Appendix 2: Two way access
Face - Contents - Bottom - Previous/Next "Telecompetition" Appendix 2: Two way access92IntroductionTwo-way access issues in telecommunication relate to situations where several phone-service providers require access to each others' networks, in order to build viable business models. More precisely, we deal with issues that arise when customers of any given network require the facility of terminating calls to customers of other networks. A core example is that all mobile-phone providers must be able to provide termination of calls in rival mobile networks as well as in the networks of fixed-phone providers. Thus, the discussion below mainly relates to termination charges and market power in termination. The companion paper (see von der Fehr (2004)) discusses one-way access, where the structure is more asymmetric, in the sense that some (potential or actual) entrant requires access to the network of an incumbent, while the incumbent (depending on the details) would simply prefer no entry. The philosophy of the approach to one-way access and the main results draw heavily on the economics literature on monopoly regulation. Hence, the natural point of departure is the "ideal" of marginal cost pricing and its various "adaptations" such as the ECPR, LRIC-pricing and Ramsey-pricing. Then, practical issues related to price calculation93 and implementation94 are touched upon. While much of this is clearly still relevant for understanding two-way access, the emphasis in the following will be changed slightly from that of monopoly regulation to one of competitive or oligopoly interaction and regulation thereof. Before continuing, we note that the formal (economics) literature on issues in two-way access is less consolidated than the literature on one-way access-price regulation. Thus, it presents less of a unified approach, and models are (in the best Industrial Organization tradition95 used to illustrate various problems in a more partial manner. That is, a series of (potential) competition problems is outlined, models to study these problems are presented, and (hopefully) sensible approaches to solve the problems are suggested based on the modelresults. The following short presentation will be relatively informal. For the underlying theoretical perspectives and models, we rely heavily on the syntheses provided by Armstrong (2002) and Laffont & Tirole (2000, ch. 5)96. Practical policy concerns are dealt with in detail by e.g. Canoy, de Bijl & Kemp (2003) and OECD (2004). The outline is as follows. Section 2 will present some preliminaries on the kinds of structures that we want to understand and a first indication of some key problems. This is meant to motivate the emphasis in the following. Section 3 then presents the first pass through two-way access in the form of a discussion of competing bottlenecks, where (perfectly) competitive phone-service providers require access to the incumbent's infrastructure, while retaining some market power over termination to their own subscribers. The main driver of results in this section will relate to the potential problems stemming from market power in call termination, which have been identified in the theoretical literature as well as by regulators and antitrust practitioners. Section 4 then moves to a discussion of network interconnection and two-way access pricing, where the setting of termination charges and the subsequent (imperfect) competition for subscribers are interlaced. A selection of problems and regulatory issues will be presented. Finally, Section 5 briefly concludes. PreliminariesAs mentioned at the beginning, we are concerned with situations where two or more service providers, each with their own network infrastructures, need to be able to link up so that customers can call97 anyone else, irrespective of whether they subscribe to the same network or a different (rival or non-rival) network. Thus, all firms or networks must buy vital inputs access - from each other. So, networks must interconnect, and we are concerned with cases where reciprocal access or two-way access is required. This access is something that has to be granted and priced, and since access must be reciprocated, there are at least two prices that must be determined. From the perspective of this presentation, the question is whether any or all of these access prices should and could be regulated, or whether the pricing should simply be left to the various networks to deal with either non-cooperatively or cooperatively through negotiation. As noted, several different problems may arise as a result of the specifics of a given situation. To prevent the analysis from getting out of hand, there is a need to deal with these various problems sequentially. Most of the recent literature as consolidated in various ways by Laffont & Tirole (2000, ch. 5) and Armstrong (2002) has approached the two-way access problems by starting out from simple models that make a few central points and then gradually introduced new aspects. Throughout, several networks will be "competing" to some extent for subscribers. If the services offered are highly substitutable, this competition for subscribers is likely to be intensive, while if the services are complementary, then the relationship between networks may be considered as an exercise in coordination rather than competition. However, even if competition for subscribers is intensive, a distinction can still be made between two different settings (following Armstrong (2002)). The basic structures The first type of structure essentially includes a monopolist controlling the fixed-telephony (backbone) infrastructure and facing several, competing providers of various services (mobile telephony, Internet access, etc.). The competitive service providers must have access to the fixed infrastructure to terminate calls, and the monopolist must have access to each service provider to terminate calls.98 Note that this type of structure is well-suited to discuss issues related to fixed-mobile interconnection, and as such it also represents the first natural step from the underlying structure used to discuss one-way access in von der Fehr (2004). In particular, this will be useful in assessing whether mobile termination charges should be regulated, in a setting where fixed termination charges are already regulated. The second type of structure is initially more symmetric. Each of several, similar networks must interface with each other, otherwise they will not be viable (by assumption). Examples include international interconnection to enable long-distance telephony and interconnection between competing mobile-phone or fixed-phone providers. Hence, reciprocal access terms are important. Different cases can be considered, in which termination charges are either set unilaterally by networks, are negotiated between networks or are regulated. In addition, the effect of the intensity and mode of competition in the final service market on unilateral or negotiated termination terms is assessed. The central issue is whether or not the setting of termination charges might dampen service-market competition or facilitate collusion. The two basic structures will be illustrated by a sequence of simple models, which we subsequently elaborate on. The simple models can be referred to as models of non-rival and rival networks, respectively. In the first section below we focus on two-way access between non-rival networks, which is meant to capture a simple setting where e.g. the (monopoly) fixedphone network does not compete with a given mobile-phone network for subscribers, while mobile-phone networks do compete (intensively) with each other. In contrast, the second section below focuses on two-way access between rival networks, which is meant to capture e.g. that several fixed-phone or mobile-phone networks must arrange reciprocal access, while at the same time they compete for the same pool of potential subscribers. This sequence of relatively simple settings allows a focus on different issues in turn, and it prevents the analysis from getting out of hand. Non-Rival Networks, Bottlenecks and Competition for SubscribersAs noted, this section is motivated mainly by markets for mobile telephony. On the assumption that mobile-phone services are largely competitive, potential competition problems arise not so much from lack of inter-brand competition as such, but more from the fact that residual pockets of market power remain, to the extent that particular providers (networks) have monopolies on termination in their nets. The topic is best illustrated by the relationship between a monopoly provider of fixed-phone services and several providers of mobile-phone services. Mobile-service provision may well be highly competitive, in the sense that the tariff competition between mobile-service providers for subscribers is intensive, yet each mobile network has a monopoly on termination in its net. Therefore, pricing of access (termination) is less competitive. The details of the analysis will then depend on various assumptions made about consumer utilities including the valuation of in-bound calls, on the presence of network externalities, on the substitutability of fixed and mobile services, etc. Benchmark In short, this is a model in which there is intensive competition for mobile subscribers, yet each network has a monopoly on providing access to its subscribers (that is, call termination). Armstrong (2002) refers to this as competitive bottlenecks to capture that service providers are in close competition for subscribers, yet (despite the competition for subscribers) they each control the vital facility (bottleneck) of providing access to their subscribers. That mobile and fixed calls are non-substitutable and that the upstream firm is not present in the downstream market imply that the upstream firm and a downstream firm are not competitors. When they are competitors, for either of the two reasons, we move towards the model of the next section, where we discuss rival network interconnection and competition. Despite its partial lack of realism, this model is very useful for analysing one set of issues related to fixed-mobile interconnection, while other issues are best approached by a different formal modelling.100 Assume that tariff schemes for mobile subscribers consist of a fixed fee (possibly, a subsidy) and a per-call or per-minute charge, which we can write as smob(q) = ƒ + pq For simplicity, in the following we shall refer to p as a per-call charge levied on the mobile subscriber and to q as the number of calls made by the subscriber. That the mobile services provided by different networks are perfectly substitutable implies that tariff competition between mobile networks for subscribers drives per-call or per-minute prices to variable cost (which include the termination charges set by the upstream monopolist, see below). Had there been product differentiation and/or imperfect competition at the downstream level, this would generally have allowed positive downstream profits. However, this is not central to the qualitative points we want to make. However, it might matter for the level at which regulated termination charges should be pegged. In addition, mobile-service providers generate revenues from terminating calls originating in the fixed net. Competition between the mobile networks for subscribers, however, drives net profits to zero. Therefore termination surpluses (if they arise) are fully recycled into the mobile subscriber fee. Hence, if (as a matter of interpretation) we include the cost of the hand-set in the total cost of mobile services, we conclude immediately that access revenues could potentially be used to subsidize hand-sets (as is frequently observed in practice). This is very useful in assessing issues of dynamic efficiency to which we shall briefly return below. The assumption that all mobile calls are terminated in the fixed net is extreme. It is made for the sake of clarity, since it ensures that the charge from terminating calls in the mobile network does not affect the cost of making mobile calls. So, this assumption fits most accurately a situation where charges for terminating calls in the fixed net are regulated (as is often the case), and where a very large proportion of calls from a given mobile network is terminated in the fixed net. Mobile-service providers incur a fixed cost per subscriber, F , in addition to a cost of originating a call, corig, which includes the charge for terminating a mobile call in the fixed net (which may be either regulated or freely set by the upstream monopolist), and a cost of terminating the call from the fixed net, cterm. Hence, if a given mobile subscriber makes q calls and receives x calls, then the total costs incurred by the network because of this subscriber are Cmob(q,x) = F + corigq + ctermx The mobile network sets a charge, t, for terminating a call to its subscribers originating in the fixed net. This is paid by the fixed network. Hence, the call-termination revenue associated with a mobile subscriber who receives x calls is Tmob(x) = tx and the ultimate profitability of this subscriber is
Πmob(ƒ,p,t) = ƒ - F + (p-corig)q(p) + (t-cterm)x(P(t)) to capture how per-subscriber profits depend on the choice variables of the mobile network: the fixed subscriber fee, ƒ , the price charged to subscribers per out-bound call, p, and the termination charged levied on the fixed network per in-bound call, t. This formulation of profits also shows immediately that to maximise profits the mobile network will maximise termination revenues, irrespective of the intensity of inter-brand rivalry between mobile networks. Thus, profits are separable, in the sense that revenues from termination are independent of the retail-tariff scheme. And, inter-brand competition relates only to the retailtariff schemes. Of course, this separability is not general. For example, if corig is related to the termination charge t through some kind of reciprocity,101 then there is a link between termination revenues and subscription revenues, and profit maximization will not be consistent with maximization of termination revenues. However, the point we want to make in this section is that the monopoly on termination will generally induce even highly competitive mobile service providers to levy termination charges that differ from the marginal cost of termination. Essentially, there are many separate relevant markets for termination in which even small and intensely competitive mobile-service providers have dominant positions (in fact, monopolies). Laffont and Tirole (2000, ch. 5) make this point forcefully by suggesting that small (mobile-phone) providers may behave more monopolistically than providers with larger market shares when (fixed-to-mobile) retail pricing cannot discriminate according to the identity of the terminating network.102 The point is simple when made in a setting where there is no competition for subscribers (but it carries wider implications). Suppose that on the access side the fixed network cannot discriminate between the mobile networks and must set uniform access prices depending on some averaging of the termination charges levied by the mobile-service providers. On the retail side, suppose that (due to regulation) the fixed-service provider cannot discriminate according to the identity of the terminating network.103 Then, if a small mobile-service provider increases its termination charge, it will have a very modest effect on the fixed-sector termination charge and a very small effect on the number of incoming calls. So, a small mobile network essentially has an incentive to maximise termination revenues. A larger mobile service provider is more concerned that raising the termination fee affects the number of incoming calls. Return to the base case where the profits defined above are driven to zero through tariff competition, that is, ƒ - F + (p-corig)q(p) + (t-cterm)x(P(t)) = 0 Further, we have marginal-cost pricing of out-bound calls, p = corig,104 and it follows that the fixed charge levied per mobile subscriber is given by ƒ = F - (t-cterm)x(P(t)) which is just the per-subscriber fixed cost minus the per-subscriber termination surplus. Thus, if t > cterm, we conclude that mobile subscribers are subsidized by the termination surplus. If t < cterm, then in-bound calls are subsidized by the mobile subscribers. However, to maximise profits a mobile-service provider will maximise termination revenue per subscriber Tmob = (t-cterm)x(P(t)) from which follows that t > cterm, and we conclude that termination surpluses will always subsidize subscriber fixed payments under the stated assumptions. Let tmonop = arg max{(t-cterm)x(P(t))} > cterm denote the termination charge that maximises termination surplus.105 We can now partially formalize the point made above on small and large mobile networks as follows. With several mobile-service providers and no discrimination based on terminating network, we could write the number of in-bound calls to a particular network, say, network i, as a function of a vector of mobile termination charges xi(P(t1,t2,…,tn)) This service provider maximises Timob = (ti-cterm)xi(P(t1,t2,…,tn)) If n is large and all networks are small (symmetry), then changes in ti will have a modest impact on xi(P(t1,t2,…,tn)) since the derivative δxi(P(t1,t2,…,tn)) / δti is small. Firm i effectively faces an inelastic demand, and it follows that there will be an incentive to set ti at a relatively high level. Suppose instead that n is small, in which case the impact of changes in ti on xi(P(t1,t2,…,tn)) is substantial. We conclude that demand is effectively more elastic, and the mobile service provider will be more moderate in its pricing of termination of fixed-to-mobile calls. The upshot of the preceding comments is that with imperfect competition in mobile call termination, the unregulated termination charges will be set above the cost of termination, t > cterm. This deviation from marginal cost pricing suggests that there might be a role for regulation to play. With this in mind, let us turn to the socially optimal mobile termination charge. This, however, will generally depend on the charge levied on out-bound calls from the fixed net. So suppose for simplicity that P(t) = C + t, where C is the marginal cost of the fixed network of originating a call and transporting it to the point of interconnection with the mobile sector. For the sake of clarity, we shall just assume that this determination of P(t) comes from the regulation of the fixed-service monopolist.106 If the profits are eliminated in both fixed and mobile service provision, welfare per mobile subscriber can be written as107 W = Vfix(C + t) + Vmob(corig) - (F - (t-cterm)x(C + t)) In this expression Vfix(C + t) captures the consumers' surplus of the fixed subscribers calling this mobile subscriber when P(t) = C + t,Vmob(corig) is the consumer's surplus of the mobile subscriber when the out-bound call charge is p = corig, while the last term is just the subscription fee of the mobile subscriber, ƒ = F - (t-cterm)x(C + t). Maximizing with respect to the mobile termination charge and noting that dVfix(P)/dP = -x(P), we immediately obtain the optimal termination charge as topt = cterm From this follows that ƒopt = F. Hence, the optimum entails marginal-cost pricing of mobile termination and no subsidization of mobile subscription fees through termination charge surpluses. Since the unregulated price is tmonop > cterm = topt with an associated subsidy to mobile subscribers embedded in ƒmonop < F = ƒopt we conclude that there is a role for regulation to play in setting the proper mobile termination charges. In the particular model under scrutiny, implementation of the optimum is very simple. Regulators should simply set t = cterm. Then the tariff competition for mobile subscribers will drive mobile profits to zero, that is, p = corig and ƒ = F, and the optimum is implemented. Recall that corig includes the charge levied on terminating mobile calls in the fixed network. Thus, under constant returns (as we have assumed), and if corig coincides with the total unit cost of origination, transportation and termination, then the optimum is first best. In contrast, if the break-even constraint of the fixed-service provider is violated (due to joint and common costs of the fixed-phone infrastructure), things are more complicated, which brings us back to the principles of optimal (fixed-net) access charges considered in von der Fehr (2004). At this point we shall not have more to say about this, and instead we turn to a brief discussion of some of the key assumptions made above. Assumptions and extensions Unregulated fixed-service provider topt < cterm and, under the stated conditions, it is optimal to let the mobile subscribers "subsidize" mobile call termination to counteract the fixed-phone mark-up. To see what is going on here, let us simply refer to the well-known double marginalization problem in a chain of monopolies.108 Two monopolies are stacked on top of each other. One supplies a vital input to the other, in the sense that outputs are produced from a fixed proportions technology where both parties supply one unit of complementary inputs.109 Then the first monopoly will charge a mark-up on its unit cost to equate marginal revenues and marginal costs. This mark-up factors into the costs of the second monopolist who will add another margin to equate its marginal costs and marginal revenue. The ultimate result is that the final price at the end of a chain of monopolies is higher than the price that would ideally be charged by a fully vertically integrated monopolist. In the most simple cases it would be simple to solve this problem by regulation. Regulators could simply require that all prices in the distribution channel coincide with unit costs. However, on the assumption that one of the parties cannot be regulated, partial regulation might still serve a useful purpose. In the particular case, on the assumption that it is impossible to regulate the fixed sector any further, it might be optimal to try to regulate the mobile sector in order to remedy problems arising from the fixed sector. Here, putting a below-cost cap on termination charges in the mobile net restores overall marginal cost pricing. Network externalities topt = cterm + s(n) > cterm where n is the number of mobile subscribers, and the function s(n) > 0 is the surcharge optimized to attract mobile subscribers. The bottom line is that compared to the case where t = cterm, it would be socially optimal to attract more mobile subscribers. Given the intensity of competition between mobile-service providers, subscribers can only be further attracted by "subsidies" to the fixed fee (e.g. hand-sets). The available instrument is to allow the competitive mobile-service providers to run a termination surplus, that is t > cterm. The higher mobile termination charge increases utility of mobile subscription, which in turn induces more to subscribe, and this is valuable for fixed subscribers. Therefore some measure of "subsidisation" of mobile subscription is socially optimal. It would be interesting to speculate (and formally analyse) what might be the likely properties of the s(n) function. For example, it might be suggested that s(n) is positive but decreasing to zero as mobile penetration increases from 0 to 100 percent. In other words, based on this type of argument and underlying model,110 it might be sensible to let the competitive mobile-phone providers enjoy their termination monopolies in the early phases of the industry, only to gradually tighten regulation of mobile termination charges as mobile penetration is completed. However, Armstrong (2002) notes that while consumption-scale-effects on the part of fixed subscribers (or all subscribers) may well imply that optimal termination charges are raised above the unit cost of termination, this does not generally imply that unregulated mobileservice providers would pick termination charges anywhere near the optimal ones. Hence, this type of network externality does not generally render regulation obsolete. Call externalities and internalization of caller welfare First, if mobile subscribers value in-bound calls as suggested in footnote 13 above, then we saw that unregulated mobile networks would reduce termination charges from tmonop(cterm) to tmonop(cˆterm) = cterm - urec < tmonop(cˆterm) where urec is the constant value attached to each in bound call by mobile subscribers. However, the socially optimal termination charge is topt(cˆterm) = cterm - urec < tmonop(cˆterm) So, despite the fact that this type of call externality puts a downward pressure on the unregulated price, the optimal regulated price also falls, and there will still be a gap between the private incentives of mobile-service providers and social values. Thus, this type of call externality does not obviate the need for regulation of mobile termination charges. Second, if mobile subscribers fully internalize the welfare of callers, then unregulated termination charges would coincide with the socially optimal charges. The only case where full internalization of the welfare of callers might be descriptively accurate is if both fixed and mobile subscribers belong to the same legal entity, so that charges levied are paid out of the same budget (e.g., households or firms), or some other type of close-knit organization (clan or club). Except for this limiting case, it makes sense to suggest that there is at most partial internalization of caller welfare on the part of the mobile subscriber, and this generally adjusts unregulated prices in the direction of the social optimum, but it does not close the gap. In this sense, the partial internalization of caller welfare reduces the need for regulation, but, again, it does not obviate the need completely. Dynamic issues First, if mobile telephony displays network externalities, in the sense that fixed-subscriber utility is increasing in mobile penetration, then, as already argued above, it is socially optimal to let termination charges rise above costs, in order to "subsidize" mobile subscription. If mobile markets are immature with few subscribers, and if further market penetration is socially valuable, but potential subscribers must in addition overcome switching costs given whatever technology they currently subscribe to, then a fortiori it is socially valuable to allow termination charges to rise above costs to "subsidize" mobile subscriptions. Both of these examples suggest that unregulated termination mark-ups are less of a problem in immature mobilephone markets than in mature markets. However, since basic mobile penetration throughout, e.g., the Nordic countries is arguably close to saturation, these arguments for light-handed regulation of mobile termination charges seem to lose weight at considerable pace. Second, and relatedly, if it is necessary and socially valuable to subsidize the introduction of new, advanced mobile technologies (such as 3G), then high mobile termination charges may follow more or less directly from Ramsey-principles. On this note, casual observation suggests that the zero profit constraints of mobile-service providers are more imminent than that of the fixed-sector incumbent. This should induce regulators to pause to think twice whether the time is yet ripe for time-invariant marginal-cost regulation of access to mobile networks from the fixed net. At the very least one should think carefully about how to regulate over time, that is, in both the mature and immature phases of various mobile-phone technologies. Of course, innovations originating in the fixed sector might also be desirable from the social perspective, and we return to some of the reasons displayed in von der Fehr (2004) for allowing the fixednetwork incumbent an access mark-up. The upshot of these brief comments on the proper dynamics of termination-charge regulation, is that, in the face of network externalities, what is optimal in the short run may differ markedly from what is optimal in the medium-to-long run.111 Receiving-party-payments Summary and conclusions In the base case we saw that separation of relevant termination "markets" and de facto monopolies on call termination will tend to lead to high termination charges, potentially far in excess of the socially optimal charges. It was even suggested that more intensive competition for subscribers would tend to increase mobile termination charges, since small operators view their returns as virtually separable into variable subscriber revenues and termination revenues. Small mobile-service providers perceive of termination charges as having little impact on the demand for fixed-to-mobile call termination. The resulting termination surpluses are then recycled into the competition for subscribers in terms of subsidies to mobile hand-sets and other fixed up-front investments. Both of these are frequently observed in practice. We can add to this that there may be a certain amount of ex post subscriber lock-in, which is not captured by the formal modelling above, but probably descriptively accurate. Markets with switching costs and ex post lock-in tend to generate even fiercer initial competition for subscribers (see e.g. Farrell & Klemperer (2001) for an extensive survey). This might push mobile-service providers further into the red in the short term, despite sizeable termination surpluses. This seems to be broadly consistent with recent empirical observations of the mobile-phone industry. It was argued that mobile termination charges are tempered somewhat by call externalities on the part of mobile subscribers and by the partial internalization of the welfare of callers by mobile subscribers.113 Neither of these, however, generally obviate the need for regulation completely. Finally, despite the static nature of the models alluded to, a few points on dynamic issues can be made. Tentatively, the models suggest that efficiency problems stemming from monopoly- on-termination in competitive networks may be less severe in immature than in mature markets, to the extent that either positive network externalities associated with mobile-phone services are significant or the (implicit) subsidization of new mobile technologies or other phone services is assessed as socially desirable.114 This might call for relatively light-handed regulation of the termination charges of competitive service providers in the immature phases of the service markets. In contrast, termination charges in the (by now) relatively mature markets for basic mobile services seem to be obvious candidates for intensified regulatory scrutiny. Rival Network Interconnection and Two-Way Access PricingThe analysis above was partial and incomplete, in the sense that it fully separated the setting of termination charges from the competition for subscribers. That is, we considered two-way access between non-rival networks. Generally, however, termination charges affect servicemarket competition, and as a result the various networks would set their termination charges with a view to how they affect their competitive position and that of their rivals in the service market. Hence, networks are rivals with respect to subscribers. Notably, it is of interest to study how termination charges that are set non-cooperatively compare to charges that might be agreed to through negotiation between the networks. Also, it is important to study how negotiated termination charges compare to the social interest. In particular, might termination agreements have competition-dampening or collusive effects on the service market? This is the main focus of the following. On the one hand, we notice immediately that a slight reinterpretation of the previous remarks on the monopoly-on-termination problem serves as a useful point of departure. Recall how we argued that even highly competitive mobile-service providers may have a de facto monopoly on termination in their nets. On the maintained assumption that fixed and mobile services are non-substitutable,115 a single (monopolistic) mobile-service provider would similarly set termination charges above the cost to maximise termination surplus. If, in addition, we assume that the single (monopolistic) fixed-service provider was also unregulated and free to set its termination charge, then it would likewise set a termination charge above costs. Essentially, under linear termination-pricing rules and in the absence of competition for subscribers, we have two monopolies selling vital inputs to each other, and the double-marginalization problem reappears. The resulting termination charges will be set too high from the perspective of the joint profits of the fixed and mobile industries and from the perspective of both fixed and mobile subscribers. In short, networks would have a strong incentive to join up and strike a deal over termination charges, and such agreements could well be in the interests of society as a whole.116 Note well the limitations of this. That an agreement might be in the social interest merely says that vertical agreements to internalize the vertical externalities ever-present in a string of uncoordinated monopolies may lower the retail price at the end of the distribution channel below the uncoordinated monopoly price. It does not say that the resulting price is anywhere near the socially optimal (cost-based) price. Vertical coordination or agreements will not undo the basic monopoly problem. On the other hand, models from the Industrial Organization literature suggest that wholesaleprice agreements may serve to sustain collusion in output markets and, therefore, be against the social interest.117 The basic argument may be illustrated as follows. Suppose two firms compete in standard Bertrand fashion in the output market, so that equilibrium output prices coincide with unit costs. Thus, final equilibrium prices are fully cost-determined. Now, if the firms could somehow agree to buy inputs at inflated prices from a commonly owned supplier, then product-market profits would still be zero, while the upstream supplier would run a surplus which is shared between the owners. In addition, if the linear input price is suitably chosen to coincide with the (hypothetical) output-market monopoly price, then total upstream profits to be shared between the two rivals would coincide with the integrated monopoly profit, and we could say that the wholesale-price agreement implements the fully collusive outcome without there being any sign of actual collusion in the output market.118 The extent to which these perspectives on coordination and agreements are relevant for twoway access between rival networks in telecommunications has been explored by Laffont, Rey & Tirole (1998a,b) and Armstrong (1998), and it is discussed by Laffont & Tirole (2000, ch. 5) and Armstrong (2002). A simple bench-mark model of two-way access between two rival networks is considered. This could represent either two fixed-phone providers or two mobilephone providers competing for the same potential subscribers. Alternatively, it could capture one fixed-phone provider and one mobile-phone provider on the assumption that fixed and mobile subscriptions are highly substitutable as seen from potential subscribers. Assume that differentiated networks are in place,119 and that they compete for subscribers using linear retail-tariffs.120 Access (termination) charges are set in reciprocal fashion either by negotiation or by regulation, and calling patterns are assumed to be balanced. Reciprocal termination charges simply mean that both networks levy the same per-call or per-minute termination charge on calls originating in the rival network. Balanced calling patterns imply that the probability that any given call is terminated on a particular network coincides with the market share of that network measured in term of subscribers. Thus, the fraction of calls from a network which will terminate off-net is proportional to the market share of the rival network. With this basic structure at hand, we shall, roughly, follow the structure of the previous section by first presenting some key results for the benchmark model, then discuss the assumptions and various extensions and finally provide a summary and some conclusions. Benchmark Retail-tariff schemes. Initially, we assume for simplicity that networks are restricted to setting uniform and nondiscriminatory retail tariffs. Thus, subscribers of network i are charged Si(q) = piq where pi is the per-call or per-minute retail tariff as above, while q is total call volume (including both on-net and off-net calls). Subsequently, we shall consider two "more realistic" retail-tariff schemes. This includes two- part tariffs Si(q) = ƒi + piq (as in the preceding section of this note) and network-based price discrimination Si(qh,qa) = ƒi + pihqh + piaqa where pih and qh are, resp., the tariff for and quantity of on-net calls, while pia and qa are the tariff for and quantity of off-net calls.123,124 Costs. The total cost of completing a call arise from origination, corig, transportation, ctrans, and termination, cterm. For simplicity it is assumed that the cost of origination and the cost of termination coincide, corig = cterm.125 Thus, under a reciprocal access charge t, we can make a distinction between on-net calls and off-net calls. We assume that the originating network carries the costs to the point of interconnection. Then, the perceived (and actual) cost of an on net call is ch = corig + ctrans + cterm = 2cterm + ctrans, while the perceived cost of an off-net call is ca = corig + ctrans + t - cterm. Thus, if the (reciprocal) termination charge, t, exceeds the actual cost of call termination, cterm, then ca > ch and off-net calls have higher cost than on-net calls. This, ultimately, forms the basis for raising one's rival's costs, but it also spells the (potential) demise of collusion as we shall see in the following. Based on this setup, two important observations can be made immediately. Observation 1: Given a balanced calling pattern, more calls originating in a given network terminate in the rival network the larger is the market share of the rival. Since termination charges have to be paid for off-net calls, this implies that if t > cterm, then the average126 marginal costs of a given network are increasing in the market share of the rival. For this reason, any given network has an incentive to build market share when there is a mark-up on termination. In contrast, if there is a mark-down on termination costs, t < cterm, then the average marginal costs of a given network are decreasing in the market share of the rival. Laffont & Tirole (2000, ch. 5) refer to these features as the endogenous-marginal-cost effect, which plays an important role in the assessment of whether large termination mark-ups (collusion) can be part of a stable market outcome. To formalize, consider network i, and assume that the rival network j has a subscriber market share of sj. Then, the (average) marginal costs of network i are ci = ch + sj(t - cterm) from which the above remarks follow immediately. In particular, if t - cterm is large – high negotiated termination mark-ups – then networks have a very strong incentive to grab market share from the rival to lower costs. This, essentially, is what will undermine large termination mark-ups. Observation 2: For given market shares, the perceived average marginal costs of a network are increasing in the termination charge. This follows immediately from the definition of (average) marginal costs above. Thus, when termination charges factor into the equilibrium prices of service-market competition, an increase in termination charges will increase equilibrium prices (for given market shares). This is the raising-the-rival's-costs effect of the Industrial Organization parable at the beginning of the section, and generally higher unit costs dampen the intensity of the competition for subscribers. Based on this model structure it is readily shown that to the extent that a unique equilibrium of the termination-charge-cum-retail-tariff game exists, it is characterized by a mark-up on termination costs, that is, tequil > cterm Essentially, the argument is that starting fromt = cterm, it is in the interest of both networks to increase the reciprocal termination charge and thereby dampen the intensity of price competition. Hence, negotiated reciprocal access charges have some competition-dampening or collusive potential as suggested by the reference to wholesale-price agreements. From the perspective of the social interest, termination charges in excess of the termination costs are clearly wasteful (given the maintained assumption of no joint and common costs of network operations). In contrast, to the extent that the services provided by the two networks are not perfect substitutes, the optimal termination charge is below costs, topt < cterm This mark-down on termination costs counter-balances the retail mark-up allowed by the imperfect retail-tariff competition. So, "subsidisation" of a key input restores balance. This is just a different version of the argument presented at the beginning of subsection 3.2. There, the mark-up embedded in fixed-phone calling-rates was optimally counter-balanced by a "subsidy" on termination in mobile networks. Here, the mark-up embedded in retail rates due to imperfect tariff-competition (product differentiation) is optimally re-balanced by a "subsidy" on off-net termination. In contrast, when t = tequil > cterm, then the networks (sub-optimally) "tax" each other. Note that as the mark-up on retail rates vanish when services become closer substitutes, then the socially optimal termination charge approaches the actual cost of termination from below. Thus, the discrepancy between tequil and topt as captured by tequil > cterm > topt suggests a scope for regulation of reciprocal access charges when service providers compete for subscribers using linear tariff-schemes. However, it should be noted that these result have abstracted from break-even concerns arising out of joint and common costs of network operation. For the usual arguments outlined in von der Fehr (2004) and repeated in the section on competitive bottlenecks, this balances the score-board somewhat. Joint and common costs of the network infrastructures put an upward pressure on the optimal, regulated termination charge, and in the abstract it is unclear whether topt ≥ / ≤ cterm Of course, it may just be the case that the need to cover common costs balances the requisite subsidy due to imperfect service-market competition so that topt = cterm But recall that tequil > cterm, and the potential need to regulate remains even in this case. The upshot is that private incentives and the social interest do not generally coincide under the stated assumptions, where networks compete in non-discriminatory, linear retail tariffs and termination charges are reciprocal. Assumptions and extensions Collusion and incentives tequil > cterm If the networks aim to maximise joint profits in the negotiations, then some fully collusive retail tariff, p*, should be attempted. This, in turn, requires some termination charge in excess of costs t*center = cterm + m* where m* solves for the collusive equilibrium price p* = p(m*) The profit-maximizing mark-up on the termination costs, m*, is positively related to the substitutability between network services and the maximum potential per-subscriber profits and inversely related to the price sensitivity of individual subscriber demands.127 In particular, we note that when services are non-substitutable,128 then the profit maximizing mark-up is zero, that is, t*equil = cterm. When, services are perfectly substitutable, as in the Bertrand example alluded to above, the profit maximizing mark-up on termination costs is maximal (since no further mark-ups are generated by the retail-tariff competition). Hence, when services become more closely substitutable, the termination mark-up is required to "do more work" to implement the fully collusive price p*. To see whether collusion is sustainable (cf. Laffont, Rey and Tirole, 1998a), we first note that with a given pair (p*,t*) = (p*,cterm + m*), the perceived marginal costs of each network are of the form c = ch + ½(t*-cterm) = ch + ½m* around the optimum. We recall the symmetry assumed, so that market shares are fifty-fifty, that is, si = sj = ½, when networks charge the same price. Further, when retail-tariffs coincide, termination payments balance. It follows that with product differentiation, there is no incentive for any of the networks to undercut the retail-tariff p* slightly, since the increase in revenue resulting from the increase in market share is just balanced by the increase in net termination payments to the rival network. A small downward deviation by one firm from the collusive retail-tariff opens a termination deficit not because of imbalanced calling-patterns, which we have assumed away, but rather because subscribers of the low-price network make more calls both on net and off net. Therefore, the net effect of a retail-tariff decrease by one network will be that more off-net calls originate in this network than in the rival network. A sizable cut in price below p* may, however, allow the deviating network to corner the market and decrease marginal costs to ct = ch since no calls are terminated off-net, if one network has a market share of one. Laffont, Rey and Tirole (1998a) show that such a cornering of the market is an optimal response to the fully collusive agreement, if services are sufficiently substitutable - this is the incentive to build market share at work. In other words, exactly when sizable termination cost mark-ups are needed to maximise joint profits will collusion be destabilized by opportunistic behavior in retail-tariff competition. A different way of phrasing this is that only moderate mark-ups, t-cterm, are sustainable. We therefore conclude that the collusive potential of termination charge agreements is somewhat limited. Note, though, that whether or not termination charge agreements can sustain termination charges far in excess of costs, the social interest requires that termination charges should be below cost in the absence of significant joint and common costs of network operations. If substantial joint and common costs have to be factored in to the termination charges, then the socially optimal, reciprocal termination charge may be above costs, and if the substitutability of services only allow modest mark-ups to be sustained, then it might just happen that it is expedient to leave networks to settle reciprocal termination charges. Two-part retail-tariffs si(q) = ƒi + piq Retaining the assumption that consumers are identical (except for their locations in the Hotelling space), it is immediate that the per-call charge of each network will be set equal to perceived marginal costs, pi = ch + sj(t-cterm) while the fixed fee will be used to appropriate consumers' surplus.129 In symmetric equilibrium, his reduces to pequil = ch + ½(t-cterm) which implies that networks make no (marginal) profits on calls. Instead, their entire profits are made from a mark-up on the fixed (monthly or quarterly) subscriber fee over and above the fixed subscriber cost F. As shown by Laffont, Rey and Tirole (1998a), competition in the subscriber fee130 implies that network profits are independent of the chosen termination charge. This is referred to as profit-neutrality of termination charges. Hence, whatever might be the negotiated reciprocal termination charge is not determined in this model. This immediately implies that termination-charge agreements between competing networks have no collusive potential when retail-tariff competition is in two-part tariffs. To see what is going on here, note that networks now have two instruments for subscriber competition given some termination charge t. Thus, to attract subscribers network i can lower fi, while keeping pi fixed. As long as pi is kept fixed, the balance of termination payments is unaltered. This implies that network i is free to use the extra instrument fi to increase market share without developing a termination deficit. In the base case with simple linear tariffs, the only instrument with which to attract subscribers was the retail tariff. When this was lowered to attract subscribers, a termination deficit arose, and this, of course, tempered the incentive of one network to undercut the retail tariff of the other. While termination-charge agreements may have no collusive potential in this version of the model, this does not imply that regulation is unnecessary, but it does imply that regulation has become a lot easier. Without regulation, the networks might agree to some reciprocal termination charge either above or below the cost of termination for reasons of habit. One interesting case might be that the firms simply agree not to bill each other ("bill-and-keep"), which is equivalent to t = 0. This implies pequil = ch - ½cterm < ch which is equivalent to below-cost pricing and significant "subsidisation" of off-net termination. This would of course be associated with high fixed subscriber-fees. From the perspective of the social interest, marginal cost pricing is optimal (as usual, in the absence of joint and common costs), that is, popt = ch This is easily implemented in the present version of the model by setting a cost-based termination charge, topt = cterm Implementation should be easy, since networks are indifferent between termination charges in a two-part tariff-regime, and they would therefore not object to the regulator's suggestion. Network-based discrimination Si(qh,qa) = ƒi + pihqh + piaqa where pih is the tariff for on-net calls, while pia is the tariff for off-net calls.131 If fi = 0 this reduces to third-degree price discrimination.132 In line with the preceding framework, networks levy call-charges that coincide with perceived marginal costs in equilibrium if fi > 0. By the symmetry and reciprocity of termination charges, for on-net calls we have p1h = p2h = ph = ch and for off-net calls we have p1a = p2a = pa = ch + (t-cterm) independently of market shares. The fixed subscriber charges coincide as well ƒ1 = ƒ2 = ƒ > 0 So, we note that the ratio of off-net to on-net call charges reduces to
Thus, if the termination charge exceeds costs, then off-net calls are "taxed", while if the termination charge falls short of costs, then off-net calls are "subsidised". Further, Armstrong (2002) notes that network profits now (again) depend on the termination charge, so that the profit-neutrality result under non-discriminatory two-part retail-tariffs no longer prevails under network-based discrimination. Hence, in principle, the collusive potential of termination-charge agreements is restored. However, to maximise profits it turns out to be optimal for networks to agree on a reciprocal termination charge of the form tequil = cterm - l < cterm where l is inversely related to the price sensitivity of individual subscriber demand. Hence, offnet calls are "subsidised" in the unregulated equilibrium - and more so when subscriber demand is inelastic. To understand what is going on here, we refer to the notion of network-based externalities suggested by Laffont, Rey & Tirole (1998b). Start by considering the case where subscribers are charged the same price for on-net and off-net calls (that is, pa = ph, as would be the case in the retail-tariff equilibrium if t = cterm. Then consumers have no particular preferences over which networks the people they call belong to. However, if t > cterm (hence, pa > ph, then callers would prefer receivers to be on the same net, and there is a positive network-based externality, since subscribers prefer to belong to a large network. In contrast, if t < cterm (hence, pa < ph), then callers would prefer receivers to be on the other net, and there is a negative network-based externality, since subscribers prefer to belong to a small network. Thus, network-based price discrimination induces network-based externalities, and the termination-charge agreement determines whether this network-based externality will be positive or negative. Effectively, when setting the reciprocal termination charge, firms decide whether the resulting subscriber competition is set against a background of positive network externalities (t > cterm), no network externalities (t = cterm) or negative network externalities (t < cterm). Since positive network externalities intensify the competition for subscribers to the detriment of network profits, in equilibrium the networks choose a structure with negative network externalities, and a termination charge of the form above follows immediately. Thus, to dampen competition, off-net calls are "subsidised", tequil = cterm - l < cterm.133 Note that private telecom networks aim to maximise profits. Thus, heuristically, at the initial termination-negotiation stage, they will aim to mutually agree to a termination charge which allows maximal profits in the equilibrium of the subsequent non-cooperative retail-tariff competition stage. Since "subsidization" of off-net calls diminishes the individual network gains from grabbing market share from the rival, "subsidization" is exactly what is called for to dampen the subsequent competition for subscribers. In other words, the mutual agreement on below-cost termination charges is like a credible and "soft" competitive posture.134 As in the previous case, it is in the social interest to set the termination charge at cost, that is topt = cterm Given the network agreement on a reciprocal termination charge of the form tequil = cterm - l < cterm there is consequently a potential role for regulators. It is interesting to note, though, that in this more realistic setting, it is the private incentives to "subsidise" off-net calls, rather than the collusive effects of high termination charges ("taxation" of off-net calls), which seem to be the main problem. In the particular setting considered, a suggested solution would simply be to ban termination-based price discrimination, which brings us back to the previous setting (profitneutrality), where networks would not object to the mere suggestion of cost-based termination charges.135 Asymmetric networks To set the scene, we first reiterate that when both networks have positive market share, an increase in the termination charge will increase the average cost of calls as above. More interestingly, the average cost of calls from the small network increases more than that of the large network, since a larger fraction of calls from the small network terminate off net. Thus, because termination costs generally feed into retail charges in favour of the large network in the subsequent rate-setting game, this could spell difficulties for reaching an agreement on termination charges. To keep the small network at a permanent disadvantage, the large network might insist on relatively high termination charges. Hence, in the asymmetric setting there is a potential additional role for the regulator, particularly in the early phases of the industry, where the small networks (recent entrants) may be very small, indeed, compared to the established network (the incumbent). Above, we concluded that with linear retail-tariffs, the networks would tend to agree to high termination charges in the symmetric setting. This qualitative result survives in the asymmetric setting (if an agreement on a reciprocal termination charge can be struck), at least to the extent that equilibrium retail-tariffs still coincide across asymmetric networks. The argument is more or less as above. As long as retail-tariffs coincide, net termination revenues are zero. Thus, an increase in the termination charge above costs will merely serve to dampen retail-tariff competition as before, and two asymmetric networks would also agree to an above-cost termination charge, if they can, indeed, agree. Turning to two-part retail-tariffs (without network-based discrimination) conclusions may change somewhat compared to the symmetric case as analysed by Carter & Wright (2003). With two-part tariffs, marginal retail charges are still cost based as above. When termination charges exceed costs, this implies that the average costs of the small network are larger than those of the large network. However, this may work in favour of the small network. When costs are relatively high for the small network, this implies that its resulting retail-tariff is also comparatively high. With comparatively high retail-tariffs, the small network generates a termination surplus. The reason is that the subscribers of the small network cut back on the number of calls (or call-minutes) compared to the subscribers of the large network. Hence, there is a net inflow of calls to the small network. As a consequence, the small network makes a positive net profit on termination. Therefore, the small network might be interested in agreeing on a relatively high termination charge. The large network, in contrast, would be interested in setting reciprocal termination charges at cost. If termination charges are set above costs, the subscribers of the small network make fewer calls than subscribers of the large network for the reason just explained. Hence, the large network generates a termination deficit. In contrast, if termination charges are set below cost, the subscribers of the small network tend to make more calls than the subscribers of the large network for similar reasons. Hence, a net outflow of calls from the small network will result. However, when termination charges are below costs this net inflow of calls to the large network will generate a termination deficit to the large network. Hence, the large network has an interest in pegging the reciprocal charge at the level of cost. This is the socially optimal level, and, provocatively, Carter & Wright (2003) suggest that the regulator simply delegate the setting of the reciprocal termination charge to the large network (viz., the old incumbent network) in this case. Intuitively, perhaps the most surprising of all the results above was that with two-part tariffs and network-based discrimination, the networks would agree to below-cost termination charges. Note, though, that this result arises in the mature and symmetric setting. However, there is reason to believe that the conclusion might well be sensitive to the symmetry assumption. If instead, the networks are, initially, highly asymmetric, a (very probable) conjecture is that the large, incumbent network would have an interest in high termination charges instead. The reason is that high termination charges increase costs and retail-rates for off-net calls, and since the small network terminates a larger fraction of calls off net, this benefits the competitive position and the profits of the larger network. Roughly similar remarks apply when either the incumbent attempts to limit the market inroads of an actual entrant or it attempts to deter entry altogether. Hence, the main regulatory concern may be high termination charges after all. Finally, let us make a few comments on entry and termination charges. An entrant must ensure access to terminate calls in the network of the incumbent. First, assume that due to switching cost, the entrant must offer lower fixed subscriber fees and lower call charges to attract subscribers. Then, it seems likely that the incumbent will attempt to insist on reciprocal termination charges at a relatively high level, since high termination charges weaken the competitive position of the entrant. To see this, note that the entrant might be doubly hurt by high termination charges. First, the small network (the entrant) has high unit costs since a larger proportion of calls from the small network terminate of net. Secondly, the fact that the small network has to set a lower price generates a termination deficit, since low retail-tariffs generate more calls per subscriber and a net outflow of calls. Hence, insisting on high termination charges is akin to a "tough" strategic posture by the incumbent, which will serve to limit the competitive position and scale of the entrant. Secondly, so far we have assumed that the two networks compete for a fixed pool of subscribers, and that both networks have full coverage from the outset. In an asymmetric (incumbent vs. entrant) setting neither of these are necessarily sensible assumptions. In particular, the full-coverage assumption is suspect as far as recent entrants are concerned. How this affects the incentives of the incumbent with respect to the level of the reciprocal termination charge is not entirely clear. For a given coverage by the entrant, the incumbent may have an incentive to insist on a high termination charge to weaken the short-run competitive position of the entrant. However, a high charge may also give the entrant a stronger incentive to build market coverage more quickly, which might adversely affect the profit prospects of the incumbent in the longer run. Summary and conclusions We started by establishing the general point that networks have a private incentive to negotiate termination charges to internalize double-marginalization externalities. If networks act in non-coordinated fashion, the resulting outcome is likely to be one in which termination charges are very high, even too high from the combined private interest of the networks. Thus, networks have an incentive to coordinate on lower termination charges. The associated internalisation of the vertical externality would also generally be in the interest of society as a whole. This is most simply illustrated by the virtues of agreements on international call charges. However, this does not generally imply that unregulated termination settlements made by competing networks will be fully (or even pre-dominantly) in line with the interests of society as a whole. This was illustrated with reference to a symmetric duopoly model where the two networks first agree to a reciprocal termination charge and then compete for subscribers by their choice of retail tariff-schemes. Several modes of competition were considered. With linear and non-discriminatory retail-tariff schemes, it was argued that networks have an incentive to agree to above-cost termination charges - that is, "taxing" off-net calls. The social interest, in contrast, dictates below-cost termination charges to counter-balance imperfections in service-market competition. Thus, it is optimal to "subsidise" off-net calls. We conclude that there is a general role for regulation of termination-charge settlements. However, it was also argued that the collusive potential of negotiated termination charges should not be overstated, since it turns out that network incentives to build market share will undermine the stability of high, negotiated termination charges. Thus, whether networkservices are highly substitutable or not, relatively modest mark-ups on termination costs are the most likely outcome of private settlements. When networks are free to employ simple two-part retail-tariffs, results differ markedly. First of all, the variable retail rates will coincide with perceived marginal costs of termination. This enables the fixed subscriber-fee to perform the task of competing for subscribers without generating termination deficits. It follows that termination charges are profit-neutral and, therefore, do not have any collusive potential. However, this leaves the "problem" that termination charges are not strictly determined. So, we cannot render a verdict on whether private settlements will lead to termination charges that are too high or too low from the perspective of the social interest. However, this leaves the simple role for regulators to suggest cost-based termination charges, since pricing at perceived marginal cost in retail competition will be aligned with the social interest, provided that the perceived costs coincide with actual costs. Networks are indifferent as to the level of the termination charge and would, therefore, not object to such a suggestion. Hence, policing of the regulatory regime should be inexpensive. This section also considered network-based tariff discrimination, which captures that retail callrates discriminate between on-net and off-net calls. On the assumption that fixed subscriberfees complete the retail tariff, all retail call-rates coincide with perceived marginal costs. Termination charges are no longer profit-neutral and, thus, have collusive potential, but due to network-based externalities, it is optimal for networks to agree to below-cost termination charges - that is, "subsidising" off-net calls. This "subsidisation" dampens the competition for subscribers and maximise network profits. The social interest dictates termination charges at actual marginal costs. Therefore, there is a role for the regulator to eliminate this "subsidy" to off-net calls. Regulation can either directly insist on cost based termination charges or it can ban network-based tariff discrimination and merely suggest cost-based termination charges. Finally, some tentative remarks were made on the sensitivity of the main conclusions to the symmetry assumption. If asymmetry is taken to refer to a situation where a large incumbent firm is facing a small actual or potential entrant, then the weight of the argument indicates that the incumbent might tend to favour higher termination charges, in order to disadvantage the small network. Thus, if the incumbent is able to force high reciprocal termination charges on the entrant, then regulators should certainly be vigilant. However, it was also suggested that cases arise where the incumbent interest as far as termination charges are concerned may be in line with the social interest. Concluding RemarksThis note has attempted to provide a partial overview of key results from the economics literature on two-way access charges. The discussion was naturally split between settings in which networks are non-rival and settings in which networks are rival with respect to subscribers. The two sections have each suggested simple modelling frameworks suitable for discussing two-way access problems, and have presented some key insights, which should be useful for regulators. In particular, the note has tried to highlight the necessary changes in regulatory emphasis as telecommunications markets pass from immature to more mature phases. Footnotes92 This appendix is drafted by Per Baltzer Overgaard, Professor, Ph.D. , University of Aarhus 93 That is, calculation of access prices based on actual costs, modelled costs and "price minus". 94 Rate-of-return, price cap and revenue cap regulation. 95 See Tirole (1988). 96 4Laeont & Tirole (2000, ch. 5), in turn, rely heavily on Laffont, Rey & Tirole (1998a,b). 97 More generally; communicate with. 98 Whether the competitive providers must also have access to each other depends on the details. We start out by suppressing this, and then re-introduce it subsequently as we move to the second structure. 99 We realize that in this section where reciprocal (two-way) access is required, it may be a little misleading to refer to the mobile sector as downstream and the fixed sector as upstream. Since both sectors buy vital inputs from each other, inputs are not simply and unambiguously moved in one direction while being processed into outputs. However, since the companion paper by von der Fehr (2004) on one-way access deals with one vital input passing from the fixed sector to the service sector, we shall occasionally continue to refer to the fixed sector as upstream and the mobile sector as downstream. 100 When applied to fixed-mobile interconnection, it is typically unrealistic to assume that the fixed-service provider is not also a competitor for mobile subscribers. Recall, through, that fixed-mobile interconnection is but one application. There might be other areas of service provision, where the fixed service provider is not a competitor either by choice or by regulatory exclusion. Also, by its simplicity the model is highly relevant for capturing the key elements of situations where the fixed-service monopolist is regulated in most of its lines of business, whereas competitive (downstream) service providers are not. 101 Recall that corig includes the cost of terminating a mobile call in the fixed net. Therefore, the termination charge levied by the mobile network could well affect the charge levied by the fixed network. For now, let us simply assume that regulation of the upstream monopoly prevents this. In the next section, we return to this issue. 102 E.g., for reason of regulation of the fixed-service monopolist. 103 Alternatively, Gans & King (2000) suggest that the demand for fixed-to-mobile calls is some function of average prices across mobile networks, since fixed subscribers may not know which mobile network they are calling. Then the charge of a small mobile network has little effect on the average price and on the demand forthcoming. 104 This is essentially by standard, Bertrand arguments. If there is a per-call mark-up, some firm would lower the per-call rate slightly, attract all customers and make a positive profit. If there is a mark-down, some firm would increase the per-call rate and decrease the fixed fee by enough to attract customers, yet have enough cost-saving per customer to make it worthwhile. 105 Armstrong (2002) makes the further point that if a mobile subscriber values in-bound calls at some number urec per call, then it is equivalent to lowering the unit cost of termination from cterm to cterm - urec. Hence, if (perfect) competition for mobile subscribers is interpreted as maximization of mobile subscriber utility subject to the break-even constraint of the mobile networks, then the outcome (equilibrium) would embody that mobile termination charges solve max{(t - (cterm - urec))x(P (t))}, which would imply that t < tmonop. Hence, appreciation of in-bound calls puts a downward pressure on mobile termination charges. We return to this below. 106 Armstrong (2002) alternatively assumes that fixed-service provision is perfectly competitive. Either way, the point of departure is that fixed subscribers are charged a price for calls to the mobile networks which coincides with the actual marginal costs of the call (which includes the mobile termination charge). We shall subsequently comment on the case where P(t) > C + t. 107 Strictly speaking, this requires some additional assumptions, including that the demand for mobile subscriptions is inelastic over the relevant range of tariffs (e.g. mobile market is covered or fully penetrated) and that mobile termination charges do not affect the profitability of fixed service provision in any other line of business (either because the fixed service provider has no other lines of business or because they are unrelated to (no substitutability or complementarity) fixed and mobile telephony or because these other lines of business are suitably regulated). 108 See, e.g., Tirole (1988, ch. 4). Laffont & Tirole (2000) alternatively refer to this as the pancaking problem. 109 17Fixed proportions is not critical to making the point, but is makes the presentation much simpler. Without fixed proportions other inefficiencies potentially arise when regulators start twisting relative prices. This latter point is also touched by von der Fehr (2004) in the discussion of one-way access and bottleneck bypass. 110 Which remains to be worked out. 111 At this point, we should caution the reader that the formal economics literature on the dynamics of optimal regulation of two-way access in telecommunications is itself very immature, indeed. 112 A detailed discussion of the virtues and vices of RPP is well beyond the scope of this short note, but see Armstrong (2002) and Laffont & Tirole (2000, ch. 5) for some discussion. In particular, with reference to the US, where the use of RPP is widespread in mobile telephony, it has been argued that this has held back mobile penetration significantly, see the comments in Canoy, de Bijl & Kemp (2003). 113 And, possibly, by receiving-party-payments. 114 It should be remarked, though, that liberal treatment of such indirect subsidies to competitive services will not generally be first best. Conceivably, there might be alternative instruments available to regulators (or, government agencies more generally). For example, as far as providing incentives to introduce new technologies, more effective instruments might be available through adjustments to the corporate tax code. 115 That is, there is no competition between providers of mobile and fixed services. 116 This, essentially, forms the basis of discussions of termination-charge agreements for international calls when national monopolists serve national markets without any interference from competition for its subscribers. Hence, the national networks are non-rival. Each network would non-cooperatively try to maximise termination surplus from in-bound international calls. Pricing of out-bound international calls would then be based on the monopolytermination charge of the other network. Thus, another mark-up is added to an already high termination charge and origination cost. Hence, the resulting international call rates could be very high indeed. The joint incentive for international call settlements is therefore strong. For further discussion of international call charges and regulation, see Armstrong (2002). 117 Note that, from a modelling perspective, access or termination charges are similar to wholesale or input prices in the more familiar Industrial Organization context. 118 This, of course, is the basis for antitrust concerns about wholesale-price agreements. The conditions under which the setting up of a jointly owned input supplier from which firms buy exclusively is part of an equilibrium of a properly specified game remain to be explored, but this is beyond the scope of this presentation. Laffont & Tirole (2000, ch. 5) refer to patent pools. Firms sink their vital patents into a jointly owned company from which outputmarket rivals license patents at inflated prices. If the license prices are suitably chosen, the patent company may appropriate monopoly profits, which are then distributed between the owners. 119 Implicitly, the base case assumes that the market is mature. 120 We shall have much more to say about various, more general retail-tariff schemes in the discussion of the model and its assumptions. We start out with the most simple, linear schemes to make a couple of basic points on the potential competition-dampening or collusive effects termination-charge agreements. 121 E.g. at opposite ends of the Hotelling line with consumers being uniformly distributed on the line. Thus, if network tariffs coincide, each network will have half the consumers as subscribers. Thus, the services offered by the two networks are differentiated in the usual sense of the Hotelling model. 122 Compared to the case alluded to above where networks set their termination charges non-cooperatively (doublemarginalization), we expect the networks to agree to a lower reciprocal termination charge. In a fully symmetric model, though, non-cooperative termination charges would also coincide. 123 h and a are mnemonic for "home" and "away". 124 In the latter case, it is also possible to make the distinction between fi = 0 and fi ≠ 0. 125 The assumed coincidence of origination and termination costs is without loss of generality. Also, the size of ctrans plays little role. In contrast, the symmetry across networks is central. 126 This means average across all calls originating in the network. 127 In the standard Hotelling formulation subscribers are identical except for some location parameter. 128 As in the example in footnote 24 on international call charges. Similarly, if fixed and mobile services are largely non-substitutable and provided by different networks, the same remark applies. 129 Note that (as above) high termination charges feed into retail tariffs. So, high perceived marginal costs still dampen retail-tariff competition. 130 In much the same way as in the previous section with competitive mobile-service providers. However, in that model, net profits were driven to zero - not so in the imperfectly competitive Hotelling framework. 131 To motivate this, note that when the termination charge differs from costs, then a network has different costs of terminating calls on-net and off-net. Thus, based on costs considerations networks would have an incentive to charge subscribers differently for on-net and off-net calls if possible. 132 Laffont, Rey & Tirole (1998b) consider both cases. 133 Laffont, Rey & Tirole (1998b) also consider the case where fi = 0 (third-degree price discrimination). Then, we are back in a linear-pricing framework as in the benchmark, where competition-dampening is aided by above-cost termination charges. It can be shown that with sufficient network-service differentiation, networks will agree to above-cost termination charges, although the collusive potential of termination-charge agreements is diminished by the on-net/off-net price discrimination. 134 For analyses of commitment and strategic posture in dynamic games, see the seminal work by Fudenberg & Tirole (1984) and Bulow, Geanakoplos & Klemperer (1985) and as well as the survey by Tirole (1988, ch. 8). 135 Hence, enforcement costs should be modest. 136 We repeat our caution that the economics literature on two-way access in telecommunications is, itself, rather immature. Therefore, the following is somewhat tentative. Version 1.0 October 2004 • © Danish Competition Authority. |
