Coasian A Alternative to Pigovian Regulation of Network Interconnection
Jay M. Atkinson and Christopher C. Barnekov*
September 2004
Working Draft Please check with authors before citing [e-mail]
*The authors are senior economists in the U.S. Federal Communications Commissions Wireline Competition Bureau. Views and conclusions are their own and do not necessarily reflect those of the Commission or of other FCC staff.
A Coasian Alternative to Pigovian Regulation of Network Interconnection
Jay M. Atkinson and Christopher C. Barnekov1 The kind of situation which economists are prone to consider as requiring corrective Government action is, in fact, often the result of Government action. R.H. Coase, 1960. 1. Economists and regulators have generally viewed network interconnection as involving significant externalities. This article proposes an efficient, non-arbitrary method of assigning interconnection costs that avoids externality problems. This facilitates efficient market outcomes by enabling subscribers to choose networks based on their true costs and benefits. Such a choice is not possible under current regulatory schemes for determining intercarrier compensation, which mingle the costs of individual networks. The current regime can be described as following what Coase describes as "the Pigovian tradition."2 That is, it involves regulatory rate-setting that is meant to offset externalities; but it actually exacerbates rather than resolves externality problems. Our proposed alternative is based on Coases recipe for resolving externalities by clarifying property rights rather than by setting Pigovian taxes. It also extends Coases analysis to deal with the presence of market power. 2. Our goal is to find a default rule that, ideally, need never actually be applied by a regulator. The point of a Coasian approach is that if rights (i.e., default rules) are well defined, the parties can and will work out efficient resolutions (more efficiently than a regulator could achieve). We develop a simple default rule that assigns interconnection costs in an efficient, non-arbitrary and competitively neutral3 manner, even when one carrier has market power. The rule is first to identify those facilities that are solely
1
The authors received very helpful feedback, corrections and suggestions from a number of FCC colleagues. Any remaining errors are our own, and the views and conclusions expressed do not necessarily represent the views of the Federal Communications Commission or of other FCC staff.
2
Coase (1960), pp. 39-42, describ ing Pigous views as expounded in A.C. Pigou, The Economics of Welfare (London: Macmillan, 1920).
3
By competitively neutral, we mean that the interconnection regime itself confers no special advantage or disadvantage on any carrier or technology. Whatever advantages or disadvantages a particular carrier or technology possessed prior to interconnection should remain undistorted by the interconnection regime. In such a system, customers can make efficient choices based on the actual costs and benefits of using particular carriers or technologies. This corresponds to the definition of competitive neutrality adopted by the FCC for Universal Service purposes. See In the Matter of Federal-State Joint Board on Universal Service, CC Docket No. 96-45, 12 FCC Rcd 8801, 47.
CoasianAlternative040901b.doc
09/08/04
incremental to interconnection, then to split the cost of providing these facilities equally between the two carriers. Each carrier would recover these and all its other costs from end user customers, not from interconnecting networks. 4 For several basic types of networks, we demonstrate below that this simple default rule results in efficient, competitively neutral interconnection. We argue that this result can also be generalized to more complex networks. We believe this rule provides a conceptual solution5 to the problem of interconnection between dissimilar networks in the presence of market power, 6 and that it provides a default that can enable interconnectors to reach competitively neutral and, with respect to interconnection, efficient outcomes. 3. Our analysis differs sharply from most of the recent literature on telephone pricing. Suppose that the structure of the industrys cost functions is quite different than that assumed in most of the recent literature, and particularly that this structure differs from the one regulators assume in setting intercarrier compensation rates. Specifically, although these rates are based largely on minutes of use, we do not believe traffic volume is an important determinant of network costs today. Suppose also, at least for present purposes, that the regulator focuses entirely on achieving efficient and competitively neutral intercarrier compensation, 7 leaving the parties to a call to resolve any call externalities 8 that may arise between them. Our results flow from these presuppositions, and we state them clearly at the outset because they differ from the presuppositions implicit in most of the recent literature. The test of assumptions and analysis is their predictive power. In the final section of this paper we present some evidence that our predictions meet this test.
4
A recent paper examines the origins of the belief that a phone company which terminates a call from another must receive compensation from the originating carrier to avoid a taking of private property, and concludes that an interconnection rule such as we suggest does not constitute a taking. Adam Candeub, Network Interconnection, Takings, and Coase, 54 Syracuse L. Rev. 369 (2004) (forthcoming).
5
In the final section, we discuss the extent to which this conceptual solution could actually be implemented in the real world.
6
Where no carrier is market dominant, interconnection need not be mandated or regulated, because it is in both carriers interest and neither can exploit the other. For a brief summary of the extensive literature on the need for regulating interconnection in the presence of market power, and also of the manner in which current access regimes artificially create market power for even small terminating carriers, see DeGraba (2000), pp. 7-8, 14, 35.
7
At the risk of belaboring the point, we do not seek to achieve optimal retail pricing through manipulating interconnection terms. In fact, our experience as staff economists for a regulatory commission persuades us that this is neither feasible nor even desirable. Once an efficient and competitively neutral interconnection regime is in place, we believe that any retail problems that remain would become much more tractable. Simply achieving such an interconnection regime would be a major regulatory accomplishment.
8
By call externalit ies, we mean one party may place a higher value on a call than the other party. Economists do not worry about the distribution of benefits in most transactions, as long as both parties are willing participants. Calls that persist more than a few moments meet this test. Recent developments such as Caller ID and enforceable do not call lists make this even more likely than before. Parties to a call may share costs in various ways, even including one party paying the other (for example, a call to a consultant or technical support service).
2
CoasianAlternative040901b.doc
09/08/04
A.
Coasian vs. Pigovian Approaches to Network Interconnection
4. The FCCs current local interconnection rules view traffic passed from one network to another as a harm (cost) imposed by the originating network on the terminating network. 9 In the Pigovian tradition, the regulator specifies the form and level of a tax imposed on the party generating the externality (considered to be the originating network) tha t is meant to internalize this external cost. The FCC expresses this as assigning network costs to the cost causer.10 The FCC requires local exchange carriers (LECs) to provide specified services to interexchange carriers (IXCs) at regulated rates set equal to its estimate of the costs imposed. Although in this case the payment is made to the terminating network rather than to the government, it is a tax in the Pigovian sense in that it is fixed by the government to offset an external cost and is not subject to negotiation by the parties. As Coase points out, however, even if the tax is exactly adjusted to equal the damage [cost] the tax would not necessarily bring about optimal conditions.11 This is because, as Coase demonstrates, the rigidly specified payments inhibit potentially efficient market adjustments. 5. Furthermore, the Pigovian tradition is based on the rather problematic assumption that an all-knowing regulator is able to discern the correct form and level of a tax that just compensates for the negative externality. If the regulator misperceives the structure or level of the costs, the fact that market adjustments are inhibited means serious inefficiencies can result. 6. In fact, regulators do appear to have misunderstood profoundly the structure of interconnection costs. In brief summary, regulators (and many economists) treat the long run marginal cost of a telephone call as equal to its long run average cost. We discuss in Section D below why we believe the number of calls or minutes of use is the wrong output variable. That is, we will argue that usage levels have very little impact on total costs, even in the long run. 12 Instead, the primary determinants of network costs today are the number of subscribers and their locations. In Section E below we discuss patterns of carrier behavior that lead us to this inference. Yet many persist in viewing the incremental cost of interconnection as equal to the average total cost per minute of use of
9
This describes local interconnection, which is made under the reciprocal compensation rules (47 CFR 51, Subpart H). Where a separate carrier provides interexchange (long distance) services, it falls under the exchange access rules (see 47 CFR 61 and 69). In this case the interexchange carrier (IXC) is viewed as imposing a cost on both the originating and the terminating LEC. The IXC is required to pay both LECs when it carries a call between them.
10
The term cost causer or a variant such as cost causation appears 49 times in the FCCs recent reworking of its exchange access rules, in which the principal underlying rationale is conforming rate structure to cost causation patterns. See Access Reform First Report and Order of May 1997 (12 FCC Rcd 15982).
11 12
Coase (1960), p.41.
In more precise terms, we will argue that, if the long run cost of constructing a network is X, the long run cost of constructing a network with the same number of subscribers (in the same locations), but twice the expected traffic volume, would be very much closer to X than to 2X. In fact, this long run total cost may be nearly indistinguishable from X, that is, small enough to be lost in the noise.
3
CoasianAlternative040901b.doc
09/08/04
the entire network, or sometimes of a portion of it. Recent experience demonstrates, as discussed below, that when regulators mandate intercarrier compensation based primarily on minutes of use, while costs are primarily a function of subscribers and their locations, massive distortions are inevitable. 7. As a consequence of this profound misperception of the industry cost structure, regulators inadvertently created opportunities to game the system that produced serious market distortions 13 and wealth transfers of billions of dollars. 14 The current Pigovian approach to interconnection transforms substantial network costs into common costs, and then allocates them among carriers in ways that generate significant externalities. It then inhibits carriers from negotiating efficient resolutions. Instead of correcting externalities, the current regime generates and exacerbates them. 8. The network interconnection problem is also more subtle and complex than a Pigovian taxation approach permits. A carrier that receives traffic from another generally benefits from the exchange. Subscribers typically value the ability to communicate with a larger callable universe, so interconnection increases a networks value to potential subscribers. 15 Because regulators have no way of ascertaining how benefits are distributed between the parties to a call, the cost causation approach arbitrarily assigns all the benefits to the calling party. 16 The Pigovian approach of taxing to offset external costs does not account for external benefits enjoyed by the terminating network. In fact, it makes it more difficult for calling and called parties to internalize these call externalities because networks costs are mingled. 9. It seems reasonable to suspect that a Coasian approach might offer a better solution. Forty- five years ago, Ronald Coase analyzed the FCCs management of radio spectrum in an article (Coase 1959) that led to what has come to be known as the Coase Theorem (Coase 1960). After only about three decades, Congress permitted the FCC to implement, in part, 17 Coases prescription for spectrum management. The public has benefited greatly as a result through more efficient spectrum licensing and Congress has collected considerable additional revenues. Like the spectrum management problem, the current interconnection problems can be viewed as a consequence of vaguely defined
13
See ISP Remand Order, 16 FCC Rcd at 9182-83, para. 68-71; see also Intercarrier Compensation NPRM, 16 FCC Rcd at 9616, para. 11.
14
For example, incumbent LECs claim that competitive LECs exploited an unforeseen aspect of the reciprocal compensation rules, producing rapid growth in dial up Internet access traffic during the late 1990s. They state that within three years incumbent LEC payments to competitive LECs rose from small amounts to over $2 billion annually (Letter of W. Scott Randolph, Verizon Communications, to Magalie R. Salas, Secretary, FCC, November 1, 2000 in CC Docket No. 99-68).
15
This is sometimes referred to as the network externality, although it is by no means the only externality involving networks.
16
This emphasis on what is called directionality of traffic has led to unending difficulties and new externalities, because it is not relevant to real costs and is fairly easily manipulated by the parties. See discussion below.
17
Hazlett (1998 and 2001) discusses the benefits arising from the limited application of Coases spectrum prescription and the likelihood of additional benefits from a further application.
4
CoasianAlternative040901b.doc
09/08/04
property rights and of rules that prevent markets from internalizing the externalities arising from these poorly defined rights. A problem of such a nature invites a Coasian solution. 10. Are Coases insights relevant for telecommunications, an industry characterized by significant market power in many local exchange markets? We argue below that they suggest a fairly simple, efficient solution to an increasingly complex and otherwise intractable set of regulatory quandaries. From a regulatory economists perspective, Coases most valuable insight is that, where transactions costs are low, clearly defined property rights enable competitive markets to resolve externalities efficiently without a regulators intervention. This suggests an alternative to the traditional (Pigovian) approach to resolving externalities. But Coase assumed competitive conditions. What if there is market power? We show that although the presence of market power means it does matter where one draws the property lines, 18 one can do so in a manner that enables the market to work. We find that if property rights are carefully defined, a simple default rule can enable the market to resolve network interconnection externalities efficiently, even in the presence of monopoly. 11. A Coasian approach, that is, regulating by defining property rights rather than by setting prices, offers at least two non-trivial advantages. The first is that regulatory errors are less harmful. As Coase (1960) shows, the parties can often rearrange rights through market transactions to achieve efficient outcomes even if the regulator errs in her initial assignment. The second advantage is that regulatory proceedings are by their inherent nature better suited to sorting out property rights than to discerning or adjusting to constantly and rapidly changing market conditions. Procedural safeguards and due process requirements are indispensable in a free society, but they would slow decisions and limit a regulators flexibility, even if she were omniscient. Thus legal proceedings are better suited to sorting out property rights than to fixing prices. Regulators are less likely to err, and their errors are likely to do less damage, if they assign property rights and leave price setting to markets. 12. The FCCs interconnection rules were developed mainly in the context of exchanging telephone voice traffic among monopoly carriers, and largely for the purpose of protecting a system of cross-subsidies from being undermined by competition. 19 As new forms of telecommunications arise and new competitors enter the market, however, the rules must facilitate efficient, competitively neutral interconnection among many technologies and many firms. We seek a simple, efficient default rule that is independent of technology and resistant to market power. 13. By promising an efficient solution to interconnection, our approach differs fundamentally from previous economic literature that proposed only second best solutions. 20 These solutions involved prescribing a vector of inter-carrier rates for
18
In Coases perfectly competitive environment with negligible transactions costs, it mattered only that the rights were clearly defined.
19 20
Brock (1994), Chapter 10. See, for example, Laffont & Tirole (2000), pp. 102-4.
5
CoasianAlternative040901b.doc
09/08/04
various interconnection services in order to produce a desired vector of end user prices. This required allocating substantial common costs, typically using a Ramsey pricing approach. Such an approach does not produce efficiency but rather attempts to minimize inefficiency. 21 This approach has never worked well, but it produced tolerable results as long as there were only a few carriers, they offered only a few services, these services were technologically similar, and efficiency in telecommunications was not so terribly important as today. 22 None of these conditions is true today. Our diagnosis is that interconnection problems arise primarily as a result of ill-defined property rights. Our remedy is not to prescribe rates, but to clarify rights. B. Two contrasting FCC responses to market power problems: one Coasian, one Pigovian
14. Our approach is informed by two FCC experiences in the 1980s that attempted to deal with market power by separating competitive and monopoly network segments. In both cases, a clear demarcation was made between more and less competitive segments. In the case of Consumer Premises Equipment (CPE), the FCC applied what one might describe as a Coasian interconnection rule. That is, it set boundaries (i.e., assigned property rights) rather than regulating rates. In contrast, after the court-ordered divestiture of AT&Ts local exchange services, the FCC applied a Pigovian interconnection rule in creating the exchange access regime. That is, it specified the terms and rates under which local and long distance carriers were to exchange traffic. The Coasian CPE deregulation has been an unqualified success, while the Pigovian exchange access system has produced ongoing problems. 23 15. The FCCs 1980 Computer II decision24 deregulated CPE by creating a demarcation point, the network interface device (NID), between the local exchange network and the internal wiring and equipment on the customers premises. Computer II gave customers complete control of (and responsibility for) the wiring and equipment on their side of the NID. This meant LECs could no longer control or charge for end users purchase or use of customer premises equipment that met FCC technical standards. This ended a long struggle over telephone company policies that strictly required that any equipment a customer wished to add or connect to her telephone must be supplied by the telephone company. The result was an explosion of innovation and new services, as well as sharp price reductions for CPE.
21
Telling policy makers that there is no efficient solution also exacerbates temptations for them to engage in wealth redistribution among rent-seekers, but that is another paper for another venue.
22
In addition, limited relief from the resulting inefficiencies was afforded by the availability of special access services that allowed business users to bypass the public switched telephone network at reduced rates. This, of course, has the side effect of diverting traffic from the public switched telephone network.
23
A Pigovian might protest that these problems stem largely from the fact that the access pricing system was used to administer a burdensome cross-subsidy scheme. We agree that this universal service funding mechanism did greatly exacerbate the difficulties, but argue below that the underlying problems are inherent in a Pigovian mechanism.
24
77 FCC 2d 384 (1980).
6
CoasianAlternative040901b.doc
09/08/04
16. Prior to its deregulation, CPE had been a profitable venue for LEC price discrimination. LECs priced CPE usage as many discrete services. They could, and did, charge usage fees for every jack and every piece of equipment the customer wished to attach. 25 Regulators could not solve the common cost allocation problems to find the cost of these items. Instead, telephone companies based pricing primarily on marketing estimates of demand elasticities for particular services. 26 State regulators also frequently pursued a Ramsey pricing policy of increasing charges for optional equipment in order to keep basic rates low. Both LEC price discrimination and these state policies drove a wedge between prices and social costs for these items and contributed to pressures for competitive entry. 27 17. As the range of items that might usefully be attached to telephones broadened and third party providers began attempting to supply CPE, it became increasingly apparent that the CPE market could be competitive. This undermined AT&Ts end-to-end control of the network, and especially its price discrimination policies. The FCC initially supported AT&Ts position, but courts kept striking down FCC decisions that were based upon AT&Ts demonstrably false arguments that permitting third party CPE might physically damage the network. 28 This led to a regulatory dilemma. If LEC-supplied CPE remained regulated, how could regulators set the appropriate prices? And how could LECs respond to competitors pricing? If CPE were deregulated, how could regulators assure that the LEC would not shift its CPE costs to the regulated network? It might then either engage in predation financed by regulated rate-payers, or simply transfer excess profits to the deregulated sector. The FCCs eventual solution was the NID, which separated the competitive CPE segment from the monopoly local exchange network. If LECs were to offer CPE, they could do so only through a fully separate subsidiary subject to strict financial separations, and could not bundle CPE with local service. 29 No regulated revenues were to flow across the NID. 18. But how would customer-owned CPE interconnect with the local exchange network, which was typically a monopoly? Brock (1995) makes the important point that Computer II was equivalent to mandating interconnection with customer-owned CPE and
25
Exclusive control of attachments facilitated LECs charging more to customers who valued the service more. Pigou noted that a monopolist can improve its profits if it can find a way to discriminate among customers effectively. Richard A. Bilas and James C. Miller III point out that this type of discrimination can exacerbate, rather than reduce as Pigou argued, the social welfare losses resulting from market power. See "Pigou's Three Degrees of Price Discrimination" Revista Internazionale de Scienze Economiche e Commerciali (February 1971).
26 27
Deregulation of CPE is discussed in considerably greater detail in Brock (1994), Chapter Six.
So-called vertical features, such as Caller ID and Call Waiting, are today priced far above their marginal cost for similar reasons. Entrants today often undercut these feature prices as a competitive tool in attracting subscribers. Thus far, this does not appear to have produced significant pressure on vertical feature prices.
28 29
Brock (1994), Chapter Six. Brock (1994), p.96.
7
CoasianAlternative040901b.doc
09/08/04
setting a zero interconnection rate for CPE. 30 A zero rate corresponds to the incremental cost of CPE interconnection, because the LECs costs are not measurably affected by the customers choice of equipment. If a customer added equipment, for example a FAX machine or a modem, that led him to desire an additional line, the LEC could charge for the additional capacity; but it could not impose a fee based on the nature of the equipment connected to the line. Thus Computer II established a demarcation between the competitive and monopoly segments and set an interconnection rate at the (zero) incremental cost of interconnection. 19. It is difficult to overestimate the impact of the decision to give customers the right to purchase CPE outright, rather than only to buy discrete CPE services from the LEC. If modems and other types of CPE were still marketed only by LECs on a usage basis, the recent development of the Internet, and of much of Information Technology, may not have happened. The blossoming of the CPE market into a highly competitive industry offering rapid technological advances and a wide variety of choice at low cost, and enabling previously unknown possibilities such as the increasingly numerous Internet-based services, is arguably a direct consequence of the deregulation of CPE. 20. The exchange access counter-example demonstrates that mere separation of more and less competitive segments does not automatically resolve problems. When the court ordered divestiture of AT&Ts monopoly local exchange services from its more competitive IXC services in the early 1980s, the FCC developed the current access charge regime to handle interconnection across this demarcation. 31 Under this regime, access charges were based primarily on the historical average total cost per minute of the originating and terminating networks. Network costs were allocated among access services using traditional accounting conventions. The consequent entangling of the costs of the interconnecting networks and sharing of property rights to facilities produced intractable cost allocation32 and rate-setting problems. 33 This also created market power problems where they had not previously existed. 34 These effects were in stark contrast to those following the clear demarcation of CPE, in that divestiture was not followed by an efficient interconnection regime. Thus an efficient demarcation must be accompanied by an efficient interconnection rule to produce efficient results.
30
Brock, Interconnection and Mutual Compensation with Partial Comp ensation, in Brock (1995), p. 2, 17; also, Brock (1994), p. 246.
31 32
Brock (1994) describes this process in considerable detail.
From an economic perspective any such allocation is entirely arbitrary (even though it may be based on generally accepted accounting principles and long-standing legal precedents). That is, any resemblance between these allocations and the manner in which a competitive market would allocate costs is purely coincidental. Thus rates based on the costs resulting from such a procedure are virtually certain to result in significant economic inefficiencies and misallocations of resources. Firms strategic responses to the distorted price signals that result are sometimes called regulatory arbitrage.
33
A survey of these problems is provided in the background section of the Access Reform First Report and Order of May 1997 (12 FCC Rcd 15982ff at 17-27).
34
See discussion at Access Reform First Report and Order, 349ff.
8
CoasianAlternative040901b.doc
09/08/04
21. These contrasting FCC experiences suggest that an efficient, competitively neutral solution to intercarrier compensation problems should resemble the successful resolution of the CPE problem. We show below that a demarcation can be made to separate interconnection from local exchange services, in which market power is often present. We also develop a default rule for assigning cost responsibility for those facilities incremental to interconnection35 that is simple, efficient, and largely selfadministering. In particular, it does not require Pigovian rate-setting by regulators. In other words, we believe there is an efficient Coasian solution to the intercarrier compensation problem. C. A Coasian property rights approach may provide a simple default rule that facilitates efficient network interconnection.
22. Carefully defining networks property rights may provide an alternative regulatory approach to the problem of inducing efficient interconnection in the presence of market power. Where there is market power, however, FCC experience described below suggests that two steps are needed. The first step is to separate competitive from monopoly segments of networks. To avoid the intractable common cost allocation problems and administrative difficulties36 that otherwise arise, this means establishing a clear demarcation between segments with and without market power. The second step is to look for an efficient rule for interconnection between these segments. Ideally, this rule would be largely or entirely self-administering, because regulatory proceedings are necessarily slow, costly, and rather clumsy. 37 To find such a rule, we first identify those facilities that are incremental to interconnection. Then we seek an efficient, competitively neutral default rule for assigning responsibility for these incremental facilities. We demonstrate below that such an approach, if successful, results in efficient interconnection despite the presence of market power. 1. Establishing a demarcation between interconnecting networks
23. What is the demarcation line between intra-network facilities and interconnection facilities? Imagine a standalone network as it would exist if it were not interconnected. It would consist of at least one switch and some lines (loops). Such a
35
We should define this term clearly at the outset and distinguish it from a similar-sounding term that refers to one part of these facilities. The term interconnection facilities describes the actual physical connections between networks: generally, the cables and entrance facilities connecting the two networks switches. The term facilities incremental to interconnection includes the interconnection facilities, but also whatever other incremental facilities may be required within one or the other network to enable exchange of traffic.
36
That is, given essential due process protections, regulatory rate setting is necessarily a cumbersome process that cannot respond quickly to changes in market conditions, and often could not legally recognize or respond to relevant market signals, even if it could discern them.
37
This has little to do with the personal qualities of particular regulators. It has much to do with the significant costs and difficulties they face gathering relevant information. This is particularly the case when parties have strong interests and are able to affect the types and quality of information the regulators can discover. The economics literature, however, often assumes a virtually omniscient regulator.
9
CoasianAlternative040901b.doc
09/08/04
network is obviously responsible for recovering all its own costs. It cannot bill another network, because it is not connected to any other network. Specifically, this standalone network is responsible for providing all the facilities required within its network to handle all calls between its subscribers. The natural demarcation is between the facilities required by this standalone network and those additional facilities required for interconnection. 24. Networks interconnect at switches, 38 which are thus a natural demarcation between standalone and interconnected networks. Fortunately, this natural demarcation corresponds very well to the boundary between more and less competitive segments. There are likely to be a number of carriers that can interconnect at any particular switch, particularly if it is a tandem. 39 It is relatively easy to purchase or lease fiber optic transport, the primary facilities required for interconnection, between virtually any two switches. This is particularly true of tandem switches, and this turns out to be a significant point below. On the other hand, the facilities between the end office switch and the wireline subscriber are often described as bottleneck facilities. This last mile to the (residential) subscriber is widely recognized as the part of the network in which competition is most difficult: it rarely makes sense for wireline entrants to build loop facilities that overlay those of incumbent networks. 40 The switch, the natural demarcation between interconnection facilities and standalone facilities, is generally the boundary between the more and less competitive segments of public switched telephone networks. 25. At present, LECs retain the exchange access rights to their customers lines. This means they can charge other carriers for access to their subscribers. They sell discrete access services to interconnecting carriers, including IXCs and local interconnectors of various sorts, who use these services as inputs in providing telephone services to their own subscribers. But this is not the only feasible arrangement. Our focus in this article is on intercarrier compensation rather than the end user market. Although we do not address the manner in which carriers retail their services, we do not exclude any particular arrangements. Subscribers might directly purchase the access rights to their own lines. Alternatively, carriers might offer a variety of packages, some of which could involve charging either calling parties or called parties on a per call basis, if some customers prefer such arrangements. In a competitive market, we would expect to see a number of different offers tailored to consumers with differing preferences. 41 Unlike the present system, however, one network could not reach across the demarcation line to
38
The principle exception to this rule is a mid-span meet, in which a cable from one carriers switch meets a cable from the interconnecting carriers switch. This form of interconnection is still from one switch to another.
39
A tandem switches traffic among end office switches, rather than to and from individual lines (loops). In the industry jargon, tandems have only trunk connections, as opposed to line connections.
40
A major exception is in dense business and extremely dense residential areas, where we do see some network overlays emerging. Another, perhaps even more significant, exception is for wireless carriers, which are becoming a significant competitive factor even in areas in which no wireline entry seems feasible. This is beginning have important implications in rural areas.
41
We already observe this development in more competitive telecommunications sectors such as the Internet and wireless telephony. See discussion in Part D below.
10
CoasianAlternative040901b.doc
09/08/04
charge another network for a call to or from its subscriber. 26. The first obvious question is whether one can distinguish between a networks standalone facilities and those additional facilities incremental to interconnection. We demonstrate in the next section that this is possible in theory, at least in our abstract, simplified networks. At the end of this paper, we return to the real world (or at least closer to it) to discuss practical difficulties of applying this theoretical concept to real world networks (and some carrier proposals for ways to do so). 2. Efficient Interconnection Rules in a simplified network
27. Real world networks are by nature so complex that it is very difficult to analyze them. Economists have always developed their principles by abstracting from real world complexities to capture essential features, and we resort to that technique here. We begin with a simplified, highly abstract representation of a network to demonstrate the principles of competitively neutral interconnection. We develop these principles through simple, graphical illustrations employing a stylized linear network and show a more general mathematical derivation in footnotes. In an appendix we treat some special cases, and in an FCC working paper (available on the FCC website 42 ), we present a somewhat more extensive development and derive the same principles mathematically for other fundamental types of networks. 28. We make a number of simplifying assumptions. For now, our networks ignore scale economies, trunking efficiencies, and the many other engineering considerations that shape any real network. We focus instead on the fundamental, underlying facilities requirements faced by each network in serving its subscribers. We assume each network has access to the same technology set and employs brilliant engineers who select cost-effective means of provisioning the underlying facilities requirements. But we are not concerned with the engineers magic, only with the abstract, underlying facilities requirements. 29. We assume initially that each network may offer only fully provisioned service to its subscribers. This means that it provides facilities sufficient to support any possible combination of simultaneous calls. 43 We relax this assumption in the appendix to allow for the possibility of call blockage, but the analysis is simplified by beginning with fully provisioned networks. 30. A simplified measure of facilities: urlinks. Imagine a small network with four identical subscribers. We want to determine the minimum facilities necessary to meet the following two requirements: First, any two subscribers to the network should be able to connect with each other. Second, there should be no call blocking. This means that any subscriber can always complete a call to any other subscriber who is not already engaged in a conversation. The call will not be blocked by inadequate facilities.
42 43
Atkinson/Barnekov (2000), http://www.fcc.gov/Bureaus/OPP/working_papers/oppwp34.pdf
Real networks, of course, are engineered so that there is a slight possibility of call blockage at peak periods, but we avoid a great deal of complexity by excluding this possibility for now.
11
CoasianAlternative040901b.doc
09/08/04
31. The most primitive network imaginable would meet our two requirements as the very first networks did, simply by using a separate line to connect each pair of subscribers. With four subscribers we would need six lines. This network consists entirely of lines and has no switching capability. For five subscribers, such a network would need ten lines, for six it would need 15. For n subscribers, this type of network requires (n2 -n)/2 lines. 32. This facilities requirement for a primitive network is an abstract measure of required facilities. Engineers, of course, do not construct real networks by stringing a line between each pair of subscribers. As the number of subscribers rises, the number of lines becomes astronomical. So engineers use various technological devices to reduce costs, for example substituting switches for some lines. We are not concerned with precisely how the engineers provision the network. We can think of the basic facilities requirement as a combination of lines and switching that is equivalent to (n2 -n)/2 lines for a network of n subscribers. We coin the term urlink to describe these line-equivalents. 33. For our illustrations, we use a linear network. To determine the facilities required, we write down all the possible simultaneous two-party conversations. For a four party network, there are three possible combinations of conversations, as shown in Figure 1.
Figure 1 Different caller combinations in a four-party network.
Combination A: 1 & 2, 3 & 4
Combination B: 1 & 3, 2 & 4
Combination C: 1 & 4, 2 & 3
34. Combination A needs only two urlinks since there is only a left side and right side conversation. The remaining possible combinations need these plus two urlinks in the center, since both the conversations cross the center of the network. In Combination B, each conversation requires two urlinks. In Combination C, the center
12
CoasianAlternative040901b.doc
09/08/04
parties require one urlink and the outside parties require three. 35. Figure 2 shows the facilities needed for our four-party network to meet our two requirements. We use a long dash to represent an urlink. The astute reader will note that our urlink includes some switching capability, enough to route calls. Thus a network of four subscribers can have a maximum of two simultaneous two-party conversations (the number of dashes in the center) and requires four urlinks (the sum of all the dashes) to allow every possible two-party conversation to take place.
Figure 2 Facilities needed for a four-party linear network.
Two links are needed across the center and a total of four links are needed.
36. Note again that this diagram is a deliberate abstraction. In the real world, a network could meet our two requirements with a ring, star, mesh or any other imaginable architecture; wireline or wireless links; and circuit or packet transmission. We also abstract from cost-saving techniques that real networks use to economize on links. Recall that initially we assume the network is engineered for zero call blocking. That is, a call attempt might fail because the called partys line is already busy, but not because of circuit blockage. Finally, this simple network does not have any redundant links to improve reliability. In other words, we are focusing on the raw capacity requirements and abstracting from engineering techniques used to provision this capacity (urlinks) economically. 37. We can generalize this depiction to any number of parties. For a linear network of n subscribers, the maximum number of simultaneous two-party conversations is n/2. (Our subscribers engage in only one conversation at a time the analysis becomes complex enough without considering three-way calling.) To ensure that any combination of n/2 conversations can occur simultaneously in a linear network requires (n/2)2
13
CoasianAlternative040901b.doc
09/08/04
urlinks. 44 38. Adding subscribers in a linear network. Before we explore interconnection between two networks, consider what happens as additional parties subscribe. If two more parties subscribe to the network, the resulting six-party network would be diagrammed as in Figure 3. Three simultaneous calls are possible, so three urlinks are needed across the center of the network. This network meets our two requirements, that any two parties can connect and that there is no call blocking.
Figure 3 Facilities needed for two additional subscribers.
Adding two more subscribers requires five more links (dotted).
New
39. Notice that the original four-party network had an average of one urlink per party, but adding the two new parties requires five more links. The average number of urlinks per party rises from 4/4 = 1 per subscriber to 9/6 = 1.5 per subscriber. As more subscribers are added, the benefits of the network increase, but the capacity needed per
44
A mathematical proof is fairly simple. In the worst case, which requires the most links, the parties in the center of the network call each other, then the next nearest pair, continuing until the pair at the extreme ends of the network call each other. The center call requires one link, the next requires three, and the ith call requires (2i-1) links. For a network of n subscribers, in which the maximum number of possible calls is n/2, the extreme call requires 2(n/2) 1 = n-1 links. The sum of all the required links is
n 2 n + 2 n/2 n /2 n 2 2 n n ( 2i 1) = 2i=1 i = 2 = . i=1 2 2 2 2
14
CoasianAlternative040901b.doc
09/08/04
subscriber also rises. 45 The network will add new parties until there are no more potential subscribers or the cost per subscriber rises to the point at which some subscribers begin dropping off the network. 46
Figure 4 JOINING NETWORKS: Interconnection of a four-party network With a two-party network.
Solid black links are intra-network. Dotted links are incremental to interconnection.
40. Assigning the incremental cost of interconnection. Suppose the two parties who joined our network in Figure 3 had instead already subscribed to another network. In Figure 4, let us assume two networks, Grayco and Whitefon, which offer identical service. Fortunately for our illustrations, however, Grayco subscribers insist on using dark gray telephones, while Whitefon subscribers are equally devoted to white phones. We see that to meet our two requirements, interconnection requires four
45
More generally, because the number of links is (n/2)2 = n 2 /4, the average number of links per subscriber is always n 2 /4/n = n/4. As the number of subscribers increases, each subscriber benefits from being able to call more people, but the raw capacity needed per subscriber rises. Recall that we are abstracting from the engineering techniques that offset this diseconomy of scale in real networks.
46
This raises the problematic issue of universal service subsidies. In this paper, we propose to focus on getting interconnection right and to leave universal service for other forums. Not coincidentally, we believe this approach is mandated by the Telecommunications Act of 1996. See 47 USC 254.
15
CoasianAlternative040901b.doc
09/08/04
additional links. 47 This is the incremental cost of interconnection. 41. It is important to note that the interconnected network of Figure 4 has precisely the same facilities requirements as the six party network of Figure 3. In this case, interconnection is precisely equivalent to subscription. A difficult problem in current intercarrier compensation regimes, and even in some bill and keep proposals, is that an end user (subscriber) whose traffic is primarily inbound may claim to be a network in order to receive reciprocal compensation or to avoid bearing transport costs. Similarly, if a networks traffic is primarily outbound, it may wish to claim it is an end user. Ideally, our rule for allocating responsibility for the incremental interconnection links should not give subscribers an incentive to masquerade as networks or vice versa, or to rearrange calling patterns, in order to exploit the rules. That is, an ideal rule will not distort behavior or encourage gaming. 48 42. How should responsibility for the four incremental links of Figure 4 be assigned? Suppose first that each network were assigned a share proportional to its number of subscribers. Grayco would pay for two-thirds of the four incremental links (2.67 links), and would have a new average of 6.67/4 = 1.67 links per subscriber. Whitefon would be responsible for 1.33 additional links, raising its average to 2.33/2 = 1.17 links per subscriber. However, a subscriber to either network receives precisely the same benefits: non-blocking access to the same subscriber list. Therefore, subscribers to Grayco would be better off if they switched to Whitefon, which gains an entirely artificial advantage due solely to this allocation rule. This is not a stable or competitively neutral solution. 43. Suppose instead that interconnecting networks split equally the responsibility for the incremental facilities required, regardless of their relative sizes. That is, Whitefon must provide the same number of incremental links as Grayco, even though Grayco has twice as many subscribers. This means, of course, that the per subscriber cost increase faced by Whitefon is greater (from 0.5 to 1.5 urlinks per customer) than that faced by Grayco (from 1.0 to 1.5 per customer). Note, however, that Whitefon customers are gaining access to four additional parties, while Grayco subscribers gain access to only two more parties. Each linear networks average links per
47
More generally, if a linear network with a subscribers interconnects with a linear network of b subscribers, the number of incremental links required is ab/2. Recall that the total links needed to meet our two requirements (enabling all possible connections and the maximum possible number of simultaneous conversations) is n 2 /4. If the size of the interconnected network is n = a + b , the total links requirement is (a + b)2 /4 = (a 2 + 2ab + b 2 )/4. Before interconnection, the two separate networks needed a 2 /4 and b2 /4 respectively. Thus the number of incremental links required for interconnection is the difference, 2ab/4 or ab/2.
48
A leading cause of regulatory difficulties (by no means exclusive to telecommunications) is that rulemakers often assume firms and individuals will not change their behavior in response to incentives generated by newly minted rules. This attitude persists despite millennia of evidence (not to mention the plain testimony of Scripture) to the contrary. See our discussion in the final section of the relative dangers of regulatory error under Pigovian and Coasian approaches to interconnection.
16
CoasianAlternative040901b.doc
09/08/04
subscriber always increases by for each subscriber on the interconnecting network. 49 44. Requiring each network to provide (or pay for) exactly half the incremental interconnection facilities yields the same average urlinks per subscriber for each network. That is, the raw capacity burden per subscriber is the same on each network. These results hold generally for any linear network as well as for other basic network forms. 50 Because full, non-blocking interconnection allows subscribers to either network precisely the same benefits, failure to balance links per subscriber does not produce a stable outcome. The burdened network will lose subscribers and the favored network will gain subscribers until the burden is equalized. Therefore the competitively neutral and efficient assignment of responsibility for incremental interconnection facilities is for each network to bear ha lf. 45. Examining Figure 4, we can see that interconnection per se does not increase the maximum number of simultaneous calls (the number of links in the center of the combined networks). Prior to interconnection, three simultaneous calls are possible: two on Grayco and one on Whitefon. The interconnected six-party network also accommodates three simultaneous calls. Thus interconnection only increases the number of possible call combinations, not the number of simultaneous calls. 51 46. Note that we have achieved this efficient, competitively neutral allocation of costs between networks without any reference whatsoever to directionality (which network caused a call by initiating it), to the directional balance of traffic between networks, or to how the calling and called end user parties bear the cost of a call. The significance of this point will become obvious below when we discuss the intractable common cost allocation problems and difficulties resulting from directionality that arise in the traditional Pigovian approach. 47. The foregoing analysis suggests a second principle, that only the costs incremental to interconnection should be split. In our example in Figure 4, dividing the total links (9) equally between the networks would give Grayco 4.5/4 = 1.125 links per subscriber. In contrast, Whitefon would have 4.5/2 = 2.25 links per subscriber. More
49
Splitting the incremental cost of interconnection equally, each network would be responsible for an additional ab/4 links (half the total increment of ab/2 identified in fn 47 above). For the network with a subscribers, the average cost rises by ab/4a = b/4. Similarly, the average cost on the other network rises by ab/4b = a/4 links per subscriber.
50
This result also generalizes to interconnection between two linear networks of any size. Recall from the [previous] footnote that the number of incremental links necessitated by interconnection of a network of a subscribers with a network of b subscribers is ab/2. If this increment is assigned equally to the two networks, their respective total numbers of links will be a 2 /4 + ab/4 and b 2 /4 + ab/4. To obtain their average links per subscriber, we divide each networks total capacity by its number of subscribers. We obtain, respectively, (a 2 + ab)/4a and (b 2 + ab)/4b. Each of these expressions reduces to (a + b)/4. Thus if the incremental links are split equally, each network ends up with the same average number of links per subscriber, regardless of their respective sizes. No other allocation of incremental links will produce this equality. This equality of links per subscriber also results if this rule is applied to mesh networks or to fiber optic rings (See Atkinson/Barnekov (2000)).
51
This point will prove significant in the appendix, where we drop the assumption of no call blocking.
17
CoasianAlternative040901b.doc
09/08/04
generally, because splitting incremental links equally produces equal average burdens regardless of the sizes of the separate networks, adding any additional costs would produce an unstable outcome. Even if the networks were the same size, but dissimilar in cost (perhaps they offer somewhat different services), an intercarrier compensation rule that included the internal network costs would generate an artificial cost distortion. 48. In a world in which many different types of networks are possible, it is important that each network recover its intra- network costs from its own subscribers. Not all subscribers have identical preferences regarding technology, service quality, additional features, price and other aspects of network offerings. A single package is not likely to be optimal for every individual subscriber. Some subscribers may place a premium on very high reliability while others may accept occasional blockage, reduced voice or data quality, or even occasional outages in exchange for a lower price. Likewise, some subscribers may be willing to pay for the benefits of mobility offered by certain technologies, while others may forego mobility for a lower price. Some subscribers may choose a lower monthly charge plus a usage charge, while others may prefer a higher monthly charge that includes unlimited usage. Splitting equally the costs that are purely incremental to interconnection and requiring each network to recover its intra-network costs from its own subscribers permits each network to offer retail packages of its choice. Each subscriber can then choose the combination of features and price that best suits her preferences without distortions generated by an intercarrier compensation rule that is not competitively neutral. 52 49. An interconnection cost allocation rule: SIIF. We can now combine our two principles to form a rule we will name Split Incremental Interconnection Facilities, or SIIF. The rule implies that each network recovers the costs of half the facilities incremental to interconnection plus all its remaining costs from end users. In particular, access rights to subscribers lines are not sold to interconnecting networks, as under current intercarrier comp ensation rules. Instead, all access rights are sold directly to end users on any retail basis the carrier chooses (subject to possible controls on dominant carriers retail rates). The end users may be a networks own subscribers or parties on other networks, but each network bears responsibility for its own billing. Under a SIIF regime, any network would have a property right to obtain non-discriminatory interconnection with any market dominant network on these terms. By nondiscriminatory interconnection, we mean that each network treats the interconnectors traffic on the same basis as that of its own subscribers. This would be the default solution, but the networks could mutually agree to another arrangement. 3. 50.
52
How much regulatory intervention would SIIF require?
An interconnection regime along these lines should require little regulatory
The FCC concluded that, even for other purposes such as promoting universal service, competitive neutrality is consistent with congressional intent and necessary to promote a pro-competitive, deregulatory national policy framework. In the Matter of Federal-State Joint Board on Universal Service, CC Docket No. 96-45, 12 FCC Rcd 8801, 48. In the 1996 Telecommunications Act, Congress had required the FCC to establish an explicit funding mechanism for universal service to replace the implicit funding through the previous intercarrier comp ensation regime. See 47 U.S.C. 254.
18
CoasianAlternative040901b.doc
09/08/04
intervention. The most difficult issue would be identifying those incremental facilities that are needed within one network, as distinguished from those between the two networks. This may be a serious difficulty, and we address this issue in the final section; 53 but let us assume for the moment that this is not a problem. Either carrier or any third party provider could bid to supply these incremental facilities. The low bidder would supply the facilities, and the interconnecting carriers would split these costs. Once a clear rule was stated, negotiations between carriers would generally produce efficient solutions, as Coase points out. Where parties disagreed, first recourse could be to an arbitrator instructed to find the lower cost interconnection proposal. We expect that, assuming the incremental facilities could be identified objectively, regulators would only rarely need to resolve disputes. 51. Our presentation of SIIF is only a rough sketch for an interconnection regime. That is, before some rule along these lines were actually implemented, its ramifications and implications need further exploration. 54 For example, we would want to be sure that incremental facilities actually can be readily identified and that the relevant facilities are available on a competitive basis. The most important facilities involved consist of inter-office transport, for which competitive suppliers appear to exist in many or most markets. The most problematic situation would be if it turns out that significant incremental facilities are required within a switch. In such a case, something like a switching equivalent of the NID might have to be developed. Even in this case, of course, SIIF would greatly reduce the scope of needed regulation, because any such facilities would comprise at most a small fraction of total network costs. 55 In contrast, todays intercarrier compensation regimes involve all (in the access regime) or most (reciprocal compensation) network costs. D. A Coasian approach offers important advantages over the Pigovian approach.
52. The SIIF default rule developed in the previous section takes an approach fundamentally different from that of the current intercarrier compensation regimes. We have described SIIF as a Coasian rather than a Pigovian approach. By this we mean that SIIF focuses on finding a clear default assignment of property rights (i.e., responsibility for facility costs and a boundary, or demarcation) between the interconnecting carriers. This differs from the current regimes approach, under which the regulator sets rates (Pigovian taxes) to be applied to particular services in order to compensate each network for the costs (harm in Pigovian terminology) imposed by traffic from
53
Note, however, that the facilities in question consist primarily of interoffice transport and ports. This means that only a relatively small proportion of the entire network is at issue, in contrast to the current regulatory regime. See further discussion in Section E below.
54
The FCC recently issued a Notice of Proposed Rulemaking inviting interested parties to comment on interconnection problems and this or other possible solutions. See Developing a Unified Intercarrier Compensation Regime, CC Docket No. 01-92, Notice of Proposed Rulemaking, 16 FCC Rcd 9610 (2001) (Unified Intercarrier Compensation NPRM). The Commission is currently considering the resulting comments. 55 All central office (switching) costs amount to only ten to fifteen percent of LEC costs, and only at most a small portion of this is likely to be incremental to interconnection.
19
CoasianAlternative040901b.doc
09/08/04
interconnecting networks. Under the Coasian approach, the regulator does not set rates. Instead, she defines initial property rights (i.e. responsibility for providing the incremental interconnection facilities). Once initial, or default, cost respons ibility has been clearly defined, the networks can negotiate an adjustment if local conditions make this desirable. 53. Perhaps more importantly, if the regulator errs in assigning the property rights (in this case, by mis- identifying the incremental facilities), the two networks can negotiate an adjustment to the default assignment. The initial assignment is important because it affects wealth distribution between the two network owners; but an erroneous initial assignment does not distort carrier behavior. 56 In contrast, in a Pigovian regime the regulator sets rates that must be charged, and an error in either the level or the structure of the rates can induce massive distortions in carrier behavior that can result in significant inefficiencies. Under a Pigovian regime, as seen above, the resulting dislocations can be sudden and sharp. 54. As we will demonstrate in the next several sections, this Coasian approach avoids many externality problems that produce distortions under the current Pigovian regimes. One particularly significant difficulty SIIF avoids is a serious third party payer problem that arises under the current regimes. At present end users choose their local carriers, but do not pay the access charges these carriers assess on interconnecting carriers. 57 Instead, access charges are billed to interconnecting networks, which have no choice but to purchase access from the end users LEC whenever they handle a call to or from this end user. As common carriers, IXCs have very limited ability to avoid serving customers, and cannot avoid delivering calls to any end user. Thus the current regimes create a wedge between cost responsibility and decision- making that generates problems of market power, externalities and opportunities for regulatory arbitrage. In contrast, SIIF aligns cost responsibility with decision- making, so that parties consider the true costs and benefits of their decisions. 1. Facilities versus services
55. The most basic difference between SIIF and the current regimes is that SIIF assigns cost responsibility for facilities, while the current regimes set prices (rates) for usage of services. Where the incremental facilities can be readily identified, we showed in the previous section that SIIF offers a straightforward solution to the cost assignment problem that is both efficient and competitively neutral. The current regimes, in contrast, require subjective assessments of cost responsibilities and can only hope to achieve
56
Under some initial cost assignments, of course, one of the networks may not be financially viable. In the absence of market power this could not happen to an efficient network. In the presence of market power, the regulator must be very careful in the initial assignment. We think, however, it would be unwise policy to attempt to manipulate intercarrier compensation to prop up inefficient carriers.
57
These inter-carrier access charges are averaged across many carriers and eventually find their way onto end users IXC bills. The IXC charges that end users eventually pay, however, bear no direct relationship to the access charges assessed by the carriers they (or the persons they call) choose. At present, federal law (47 USC 254(g)) prevents IXCs from charging higher rates for calls to or from customers in high access charge areas.
20
CoasianAlternative040901b.doc
09/08/04
second best efficiency. 58 An objective 59 assignment of cost responsibilities is not necessarily easy, as we discuss in Part D, but at least the difficulty is limited to a relatively small proportion of the network. 60 Under the current regimes, costs must be assigned for a much larger portion of the network 56. The current Pigovian rules cannot produce an objective or efficient solution because they attempt two impossibilities. They try to allocate costs first among the various services that carriers provide, then between the parties to a call (in order to charge their respective networks). 61 Because most network services share the same inputs, however, most network costs are common to the various services. Economists have never found a satisfactory way to allocate common costs efficiently other than through market interactions. As Hayek explained, 62 regulators cannot possess the requisite specific knowledge of market conditions to make such an allocation. Nor can regulators discern how to split the costs between the parties to a call efficiently, because they cannot know the relative distribution of benefits. Instead, for reasons of practicality, the current rules assume that the caller receives all the benefit, so the callers network should pay all the costs. 63 The effect of these rules is that each network pays an economically arbitrary share of other networks common costs. Customers cannot subscribe to one network without paying part of anothers costs. Thus the current rules co-mingle the costs of interconnecting networks, which makes achieving efficient outcomes problematic. 57. By focusing on facilities rather than services, SIIF produces cost assignments that track, rather than distort, the structure of network costs. Most network costs are a function of the number of subscribers (and, to a much lesser extent, of capacity) rather than of the level of usage. 64 Telephone companies have no cost of goods
58
They cannot actually achieve second best efficiency, they can only hope to do so, because the conditions required actually to attain this can never be met. See Laffont & Tirole (2000).
59
By objective we mean that, given the configurations of the networks to be interconnected, both parties, or an outside arbitrator, would likely reach the same assignment.
60
We note also that networks have had little problem identifying interconnection facilities to date. That is, although networks are by no means reluctant to bring disputes to federal or state regulators for resolution, they rarely bring disputes over ascertaining what facilities are required for interconnection. The problem, however, is a bit more subtle than simply identifying interconnection facilities as noted previously, the facilities incremental to interconnection include these but may also include some other facilities that are internal to each of the networks. This latter point is explored a bit further in the appendix.
61
A brief description of the current system is provided in the background section of the Access Reform First Report and Order of May 1997 (12 FCC Rcd 15982ff at 17-27).
62
See Friedrich A. Hayek, The Use of Knowledge in Society, American Economic Review, XXXV, No. 4; September 1945, 519-30.
63
This is the rule for wireline carriers. Wireless carriers in the U.S., however, typically charge their own subscribers for both originating and terminating calls.
64
By this we mean that the long run total cost of Network B, designed to handle twice the traffic volume of (an otherwise identical) Network A, is much closer to the cost of Network A than to twice that cost. In fact, in Part D we discuss some evidence that the cost of Network B may be very close to that of Network A. In contrast, the cost of Network C with ten times the subscribers of Network A (holding subscribers
21
CoasianAlternative040901b.doc
09/08/04
sold. Switches and circuits do not even use measurably more power when they carry more traffic. To produce usage-based rates requires some averaging procedure, but in doing this we are not averaging over the right variable. Regulators have seriously misperceived the cost structure. As a result, they have created a pricing structure that is inconsistent with it. The current regimes do not even distinguish between peak and offpeak usage, although the marginal cost of the latter is approximately zero. By allocating network costs to various services on a usage basis, the current rules tend to transmogrify network cost structure from a flat (connection- or capacity-based) to a per minute of use basis. This discourages efficient use of network facilities by overpricing off-peak (and underpricing on-peak) use. SIIF avoids this distortion because intercarrier payments are minimized and, more importantly, have nothing directly to do with traffic volume or direction.65 58. Because SIIF must assign responsibility only for those facilities incremental to interconnection, 66 rather than for all network costs, it greatly reduces the problem of cost allocation. These incremental facilities consist mainly of interoffice transport facilities. For telephone networks, total interoffice transport accounts for only ten to fifteen percent of total network costs, and only a fraction of this is actually incremental to interconnection. Therefore a SIIF approach greatly reduces the scope of the cost allocation problem. Even if there is controversy, it is likely to be over a small fraction of network costs. 2. Treatment of market power
59. Market power is the primary challenge in devising an economically efficient interconnection regulatory regime. But for the problem of market power, it is not obvious that regulation would be needed at all to achieve efficient interconnection, or that it would have to be mandated. 67 Because potential subscribers generally value access to a broader callable universe, it is normally in communications networks interests to interconnect. As discussed in Section E, the Internet is an example of apparently efficient interconnection that is mandated by customers rather than regulators. Where a network might be able to exploit market power, however, interconnection may have to be mandated and its terms regulated. 68 60. A Coasian approach such as SIIF treats market power in a fundamentally
per square mile constant), regardless of its expected traffic volume, appears to be close to ten times that of Network A. One might say that there are powerful scale economies with respect to traffic volume, but not with respect to subscribers. A more relevant way of putting this, however, is that costs are primarily a function of subscribers rather than of traffic volume.
65
Volume affects SIIF costs only by affecting the requirement for incremental interconnection facilities, and cost assignments are not affected at all by the direction of traffic flow.
66 67
All other network costs are assigned to the network that incurs them.
Regulation might also be employed for other purposes, of course, such as transferring wealth among customers or firms.
68
For a concise summary of the literature on the need for regulating interconnection in the presence of market power, see DeGraba (2000), p. 14.
22
CoasianAlternative040901b.doc
09/08/04
different manner than the Pigovian approach of the current interconnection regimes. The essential difference is that the Coasian approach defines property rights and facilitates market adjustments, while the Pigovian approach sets prices. Under SIIF no payments flow between networks. By establishing a demarcation between networks, SIIF prevents a LEC from using any market power it may possess to exploit interconnecting networks. Under SIIF, a monopoly LEC can exploit only its own subscribers. Where there is a potential entrant, of course, attempting to exploit ones subscribers may induce them to become former subscribers. In the presence of local monopoly power, regulators may exercise control over LEC prices. 61. A SIIF regime may significantly reduce the LECs ability to exploit even its own subscribers. The analysis of the previous section demonstrates that, if the networks face the same cost functions, interconnection on a SIIF basis results in the same cost per subscriber. Recall from above that, under SIIF, interconnected traffic receives the same treatment as on-network traffic. Thus, if a LEC charges more than its average cost per subscriber, an efficient entrant could interconnect on a SIIF basis and offer the same service at a lower price. Even if the entrant faces some disadvantage, cost perhaps related to scale economies or because it offers a somewhat different type of service, the potential of entrance with SIIF interconnection may exert significant restraint on the monopoly LEC. If regulatory authorities are still concerned about the LECs end user pricing, the task of regulating these rates is greatly simplified under a SIIF regime. 69 Thus a Coasian approach such as SIIF at least contains any market power and may reduce its effects. 62. In contrast to establishing a demarcation point, the current (Pigovian) schemes require interconnecting networks to purchase services from each other at rates set by the regulator. An interconnecting network must accept or deliver traffic, and pay access charges, to whichever network is able to sign up a particular subscriber. This extends any market power one LEC may have to interconnecting networks. 70 In addition, this system gives every LEC automatic monopoly power over traffic to or from its subscribers. The problem is particularly severe for traffic to subscribers, and the resulting terminating access monopoly problem has been extensively discussed in FCC decisions. 71 Under this system, even small, competitive LECs possess monopoly power over calls to or from their subscribers. In fact, the LEC has something greater than monopoly power: a monopolists customers may limit their purchases in response to a supracompetitive price, but interconnecting networks are forced by mandatory interconnection rules to purchase access for as many calls as their subscribers choose to make. Furthermore, subscribers choices are unaffected by the monopolists higher price, because, as described above, they do not face these prices. Thus this system is incompatible with deregulation, because there is no market constraint on the access rates
69
Of course, regulators may also lose opportunities that exist under the current regimes to subsidize some end users at the expense of customers of other networks.
70 71
Brock (1995) See, for example, Access Reform First Report & Order, CLEC Access Charge Order.
23
CoasianAlternative040901b.doc
09/08/04
these companies may charge. 72 Rather than resolve or limit market power concerns, the Pigovian approach expands them. 3. SIIF avoids problems of regulatory arbitrage
63. SIIF differs from the current regimes in that the cost assignments it generates depend only on the incremental facilities required for interconnection. In contrast, cost assignments under the current regimes are based on such factors as the nature, volume and direction of flow of traffic, classification of the carriers, and identity of the subscribers involved. These are arbitrary distinctions in the sense that they do not actually reflect the underlying economic costs of interconnection. They only affect the regulatory formulas by which regulators have chosen to assign costs. Networks and end users can manipulate these factors to a large extent by re-arranging their traffic flows or misrepresenting their identities. Because these potential manipulations can have a tremendous effect on payments by or to a carrier, powerful incentives for regulatory arbitrage arise. Several serious problems of this nature recently caused the FCC to take palliative actions and to seek comments on a new approach to intercarrier compensation. 73 The underlying problem is that the Pigovian interconnection system is inherently unstable. In contrast, SIIF cost assignment is unaffected by any of the factors that have so much effect under the current regime. Thus SIIF eliminates the potential for regulatory arbitrage. E. Market evidence appears to confirm the relevance of our model
64. If the best test of a theory is its predictive power, then current behavior of telecommunications carriers suggests that our Coasian analysis is relevant and that its proposed solution is useful and competitively neutral. We find this evidence in two relatively competitive telecom markets and even in the more regulated sector. The U.S. markets for Internet services and for wireless telephone services are widely recognized as largely competitive. Although their interconnection practices are not well documented, because they are largely shaped by confidential commercial contracts, the broad outlines we can observe seem to support the relevance of our Coasian model. Furthermore, recent behavior in even the more regulated wireline sector also suggests that our analysis has predictive power. We do not attempt here an exhaustive description or a definitive analysis of these telecommunications sectors, but we do sketch briefly how we think the limited evidence we have suggests that our model is relevant, how it contradicts some underlying assumptions of the current regulatory system, and why we believe further research along these lines would likely be fruitful.
72
The FCC recently found it necessary to limit access charges by small entrant LECs because in many cases they had been set at extremely high rates. See CLEC Access Charge Order (In the Matter of Access Charge Reform, Seventh Report and Order, released April 27, 2001, CC Docket 96-262).
73
Unified Intercarrier Compensation NPRM; CLEC Access Charge Order; ISP Remand Order.
24
CoasianAlternative040901b.doc
09/08/04
1.
Interconnection among Internet Service Providers 74
65. The Internet Service Provider (ISP) market appears to be quite competitive and in the United States is mostly unregulated, subject only to the usual commercial law framework. There is mandatory interconnection in the ISP market, but this mandate does not emanate from a regulator. Instead, customers mandate interconnection by demanding global access from their ISPs. Because this interconnection results from private contracts, we have only limited evidence of interconnection arrangements, but some features are observable. Many papers have been written about the Internet, and we do not attempt a thorough description here. We do, however, touch on certain aspects of the interconnection pattern that appear to support the broad thrust of our models predictions. a. Inferences from published Internet interconnection terms
66. The Internet is a useful example of commercial interconnection in a highly competitive environment. Even the largest ISPs comprise only a small fraction of the Internet universe. They typically provide Internet services within a certain area (which may be small or large, regiona l or national, or even multi- national). These individual ISPs interconnect with all other ISPs throughout the world through myriads of commercial agreements. Smaller ISPs connect to the Internet backbone through commercial agreements either directly with an ISP that is an Internet backbone provider (IBP), or indirectly by becoming customers of larger ISPs that are in turn customers of an IBP. No IBP can extract a supracompetitive payment, 75 because there are many competing backbone providers. Nor can large r ISPs extract supracompetitive payments from smaller ones, because even a small ISP generally has multiple options to get to a backbone provider. We are not attempting an exhaustive description, but simply making the point that a system of commercial interconnection agreements has developed in a largely competitive environment largely free of regulatory intervention. We would therefore expect global access to be priced roughly at its cost. Perhaps more significantly, we would expect the price structure to conform rather closely to the cost structure. Thus this market is a useful example of competitive interconnection that has developed in a largely unregulated context. 67. Kende (2000) and Laffont, Marcus et al. (2001) describe ISP and IBP interconnection patterns with respect to transit and peering. Oversimplifying somewhat, these two terms refer to interconnection arrangements that, respectively, do and do not involve payments from one network to the other. In transit arrangements, one (smaller) ISP is a customer of another and pays for global interconnection that enables it to send and receive traffic worldwide. Under peering arrangements, larger ISPs exchange traffic on a bill and keep basis without intercarrier payments. In
74
This section benefited greatly from conversations with J. Scott Marcus; however, he should not be implicated in any remaining errors of which he was unable to disabuse us.
75
We do not claim that such a situation could never develop, and there is an extensive literature exploring conditions under which it might. We simply note that Internet interconnection has developed thus far in a highly competitive environment.
25
CoasianAlternative040901b.doc
09/08/04
agreeing to peer, Network A agrees to accept and deliver (terminate) peer Network Bs traffic that is addressed to As customers. In exchange, Network B does the same for traffic A originates addressed to Bs customers. Each network agrees to accept this traffic at whichever interconnection point its peer finds convenient, which is normally at the first possible point of interconnection (this is often called hot potato routing). 68. Although the precise terms of peering arrangements are generally proprietary, several large IBPs have published the general terms under which they will agree to peer with competing networks. 76 These general terms appear consistent with our model. Networks peer when it is mutually advantageous, 77 so a network evaluating a potential peer wants to be sure that the costs it will save by being relieved of delivering traffic destined for the peers customers are greater than the additional costs it incurs by accepting the peers traffic for delivery to its own customers. Among the costs saved are those related to measuring, tracking, and billing for traffic, and related dispute resolution. 78 Because these billing-related costs can be substantial, both networks can often reduce their overall costs through peering arrangements. Published peering terms seem aimed primarily at making sure net cost reduction will result. 69. We believe this pattern contradicts the widely held view, on which the current regulated intercarrier compensation regime is based, that the incremental cost of traffic equals its average cost. Although ISP retail pricing is sometimes based in small part on expected traffic volume, 79 we see only a limited interest in traffic volume in published peering terms for inter-network interconnection. That is, the terms focus first on requiring a large minimum volume for a peering candidate. 80 In contrast, there is less concern for directionality of traffic. One IBP explicitly states that minimum traffic
76
As of April 2004, peering terms were posted on the websites of Cable & Wireless (U.S. and Europe), WorldCom (UUNet) and Level 3.
77
Cable & Wireless, for example, states on its website that it will peer where such exchanges offer equitable benefits and are cost-effective, and then sets out general criteria that make this likely. In evaluating a potential peer, a network looks for such things as an equivalent geographic configuration, adequate transport capacity, and some sense of whether each peer will haul a roughly equivalent amount of traffic. The potential peer must also been technically well-managed, because interconnecting with an unstable network can cause costly problems on ones own network.
78
Dispute resolution costs are not negligible. We have no precise data, but based on informal carrier comments, disputed bills for intercarrier compensation among telephone carriers amount to very substantial sums, particularly relative to CLEC total revenues. These disputes arise largely from regulatory distinctions among technologically indistinguishable traffic. Such disputes seem largely absent from the Internet,
79
Web-hosting rates typically depend on both storage capacity and maximum monthly traffic. Dial-up ISP access packages often include some monthly number of hours with additional access priced per hour. In both cases, however, unlimited usage is typically available at a modestly higher flat rate. These pricing schedules imply that volume is not a significant cost factor, because the rate for unlimited traffic is typically not much higher than the rate for a relatively low monthly volume.
80
Routers currently face limits on the number of other routers with which they can directly connect, so direct interconnection for handling small traffic volumes is impractical. We are indebted to J. Scott Marcus for this explanation.
26
CoasianAlternative040901b.doc
09/08/04
volume can be measured in either direction. 81 Published terms allow the ratio of inbound to outbound traffic to be as great as 2:1. This appears designed to exclude pure one-way networks such as website hosts 82 and potential free riders,83 rather than to reflect a significant incremental cost of traffic volume. After all, a 2:1 ratio allows for vast differences in volume in each direction. The volumes at issue amount to hundreds of terabytes per month, and even half these amounts would represent vast cost differences if volume were actually a primary determinant of cost. Carriers in this competitive environment do not behave as though the incremental cost of traffic were equal to the networks average cost per unit of traffic. 70. The peering terms demonstrate that the concept of directionality has little meaning in the Internet context, as we believe is true in telephony. Such concern for directionality as there is appears aimed at excluding free riders. Unlike telephony, however, there is no regulatory framework that mandates pricing on this artificial basis. As Kende points out, when an individual visits a website, he typically sends a small packet of data to the website and receives a much larger volume of data in response. Which party causes this traffic? The question is irrelevant. This exchange is in both parties interest, and both parties purchase their own access to the Internet. The interconnection that makes this exchange possible makes both parties networks (ISPs) more valuable to their customers. This is why customers subscribe to ISPs. 71. In the Internet transit service market, pricing is typically entirely independent of directionality. It depends, rather, entirely on the capacity of the interconnection port. In this case, the transit purchaser is buying not only the port service, but also the services provided by the network selling the transit. The price thus covers not only the costs of the port itself, but also whatever incremental costs are incurred by the transit provider. Port capacity is, of course, a proxy for the total volume to be passed between the interconnected networks, because few transit purchasers will buy OC-3 service to handle the traffic for which a T-1 port would suffice. But the prices of various ports are much less than proportional to their capacity. 84 This implies that incremental traffic costs cannot be anywhere near average cost per unit of traffic volume.
81
Cable & Wireless, Settlement-Free Peering Policy, USA posted on the Internet (as of April 2004) at http://www.cw.com/docs/network/USPolicy.pdf.
82
From our perspective, an ISP that is primarily a website host at a single location is more of an end user than a network. If it interconnected on a peering basis, it would likely not be supplying half the facilities incremental to its interconnection. In contrast, a website host that had its website server farms at geographically dispersed locations and supplied substantial transport among them would look much more like a carrier and might qualify as a peer, or might come close to qualifying but pay some additional compensation to its peers.
83
If a network can shift some of its costs to a competitor without relieving that competitor of comparable costs, it can be said to be getting a free ride that it can use to gain a competitive advantage over its competitor in attracting subscribers.
84
An OC-3 provides 100 times the capacity of a T-1. In April 2004, a typical published price (for a three year contract) was $17,125 for an OC-3 and $458 for a T-1 port. http://www.ecttelecom.com/ t1.html
27
CoasianAlternative040901b.doc
09/08/04
b.
Interconnection between networks of greatly differing size
72. What we see in the Internets interconnection arrangements resembles what our model would lead us to expect. As explored in the appendix, our model implies that where networks of substantially differing size interconnect, a disproportionate share of the incremental facility requirements occur within the larger network (see Figure 6 in Appendix). As our earlier analysis and that in the appendix demonstrate, the greater the disparity in network size (measured by number of subscribers, nodes, or an analogous measure), the greater the proportion of incremental facilities required of the larger network. This is consistent with the pattern we observe in Internet interconnection. Where networks of unequal scope interconnect, it is the smaller network that purchases transit service from the larger. 73. Peering and transit seem to reflect conditions under which incremental facilities required because of interconnection are roughly balanced (peering) or occur predominantly on one network (transit). When roughly equivalent networks interconnect, they use peering: bill and keep or settlement- free interconnection involving no intercarrier payments. When a small network interconnects with a large one, however, transit payments compensate the large network for the greater burden of incremental facilities required within the larger network. 85 74. Internet interconnection arrangements seem to be a useful illustration of how carriers might actually apply a default interconnection rule similar to SIIF. The greatest difficulty that commenters raised in response to the FCCs NPRM on this topic 86 was that it would be difficult to identify precisely which facilities within interconnecting networks are actually incremental to interconnection (as distinguished from the readily identifiable interconnection facilities between networks). As a result of this difficulty, ISPs seem to have found a work-around that has the same efficiency advantages. In comments on an earlier draft of this paper, a colleague pointed out that interconnecting Internet backbones do not reveal to each other the intimate details of their networks that would be necessary to identify incremental facilities within the networks. Instead, he notes, [t]he practice is to so arrange the terms of interconnection that the costs that would otherwise be difficult to [identify and] split
85
We have seen a recent example of this relationship in an international controversy over Internet connections. Australian ISPs have complained that they must bear the entire cost of transpacific cables to connect with U.S. Internet backbones. See discussion in Kende (2000) at p. 33ff. But the U.S. Internet is many times the size of Australias. Our model suggests that if the Australian and U.S. Internets interconnect, the incremental facilities requirement within the U.S. networks is much larger than in Australia. Thus having the Australian network bear the full cost of the interconnection facilities (the actual cable between the two networks, a portion of the total incremental facilities required for interconnection) may simply approximate splitting the incremental facilities equally between the two networks. We see the refusal of the U.S. networks to split the cost of this transpacific cable as consistent with the predictions of our model. Both Australians and Americans benefit from this interconnection, but Australians gain access through this cable to far more of the global Internet than do American subscribers. If at some point the Pacific Internet grows to be larger than that within the United States, we would predict that the U.S. firms would bear the costs of the transpacific cable.
86
Unified Intercarrier Compensation NPRM.
28
CoasianAlternative040901b.doc
09/08/04
(behind the point of interconnection) should in principle be roughly comparable.87 As he further notes, Arguably, this is just another Coasian example of parties negotiating their way to a better outcome than the regulators might otherwise have prescribed. The point of proposing a rule like SIIF is not that a regulator should actually identify incremental facilities and split the responsibility for them. The point is rather to define rights so that interconnectors can negotiate an efficient outcome even where one carrier is market dominant. c. Evidence of competitive neutrality
75. There are at least two types of evidence tha t the market-driven system of Internet interconnection is competitively neutral. First, we see large and small ISPs coexisting, at least in the 1990s. A Stiglerian survivorship analysis approach88 would suggest that whatever scale economies larger ISPs might enjoy are roughly balanced by advantages smaller ISPs enjoy in retailing to particular subscriber groups. One of the typical blessings of a competitive market is the availability of a wide variety of offerings and packages to appeal to customers with varying preferences and requirements. 89 This survivorship evidence is less than conclusive, however, because there were many unusual market circumstances90 that may have accounted for the survival of small ISPs during the late 1990s. Recent changes in market conditions and in regulations may or may not explain what appears to be a sharp reduction in the number of smaller ISPs more recently. It is too soon to be sure whether the survivorship evidence is persuasive. 76. A second, perhaps more significant type of evidence of competitive neutrality is the willingness of larger ISPs to sell either retail or wholesale Internet access. That is, they typically offer their own retail packages to end users and also offer transit arrangements that enable smaller ISPs to connect to IBPs. Smaller ISPs typically seem able to choose among several, or many, transit providers to obtain access to the global Internet. The interconnection model is apparently competitively neutral, because ISP size is evidently more a function of the ability to craft or tailor retail packages for particular end user customers with varying interests than of any market power in interconnection.
87 88 89
E-mail from J. Scott Marcus, April 20, 2004. Stigler, G. J. 1958. The Economies of Scale. Journal of Law and Economics 1: 54 - 71
In contrast, a curse of regulation is that the regulator typically requires a uniform, one size fits all approach. This is, for the most part, because it is quite difficult for the regulator to evaluate compensating variations in various dimensions of service quality and price, particularly within the necessary legal constraints requiring the regulator to articulate a reasonable basis for decisions. Put another way, given the prohibitions against discrimination in common carrier law, it is nearly impossible to prove to a courts satisfaction that joint and common costs have been allocated correctly among several variations in service offerings. The more legally defensible approach is one size fits all.
90
We note in particular the dot.com boom and FCC rule changes that generated large amounts of reciprocal compensation paid on Internet dial-up traffic, inducing some LECs to offer to serve ISPs on very favorable terms.
29
CoasianAlternative040901b.doc
09/08/04
77. Although we have limited detailed evidence of Internet interconnection patterns, the broad outlines we can observe seem to confirm the validity of our model in this highly competitive market. They seem to confirm that a competitive market would settle on a system of this general nature. Finally, although the evidence on this point is not definitive, it also suggests that interconnection on such a basis is, in fact, competitively neutral. 2. Interconnection Companies among Wireless Telecommunications
78. Wireless telecommunications is another highly competitive sector within which the interconnection arrangements appear to support the relevance of our analysis. When wireless carriers (CMRS providers) interconnect with regulated LECs, the terms are largely subject to regulation. In contrast, when wireless carriers interconnect with each other, interconnection is basically unregulated. As with ISPs, interconnection among wireless carriers is mandated by customers, rather than by regulatory fiat. A wireless carrier that would not terminate calls to or from other wireless carriers would find its appeal to potential subscribers rather limited. Also as among ISPs, details of wireless-towireless interconnection are sketchy because it is largely accomplished through confidential commercial agreements. What we can observe, however, seems broadly consistent with the predictions of our model. a. Inferences from wireless interconnection arrangements
79. What we observe is that wireless carriers typically interconnect with each other on a bill and keep basis. Typically, no money is exchanged between the wireless carriers and each carrier provides its own incremental interconnection facilities. 91 In contrast to interconnection involving wireline carriers, we see no terminating access or reciprocal compensation payments between wireless carriers. 92 These interconnection arrangements seem to meet the competitive neutrality test, because they have permitted the long-term survival of several wireless carriers in each market. 80. Although we are not directly concerned with retail pricing in this paper, broad trends in wireless retail pricing provide intriguing clues as to the nature of network and interconnection costs. Several years ago, wireless carriers charged their customers high per minute rates. Today the trend is toward what amounts to a two-part pricing system (customers choose among packages that offer various volumes of minutes at declining prices per minute, with penalties for understating actual usage). Because market discipline in this competitive sector is likely forcing the pricing structure to conform to
91
It is worth noting that when one wireless customer calls another, each customer pays for his own connection to the network, just as each Internet user purchases her own Internet access. We are not directly analyzing retail pricing in this paper, but we do find it quite interesting that retail arrangements have developed along these lines in both of these highly competitive telecommunications sectors.
92
Increasingly, wireless carriers also seem to be arranging even roaming agreements that involve little or no intercarrier payment.
30
CoasianAlternative040901b.doc
09/08/04
the cost structure, this suggests that a substantial portion of these carriers costs are driven by the number of subscribers and that they face modest incremental costs per minute of use. The fact that this pricing pattern can persist in a competitive environment suggests that it reflects the actual cost pattern. 81. The wireless network costs that vary with minutes of use may not even be primarily costs of the wireless network itself, but rather may largely reflect intercarrier compensation obligations arising from the current wireline regulations. One clue that suggests this is the increasingly common offer of unlimited intra-wireless-carrier calling free of per minute charges. 93 This offer is typically valid even during peak hours. This suggests that the principal source of costs that vary with minutes of use is intercarrier compensation, rather than costs of handling incremental traffic within the network. 82. This offer of zero-priced on-net usage is in addition to the increasingly common offers of unlimited or extensive off-peak (nights and weekends) usage free of per minute charges. This inexpensive or free off-peak usage has become possible only as intercarrier compensation obligations have diminished. In the early years, wireless carriers had to pay substantial access charges to wireline carriers on both originating and terminating calls. Before 1996 wireline access charges were intentionally marked up to create cross-subsidies to keep monthly wireline subscription rates low. The 1996 Act mandated a policy of reducing these charges toward (ave rage) cost (per minute). Although wireless carriers do not report their costs publicly, we suspect that the bulk of wireless operating costs in the early years represented intercarrier compensation payments. 94 The 1996 Telecommunications Act fundamentally changed these relationships, lowering access charges and replacing them with reciprocal compensation for most wireless traffic (i.e., local traffic). Although neither access nor reciprocal compensation rates reflect peak/off-peak differentials, reciprocal compensation rates were a small fraction of access rates. Furthermore, at least in theory, wireless carriers could collect reciprocal compensation from wireline carriers for calls from wireline subscribers. These developments empowered wireless carriers to reach interconnection agreements with wireline carriers that drastically reduced their intercarrier payments and made off-peak pricing (including offers of free off-peak minutes) feasible. b. The Sprint petition on reciprocal compensation
83. Sprint recently obtained a ruling from the FCC that it was entitled to seek reciprocal compensation based on its own wireless networks traffic-sensitive costs rather than the wireline carriers costs. 95 Sprint produced studies purporting to show that its
93
See, for example, AT&T, Verizon Go Unlimited As Wireless Battle Accelerates, Wall Street Journal, February 2, 2004.
94
We have no data, but base this inference largely on the frequency and intensity of wireless carrier pleas for relief from this burden.
95
That is, in the course of arbitrations by state commissions under Section 252 of the 1996 Telecommunications Act, Sprint is entitled to present evidence attempting to rebut the presumption that its costs are the same as those of the incumbent LEC. See Letter from Thomas J. Sugrue, Chief, Wireless Telecommunications Bureau, and Dorothy T. Attwood, Chief, Common Carrier Bureau, to Charles
31
CoasianAlternative040901b.doc
09/08/04
eligible costs per minute are several times those of wireline carriers. 96 Under traditional analysis, almost all of Sprints costs could appear to be traffic-sensitive, and thus eligible for recovery under reciprocal compensation rules. 97 84. Although Sprint has presented such stud ies in negotiations and arbitrations in various states, it has not actually collected these higher rates for local traffic originating on wireline carriers. Instead, Sprint has generally settled with the incumbent LECs on a bill and keep basis. 98 Sprint thus gave up an opportunity to recover high payments from wireline carriers. This contrasts with the eagerness with which CLECs exploited somewhat similar incentives with respect to ISP dial- up traffic. 99 One explanation is that reciprocal compensation rates in this range would not be sustainable or acceptable to state regulatory commissions. Our conjecture is that in addition Sprint knows that its own traffic-sensitive network costs do not actually vary much with minutes of use 100 and that its own long run interests are better served through a more sensible bill and keep arrangement. If Sprints costs really did depend mainly on minutes of usage, its behavior would likely have been different. 85. It would be fascinating to look at the confidential details of wireless-towireless interconnection agreements for further clues as to the nature of interconnection costs. Perhaps some researcher will find a way to do this. In any case, we find it significant that the interconnection arrangements that have developed in this highly competitive telecom market sector appear to resemble what our model would predict. Whether or not these arrangements are responsible for the ongoing competition, they are clearly consistent with it. That is, these arrangements appear to be competitively neutral.
McKee, Senior Attorney, Sprint PCS, CC Docket Nos. 95-185 and 96-98, and WT Docket No. 97-207, 16 FCC Rcd 9597 (2001), affirmed by the Commission August 27, 2003.
96
These studies followed the FCCs TELRIC methodology. They therefore estimate (forward -looking) average cost per minute, not the incremental cost of additional calls or minutes.
97
The FCC has viewed the costs of facilities that are shared by multiple users as traffic-sensitive (or usage-sensitive), as distinguished from facilities that are dedicated to a single user. In the Access Reform Order (741ff), the FCC generally required costs of the former type of facilities to be recovered through usage-based (i.e., traffic-sensitive) rates and the latter by monthly (flat) rates. From this perspective, most wireless network elements (e.g. spectrum, cell towers, transport) are arguably traffic sensitive, and thus eligible for reciprocal compensation payments under FCC rules. Actual behavior of wireless carriers, as discussed above, seems to tell a different story as to whether these costs are actually incremental to calls.
98
At last count, bill and keep settlements have been reached with the incumbent in all nine BellSouth states and in 13 of 14 Qwest states. Source: Conversation with Stacy Jordan, FCC Wireless Telecommunications Bureau, April 15, 2004.
99
See ISP Remand Order, 16 FCC Rcd at 9182-83, para. 68-71; see also Intercarrier Compensation NPRM, 16 FCC Rcd at 9616, para. 11.
100
The term traffic-sensitive is somewhat misleading in that these costs do not necessarily vary in proportion to traffic volume. In fact, we contend that traffic volume has very little effect on the level of most of the costs that are called traffic-sensitive. These could be more accurately termed costs that are recovered through traffic -sensitive rates, but this is much too wordy. It would be less confusing if the FCC had instead chosen to call them what they actually are: shared facility costs, as distinguished from dedicated facility costs.
32
CoasianAlternative040901b.doc
09/08/04
3.
Support for SIIF-like solutions among Telephone Companies
86. The large U.S. telephone carriers have supported moving toward one or another form of bill and keep to replace the current jumble of intercarrier compensation regimes for some or all types of traffic. 101 A broad-based collection of carriers of almost every description met during 2003-4 attempting to achieve a consensus proposal to replace the current system. This group included very large carriers, some smaller wireline carriers including some competitive LECs, and some wireless carriers and rural carriers. These negotiations were confidential, but the consistent pattern of descriptions in press reports was that the discussions focused on some variation of a bill and keep solution. 102 87. Of the 25 companies that participated at one or another stage of the negotiations, nine agreed on a consensus proposal submitted to the FCC in August 2004.103 The rest, including all wireless carriers, dropped out because of dissatisfaction with various details. The proposal submitted could be loosely described as moving a long way toward a bill and keep solution, in that termination rates move to zero for all but rural carriers. The proposal, however, mingles issues of support for rural carriers with interconnection matters, and is thus difficult to summarize briefly. 104 88. The difficulty in these negotiations, of course, is that the various types of carriers have sharply divergent starting positions under the current intercarrier compensation scheme. The small, rural carriers, in particular, are substantial net recipients of various subsidies built into the intercarrier compensation system and are dependent on these subsidies. 105 At the same time, it is obvious, in varying degrees, to almost all parties that the current scheme is subject to great pressure. This scheme was designed to operate under monopoly conditions, and increasing competition in telecommunications is eroding its viability. How much longer this system can survive is largely a matter of the pace of technological developments that facilitate competition, especially intermodal competition. 106 The increasing recognition that the current scheme cannot long survive is arrayed against the resistance to change of those who most benefit
101
See ex parte letters in FCC Docket 01-92 from BellSouth (June 2, 2002), Qwest (August 2, 2002) and Verizon Wireless (May 3, 2002 and January 27, 2003).
102 103 104
See, for example, Telecommunications Reports, TR Daily, February 27, 2004 and March 16, 2004. Ex parte letter in FCC Docket 01-92 from the Intercarrier Compensation Forum, August 16, 2004.
As this proposal is now formally before the FCC in an open docket, it would be inappropriate for the authors to comment on its merits. The reader may view the proposal on line at http://gullfoss2.fcc.gov/prod/ecfs/retrieve.cgi?native_or_pdf=pdf&id_document=6516287716 .
105
A summary showing the extent of this dependence is provided in a PowerPoint presentation available on the website of the National Telecommunications Cooperative Association at http://www.ntca.org/content_documents/2003.Legal.Seminar.DM.11.07.03.ppt (viewed August 31, 2004).
106
For example, wireless carriers and now VOIP providers are gaining market share and eroding the access charge revenues of, particularly, the small, rural carriers that have the highest access rates and are most dependent on access revenues. Even if competitive wireline entrants never appear in the rural areas, the increasingly pervasive intermodal competition may make the current system non-viable, and perhaps sooner than later.
33
CoasianAlternative040901b.doc
09/08/04
from it. 89. What makes this process so interesting is that no party to the negotiations can predict with much confidence what the future holds. No party is entirely sure, for example, of the impact of VOIP (or of the next innovation to come along), of the spread of broadband services, of the feasibility of telecommunications via power lines, or of dozens of other technological developments. These impacts will depend largely on legislative and regulatory decisions that cannot be well predicted by any party. In the American system, laws and regulatory rulings are subject to review by courts. Although various parties might believe or hope that they can successfully influence legislators or regulators, the history of court rulings, especially in the telecommunications arena, makes any prediction of the ultimate outcome quite a chancy proposition. 90. What we seem to be observing in the presence of this uncertainty is something loosely approximating a Rawlsian convention107 (in addition to the usual struggle of interest adjustments and defenses of economic rents). Rawls imagined a constitutional convention in which individuals negotiated over the laws that would apply in their lives; but Rawls convention occurred among unborn individuals, before any of them knew what lot they would draw at birth. He surmised this would lead to adoption of fair rules. Somewhat similarly, there is some suggestion that carriers may be seeking a solution that would reduce their risks in the face of great uncertainty as to the legal and technological developments that may occur. 108 That is, they may be seeking a competitively neutral solution. We find it significant that the chief proposals appear to resemble, at least loosely, our models solutions. 91. A recurring comment from carriers in the FCCs Unified Intercarrier Compensation NPRM is that it would be very difficult to identify and measure those facilities that are not interconnection facilities, but are nonetheless actually incremental to interconnection. Interconnection facilities are the actual connections between networks: the cables and entrance facilities connecting the two networks switches. The facilities incremental to interconnection include these, but also whatever incremental facilities may be required within one or the other network to enable exchange of traffic. These internal incremental facilities can be easily identified on our simple diagrams such as Figure 4 as the incremental facilities that are within one or the other network. Identifying them in real world networks is rather more challenging, and would likely lead to challenges between carriers. 109 Some carriers have offered various proposals that we (choose to) interpret as practical ways to implement our conceptual solution in the real
107 108
John Rawls, A Theory of Justice, The Belknap Press of Harvard University Press, 1971.
A significant difference from Rawls, of course, is that these parties already know their starting positions.
109
It is worth pointing out, however, that the facilities in question likely consist primarily of interoffice transport and ports. This means that only a relatively small proportion of the entire network is at issue. In contrast, under the current regulatory regime, the entire network is at issue in the access regime, and almost all but loops is at issue in the reciprocal compensation regime. Identifying the relevant facilities may be difficult, but the room for error and controversy is much less than if all or most of the network is at issue.
34
CoasianAlternative040901b.doc
09/08/04
world. These proposals bear at least vague resemblance to our SIIF rule in that they (1) typically involve bill and keep rather than reciprocal compensation payments and (2) divide responsibility for incremental facilities in various ways that may approximate an even split, given the difficulty of identifying incremental facilities within networks. 110 92. We find it quite interesting that many carriers seem to be considering a solution at least broadly along the lines suggested by our model. It appears to us that carriers are seeking a solution to the intercarrier compensation problem that lowers risk regardless of what technological and regulatory developments arise. Pending further information, this may be strong evidence that the solution we propose is actually competitively neutral. F. Conclusion
93. A Coasian property rights approach appears to yield a more efficient and competitively neutral alternative than Pigovian-style rate regulation. Furthermore, there is less risk of damage due to regulatory error because carrier behavior is less likely to be distorted. In addition, regulatory processes are better suited to property rights adjudication than to price-setting, and it is easier for the parties to correct any initial error. We have seen that there is intriguing evidence that telecommunications markets seem to move in this direction under competition in absence of regulation. Regulation and debate concerning intercarrier compensation has long been dominated by accountants, lawyers and economists who do not have a clear understanding of network costs. We have only sketched a broad approach to a new way of thinking about the interconnection issues that have proved troublesome as competition has begun to spread in telecommunications and new technologies have altered traditional relationships. Perhaps it would be fruitful if economists were to examine these issues from this fresh perspective.
References:
Atkinson, Jay M. and Barnekov, Christopher C., 2000, A Competitively Neutral Approach to Network Interconnection. FCC, Office of Plans and Policy Working Paper Number 34, December 2000, <http://www.fcc.gov/Bureaus/OPP/working_papers/oppwp34.pdf > Brock, Gerald W. Telecommunication Policy For the Information Age: From Monopoly to Competition, Harvard University Press, 1994. Brock (1995) Gerald W. Brock, The Economics of Interconnection, Teleport Communications Group (1995). The three component articles in this publication were placed i...