for each operator i, p is the 2 2 i solution of the maximization problem of operator i. Operator 1: Given the new access price a0 , the incumbent s pro t can be written as 1 (p1 , p2 ; a0 , p0 ) = 0 (p0 0 c) + [s1 (p1 , p )(p1 0 c) + (1 s1 (p1 , p )) 0 (a + a c). 2 2 Provided that the competing operator sets p = p + 0 a, the incumbent s market share 2 2 satis es s1 (p1 , p ) = s1 (p1 0 a, p ) because they only depend on price di erences. Hence 2 2 operator 1 s pro t can be rewritten as 0 (p0 0 c) + [s1 (p1 0 a, p )(p1 0 a 0 c) + 0 (1 s1 (p1 0 a, p ))(a c) + 0 a]. 2 2 With a change of variable p1 p1 0 a, the incumbent s maximization problem becomes e p 2p p2 max 0 (p0 0 c) + [s1 (e1 , p )(e1 0 c) + 0 (1 s1 (e1 , p ))(a c)] + 0 a p1 (5) Clearly, p is the solution to this problem because, apart from the constant 0 a, it is 1 exactly the same as problem (3). Since p1 p1 0 a we have shown that p = p + 0 a, e 1 1 to an increase in the incumbent s pro ts by 0 a. Operator 2: Given the new access charge a0 , operator 2 s pro t can be written as 2 (p1 , p2 ; a0 ) = s2 (p , p2 )(p2 0 (a + a)) 1 Provided that the competing operator sets p = p + 0 a, the entrant s market share 1 1 the maximization problem of the non-integrated network s pro t can be written as max s2 (p , p2 )(e2 0 a) 1e p p2 provided that p = p + 0 a. Moreover, notice that the increase in the access price leads 2 2 satis es s2 (p , p2 ) = s2 (p , p2 0 a). Hence, using the change of variable p2 = p2 0 a e 1 1 (6) Clearly, p is the solution to this problem because it is equivalent to problem (4). Since 2 p2 p2 0 a we have shown that p = p + 0 a, provided that p = p + 0 a. e 2 2 1 1 Hence, we have established the following result: Result 3.1. Consider an increase in access price for call termination on the PSTN network. As a consequence, the incumbent s pro ts increase, while the entrant s pro ts are una ected. 13 Both operators pass on the access price increase to consumers by charging a higher retail price for VoIP telephony. Market shares remain the same. The result says that rents are redistributed from consumers using the new technology to the bottleneck owner of the old technology. It is instructive to take another look at the above result. An access price increase by a works a ects prices in the same way as a per-user cost increase of the new technology (of magnitude 0 a). This can be seen as follows. The pro ts of operator 2 are equal to s2 (p1 , p2 )(p2 0 a + 0 a), that is, the pro t function has the same form as with access price a and costs 0 a. The pro t function of operator 1 becomes 1 (p1 , p2 ; a + a, p0 ) = 0 (p0 c) + [s1 (p1 , p )p1 + 0 (1 s1 (p1 , p ))(a + a) 0 c]. 2 2 The pro t-maximizing price p1 when p2 is given, is determined by the rst-order condition of pro t maximization: s1 (p1 , p2 ) s1 (p1 , p2 ) p1 0 (a + a) + s1 (p1 , p2 ) = 0, p1 p1 which is equivalent to s1 (p1 , p2 ) s1 (p1 , p2 ) (p1 0 a) 0 a + s1 (p1 , p2 ) = 0. p1 p1 This equation is also the rst-order condition of pro t maximization given access price a and per-used costs 0 a for the new technology. Hence, a access price increase for accessing the old technology is passed on to consumers (which increases the expected cost of providing service for rm 2 by 0 a) in exactly the same way as a cost increase for a the new technology by 0 a. The only di erence between an access price increase and a cost increase is that the owner of the essential facility, that is, rm 1, bene ts from an access price increase because the associated downstream cost increase generates revenues upstream at the essential facility. All consumers using the new technology su er. In terms of consumer behavior this suggests that providing an access rule that is bene cial to the network owner of the old technology is likely to discourage consumers to move to the new technology. We will return to this issue when we analyze a model with endogenous consumer migration. While a high access price is not desirable for consumers, rm 2 s after entry pro ts are neutral with respect to the access price. Hence, in this simple model, entry incentives are not a ected by the level of the access price. However, for high retail prices (which are due to a high access price), at some point the participation or incentive constraint becomes binding 14 (for some consumers). When that happens, there is no full pass-through of access payments to the consumers of rm 2. Rather rm 2 will have to reduce its pro t margin, making entry less attractive. See also the next section, which explicitly considers the incentive constraint of consumers to adopt the new technology. In the present context an analysis of total surplus is straightforward. Provided that the market is symmetric, the socially desirable market share for each operator is 50%. This is an equilibrium outcome for any access price such that the participation constraint of consumers is not violated (and the technology choice by consumers is exogenous). However, if the market is not fully symmetric, strategic behavior between rms typically does not lead to an implementation of a socially optimal outcome. In particular, if one network is more attractive than the other on average, then the equilibrium market share of the less attractive network is socially excessive.14 4 4.1 Consumers choosing between PSTN and VoIP The model In the previous section, we assumed that a fraction of 0 consumers stay with the PSTNbased technology, while the remaining fraction of consumers adopt VoIP; these fractions were exogenously given ( 0 + = 1). In this section we look at the case in which consumers can decide to switch from PSTN to VoIP. If they do so, they can choose between the VoIP o erings of the incumbent and the entrant. As before, the marginal cost of call termination on the PSTN network is c > 0, while the incumbent charges a termination charge a for call termination to its PSTN customers. Consumers have identical, inelastic demand for one unit of telephony services, while calling patterns are balanced. Market shares in VoIP services are denoted by si (p1 , p2 ), i = 1, 2. We extend on our earlier model speci cation as follows. Consumers utility functions: Consumer tastes are described by types (y, t), uniformly distributed on [0, 1] [0, 1]. The y dimension describes preferences for operator 1 versus operator 2 (or their brands), and the t dimension re ects consumers inclinations towards VoIP versus PSTN. A straightforward interpretation is that y captures consumers loyalty towards operator 1, independent of the service that they purchase. With regard to the other 14 However, in di erence to Armstrong and Vickers (1998) and Lewis and Sappington (1999), the access price does not a ect market shares and therefore is ine ective. See also the model in the following section. 15 dimension of a consumer s type, if a consumers has type t close to 0 this means that he is more inclined to adopt VoIP, whereas a consumer with t close to 1 is rather reluctant to adopt VoIP. The distance between the addresses of the products and consumer types give the disutility of consumers for the particular o erings, as will be speci ed below. VoIP services are located at points (0, 0) and (1, 0), and the PSTN service at (0, 1) (in fact, with a properly adjusted U0 the latter could be any point for which the second coordinate is 1). Note that in our setting y not only plays a role when consumers choose between VoIP services, but also when consumers decide whether purchase PSTN or VoIP services.15 Consumers either subscribe to the PSTN service o ered by the incumbent rm or to one of the two VoIP o erings. A consumer who purchases PSTN services derives utility r + U0 (1 t) y p0 , where r is the basic utility from telephony and U0 R is interpreted as a technology-speci c utility of PSTN-services relative to VoIP services (which may also include the rm-speci c utility, see below). Parameters and measure the degree of heterogeneity among consumers: a large corresponds to a low substitutability between PSTN and VoIP, and a large corresponds to a large degree of di erentiation between the operators. p1 where U1 R can be interpreted as a brand or rm-speci c utility that captures the A consumer who purchases VoIP services from rm 1 derives utility r + U1 t y asymmetry between operators. Similarly, a consumer who purchases VoIP services from rm purchase; hence, parameters are such that there is always a technology available that delivers su cient gross utility. 2 derives utility r t (1 y) p2 . We will implicitly assume that all consumers make a Ful lled expectations: We will be assuming that before consumers learn the prices of VoIP services, they have certain beliefs about the prices that they can expect, and based on these beliefs, they gure out whether to go for VoIP or stick with PSTN. This may correspond to a situation in which consumers before they actively start searching for information about a recently introduced product that they are interested in have already had some exposure to some information about that good, for instance through friends and relatives, articles in newspapers and magazines, and advertisements that they may have passively observed. Hence they are aware of the existence of the product, and, based on the various pieces of information that they have received, they form expectations about its price. Now suppose that based on her beliefs and preferences, a consumer decides that she wants to buy VoIP services. She 15 This does not a ect our results in any important way. 16 will then start searching more actively, in order to learn actual prices. Also, she will make comparison between the incumbent s virtues compared to those of the competitor. We may allow for consumers to return to PSTN if they nd out that VoIP is too expensive compared to its bene ts. However, if beliefs concerning prices are correct, this will not happen. In our model, we do not explicitly incorporate underlying processes of advertising, belief formation and search behavior, but we capture the essence by requiring that in equilibrium, beliefs must be ful lled. Accordingly, consumers make their migration decision, that is, whether or not to adopt VoIP, before VoIP prices are searched.16 An alternative way of understanding this speci cation is to argue that the decision to migrate to VoIP involves a certain level of commitment, as the e ort to make a rst comparison between PSTN and VoIP services has been sunk (and possibly some equipment has been replaced), whereas prices can be adjusted in a more exible way. In other words, due to search and learning costs, consumer migration to the new technology involves more commitment than setting prices.17 Consumers have identical beliefs about VoIP prices. Moreover, since we restrict the analysis to pure strategies, a belief function can be described by a function that, for each rm i, attaches probability 1 to one particular price level pi , and probability 0 to all other b b prices pi 6= pi . To simplify the notation, we will not explicitly de ne these belief functions, b but more simply summarize beliefs by p1 and p2 . b Pro t functions: The pro t functions have the same structure as speci ed in the previous section. Because of endogenous consumer migration, they have to be adapted as follows: bb bb bb 1 (p1 , p2 ; a, p0 ) = 0 (p0 , p1 , p2 )[p0 0 (p0 , p1 , p2 )c] + (p0 , p1 , p2 ) 2 (p1 , p2 ; a, p0 ) = (p0 , p1 , p2 )s2 (p1 , p2 )(p2 0 (p0 , p1 , p2 )a), bb bb [s1 (p1 , p2 )(p1 0 (p0 , p1 , p2 )c) + s2 (p1 , p2 ) 0 (p0 , p1 , p2 )(a c)], bb bb of Structure the game and equilibrium: The model that we analyze then has the following 16 Another possibility would be that consumers decide after observing prices and their taste parameters. While we consider such a speci cation a valid alternative, such a model becomes very cumbersome to work with. 17 An alternative speci cation would be to assume that consumers do have price information before making their migration decision (but do not yet know their taste parameter with respect to operators). This would give the entrant more possibilities to penetrate the market with its VoIP services and would introduce an additional strategic dimension into the problem. Namely, the incumbent could make the VoIP segment on average less attractive by increasing its price. 17 structure: t = 0: The regulator sets access price a and PSTN price p0 , observed by all.18 ual s inclination towards PSTN versus VoIP. All consumers form expectations about VoIP b prices p1 and p2 . b t = 2: Given their preferences and beliefs, consumers decide whether to go for PSTN or VoIP. t = 3: Each consumer learns his or her preference parameter y [0, 1], re ecting an indit = 1: Each consumer learns his or her preference parameter t [0, 1], re ecting an individ- At the same time, the operators (simultaneously) set VoIP prices p1 and p2 . vidual s inclination towards operator 1 versus operator 2. Consumers observe prices p1 and p2 and make purchase decisions, that is, they choose VoIP telephony from the incumbent or from the entrant if they opted for VoIP at t = 2. Otherwise, they choose PSTN telephony from the incumbent. We solve for ful lled expectation equilibrium, that is, (i) each rm maximizes its pro ts while taking consumers beliefs and its rival s strategy as given; (ii) based on their beliefs p1 , b p2 consumers choose the utility maximizing technology. Subsequently, at stage 2, given the b prices set by the rms, they choose the utility maximizing operator provided they adopted the b b prices p and p satisfy p = p1 and p = p2 .19 1 2 1 2 new technology; and (iii) in equilibrium, consumers beliefs are ful lled, so that equilibrium Alternative representation. It is important to note that the set-up with ful lled beliefs is not necessary to keep the model tractable. We would obtain the same results if we solved for subgame perfect equilibria in the game in which stage t = 2 is split into two separate stages: t = 2a: Given their preferences consumers decide simultaneously whether to go for PSTN or VoIP. t = 2b: The operators (simultaneously) set VoIP prices p1 and p2 . In this alternative formulation, one does not need to introduce consumer beliefs about VoIP prices. At stage 2, consumers maximize their utility given the decision of all other consumers. At this stage consumers utility levels depend indirectly on the decision of the other consumers, because these subsequently determine the equilibrium prices that are charged. 18 In an extension of the game, we will later consider the case in which the incumbent chooses the price for the PSTN service. 19 In their seminal paper Katz and Shapiro (1984) solve for ful lled expectations equilibria in a market with network e ects. 18 Surplus levels. To be able to discuss the e ects of regulation on consumer surplus and welfare, we provide the formulas for calculating various surplus levels in the model. The aggregate surplus of PSTN users is equal to: Z 1Z 1 (r + U0 (1 t) y p0 )dydt CS PSTN = 0 1 1 = [(r + U0 p0 )(1 ) + (1 ( )2 )]. 2 2 The aggregate surplus of subscribers to operator 1 s VoIP service is equal to: Z Z s1 (p ,p ) 12 VoIP = (r + U1 t y p )dydt CS1 1 0 0 1 1 = [(r + U1 p )s1 (p , p ) s1 (p , p )2 ] s1 (p , p )( )2 . 1 12 12 12 2 2 The aggregate surplus of subscribers to operator 2 s VoIP service is equal to: Z Z 1 VoIP = (r t (1 y) p )dydt CS2 2 0 s1 (p ,p ) 12 1 1 = [(r p )(1 s1 (p , p )) + (1 s1 (p , p )2 )] (1 s1 (p , p ))( )2 2 12 12 12 2 2 1 1 = [(r p )s2 (p , p ) s2 (p , p )2 ] s2 (p , p )( )2 . 2 12 12 12 2 2 VoIP + CS VoIP Aggregate consumer surplus is equal to CS = CS PSTN + Let CS VoIP = CS1 2 VoIP + CS VoIP . Producer surplus is equal to aggregate pro ts: P S = + . Welfare is CS1 2 1 2 then de ned as the sum of consumer and producer surplus: W = CS + P S. 4.2 Equilibrium analysis We start by looking at consumers choices at the last stage, t = 3, for those consumers who have chosen to adopt VoIP. The consumer who is indi erent between the two VoIP services is parameter y < y subscribe to operator 1 s service, and all others to operator 2. Accordingly, if a fraction demands VoIP services, then the total demand for VoIP o ered by rm 1 is 1 U1 p2 p1 + + . s1 (p1 , p2 ) = 2 2 2 Note that if U1 , operator 2 must price below operator 1 to capture any market share. This corresponds to a situation in which there is vertical quality di erentiation between the two 2 o ers higher quality. located at location y, given by U1 y p1 = (1 y) p2 . All consumers characterized by operators and where operator 1 o ers higher quality. Correspondingly, if U1 operator 19 locations t but they do not yet know their addresses y. Hence, the expected utility of a consumer of type t who intends to migrate to VoIP is as follows: Z s1 (p1 ,p2 ) At t = 2, consumers expect prices p1 and p2 . At this stage, they have learned their b b Z 1 0 [r + U1 t y p1 ]dy + s1 (p1 ,p2 ) [r t (1 y) p2 ]dy pb pb pb = s1 (b1 , p2 )r + s1 (b1 , p2 )U1 s1 (b1 , p2 ) t +s2 (b1 , p2 )r s2 (b1 , p2 ) t pb pb where pb s2 (b1 , p2 )2 s2 (b1 , p2 )b2 p bp 2 = r t s1 (b1 , p2 )(b1 U1 ) s2 (b1 , p2 )b2 [s1 (b1 , p2 )2 + s2 (b1 , p2 )2 ] pbp p bp pb pb 2 = r t 2 pbp p bp pb pb s1 (b1 , p2 )(b1 U1 ) s2 (b1 , p2 )b2 [s1 (b1 , p2 )2 + s2 (b1 , p2 )2 1] 2 ep b = r t p(b1 , p2 ), 2 pb s1 (b1 , p2 )2 p bp s1 (b1 , p2 )b1 2 pared to the average price for VoIP services, it is adjusted in order to take into account the potentially asymmetric utility level U1 as well as the expected reduction in utility from not simpli ed into This function p(b1 , p2 ) will be called the adjusted average price for VoIP services. Comep b pbp p bp pb pb p(b1 , p2 ) s1 (b1 , p2 )(b1 U1 ) + s2 (b1 , p2 )b2 + [s1 (b1 , p2 )2 + s2 (b1 , p2 )2 1]. ep b 2 consuming the ideal product speci cation. It is straightforward to show that p(b1 , p2 ) can be ep b pbp p bp pb pb p(b1 , p2 ) = s1 (b1 , p2 )(b1 U1 ) + s2 (b1 , p2 )b2 s1 (b1 , p2 )s2 (b1 , p2 ). ep b Accordingly, at t = 2, the location t of the consumer who, given his beliefs about VoIP prices, is indi erent between PSTN and VoIP services, is implicitly de ned by r + U0 (1 t) at t < t, is given by bb (p0 , p1 , p2 ) = ep b 1 p0 U0 p(b1 , p2 ) + . 2 2 20 p0 = r t p(b1 , p2 ). ep b 2 2 The expected utility derived from staying with the PSTN service is r+U0 (1 t) p0 . 2 Therefore, the fraction of consumers opting for VoIP services, that is, all consumers located PSTN 1 VoB 1 VoB 2 s1 Figure 2: Segments and market shares for PSTN and VoIP telephony. The fraction of consumers staying with the PSTN network is then, by de nition, equal to bb bb 0 (p0 , p1 , p2 ) = 1 (p0 , p1 , p2 ). Figure 2 illustrates the endogenous segmentation of the market, and the division of the VoIP segment among the incumbent and the entrant. pli es the pro t functions. Note that at this stage, again because expectations are given, e ep b also function p(b1 , p2 ) can be treated as a constant. Thus we write p = p(b1 , p2 ), and for ep b equilibrium at t = 2 is characterized by the following prices: e p1 (p0 , p) = e 3 + U1 a p p0 + U0 ++ a, 3 2 2 e 3 U1 a p p0 + U0 ++ a. 3 2 2 (7) (8) given consumer choices with regard to PSTN versus VoIP, it can be shown that the Nash bb bb At t = 2, for given consumer beliefs, (p0 , p1 , p2 ) and 0 (p0 , p1 , p2 ) are xed. This sim- Some interim observations can be made from equations (7)-(8) under the assumption that result. Brand loyalty or superior performance of rm 1 s VoIP services (U1 > 0) translate into p is xed. Clearly, if the VoIP services are closer substitutes ( smaller), then lower prices e p2 (p0 , p) = e a higher price p1 . Finally, provided that the last term in the pricing equations is su ciently small, a higher access price translates into higher prices. Furthermore, rm 2 s price-cost 21 margin is not a ected by the access price since p2 = 3 U1 3 expectations the neutrality result, which was derived in the previous section, still holds. For given expectations, the present model is a special case of the model analyzed in the previous section. segment translates into lower prices for VoIP services. This is due to the cost e ect that a higher p0 will lead to less demand for PSTN services, which reduces the likelihood that Still given the assumption that p is xed, we also observe that a higher price in the PSTN e + 0 (p0 , p1 , p2 )a. Hence, for given bb subscribers to operator 2 s VoIP service make use of terminating access to the PSTN network. This corresponds to lower perceived costs for operator 2, and hence, a more competitive outcome. The reverse holds for the adjusted average VoIP price p, and for the xed-utility e advantage of PSTN compared to VoIP services, U0 . e b beliefs must be con rmed in equilibrium, that is, the above solution must satisfy p1 (p0 , p) = p1 and p2 (p0 , p) = p2 . If we de ne e b e e e e e e g(e) s1 (p1 (p0 , p), p2 (p0 , p))(p1 (p0 , p) U1 ) + s2 (p1 (p0 , p), p2 (p0 , p))p2 (p0 , p) p s1 (p1 (p0 , p), p2 (p0 , p))s2 (p1 (p0 , p), p2 (p0 , p)), e e e e The prices in (7)-(8) do not yet characterize an equilibrium outcome. To be an equilibrium, that g( ) is linear in p, so that there exists a unique xed point: e p = e then the equilibrium value p is de ned as a xed point of g( ). It is straightforward to verify e 2 18a ( p0 + U0 ) + (27 2 18U1 U1 ) . 18 (2 a) The interpretation of this solution is that the quality-adjusted average price for VoIP services is decreasing in the utility of rm 1 s VoIP services U1 (for not too small). For a > 0, it is increasing in the utility of PSTN services U0 , and decreasing in the price of the competitive segment p0 . The latter two properties can be explained by the fact that explained above. Therefore, in an equilibrium outcome, pricing in the VoIP segment becomes more competitive (and therefore, p falls). e an decrease in U0 p0 makes migration to VoIP more attractive, everything else equal, as We restrict our analysis to moderate levels of the terminating access price. This is a reasonable restriction in the light of the fact that the underlying marginal costs of call ter- mination are very small or even negligible in reality. Assumption: a < 2 . 22 segment: Substituting the constant p into (7)-(8), we obtain the equilibrium size of the PSTN e = 0 (p0 , p , p ) = 0 12 2 9 [4( p0 + U0 ) + 3 ] U1 18U1 ) . 36 (2 a) (9) Next, we obtain equilibrium prices for VoIP services: p = + 1 p 2 U1 + a 0 3 2 a{9 [4( p0 + U0 ) ] 30 U1 U1 } + 24 (3 + U1 ) = , 36 (2 a) U1 + a = 0 3 2 a{9 [4( p0 + U0 ) ] 6 U1 U1 } + 24 (3 U1 ) = . 36 (2 a) (10) (11) Note that if an equilibrium exists, it is unique (given by (10)-(11)). One can check that in order for p and p to be pro t-maximizing prices, the following second-order condition 1 2 for pro t maximization has to be satis ed: 2 U1 + 18U1 + 9 (4 4a + 4p0 3 4U0 ) < 0. 36 2 (2 a) 1 2 36 U1 This condition is equivalent to > 0, and will therefore always be satis ed in an interior equilibrium outcome.20 Using that a < 2 , it can be rewritten as a < U0 3 + . 4 + 1 U1 + p0 2 The insight from the previous section with respect to prices in the VoIP segment is still valid. In the Hotelling speci cation we obtain the markup due to product di erentiation corrected by a term that re ects the asymmetry that is introduced due to U1 plus the marginal cost for operator 2 due to termination on the PSTN segment. Operator 2 s pro ts in an equilibrium are equal to: = 2 2 (3 U1 )2 [9 (4 + 4p0 4U0 4a 3 ) + U1 + 18U1 ] 648 2 (2 a) For U1 = 0 the expression reduces to = 2 (4 + 4p0 4U0 4a 3 ) 8 (2 a) The equilibrium expression for rm 1 s pro t is somewhat more involved. 20 In what follows, we implicitly assume that the solution to the system of rst-order conditions for pro t maximization characterizes the equilibrium outcome that we discuss. 23 Remark: Since we are interested in the migration from PSTN to VoIP we analyze only interior equilibria. Here we comment on the possibility that full migration to VoIP is an equilibrium. We will see that for a range of parameter constellations there exist multiple equilibria. For a given price p0 , a consumer of type t = 1 has expected utility from PSTN-telephony equal to r + U0 /2 p0 . In equilibrium his expected utility for VoIP telephony would be r ((s )2 +(s )2 ) /2+s (U1 p ) s p . Note that in an interior equilibrium a consumer of 1 2 1 1 22 type t = 1 must strictly prefer PSTN. However, in such an equilibrium VoIP prices are higher than in a situation in which all consumers have migrated. Therefore, denoting equilibrium values for a = 0 with superscript 0 the condition for the existence of an equilibrium in which s 0 (U1 p 0 ) s 0 p 0 . 1 1 22 all consumers have migrated to VoIP is r + U0 /2 p0 < r ((s 0 )2 + (s 0 )2 ) /2 + 1 2 4.3 4.3.1 Comparative statics The PSTN terminating access price The focus of this paper is to understand the impact of regulatory decisions in the PSTN segment on market outcomes in the VoIP segment. For this we derive comparative statics results in the regulated access price a. Note that for a given number of VoIP customers a higher access charge implies that the entrant faces higher perceived marginal costs and the incumbent a higher opportunity cost to attract customers in the VoIP segment. This shifts the reaction curve of both operators outward. Since products are strategic complements retail prices are in ated.21 In the previous section, we showed that the neutrality of rm 2 s pro ts resulted from the property that both rms equilibrium prices increase by the increase in opportunity costs due to a higher a. This cost increase (compared to the case a = 0) is equal to 0 a, which is the expected access payment incurred by the entrant. Since p1 p2 was not a ected if a increased, market shares si remained the same. Note that, in the present model with endogenous segment size, higher perceived costs are also passed on to consumers. In particular, p 0 a = U1 /3, which, again, is independent of a and market shares s are 2 i independent of a. However, rm 2 s pro ts are not neutral to the access price. The reason is 21 that consumers, anticipating higher VoIP prices, become more reluctant to migrate to VoIP, Note that in standard models of price competition with di erentiated products rms o er strategic complements. This gives rise to monotone comparative statics properties (see e.g. Vives, 1990, and Milgrom and Roberts, 1990). For a recent overview of the literature on strategic complementarites see e.g. Vives (2005). For a rst application of the theory to telecommunications markets see Peitz (2005). In our simple price competition model, VoB prices are indeed strategic complements. 24 that is / a > 0.22 Formally, taking derivatives of the expressions reported in equations 0 (10) and (11), we obtain for i = 1, 2, p i a 0 a a a 0 = + 0 2 a > 0. = + 0 Note also that prices respond more strongly to changes in the access price if the access price is already high. Formally, p is convex in a: i ( + 2 p 0 0 i + = a2 2 a a 0 2 a )(2 a) + a 0 > 0. (2 a)2 Consider now the change of rm 2 s equilibrium pro t in response the change in the access price. The equilibrium pro t of rm 2 is decreasing in the access price because 2 = [s2 (p , p )(p a)] < 0 12 2 0 a a Our main comparative statics result can be summarized as follows. Result 4.1. For a given PSTN price, a higher access price for call termination on the PSTN network leads to (i) a smaller customer base for VoIP telephony. (ii) higher prices for VoIP telephony, and (iii) lower pro ts for operator 2. This result not only holds for our particular speci cation but more generally. Consider pro t functions as in equations (1) and (2) that depend on pi , and . The required properties are that (P1) for given a and , prices are strategic complements; (P2) for given , a higher access prices increases marginal pro ts; and (P3) higher retail prices lead to a lower penetration of VoIP. Properties (P1) and (P2) have been shown in the model with exogenous shares 0 and . Thus, for given 0 and an increase in a leads to higher prices. Since this is anticipated, given these higher prices more consumers stay with PSTN. This feeds back into higher expected costs for operator 2 and marks a new round in which rms increase their price. Hence, in equilibrium (provided that it exists), prices increase and the market penetration of VoIP decreases. 22 From equation (9) it can be directly seen that indeed 0 / a > 0 whenever 0 > 0. 25 pi , i=1,2 1.5 1.4 1.3 1.2 1.1 0.1 0.2 0.3 0.4 0.5 a 0.25 0.2 0.15 0.1 0.05 0.1 0.2 0.3 0.4 0.5 a 2 0.12 0.1 0.08 0.06 0.04 0.02 0.1 0.2 0.3 0.4 0.5 a 1 0.24 0.23 0.22 0.21 0.1 0.19 0.18 0.2 0.3 0.4 0.5 a welfare 9.15 9.1 9.05 0.1 8.95 8.9 0.2 0.3 0.4 0.5 a Figure 3: Illustration of the equilibrium outcome when the PSTN retail price is exogenously given. We will illustrate the equilibrium properties with some diagrams based on a numerical example.23 Suppose that p0 = 0.25, = = 1, r = 10, U0 = U1 = 0, and c = 0.1. The condition for an interior solution then requires that a < 0.5. We will therefore look at the implications for a [0, 0.5].24 Figure 3 contains various illustrations of the equilibrium properties. As illustrated in gure 3, rm 1 s pro t may be partly increasing and partly decreasing in the access price a. This is the case for p0 su ciently large. For small values of access price a, a PSTN consumer is then in expectations more valuable for rm 1 than a VoIP consumer. Thus, an increase in a which shifts consumers from the VoIP to the PSTN segment is pro t increasing. This explains why rm 1 s pro ts are initially increasing in a. This no longer holds for larger a. The reason is that for larger a, competition in the VoIP segment is more relaxed so that, for retail prices in the VoIP segment above a certain level, a consumer in the 23 24 Obtained by using Mathematica software. We have checked the robustness of the e ects on surplus levels by varying parameter levels. This con rmed that qualitatively, the observations discussed above do not seem to depend on the parameter levels (within reasonable bounds). 26 VoIP segment is in expectations more valuable than a consumer in the PSTN segment. Firm 1 may therefore obtain a larger pro t with a lower access price since this implies a larger VoIP segment. This result suggests that it is not necessarily in the interest of rm 1 to lobby for a high access charge. In particular, if p0 is su ciently small, then rm 1 s pro t is globally decreasing in a (note: this is not illustrated in the gure). The reason is the following: With a higher access price a consumers expect the VoIP segment to be less competitive. Therefore only few consumers decide to migrate to the VoIP segment. Since the PSTN-segment is not very pro table, rm 1 would be better o if many consumers would migrate. To the extent that rm 1 can in uence U0 it has no incentive to improve the quality of PSTN services. Rather the opposite is true, since it would like to convince consumers to move to the VoIP segment. More generally, one can observe that the larger p0 , the larger the pro t-maximizing access price. Again the argument is that relaxed competition in the VoIP segment (and thus a smaller market share of VoIP) is in the interest of rm 1 if retail price regulation in the PSTN is less strict (to the e ect that PSTN customers are more valuable). Figure 3 shows that total welfare is decreasing in a, which is somewhat surprising as the policy implication is that a should be lower than marginal cost. However, note that we assumed that the demand for calls is perfectly inelastic. This implies that retail prices above or below perceived marginal costs do not a ect participation. This, in turn, implies that all welfare results are completely driven by the division of the market among PSTN and VoIP, and in the VoIP segment, the division between the two operators. For the speci c parameter values that we chose, it turns out that the welfare-maximizing size of the VoIP segment is as large as possible (under the restriction that a 0). Note that in general VoIP prices are of the access price. 4.3.2 The PSTN retail price increasing in a. Therefore all consumers are necessarily (weakly) worse o after an increase Recall our assumption that the retail price for PSTN services is regulated. While this is no longer an appropriate description in those countries in which retail regulation has been phased out, it still can be used as a useful benchmark since various forms of wholesale regulation a ect retail prices in the PSTN segment (e.g. resale competition limit the incumbent s market power in the retail market). Thus the xed price for PSTN that we assume in our model, can be seen as a simpli cation of situations in which the PSTN price is less exible than VoIP 27 prices, for instance due to regulatory measures that lead to unbundling and resale-based competition in the PSTN segment. It is interesting to see how our results depend on the level of the PSTN retail price. We can make a number of observations, mostly based on (10)-(11): Result 4.2. (i) A higher price for PSTN telephony leads to a larger customer base for VoIP telephony. (ii) Provided that the PSTN access price is positive, a higher price for PSTN telephony leads to lower prices for VoIP telephony. (iii) Provided that the PSTN access price is zero, a higher price for PSTN telephony does not a ect prices for VoIP telephony. (iv) A higher price for PSTN telephony increases the entrant s pro ts. Let us discuss these observations in some more detail. Observation (i) is not surprising. If PSTN telephony becomes more expensive, more consumers will switch to VoIP. Observations (ii)-(iii) can be explained as follows. The mechanism behind the e ect p 0 i = a < 0, i = 1, 2, p0 p0 is that an increase in p0 reduces the size of the segment of PSTN customers, which in turn reduces the probability that a customer of operator 2 makes a call to the PSTN network. Hence, because of the reduction in expected access payments to the incumbent, operator 2 s perceived marginal cost is reduced. The result is a more competitive outcome in the VoIP segment, as has been explained before. Thus a higher price in the PSTN segment leads to a lower prices in the VoIP segment so that products across segments are strategic substitutes. Note, however, that if the incumbent s access price for termination on the PSTN network is zero, then the entrant s perceived marginal cost remains una ected if the number of PSTN customers decreases. We remark that an increase in p0 has the same e ect on prices in the VoIP segment as an increase in the xed utility of PSTN telephony. More precisely, a larger value for U0 increases the customer base for PSTN services, and hence in ates the entrant s perceived marginal cost. Therefore, p i > 0, i = 1, 2. U0 To understand observation (iv), note that the equilibrium pro t of rm 2 is increasing in 28 the price of the PSTN-segment, d d (3 U1 )2 2 > 0. = 2 = dp0 dU0 18 (2 a) The reason is that such a higher price leads to more migration to the VoIP segment. This e ect is reinforced because such migration leads to lower perceived costs of rm 2 and thus, with ful lled expectations, makes the VoIP segment more attractive. Due to the expansion of the VoIP segment, the entrant bene ts from such a change. Finally, we turn to the comparative statics properties of rm 1 s pro ts. Firm 1 s pro t is increasing in p0 for p0 small. This is hardly surprising since a high retail price in the PSTN segment directly feeds into pro ts. A possibly countervailing e ect is that rm 1 loses market share in the retail market. However, as long as a consumers in the VoIP segment is in expectations more valuable than a consumer in the PSTN segment, migration from PSTN to VoIP is good news for rm 1. For large p0 the e ect is reversed. Thus, for a given a there is a nite pro t-maximizing retail price for PSTN telephony. In principle, the regulator can refrain from regulating the retail price p0 in the PSTN segment even if, as in our model, rm 1 maintains a monopoly position in that segment. The reason is that although rm 1 wants to milk its PSTN customers it cannot price too high in order not to loose consumers to VoIP. This con rms that VoIP o ers by rm 2 give rise to some disciplining e ect on the rm 1 s PSTN o ers. We will return to this situation below. 4.4 Access regulation when the PSTN retail price is endogenous As an extension of the model, consider two ways in which the PSTN price p0 may be endogenous. Resale competition in the PSTN segment: Suppose that due to regulation of the incumbent s originating access price (equal to a0 ), there is intense competition in the PSTN segment.25 The calls that the competitive fringe of PSTN entrants without local networks generate, terminate either on the incumbent s PSTN network, or on one of the VoIP networks. Hence each of these entrants faces a perceived marginal cost of 0 a + a0 . Perfect competition in the PSTN retail segment then drives down the incumbent s retail price to p0 = 0 a + a0 . In the model, the determination of p0 takes place as follows. For given access prices, equilibrium prices p and p have been determined. Then p0 is obtained as the xed point 1 2 25 See also De Bijl and Peitz (2002), ch. 5. 29 of p0 = (p0 , p , p )a + a0 . Notice that the operators, when choosing prices in the VoIP 0 12 segment, do not take into account that p0 a ects the size of the VoIP segment. The reason is that at the moment they choose VoIP prices, consumers already have made the decision whether to migrate to VoIP or not. The e ect of access price a is now as follows. The access price a does not a ect the size of the VoIP segment (nor does it a ect market shares within that segment). An increase in a leads to higher retail prices in both segments and thus reduces both CS PSTN and CS VoIP . It increases the incumbent s pro ts while leaving the entrant s pro ts una ected. Overall, welfare W is not a ected. Accordingly, the model s outcomes are similar to the outcomes of the model in section 3, where we assumed that there was no migration between technologies (but distributional e ects are typically di erent). Sequential price setting: As an alternative to a situation of resale competition in the PSTN segment, consider the case in which operator 1 is be a monopolist in the PSTN segment and is able to set a pro t-maximizing price p0 . In particular, we assume that operator 1 chooses p0 after all parameters, including the access price, are set, but before the rest of the game evolves. Hence the incumbent takes into account the equilibrium prices from competition in the VoIP segment by using backward induction. This sequential timing of moves can be motivated by the fact that the incumbent is less exible in setting PSTN prices than in setting VoIP prices. For instance, because of universal service obligations, the incumbent may have to notify the regulator, or will need regulatory approval, for changes in PSTN prices. Given the PSTN access price and the equilibrium outcome in the VoIP segment, operator 1 chooses p0 to maximize pro ts (p , p ; a, p0 ). One can show that this pro t function is 12 concave in p0 . More precisely, it is quadratic, so that nding the pro t-maximizing PSTN price is relatively straightforward. It can be shown that p = 0 1 2 [36a2 3a(U1 + 10U1 3 ( 4c + 4U0 + 4 )) 144 2 +2 (U1 6U1 + 9 (4c + 4U0 + 5 + 4 ))]. The following result, which we state without proof, con rms our Result 4.1 for the case that the PSTN retail price is not xed: Result 4.3. For endogenous p0 , a higher access price for call termination on the PSTN network leads to 30 p0 1.17 pi , i=1, 2 1.25 1.2 1.16 1.15 1.15 1.14 0.1 0.2 0.3 0.4 0.5 a 1.1 1.05 0.1 0.2 0.3 0.4 0.5 a 0.7 0.68 0.66 0.64 0.62 0.1 0.2 0.3 0.4 0.5 2 0.35 0.34 0.33 a welfare 9.355 9.354 9.353 0.1 0.2 0.3 0.4 0.5 a 1 0.85 0.8 0.75 1 9.352 0.1 0.2 0.3 0.4 0.5 a 0.1 0.2 0.3 0.4 0.5 a Figure 4: Illustration of the equilibrium outcome when the PSTN retail price is endogenously determined. (i) a smaller customer base for VoIP telephony. (ii) higher prices for VoIP telephony, and (iii) lower pro ts for operator 2. We illustrate the equilibrium properties with some diagrams based on numerical calculation, when = = 1, r = 10, U0 = U1 = 0, and c = 0.1.26 See gure 4. The PSTN price p may be decreasing or increasing in access price a. For the chosen parameter values, it is 0 convex in a: rst decreasing, but increasing for relatively high values of a. The observation that p (a) is U-shaped can be explained as follows. Note that the in0 cumbent makes pro ts from selling wholesale access in the PSTN segment, and from selling retail services both in the PSTN and the VoIP segment. First, recall that earlier we saw that an increase in the access price (for given p0 ) led to higher prices in the VoIP segment. This implies that for a given p0 , a higher access price leads to a smaller VoIP segment. However, the incumbent is free to adjust its price. Note that for low levels of the access price, an increase feeds only weakly into higher VoIP prices. Since PSTN customers are very valuable at low levels of a, it is pro t maximizing for the incumbent to reduce p0 . At higher levels of 26 Again obtained by using Mathematica software. 31 a an increase feeds strongly into higher VoIP prices, as VoIP consumers are, in expectation, rather valuable for the incumbent. He therefore reduces the reduction of the size of the VoIP segment (the reduction that would occur with a constant p0 ) by increasing its PSTN price. Note that in general, a higher terminating access price a leads to larger PSTN segment. For relatively low levels of the access price, the incumbent decreases its PSTN price to bene t from an even larger PSTN segment. This implies that PSTN consumers bene t from a moderately high access price. VoIP customers su er and there is less migration to VoIP. To the extent that our model approximates current telecommunications markets in which the incumbent enjoys market power in the PSTN segment, our ndings run counter to the view that a higher access charge would lead to a decline of the PSTN segment.27 Figure 4 shows that total welfare is concave in a. To understand this, recall that all welfare results are driven by the division of the market among PSTN and VoIP, and in the VoIP segment, the division between the two operators.28 The regulator could in theory use the access price to implement the optimal split between the PSTN and VoB segment, which explains why the optimal access price is, in general, not equal to marginal cost. The optimal split depends on the parameters of the model. With the parameters of our numerical example the welfare-maximizing allocation does not have the property that = 1/2 for two reasons. First, the PSTN segment is more costly to operate, and second, since there is more variety in the VoIP segment, VoIP o erings tend to better t consumers tastes. In our example, the welfare-maximizing allocation can be shown to be = 0.675. 5 Conclusion In this paper, we explored competition between an incumbent o ering both PSTN and VoIP telephony, and an entrant active only in the VoIP segment. Our analysis has shed light on the e ects of regulation in one segment on competition in another, unregulated segment, and has focused on cost e ects of access price regulation. Given the publication of reports such as OECD (2006), this type of analysis is urgently called for. We looked at two di erent settings. In the rst setting, which can be seen as a benchmark case, we assumed that the size of the customer segment interested in VoIP was exogenously given (hence consumer migration between the segments was not possible). In this simple 27 For instance, a recent OECD report (OECD, 2006, p. 28) contains a statement that can be interpreted as such. 28 Welfare in a model with full participation only depends on price di erences and not on price levels. 32 model with inelastic demand an increase of the PSTN terminating access price increases the incumbent s pro ts, but as it does not a ect the entrant s pro ts, market entry is independent of access regulation. In the second setting, we endogenized consumers choices for PSTN versus VoIP services. We used a set-up with the interpretation of ful lled expectations about retail prices in the VoIP segment (this interpretation was not necessary, though). For instance, before consumers learn the prices of VoIP services, they already have certain beliefs about expected prices, which help them to gure out whether to go migrate to VoIP. Accordingly, consumers are already aware of the existence of a new technology, and make their migration decision before prices are actively searched. Another interpretation is that the decision to migrate to a new technology requires some commitment, as the e ort to do so is sunk before the process of actively searching prices starts. Focusing on ful lled-beliefs equilibria allowed us to keep the analysis tractable, while at the same time incorporating consumers migration decisions in a realistic way. Below we shortly discuss our main results and make some comments on the application of CPP versus RPP. We then conclude by addressing a couple of limitations of our analysis which suggest avenues for further research. Regulation of the PSTN retail price: The PSTN retail price only a ects competition in the VoIP segment if the consumers technology adoption decision is endogenous. An important result of the analysis is that, as long as the PSTN access price is positive, a higher price for PSTN telephony leads to lower prices for VoIP telephony. Only for an access price equal to zero, the retail price level of PSTN telephony does not a ect retail prices for VoIP telephony. These results illustrate the links between di erent telephony networks links that should not be ignored by regulators. These links have been explored in more detail. Suppose that an entrant in the VoIP market faces a positive access price. This access price may or may not re ect marginal cost levels; it is only important that this access price is positive. Then a lower regulated PSTN price leads to a smaller customer base for VoIP telephony and softens price competition among VoIP operators. Note that if a regulator allows an integrated incumbent to include a mark-up for common costs in its access charge, which is typically the case, then the access price will be above the marginal cost level. The result is less migration to VoIP. In addition, if a universal service obligation forces the incumbent to price PSTN telephony at a low level, VoIP retail prices become in ated and the adoption of VoIP will be slowed down even more. 33 Access regulation: Access regulation on the PSTN network a ects the VoIP market. For instance, if the incumbent charges for call termination on the PSTN and VoIP networks use bill-and-keep, then a higher access price for call termination on the PSTN network leads to a smaller customer base for VoIP telephony. In the context of access price regulation it is important that regulators take into account these linkages between di erent market segments, and that regulation within one segment may have spillover e ects to other segments. In markets in which the PSTN retail price is not regulated (and in which an incumbent enjoys market power), a higher access price leads to higher VoIP retail prices (as in the regulated case) but tends to lead to lower retail prices in the PSTN segment. This suggests that regulation has winners and losers: consumers of the old technology are the winners from a high access price (which can be seen as a protective measure for the old technology) and consumer of the new technology are the losers. RPP versus CPP: Our analysis has been carried out under the calling party pays (CPP) principle. If, however, the receiving party pays (RPP) principle is applied, then the user of the PSTN network has to pay for terminating access of the calls that he receives. This implies that the perceived cost of the VoIP operator no longer contains any payment for terminating access. In that case, if the PSTN operator charges a higher price to his PSTN customers for receiving calls, he makes PSTN telephony less attractive and will therefore reduce his PSTN customer base. Thus the comparative static results with respect to market share are reversed. Furthermore, under ful lled expectations the price level of the VoIP segment is independent of the relative success of VoIP. This suggests that with the appearance of VoIP the rationale for applying the case for applying RPP is strengthened. More generally, with the coexistence of di erent technologies, technology-speci c costs have to be attributed to users, ignoring the issue of markups. Under RPP, costs of PSTN telephony are attributed to PSTN users whereas under CPP, they are partly borne by VoIP users. However, to the extent that some of the costs arise due to social obligations, in particular, universal service obligations, the application of RPP may put an excessive burden on PSTN users. This burden may become unbearable if the PSTN segment shrinks drastically in size. Our modelling strategy has been to isolate the cost e ects of access price regulation abstracting from two important aspects. First, integration in our model with a given PSTN price is neutral to competition. Thus cannibalization is not an issue. This result is due to the particular timing, because an integrated rm with a regulated PSTN price cannot commit to a 34 high VoIP price (which would avoid cannibalization). However, the rm can possibly commit not to o er VoIP services at all. If we introduce this possibility in our model, there is a range of parameters where this is indeed the pro t maximizing solution for the incumbent. By not o ering VoIP services it relaxes competition in the VoIP segment thus making consumers reluctant to migrate to VoIP. Also note that if we endogenize the PSTN price, our model no longer has the property that integration is neutral to competition. An integrated rm takes pro ts from retail in the VoIP segment into account and adjusts the PSTN price accordingly. Second, in our model predation is not an issue. Predation tends to make the incumbent more aggressive in the VoIP market as an attempt to maintain its customer base. Such predatory behavior arises in dynamic models, in particular with consumer switching costs. We leave it for future research to analyze predatory behavior in the context of VoIP. 35 References [1] Analysys (2004), IP Voice and Associated Convergent Services , Final Report for the European Commission, 28 January. [2] Armstrong, M. (1998), Network Interconnection in Telecommunications , Economic Journal 108, 545-564. [3] Armstrong, M. (2002), The Theory of Access Pricing and Interconnection , in: M. Cave, S. Majumdar , and I. Vogelsang (eds.), Handbook of Telecommunications Economics, Amsterdam: North Holland. [4] Armstrong, M., Doyle, C. and J. Vickers (1996), The Access Pricing Problem: A Synthesis , Journal of Industrial Economics 44, 131 150. [5] Armstrong, M. and J. Vickers (1998), The Access Pricing Problem with Deregulation: A Note , Journal of Industrial Economics 46, 115 121. [6] Baumol, W. (1983), Some Subtle Issues in Railroad Regulation , International Journal of Transport Economics 10, 341-355. [7] Canoy, M. and M. Peitz (1997), The Di erentiation Triangle , Journal of Industrial Economics 45, 305-324. [8] De Bijl, P.W.J. and M. Peitz (2002), Regulation and Entry into Telecommunications Markets, Cambridge University Press, Cambridge, UK. [9] De Bijl, P.W.J. and M. Peitz (2005), Local Loop Unbundling in Europe: Experience, Prospects and Policy Challenges , Communications and Strategies 57, 33-57. [10] De Bijl, P.W.J. and M. Peitz (forthcoming), Local Loop Unbundling: One-Way Access and Imperfect Competition , in: R. Dewenter and J. Haucap (eds.), Access Pricing: Theory and Practice, Elsevier Science. [11] Ebrill, L. and S. Slutsky (1990), Production E ciency and Optimal Pricing in Intermediate-Good Regulated Industries , International Journal of Industrial Organization 8, 417-442. [12] Foros, O. (2004), Strategic Investments with Spillovers, Vertical Integration and Foreclosure in the Broadband Access Market , International Journal of Industrial Organization 22, 1-24. 36 [13] Katz, M. (1984), Firm Speci c Di erentiation and Competition among Multiproduct Firms , Journal of Business 56, S149-166. [14] Katz, M. and C. Shapiro (1985), Network Externalities, Competition and Compatibility , American Economic Review 75, 424 440. [15] Lal, R. and C. Matutes (1989), Price Competition in Multiproduct Duopolies , RAND Journal of Economics 20, 516-537. [16] La ont, J.-J., P. Rey, and J. Tirole (1998), Network Competition: I. Overview and Nondiscriminatory Pricing , RAND Journal of Economics 29, 1-37. [17] La ont, J.-J., and J. Tirole (1994), Access Pricing and Competition , European Economic Review 38, 1673-1710. [18] La ont, J.-J., and J. Tirole (1996), Creating Competition Through Interconnection: Theory and Practice , Journal of Regulatory Economics 10, 227-256. [19] La ont, J.-J. and J. Tirole (2000), Competition in Telecommunications, MIT Press, Cambridge, MA. [20] Lewis, T.R., and D.E.M. Sappington (1999), Access pricing with unregulated downstream competition , Information Economics and Policy 11, 73-100. [21] Manez, J. A. and M. Waterson (2001), Multiproduct Firms and Product Di erentiation: A Survey , University of Warwick Working paper no 594. [22] Milgrom, P. and J. Roberts (1990), Rationalizability, Learning, and Equilibrium in Games with Strategic Complementarities , Econometrica 58, 1255-1277. [23] OECD (2006), The Policy Implications of Voice over Internet Protocol , report to the Working Party on Telecommunication and Information Services Policies, DSTI/ICCP/TISP(2005)3/FINAL. [24] Ofcom (2006), Regulation of VoIP Services , Statement and further consultation, 22 February. [25] Peitz, M. (2005), Asymmetric Access Price Regulation in Telecommunications Markets , European Economic Review 49, 341-358 . 37 [26] Vives, X. (1990), Nash Equilibrium with Strategic Complementarities , Journal of Mathematical Economics 19, 305-321. [27] Vives, X. (2005), Complementarities and Games: New Developments , Journal of Economic Literature 43, 437-479. [28] Vogelsang, I. (2003), Price Regulation of Access to Telecommunications Networks , Journal of Economic Literature 41, 830 862. [29] Willig, R. (1979), The Theory of Network Access Pricing , in: Trebing, H. (ed.), Issues in Public Utility Regulation, Michigan State University, East Lansing, MI. 38

Find millions of documents here - Study Guides, Homework Solutions, Papers, Exam Answer Keys and more. Course Hero has millions of course related materials that will enable you to learn better, faster and get an A in all your courses.
Below is a small sample set of documents:

626.Management of Spec.doc
Path: Michigan >> LAW >> 626 Fall, 2008

628-F2002 SPP728 .doc
Path: Michigan >> LAW >> 628 Fall, 2008
rr07-628.pdf
Path: Michigan >> LAW >> 628 Fall, 2008
Powell_CWN as Civic Participation final.pdf
Path: Michigan >> LAW >> 631 Fall, 2008
632 syllabusw03.doc
Path: Michigan >> LAW >> 632 Fall, 2008
Bauer-Net-Neutrality-TPRC-2006-fin.pdf
Path: Michigan >> LAW >> 633 Fall, 2008
rr08-635.pdf
Path: Michigan >> LAW >> 635 Winter, 2008
TPRC2006BarriersToEntry.pdf
Path: Michigan >> LAW >> 636 Winter, 2008
rr08-638.pdf
Path: Michigan >> LAW >> 638 Fall, 2008
xm-f96.pdf
Path: Michigan >> LAW >> 641 Fall, 2008
ps1a.pdf
Path: Michigan >> LAW >> 641 Fall, 2008
rr08-641.pdf
Path: Michigan >> LAW >> 641 Fall, 2008
COsorioTPRC2006Draft.pdf
Path: Michigan >> LAW >> 641 Fall, 2008
rr08-642.pdf
Path: Michigan >> LAW >> 642 Winter, 2008
TPRC_DNSSEC_revised.pdf
Path: Michigan >> LAW >> 642 Winter, 2008
647.Info.Res.Serv.doc
Path: Michigan >> LAW >> 647 Fall, 2008
UnderstandPrivacyBreach.pdf
Path: Michigan >> LAW >> 647 Fall, 2008
40857.0001.001.pdf
Path: Michigan >> LAW >> 647 Fall, 2008
feqma.pdf
Path: Michigan >> LAW >> 647 Fall, 2008
648-748-InfoCulture Syllabus 2003.doc
Path: Michigan >> LAW >> 648 Fall, 2008
rr08-648.pdf
Path: Michigan >> LAW >> 648 Fall, 2008
feqma.pdf
Path: Michigan >> LAW >> 648 Fall, 2008
650-HW1.doc
Path: Michigan >> LAW >> 650 Winter, 2008
PubPol 650 Nov 2008 draft.doc
Path: Michigan >> LAW >> 650 Winter, 2008
2.ppt
Path: Michigan >> LAW >> 650 Winter, 2008
8.ppt
Path: Michigan >> LAW >> 650 Winter, 2008
rr08-656.pdf
Path: Michigan >> LAW >> 656 Winter, 2008
656.txt
Path: Michigan >> LAW >> 656 Winter, 2008
cooper video.pdf
Path: Michigan >> LAW >> 657 Fall, 2008
657.txt
Path: Michigan >> LAW >> 657 Fall, 2008
rr08-658.pdf
Path: Michigan >> LAW >> 658 Winter, 2008
Comparative NII.pdf
Path: Michigan >> LAW >> 658 Winter, 2008
658.txt
Path: Michigan >> LAW >> 658 Winter, 2008
660 Seminar Information 03.doc
Path: Michigan >> LAW >> 660 Fall, 2008
RationalizingInternetSafeHarbors.pdf
Path: Michigan >> LAW >> 660 Fall, 2008
TPRC Gruber.pdf
Path: Michigan >> LAW >> 662 Fall, 2008
663-syllabus.pdf
Path: Michigan >> LAW >> 663 Winter, 2008
rr08-664.pdf
Path: Michigan >> LAW >> 664 Winter, 2008
DYanich TPRC paper.pdf
Path: Michigan >> LAW >> 664 Winter, 2008
38498.0001.001.pdf
Path: Michigan >> LAW >> 664 Winter, 2008
Commons Theory Paper (TPRC).pdf
Path: Michigan >> LAW >> 667 Fall, 2008
669-769-02-Syllabus.pdf
Path: Michigan >> LAW >> 669 Winter, 2008
Syllabus-Final.doc
Path: Michigan >> LAW >> 669 Winter, 2008
39714.0001.001.pdf
Path: Michigan >> LAW >> 669 Winter, 2008
Neither Fish Nor Fowl.pdf
Path: Michigan >> LAW >> 670 Fall, 2008
39673.0001.001.pdf
Path: Michigan >> LAW >> 670 Fall, 2008
Econ-671-Syllabus.pdf
Path: Michigan >> LAW >> 671 Fall, 2008
PubPol 671 Prior Syllabus - 2008.pdf
Path: Michigan >> LAW >> 671 Fall, 2008
InfanteOliverMacianTPRC07.pdf
Path: Michigan >> LAW >> 671 Fall, 2008
39674.0001.001.pdf
Path: Michigan >> LAW >> 671 Fall, 2008
673 Legal Info Leary.doc
Path: Michigan >> LAW >> 673 Fall, 2008
rongilg_92308_152138NERS673Lec_4.pdf
Path: Michigan >> LAW >> 673 Fall, 2008
schejter universal service TPRC.pdf
Path: Michigan >> LAW >> 673 Fall, 2008
40993.0001.001.pdf
Path: Michigan >> LAW >> 673 Fall, 2008
The_Song-of_Roland_Essay
Path: Cornell >> HIST >> 1510 Fall, 2008
41010.0001.001.pdf
Path: Michigan >> LAW >> 675 Fall, 2008
Spulber & Yoo TPRC draft 8-16-07.pdf
Path: Michigan >> LAW >> 677 Winter, 2008
35679.0001.001.pdf
Path: Michigan >> LAW >> 681 Fall, 2008
36687.0001.001.pdf
Path: Michigan >> LAW >> 683 Winter, 2008
SamarajivaGalpayaRatnadiwakara_TPRC.pdf
Path: Michigan >> LAW >> 683 Winter, 2008
Seramar_684_final.doc
Path: Michigan >> LAW >> 684 Winter, 2008
Net_Neutrality_Jamison_Hauge_8_16_07.pdf
Path: Michigan >> LAW >> 686 Fall, 2008
TPRC Payal Malik.pdf
Path: Michigan >> LAW >> 687 Winter, 2008
Netneutralitymarsdencavetprc2007c.pdf
Path: Michigan >> LAW >> 689 Fall, 2008
Bauer-Bohlin-TPRC-2007.pdf
Path: Michigan >> LAW >> 691 Winter, 2008
hammon52A02-0308-cr-693.pdf
Path: Michigan >> LAW >> 693 Fall, 2008
37476.0001.001.pdf
Path: Michigan >> LAW >> 693 Fall, 2008
Taking up Coase's challenge 50 years on_final.pdf
Path: Michigan >> LAW >> 693 Fall, 2008
37475.0001.001.pdf
Path: Michigan >> LAW >> 694 Fall, 2008
bviani_08.16.07.pdf
Path: Michigan >> LAW >> 694 Fall, 2008