;.":!: !'•• .;',-. «::V s; MJ.T. LIBRAR5ES - DEWEY Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium Member Libraries http://www.archive.org/details/accesspricingcom00laff2 Cewey working paper department of economics ACCESS PRICING AND COMPETITION Jean-Jacques Laffont Jean Tirole 95-11 July, 1994 massachusetts institute of technology 50 memorial drive Cambridge, mass. 02139 ACCESS PRICING AND COMPETITION Jean-Jacques Laffont Jean Tirole 95-11 July, 1994 MASSAC HUSE.7 z, MSTiTLi OF TECHNOLOGY 1 MR 2 9 1995 LIBRARIES <"! 95-11 Access Pricing and Competition* Jean- Jacques LaiFont^and Jean Tirole* July *We thank ^Institut Jerry 6, Hausman and John Vickers Universitaire de France, and 1994 comments. d'Economie Industrielle, Toulouse, for helpful Institut France *Institut bridge, d'Economie MA USA Industrielle, Toulouse, France; CERAS, Paris; and MIT, Cam- °t 5--// Abstract In many industries (electricity, telecommunications, railways), the network can be described as a natural monopoly. A central issue is how to combine the necessary regulation of the network with the organization of competition in activities which use the network as an input and are potentially competitive (generation of electricity, value- added services, road transportation). In this paper we derive access-pricing formulas in the framework of an optimal regulation under incomplete information. how First, we study access-pricing formulas take into account the fixed costs of the network and the incentive constraints of the natural monopoly over the network. Second, we examine the difficulties coming from the ac- counting impossibility to disentangle costs of the network and costs of the monopoly's competitive goods. Third, the analysis is extended to the cases where governmental transfers are prohibited, where the competitors have market power, and where competitors are not regulated. is Fourth, the possibility some consumers bypass the network taken into account. Finally the role of access pricing for inducing the best market structure results is assessed. The conclusion summarizes our and suggests avenues of further research. Introduction 1 many In industries (electricity, telecommunications, railways), the network can be de- scribed as a natural monopoly. However, many activities which use the network as an input are potentially competitive (generation of electricity, value added services, road A transportation). central issue is how to work with the organization of competition Two different types of policies combine the necessary regulation of the netin those activities. can be observed. In the USA, the local network mo- nopolies in the telephone industry have been prevented from entering the value added markets as well as the long distance market because of the Department of Justice's belief that in it is impossible to define access rules to the network which create those markets 4 . The argument here is that the network to provide unfair advantages to it is its R&D subsidies...) superior quality of access, cost auditing. In the language of competition too easy for the monopoly in charge of own products even if (favorable access charges, subsidiaries are created to improve modern regulatory economics, asymmetries between the regulator and the monopoly are so large that tion fair of informa- competition fair is not possible. The alternative policy of defining access charges common. The long also (BT) to reach market. The and letting the regulatory choice of the access charges regulatory review of 1991, its European Community BT is BT in the long distance complex. For instance, in the has argued that the current setting of access charges creates competitor (Cave (1992)). Note also that it is the goal of the to define "reasonable" access charges to organize the competition of electricity generation in Europe. Other examples of computer reservation systems and gas It is difficult activities with regulated access are pipelines. to appraise the difficulties of regulating access if all the imperfections of regulatory processes are taken into account simultaneously. In this paper, we start an optimal benevolent regulation of a network monopoly facing a competive fringe potentially competitive markets focus incorporated into our modeling. a first of a The modeling step, we choose of the costs to focus more general theory in from in the and we successively introduce various imperfections. To on access we neglect the relevant policy. is distance operator Mercury pays access charges to British Telecom consumers through the local loops and compete with unfair advantages for monopoly compete issue of network externalities, which could be easily Perhaps more importantly, we ignore the divestiture and benefits on the of vertical integration is complex, and, in vertically integrated structure as a building block which breakups might be desirable. '"'DO J... Jid not believe that judicial and regulatory rules could be effective in preventing a monopolist affiliate that provided competitive products and services" Noll (1989). from advantaging an The existing theoretical literature on network access pricing includes two papers build- on contestable markets theory. Willig (1979) studies interconnection by competing ing He suppliers. may stresses the fact that strategic behavior deter socially beneficial entry (for example to expand regulation) and analyses the extent to its lead the network operator to rate base under rate of return which Ramsey pricing market structure. Baumol (1983) discusses similar compatible with the best is issues in the framework of railroad ac- cess pricing. Vickers (1989) addresses the access pricing issue within the Baron- Myerson model of regulation under incomplete information. monopoly (network integrated sector of services with commodity in particular a vertically plus services) with the case of an imperfectly competitive and without participation Section 2 describes the model. olized He compares A of the network operator. monopoly operates a network. (for concreteness, local telephone), as well It produces a monop- another commodity (long distance telephone) which faces a competitively produced imperfect substitute. The long distance operators require access to the local network in order to reach the final consumers. That the potentially competitive market needs to use the monopolized is, The purpose an input. access. We start of the paper is commodity normative theory of how to price to develop a by postulating that the regulator has complete information and faces a social cost for public funds associated with the deadweight loss of taxation. price must then exceed marginal deficit. Equivalently, access for a further any, is meant The access cost in order to contribute to the reduction of the overall can be priced at marginal cost and a tax levied on access. Sec- tion 3 studies the impact of the regulator's incomplete information cost structure as on the monopoly's informational about the monopoly's Incomplete information rents. departure from marginal cost pricing of access. The may call further distortion, to reduce the monopoly's rent. Section 4 explores conditions if under which the observability of the monopoly's subcosts, namely the cost related to the monopolized good and the The analysis ceive any its cost related to the competitive is extended good respectively, improves social welfare. in Section 5 to the case where the monopoly does not monetary transfer from the regulator and therefore must cover revenue. The access charge is cost with then higher or lower than in the absence of a firm-level budget constraint, depending on whether the monopoly's fixed cost 6 generalizes our access pricing formula to the case in is its re- is high or low. Section which the substitute commodity produced by a competitor with market power rather than by a competitive industry and Section 7 considers the case in competitive segment at all which the regulator has no mandate to regulate the (including the production of the competitive good by the monopoly.) Sections 6 and 7 emphasize the complexity of regulating a competitive segment with limited instruments. when consumers Limitations on regulatory instruments are particularly penalizing of the competitive commodity can bypass the use of the monopolized good (local telephone). The determination of the access price conflicting goals of limiting inefficient bypass it and must then trade off the of contributing to budget balance (be the State's or the firm's). Unless the regulator is able to simultaneously tax the long distance market to raise funds to cover the firm's fixed cost and use the access charge to approximate the efficient level of bypass, only a rough balance between these objectives can be achieved. The separation of the regulatory function from the taxation authorities creates serious difficulties in the presence of bypass (Section 8). Section 9 links the determination of the access charge with the choice of a market structure and with the promotion of entry in the competitive segment. ponent pricing rule recently advocated by some economists is The efficient com- assessed in the light of our normative treatment. Section 10 compares the theoretical precepts with two regulatory practices, namely the British Telecom. fully allocated cost method and the Oftel The Conclusion provides a summary brief access pricing rule for of our analysis and discusses avenues of further research. The Model 2 A monopoly operates a network with cost function Co Q where of co, Q) Co0 > , effort is F(.) on [3, J3] C0eo < , with a strictly positive density function /. 0]). With the network the firm produces a quantity q of a let (1) private information of the firm is > telephone). 0, exerted by the firm in operating the network. rate assumption {jg[F(3)/ f(3)] monotone hazard Coq > , a parameter of productivity, and eo a level a parameter of adverse selection which c.d.f. and (/3, the level of network activity, 3 nonmonetary ,3 is a is =C : Let us assume that q units of this also a good 1 the classical monopolized commodity commodity require S(qo) be consumers' utility for this good, with S' The monopoly produces We make 3 has ; > 0, S" < q (local units of network 0. (long distance telephone) in quantity q x . This other production has an imperfect substitute produced in quantity q2 by a competitive fringe. Let V' ( <?i ^2 ) , be consumers' utility for these two commodities. In addition to the network input the production of C\ where t\ is = good 1 also generates a cost Ci{3,e x ,qi) the effort exerted to reduce the cost of producing (2) good 1. The disutility of effort is + v{e ex with ) y>' > 0, 0" > > y'" and 0. good 2 generates a In addition to the network input the production of c is common knowledge. Because we wish to focus on the network monopoly's incentive we need not to provide access, Total network activity investigate incentive issues in the fringe. Q= therefore is qo + qi + q2 FT Under complete information, the effort levels. lator. We make Let t = - p 2 q2 Let a denote the access unit charge . The paid by the competitive fringe to the monopoly. and - cq2 fringe's level of profit aq2 is then (3) . utilitarian regulator observes prices, quantities, costs, denote the net transfer received by the monopoly from the regu- from the sale of the competitive good and network good to the con- sumers, and that the firm receives the access charges. This obviously involves no The monopoly's utility level U = tThe 1 + regulator must raise A (where A S(q ) > + V(q t ) + is then + ip(e + C + Ci — p ex ) + — piqi q x .q 2 ) - p <?o ~ Pl?l V(qi,q 2 ) + [t p qQ - - ii>(e ~ P2<?2 - aq2 (4) . with a shadow price of public funds + (1 'M(* utilitarian regulator piqi - p 2 q2 - loss of : because of distortive taxation). The consumers'/taxpayers' Under complete information an S(q : the accounting convention that the regulator reimburses costs, receives directly the revenue generality. where cost cq 2 (1 + e + aq2 + t) + A)(f p 2 q2 + CQ + C - p X q - utility piqi). is (5) would maximise + C + C\ - p - cq2 - q - piqi) aq2 (6) under the individual rationality (IR) constraints of the monopoly and the fringe U = -0(e o + t EI = p 2 q2 — ei) cq 2 — + aq 2 aq 2 > >0 (7) 0. I Since public funds are costly, the monopoly's IR constraint can be rewritten is $ binding and social welfare S(q - + (1 Similarly, raising 4- ) V(ft, ft) + A) (V(e ci) constraint +C money through the term Xaq2 in social welfare). + + Apxft (/3, co, ?o + ft + ft) + (9), social S(q - + (1 A) (V(e Assuming that 5 and The access charge is ) + V + ex) = + V(ft, ft) +C (P, e are concave valuable (see (10) , A(p ft + Pift + P2ft) <7o + + ft Ci(j3, and Co and C\ convex ^^ 1 A = ?t 2 optimal we can write (12) + j_ A ft 1 + A (14) 772 are (respectively) the price superelasticities of goods 0, 1,2, and v'(e + ex) = -C0eo = -Clei Straightforward computations show that f/ = 770 (ftft~ft2ftl) ft=ft 772 = pi Opi </,- . (15) and ^ < ft \ nfi lb l ^ < 772 n-i 111) ) ftft2 (ftft~7i2fti) 772 ; ft ft t + . ; ftft = — —0q in (eo, ex, ft, Q), (11) 11. m P2 rj cq2 ) 7/o 1 n-G»-c u ft) + r±rl +A = = Pi-C g-Cx Lim ft, + ft) Pi 77,- is (9) chosen to saturate the fringe's IR is characterized by the first-order conditions, that £i , Ci, ft)). p2 -c. Po where Ci(0, welfare takes the final form U= 770, cq2 : Substituting (10) in where - Aag2 the access charge from the fringe a regulation + ft Ap + ftftl = 7.-i i*' jt- The benchmark J ' = 1 ' 2 = z 1 < 2 - - pricing rules are Ramsey pricing rules because A > 0. The access charge can be rewritten 1 Access °2 (1 2 is 7/ 2 priced above marginal cost because deficits are socially costly. low or when the social cost of funds issues raised by If (18) A can be viewed as a tax used to raise money. - , , is + is high It is when the elasticity of good high. In the next section we address various asymmetric information. the regulator has two instruments, the access price and a tax on good (14) can The term be rewritten a + marginal cost of access. r2 = Cqq + y+x^ Furthermore, it is 2, r2 equation , an d a can obviously be taken equal to the clear that if the regulator had access to a nondistortive (lump sum) tax, marginal cost pricing of commodities and access would be optimal. Access Pricing and Incentives 3 The first pricing. issue we consider is the extent to which asymmetric information affects access Let us consider the case where accounting rules enable the regulator to observe separately Co and C\. Let Eo((3,Co,Q) denote the solution in the solution in e\ of C = x t Ci(0,ei,q{). e The of = Co Co{{3,eo,Q) and firm's utility + aq2 -ij,(Eo(P,Co,Q) + E 1 let Ei(j3,C\,qi) can be written (19) (l3,Cl ,ql j). Appealing to the revelation principle we consider a revelation mechanism {*(£), which C (/3), CM, qi 0), q2 0), specifies the transfer received, the sub-costs to produced, and the access price to be received if Q0), a(0)} (20) be realized, the quantities to be the firm announces a characteristic Neglecting momentarily second-order conditions, and using the rent variable U(j) = t(,3) + a(/3)q2 (<3)-T!>(EQ (0, CM, Q(0)) + Ei(0, CM, ?i (/?))), j3. monopoly can be written the incentive constraint of the ^=-^^Hw + w)- c Since -^ > ^- > and > (IR) constraint {U(3) 0, the rent for all down boils /?) decreasing in is U{0) Optimal regulation +\[po(0)to(po{fi)) -(1 + A) {v(e + (j3) > 0. (22) + V( qi (pM,P2(P)), (/?, e Q ((3), rationality to + M0)qi(pM,P2(P)) +C ei (0)) and the individual from the maximization, subject to (21) and results {s(qo(po(P))) \l /? (21) fc(po (0)) (22), of fc(Pt(fl, P»{0))) +P2{P)9i(pi{P),Pi(fi))] + qMPlPiifl)) + toMfiMP))) + C (^eMMPi(0),P2(0)))+cq2iPi(0),P2m)]-XU(0)}dF(0). t We obtain (see appendix Po Pi ~ A " Cqq l — + C\ qi Xf, l p2 + 1 A £ tf7_0_f +A / Pl + 1 A _ F A 1 ?\ T (23) 1). - C0Q Po : f A / Po dQ\ -Co Q -c = _A_ i + A p2 A \ [ ,Q)/ j?i _J_ F 1_ " fj ,Q) i C03 (/3,eo,Q) \ C0eo (^eo,g)J Ugl C03 (P,e C0eo (/?,e d y/ i 2 + A d <9?i ry_5_ / p2 f I f agl Clg (y3, ei gl n CUl {P,euqi)}J ) , _ Cofl (/?,e ,g) ) c0eo (/?,eo,g)/' All prices are modified by an incentive correction related to the sub-cost function Co and pi has The an additional correction associated with C\. now be written can access charje 3 ' J =c A p2 is F , v d f C'OB ( ^^TTA^ rTA7 ^i" c^r Let us analyse the new term, which that there A + no correction for the is most ( ' due to incentive constraints. efficient type as F((3) = 0. First, we observe Second, -jf- = — Co/j/Co eo same the to is the rate at which the firm must substitute effort and productivity to keep level of cost for the determine the firm's network ar "ivity. From The rent. conversely The if it is negative. we see that it is a crucial term regulator wants to limit this costly rent. and positive, he raises the access price is (21), If -Sni^f) therefore reduces quantities to extract rent and on the access charge thus depends on effect of incentives fine characteristics of the cost function. Access charges must be increased (decreased) for incentive reasons network of the C is — > Co0gCoe o is obtained if (<) > "Spence-Mirrlees"' condition on cost, Cqqq clear positive sign Intuitively, production marginal cost. However, effort increases the about lie if, as is more ambiguous. if an increase its in the level of Q characteristics (in elasticity the quantity elasticity of the marginal effect on cost of the productivity charac- if teristics is must be decreased (increased) increases (decreases) the "ability" of the firm to 0. and our previous assumptions, a reasonable, effort decreases the marginal cost, the effect terms, the cost function such that t-OeoQCoQ With the if j3 is larger (smaller) than the quantity elasticity of the marginal effect on cost of effort). From we know 5 Leontief's theorem charge (and on good 1) disappears iff Co If there furthermore C\ is is separable in in addition that the incentive effect on the access there exists a function £ such that =C (/?, (£(/?,e ),Q). ei) (23) on the one hand, and no incentive distortion on any commodity : qi on the other hand, the dichotomy between pricing rules (unaffected) and the cost reimbursement rules (see below) holds. The access pricing rule can then be rewritten 1 Note that Pi — Cqq — is T}2 la 7iJ also possible to find conditions conclusions for the incentive correction (29) A C\ qx Pi It + : on the when it -aOQ Pi cost function that yield exists. nonambiguous For instance, the following class of cost functions leads to a well defined sign of the incentive correction in the access charge 3 From Leontief's theorem we know that ^- independent of (-Oe„ function £(.) such that (29) holds (see Laffont-Tirole (1990a)). 10 Q is : equivalent to the existence of a Co If d > b(d Let us < now 6), = (3Q the incentive correction - „c/->& c eQ Q positive (negative). is consider the effort levels (in the case of dichotomy) obtained by maximizing under (21) and (22) with respect to (23) d 60 , ei). We obtain '/''(eo + ei) = -C0eo ^'(eo + ei) = -Clei 2\F (iTwr (e ei) (w + w) + 0,(eo + ei) Wdc: Cui Equations (30) and (31) can be interpreted as cost reimbursement (31) rules. If the con- ditions underlying the irrelevance of subcost observability (see Section 4) hold, one can obtain sufficient conditions for the optimal contract to be implementable through a single menu of linear sharing rules share of the total cost which is t = A — B(Cq + C\) where B = From (30), (31), and reimbursed. the most efficient type receives from the regulator a tract) and equates its disutility lump sum ^'fc F({3) = 1 = 1 ^'fc we 0, " 1S t ^ie see that transfer (fixed price con- of effort to the marginal cost reduction as under complete information. REMARKS 1. - The most relevant cost functions seem to be such that higher production higher effort levels and therefore induce higher rents. for marginal costs and therefore indirectly call for a general Incentive considerations raise reduction of quantities. Access prices are increased for the competitor but simultaneously the high prices for its own product and own misrepresentation. efficiency of its this link 2. is - If is products. Because access costs are the the competitor's, the current model The levels call may monopoly same penalized by is for the monopolist understate the incentives s for firm cannot claim simultaneously high access costs and high own products (see Laffont-Tirole (1993, Chapter 5) for an example where suppressed). the fringe is not regulated but is competitive, the same results obtain ; P2 = o. — c then ensured by competition rather than by the existence of a shadow cost of pubiic funds. 11 3. Let us consider implementation of the optimal revelation - ing specification C = now includes t qi +q2 .32) ) (33) monopoly can be rewritten objective function of the i(fi) + ){q i/ 1 (/?-e 1 )<7i- 1 where follow- : C = H (p-e The mechanism with the -^20- HQ- Co l ( the access charge. H?(Z)) - ) The second-order (34) condition of incentive compat- ibility is d[3{ and the revelation mechanism r( The price firm now is free to + qi + \q ° equivalent to a nonlinear reward function is v f_&_) Co choose and the quantity of good its An cost, (as well as 1 which determines alternative implementation Co, C\, production levels qo, the conditional demand \ is . good it is TJ-\(C\ = D 2 (qi, chooses. example) which service (good 1) selects the level of raises the cost for the is C\ = = all prices leading to the jQ + Co(p,e + a) and substitute in (36). ,Q) It decreasing in the access charge i (an network, but decreases the cost of -iqi +Ci{/3,e l ,q 1 ). 12 c For this purpose consider some unobservable action : Co will affect the can be advised to choose the picks. it it for it obtained by writing the transfer as a function of costs easy to check that the transfer received by the firm vestment knows that its rent. function of good 2, q2 Suppose that the monopoly (36) among indifferent is 4. - it It 2). and access charge a qi : +iir i(£i)) access charge but access price defined in formula (27) since same weighted \qiJ) q2 J inits Under complete information the optimal investment is = i level, which minimizes total cost, qi/Q. Under incomplete information, the moral hazard constraint on investment dE - .dE l is x Q+ dc; qi=0 dco or .^qi dEildCx QdEo/dCo' 1 To decrease the C and C\, may use (implying exist if j^P- ^ who rent of asymmetric information the regulator, cost reimbursement rules with different rates j^-) and this may observes both for sharing overruns lead to a distortion of investment (which would not the regulator were not observing separately sub-costs Co and C\ as in the next section). If 1 i too low (too high), the access price is (favor good 2) and may be viewed by is to foster is increased (decreased) in order to favor good the firm as a nonrecognition of its On low access charge investment costs in the network. indeed low to decrease those types of investment which 4 A an increase (decrease) of investment. may be It excessive. Sub- Cost Observability Section 3 derived the optimal regulation under the assumption that the regulator could audit the costs of the regulated monopoly well from the cost of the commodity produced Using Ck instead of e k in (l A X)rp' (1 + + A)/ XFr + cost of the network in the competitive sector. the optimization leading to (30) + = enough to separate the d{E ° dEk dCk + El) 00 W Hj' and " = ' (31), we have \ l (37) high-powered scheme on activity k corresponds to a case where the transfer received by the firm is highly dependent on cost Ck , i.e. where ( — §§*•) is Indeed high. if ^ is high in absolute value, marginal cost reductions on activity k per unit of effort are small and therefore the firm is already induced to exert a large effort on activity for further reference that if the cross If terms gC ^ vanish the terms sub-costs are not observable, the firm minimizes chooses {e k } or, equivalently {Ck }, its effort for so as to minimize T.I=q 13 ( — ff^) k. are equal. any cost Ek {/3,Ck ,Q k ) Also, note target. subject to C It -f = C C\ Q Q = Q,Q\ = (with the convention Hence, in the absence of sub-cost qi). observation the marginal effort levels to reduce sub-costs (— -Sf-) are equalized 6 Whether the regulator wants in to give differentiated incentives on different activities the case of sub-cost observability hinges on the cross-partial derivatives that this expression absolute value proportional to in is which the firm can substitute 3 and determinant of the firm's rent. no point is of effort to minimize cost + =C C\ solution to the system in order to |^ = -^ { observability is useless if This result d 2 tells Section 3) and it is at the main total effort Yi\=Q Ek(3, Ct, = Ek (3,C,Qo,Q\), and Co + C\ = C }. for all (k,£) we obtain which (see appendix 2) is Qk) the unique : : the firm faces uniform incentives ii) measures the rate has a unique solution e k the firm faces higher incentives i) (note reduce rents. Solving for the optimal regulation Under sub-cost h the level of cost or the level of effort does not affect this assume that the minimization of formally, let us subject to Co effort in activity k (see -& ). E QC in using sub-cost observability to try to affect the firm's allocation rate, there More If A de . E k /dCk d3 = on those on activities and therefore sub-cost observability all activities K for all k for which low costs reduce rents and some constant K. us that the incentive constraints of the firm are identical with or 2 without sub-cost observability when d Ek/dCkd3 = = constant for k the optimal pricing rules and in particular the optimal access pricing 0, 1. Consequently unaffected in this is case by sub-cost unobservability. A characterization of the subclass of cost functions satisfying the sub-cost-irrelevance property with that K = is easily obtained. = for A: = 0Ck d3 some functions On H k and 0,1 <* ek the subcost functions must be such = Ek {3,Ck ,Q k = Hk {3,Q k + G k {Ck ,Q k G k for k = 0, 1 ) ) and, since C < ejk 0, ) for A: = ) Ck = Ck (Q k Hk (3, Q k )-e k , the other hand, recall that the dichotomy property holds Ck = Ck (Zk (3,e k ),Q k at all k, : d2 Ek for For when ). : 0,1. 'Then, the theory of access pricing is similar to the one of Section 3 but for marginal costs evaluated the effort levels induced by the incentive scheme denned when sub-costs are not observable. 14 These related conditions are d dQ k The dichotomy different. fdEk \ d 2 Ek 33 dCk d/3 V J requires d 2 Ek dCk dQ k 8Q k d0 while the irrelevance of sub-cost observation requires d2 E = dCk d(3 The tfforallJfc. class of cost functions Ck (0,e k ,q k = ) satisfies the sub-cost irrelevance property for any d k = property for d k On 3q k " , bk e k qk (with bk Ckqk > 0) and the dichotomy b k only. Ck (3,e k ,qk = ^t the other hand, satisfies ) the dichotomy property but not the sub-cost irrelevance property. Optimal Access Pricing ernment Transfers 5 We now assume us for that the regulator prohibited from transferring is example consider a constant returns to irrelevance of subcost observability hold Co = H {3 l let c = H (f3 —e ) cj , = H\(3 — e - e Q )(q Q l x ) scale case + qi + qi{Pi,a where the dichotomy and the q2 ) (we omit the fixed cost for notational simplicity). PoqoiPo) +Piqi(Pi,* + to the firm. Let (/3-e l )q1 Under symmetric information, the budget constraint -<-'o(<7o(Po) money : C =H and Absence of Gov- in the + c) + aq 2 of the (p\,a monopoly is + c) + c) + q2 (pi,a + cj) - c^p^a + c) - t > 15 0, (38) where must now be interpreted t consumer charges The U= — (so as the firm's v(eo t + regulator wishes to maximize, for each value of subject to constraint (38) (where U is leads to the incentive = -V(e is, -co (qO ( Po(0)) max > + : e 1 ((3)) (39) (40) 0. : 7i (pi(i3), c iqi ( Pl (;3), a(/3) [S(q (p (3))) -Pi(0)qi s.t. + + V( qi + c) program regulator's optimization / : + Pi((3)qi (pM, a(0) + c) + a(0)q2 (Pl (/3), a(0) + c) Po(/3)qo(po(i3)) - each 3 for social welfare + ei (/?)), (/?) -iP'(e (3) U(0) constraint t, namely and individual rationality constraints = C\, )-p l q l (p l ,a. + c)-(a + c)q2(pi,a + c) + [f q (p firm's rent, t(3) U(f3) The budget Cq, the firm's welfare). Under asymmetric information, the U{B) paid from is ei)). S(qo{Po))+V(qi{pi,a + c),q2 {pi,a + cfj-p The manager's compensation and ( a(0) = is Pl (3), a(8) + c) U((3) + q2 (P i(/3), a(0) + + 0(«o(/?) c)) + eM). (41) : + c),q2 ( Pl (0),a(0) + cj) - Po(3)q [pM, <3) + c) - P2(3)q 2 (pM, a(8) + c) + (po(3)) U{3)\ dF(3) (39), (40), (41). Let fi(3) be the Pontryagin multiplier of the state variable U and (1 + multiplier of the budget constraint. Optimizing over prices we obtain the as in Section 2 with \{3) replacing sality condition fi(3) = 0, A. \(3))f{3) the same equations Using the Pontryagin principle and the transver- we have 16 Optimizing over 0'(eo(/9) + ci(/3)) effort levels = we get finally v / H'^qM + qi ((3) + q : 2 ((3)) ~ Si k0)S0)d(3 ,f (i \'\'j "(eo(0) + + Hp))S{p) e 1 (0)) and neoifi) When with ( + ei(fl) = SiMMiPW / - #{*(/?) - . ^ {eatf) + ei(fl). the dichotomy does not hold, formulae similar to those of Section 3 are obtained 1+A) /L) replaced by Sg\0)f0W In particular the access pricing equation a "m c0Q + , 1 Under the assumption + P2 A(/?) + becomes www ss , (i 2 + f m)m dQ\ c0eo on how binding the budget constraint is. is shifted by the upwards or downwards depending Consequently, the access charge than in the absence of budget constraint when the fixed cost is is higher (lower) high (low). Access Pricing and Competitor with Market Power Competition for (42) / of dichotomy, the ratios of Lerner indices are unaffected lack of transfers, but the whole price structure 6 -coa] a in the competition electricity in markets using the network as an input in long distance often imperfect, as is the case telecommunications, or competition in the generation of England. To account analysis of Section 5 with the is for this market power we pursue the balanced- budget same technology (guaranteeing the dichotomy and the irrelevance of sub-cost observation). 17 If the competitor has market power and is unregulated, the maximization of expected welfare must be carried under the constraint that the competitor chooses price so as to maximize its profit : + ft#(ft,p2 - ttfri.ft) From ) ap2 + a) Aft,;*) = (c 0. r (43) tfp 2 we the first-order condition of this maximization program (see appendix 3) obtain the access pricing rule : 09 1+ V2 X(/3) I T/2 '/l^ (44) - TJ12V21 with V2 = 7/i7/ 2 V2] ' Equation (44) is superelasticity and only r\\ demand if 7/217/12 corrected for market power. > — 7/1(7/2 1). > budget (reflected in A power. To explain 7/ 2 i7/i 2 good 2 describe the More to offset the demand for 0) calls for a higher final price increase in p 2 (that monopoly case good than (t/ 2 i 2), interesting = t/ 12 = in the However, with dependent demands, there the expression of t/j ) : pi is is monopoly r/ 2 if < 0), n\ 7/2 absence of monopoly to a social loss equal Because marginal revenue in the access price) reduces this is, the is the need to balance the monopoly power amounts distortion. changes exceeds the regular superelasticity demand elasticity notice that the this, It -). effects of times the monopoly's profit p 2 <l2l Tl2i once the standard subsidy p 2 q 2 /V2 made the for In the independent and therefore (assuming a constant in 7/1 and p 2 on the mark up or monopoly power 1/(1 in pi to A - T] complex, but can be given a natural interpretation. The two terms the derivatives of the elasticity of in is 27/17/2 - U T]21 + 7/217/2 - ( see (44)) decreasing, an is Hence profit. 7/2 a second effect (described by the term < t/ 2 . 7/217/12 reduced to lower the monopoly profit p2?2/ r/2- To rebalance consumers' choice between the two substitutes. P2 must also be reduced, and so must be the access charge. If the social welfare function did not include the competitor's profit (an extreme of depicting the redistribution problem), similar calculations a "An = C Q + - \(3) P2 t~t:^ would yield P2Vij;{^)+PiVx2i: {^) + : ^_ P2> l+A(J)7/ 2 alternative formula would be obtained way 7/17/2-7/127/21 if a nonlinear access pricing rule was used. conclusion. 18 See the Restricted Regulation 7 In practice, the regulator may let (or be instructed to let) the natural monopoly select his and regulate only the price of the monop- pricing behavior in the "competitive market" olized good and the access set its charges on value added services. This section studies the design and consequences price to the network. For instance, France of such restricted regulation a competitive fringe. We We come under simplifying assumptions. is free to back to the case of consider the class of regulatory mechanisms that associate with an announcement $ prices po{P) and The managerial compensation a((3). defined as a residual by the balanced budget constraint, once efforts e Pi Telecom and t((3) is then and price t\ have been chosen. The monopoly has an additional analysis proceeds as in Section 5 except that the moral hazard variable pi which leads to the first-order equation, that the monopoly's marginal profit be equal to zero : 9qi, ~ co ~ MP = qi + ^-[pi i Cl] dpi dq 2 + ^\P2 - Co dp! , c] = 0. we make the following assumptions Concavity of profit : dm < *&£ dpi Generalized strategic complements The standard : ^^- > 0. condition for strategic complementarity d ~ W + T^tPl opi °0 L ~ c l]]/ dP2 is > 0. J Here, the monopolist also receives income from giving access. for generalized strategic complementarity is A sufficient condition strategic complementarity plus the condition that the marginal profit on access increases with p 2 (condition that holds with linear demand and constant Appendix 6 returns to scale). shows that for a given shadow price, the access price is lowered relative to the unrestricted control case, in order to reduce the complementarity for a given p x strategic . On monopoly price through (generalized) the other hand, monopoly power on good 1 To rebalance the consumers' choice between the two goods, p 2 must be raised. The effect of restricted regulation on the access price is therefore ambiguous in general. raises px . However, in the case of linear demand, for the optimal remains the formula of full regulation. 19 pi, the access pricing formula Access Pricing and Bypass 8 A practical problem encountered ing access to the network may in the telecommunications sector useful as is is the following giv- : expands the space of commodities offered and it decrease the rents of asymmetric information by shrinking the incumbent's scale of More operation. the incumbent cations for far may it be useful for yardstick reasons also if the technology of correlated with the monopolist's technology. However, in telecommuni- is example giving access possibility of So directly to the network to long distance operators leads to the bypass of the local loop by large consumers to connect with these operators. we had no need of long distance, to distinguish between producer prices and consumer prices and we could assume without was loss of generality that access pricing the only tool used to oblige the competitor to participate in the fixed costs of the local loop. However, when bypass is feasible high access prices required for funding these costs induce inefficient bypass. Nonlinear prices for local telecommunication there is a much more efficient way of dealing with this necting consumer prices and producer prices The monopolist has marginal The fringe has marginal cost C2 10 b is the same is Let t2 be a tax on good is 2. price of good With a competitive = 2 charged to those price of good — a [0,1] of [0, + c2 + t2 a = The cost b. c2 + t2 = p 2 + b- good 2 is then : . the local loop is . through bypass has paid a fixed cost is de facto a 9. See Laffont-Tirole (1990b) for a different example of the use of nonlinear prices to 9 However, this option is consumers and a (low) marginal 00) i 10 long distance. Ci for fringe the price of who bypass 2 obtained p2 given that consumer But, distributed according to a c.d.f. G(9). p2 The marginal network and Let V(pi,p 2 ) be the indirect utility function. p2 The . problem which amounts to discon- Consider the case of a continuum . 8 . characterized by a fixed cost 9 € for everyone, 9 help mitigate this dilemma 9 cost cq for the local with identical preferences V(qi,q 2 ). bypass technology may fight bypass. rarely available to regulators. The dichotomy property is assumed so that we can study pricing issues independently of incentives who bypass For given prices (p u p2,a) consumers are those with 9 lower than 9' defined by V(pi,p2 ) Leaving aside good r Max 9' r* [V(p u p 2 / + where J 8m = -9 m + V(Pl ,p2 optimal pricing and access pricing are then the solution of 0, + b- a) - 9 + [V(pi,p 2 ) + (1 + A) (1 + r - c - Co)ft(a) + (p 2 - a - Pl x ( -d- A) [(p, ?i Let 77 x (a), 772(a), demand with the = = ?i(Pi,P2 + 6 - a) g 2 (a) 9l(?l,P2) 77 12 Let AR 2) AR = is -d- (p x an income the switch occurs. Co)(qi(a) effect. AR < It is - qj + (p 2 switches to bypass. for raising funds, means that less c2 - co)q2 ]]dG(9) dG{9) (45) seems AR ~ difficult to predict 11 . the sign of Then, from appendix p2 ?2(Pl,P2)elasticities demand - a - and revenues associated t}i,t) 2 ,t]i2,t}2i,Ri,R 2 functions qi,q2 c 2 )q 2 (a) If we think it is - (p 2 - c2 . - co)q 2 . of the difference between prices the difference of "collected taxes" taxes are collected 5 AR. So we we have when when the switch occurs. (see equation A5.3 in - c2 Tf 2 "Co if the complete results only + A 1 XtJi 1 l+\rj 2 are superelasticity formulas for the hybrid population of consumers bypass and those Note that A 1 - will give : Co p2 , + b - a) when no switch occurs Pl rf l ?2(Pi,P2 5). Pl-Ci-- 1 - n c 2 )q 2 {a) the loss of profits (including access charge and tax on This should favor a lower access price than where (pa functions qi(a),q 2 (a) and the prices Pi,p2 and let and marginal costs as taxes for = be the (a), /721(a), i£ 1 (a), i? 2 (a) when a consumer appendix = 72 the elasticities and revenues associated with the It + co) 9l : for ease of notation ?i(a) good + b-a). in who do who not. addition a direct subsidy (or tax) for bypass was available this effect would disappear. 21 The deviation of the access price from the marginal cost Q - where is 772 mation f 1 are constant, if 7/2l(a) 772(a) f the proportion of those the classical super-elasticity and similarly L 1 r72(a)/' who bypass ft : ss J71) 12 , is small (then we obtain the f2 * 72 approxi- : ~ a Pricing access at marginal cost We _ 1 1+Alf2 p2 If elasticities A cp _ then is is Co. then appropriate. have assumed here the existence of transfers. If there is no transfer, similar results hold as long as the tax revenue on good 2 goes to the monopoly. If not, access prices participate strongly in raising funds to pay for the fixed costs of the local loop and very bypass inefficient is be expected. to Access Pricing and Entry nent Pricing Rule. 9 : The So far we have focussed on pricing access when there exist competitors The concern was the which need access. firms. In other cost. The optimal now that the competitor access pricing rule for some services efficiency of supply of services given existing words we have assumed that the regulated firm's the access price. Suppose Compo- Efficient may be must pay a affected rival enters regardless of fixed entry or operating by the existence of this fixed cost, because the entrant does not internalize the consumer surplus created by the introduction of good Ideally one 2. would like to use direct entry subsidies to deal optimally with the entry decision. [Under incomplete information about the entrant's cost, one would need a nonlinear transfer function of the quantity to be produced is about the variable fixed cost if cost, it is may be optimal induce a better entry decision about the fixed cost]. In the absence of such to lower the access price obtained so far to 13 . Also, concerned with entry decisions, :: the asymmetry of information and a stochastic decision of entry as a function of the announced the asymmetry of information alternative instruments if Baumol has proposed a simple and In the absence of these assumptions different taxes for those who bypass and those influential who do not would be desirable. :3 The are a relatively favorable interconnect prices levied dynamic version market structure. A on Mercury to access British Telecom's network widely believed that they were designed to realise a particular different but related reason for low access prices is the high switching costs of of this point. It is consumers. 00 u access pricing rule , In our notation, this rule and the monopoly's price its recommends an access rule) which The idea and only is : = pi-ci. (46) that an entrant with marginal cost c on the competitive segment will enter if it is more efficient p2 ECP price equal to the difference between marginal cost in the competitive market a • (ECP rule" from the precepts of contestability theory. follows if component pricing called the •'efficient = a : +c= (pi - Cj) rule under perfect substitutes +c< The : pi &c< c\. make following assumptions this rule optimal. First, the monopoly's and the entrant's goods are perfect substitutes. Otherwise entry by a less efficient entrant may be Second, the regulator observes the desirable. monopoly's marginal cost on the competitive market. Third, the entrant has no monopoly power. Otherwise, the choice of p\ not only determines the access price, but also constrains the entrant's monopoly power. Fourth, and relatedly, the technologies exhibit constant returns to scale. Fixed entry costs for instance would create several difficulties with the Baumol On rule. On the one hand, the no-monopoly-power assumption would be streched. the other hand, the entrant might not enter because he does not internalise the increase consumer surplus created by in [Incidentally, pricing. But, p 2 = a entry. Fifth, the Baumol seems + c is lower + ci).] On to benchmark pricing recommend Ramsey than the Ramsey price for cost (cq rule is marginal cost pricing as the benchmark. + c) if p\ the is Ramsey the other hand, if those five conditions are satisfied, argued that the access pricing rule (46) is irrelevant. for cost (cq should supply only access (p\ is irrelevant) ; or c > C\ Either c < cx it price can be and the monopoly and no access should be given (a is irrelevant). • ECP rule under imperfect substitutes differentiation (as we have done the competitive segment. To we assume constant returns entrants, until : now) to In the rest of this section justify the presence of several firms in retain the spirit of the for the much as possible. known marginal cost for the Baumol to scale, competitive entrants, and the dichotomy property we introduce rule as monopoly's cost function. For instance, we can take the sub-cost functions assumed in Section 5. The dichotomy ensures that the pricing rules will not be affected by incentive corrections. • 4 is offered by a single supplier who also competes the single-supplier component's price should component, offering the remaining product See Baumol (1993) with others cover its in : "If a component of a product incremental cost plus the opportunity cost incurred when a rival supplies the 23 final product". our model the Baumol rule amounts to In a(0)=p (!3)-c 1 1 Competition =p 1 (f3)-Q-. competitive segment implies that in the = Pi(fl) + c — C\. />2(/?) Optimizing expected social welfare under constraint (46) and under the incentive constraints (see appendix we obtain the 6), = a access pricing rule + Coq 1 where is ife an "average V2 = — 9i T>2 : + . , Vi + 92 to the formula a —+ P2 92 ; 9i = 772 + 92 Pi 9i 92 is 72i ; = Coq + ^^L ^ 772 92 72 here of dichotomy), the elasticity used The (47) " M/3) 72 elasticity" Pi 9i Compared + 9i 9i + obtained in Section 5 (in the case assumed not the correct one, namely ~ 7712*721 7l72 + 72 721 771*72 difference between the two formulas (assuming that there centive correction) stems from the fact that the Baumol is rule ties the no need Symmetric demands and p2 - c - Co = pi - ex - Co, costs. One then has for r/ 2 and c = c\. an in- : These imply that or a Thus = fji for two prices on the competitive segment. Let us compare the two rules in some specific cases - (48) »7l2- 92 symmetric demands and = - pi cx . component pricing costs, the efficient rule is consis- tent with optimal regulation. - ous Linear demands, symmetric symmetry assumption in costs, and captive customers. one respect the monopoly has captive customers while : competitors do not. Under linear demands, the with d < b and a 2 < at . demand 9i = ai - &Pi + dP2 92 = a2 — bp2 + > pi and 24 a < functions are dpi Marginal costs are identical Pi Let us alter the previ- px : — Ci = C\. c. Our formulas then yield The intuition that the is monopoly must charge a higher price than because captive customers make markup relatively markup on good turn, this high component efficient efficient in can be viewed as an access subsidy relative to the 1 pricing rule. previous example that ai = a 2 but c x , Pi The intuition now is < < p2 Then our formulas c. and that under linear partially reflect the cost differential c 2 — < a demand C\. price cap Price Caps : — pi Let us assume in the yield : c\. — p\ must the price differential pi There The "cost absorption". is must therefore be lower than that recommended the Remark on competitors covering the fixed cost. In Linear symmetric demands, cost superiority of monopoly. - its efficient Suppose that the monopoly is component only access price pricing rule. regulated according to the : a with the access pricing rule po + aipi < a = pi- cx (49) p, : or p2 This of course lator If makes use (see l not pure price cap regulation (neither a and a x appendix Pi - Cqq _ (1 ~ Cqq - a '' J \ . On - Cigi _ 1 effort level (? -f qx q<i respectively ~ v_ ft we = Lqq H + notice two differences. First q2 and therefore Cqq than under optimal regulation. 25 i\\ is given by (49). Then — the other hand, under a price cap, effort production level + 7o the multiplier of the price cap constraint and (48) qx v) P\ Comparing with and 7) Po is practice) because the regu- are chosen approximately equal to q Po where v is of the cost information c\. the weights we obtain is =Pi-c + c. is is 1 — u will in general differ from optimal conditionally on the total evaluated at a (conditionally) higher Access Pricing 10 To assess current practice Co in Practice = CqQ + Fully distributed costs a) in allocating : = kQ C\ , C2 = and Ciqi cq2 The most popular method . of pricing access consists the fixed cost to the firm's products in a mechanistic way. For instance, a markup above marginal product's model, assume that in the light of the theoretical = Cqt —, Po may be uniform cost pi = (c x + Co) + — This accounting rule generates "excess" revenue ^4°. : a , = Cq 4. *ap. 4. + — = bm. ^ , that covers the fixed cost. It component interesting to note that this accounting rule satisfies the efficient is pricing rule, as Px This of course need not be the case - cx = a. Po where 5 is = Co(l + 6), pi - (co + ci)(l + S), There is demand and proper cost, The Oftel rule and B2 = long distance, access). The co(l + S), a. is (a entry considerations. : — The idea of the rule fully distributed costs. Let arbitrary and has no reason to reflect the The access rule designed Telecom's (BT) network can be sketched as follows Ci)li, = no point dwelling on the conceptual drawbacks of us just recall that the fixed cost allocation b) a chosen so as to ensure budget balance, yields Pi-cl > — of distributing the fixed For example, a markup proportional to marginal cost, namely cost. A) methods for alternative is Co)q 2 denote BT's AD "access deficit" : by Oftel Let B = for the access to British (po profits in its various is B\ co)qo, product cost of giving access is BT's = (p\ — lines (local, here equal to the fixed cost to set a usage-based price of access to Namely the margin above the — ko. local network. proportional to the access deficit per unit of BT's output in this activity and to the share of BT's variable profits provided by this activity. For the long distance a - c activity, = AD qx one thus has B B + B + B2 x x 26 : l')0) B Xote that 2 depends on the access price and that itself mination of access prices under fully distributed costs) (as is the case for the deter- formula (50) involves a fixed point. In practice (again as in the case of fully distributed costs), the access price outcome of a dynamic tatonnement, whose path ought If the budget is balanced (that is, if to be studied in AD = B +Bi+B 2 ), must be the more detail. the Oftel formula interestingly reduces to = pi-ci a which yields To there exactly the "opportunity cost" of is under budget balance the is is Oftel formula thus rule. usage -and not only cost- based, suppose that competition not only on the long distance domestic market (goods q x and q 2 ), but also on the international market (good q$ produced by competitors). its The in the activity. component pricing efficient see that the access pricing rule BT The BT and good q4 produced by Oftel rule then defines different access prices for competitors, a 2 on the domestic long distance market and a 4 on the international market despite identical access costs 15 With obvious . a2 ~ cO = notation, AD B TT HLoBi x aQ d a 4 ^TJ 1l ~ <k> = AD B3 93 Ef=0^i' and so a4 The markups The Oftel rule is - = Co -5-7— n 3 /q3 (51) • are thus proportional to BT's unit revenues on the competitive segments. therefore related to our access pricing formula, in that both are usage- revenue imposed by competitors on the network provider. The based and reflect the loss of difference between the Oftel rule and our formula (29) is that the unit revenues replace the superelasticities. 11 Conclusion In a first best world, access pricing to a network (like any other pricing) should be marginal cost pricing. In practice the appropriate rule departs from the first best and depends on the constraints and on the available instruments. First, the provision of the nondistortive ;;; lump sum taxes. For the sake of the argument. network imposes fixed costs which cannot be financed by Competitors should then contribute to the fixed cost of In practice, a domestic therefore network costs may call uses two domestic local loops while an Further, the access price peak load structure can only be coarse, and differ for the two types of calls if their timing stuctures differ within pricing international call uses only one. bands. 27 the network. This contribution, which takes the form of an access price above marginal cost, reduces either the financial burden of the regulator (in the case of transfers) or the price distortions associated with the firm's budget constraint. Second, asymmetric information of the regulator about the firm's costs may introduce other standard distortions, because informational rents must be taken into account defining proper marginal costs. may be Further distortions desirable if when network and competitive segment subcosts cannot be disentangled. When taxation of the competitive goods is feasible, access can be priced at marginal cost while taxes contribute to covering the fixed cost. [Note nevertheless that the marginal cost referred to firm, and is is the marginal cost which arises from the optimal regulation of the in general different from the first best marginal cost because effort This disconnection of consumer and producer prices very handy is competitive segment can bypass the network. By and taxation mandates reduces the number price regulators' if consumers on the contrast, a separation of regulatory of instruments, must then arbitrate very imperfectly between the goals of limiting and participating pricing rules to the coverage of fixed costs. become more and more complex to use access prices, rather lower]. is It is clear and the access inefficient bypass more generally that access as well as inefficient as the regulator tries than complementary instruments, in order to meet various market structure goals such as inducing proper entry or reducing monopoly power in the competitive segment or to achieve distributive objectives. Last, we have shown that our analysis can shed substantial fight on the effectiveness and of existing policies Looking forward, our analysis can be pursued in several directions. With a proper component pricing of policy proposals such as the efficient rule such as the Oftel rule for British Telecom. regulatory model of predatory behavior, one should be able to investigate the conjecture that access pricing products. The is a intuition more is efficient tool to the following : A prey than the pricing of the competitive low price for the competitive product signals a high efficiency. This information can be used by the regulator to extract the monopoly's rent in the future. On the other hand, a high access price seems immune to ratcheting. Next, one could substitute to the optimal regulation various types of current regulations (price cap, rate of return) and analyze how the deviations from optimal regulation impact on the access pricing formulas (see for example the Remark could even depart from the regulatory context and see in Section 9). how competition One policy alone would deal with access pricing. Excessive access prices could be considered as a predatory tool to eliminate competition as has been claimed in the Our framework AT&T also enables us to discuss a current case before divestiture. debate in Europe. Telecommuni- cation companies are required to offer universal service of local telephone (here good zero) 23 at a low price. Their budget constraint would normally force them to increase the price of their other services as well as the price of access to the network. A concurrent low-access- Community might demand marginal charge constraint (for example the European cost pricing in order to increase competition) then forces the network firm to further increase the price of its other services and and to lose its competitive edge. We are not then very far from the solution adopted by the U.S. courts, which prohibits the network from offering competitive services. There, universal service charges to the network instead of the profits made by is financed by appropriate access the network firm on its additional services as in Europe. We have not discussed nonlinear access pricing rules, for long as the fixed part of the tariff plays the role of a money needed to raise the to pay lump sum tax, it is for the fixed cost of the network. take in account the fact that beyond rate (see for example two-part some example Laffont-Tirole (1993) level the fixed ch. 2). tariffs. As an excellent tool However, one must charge lowers the connection Similarly we have not discussed the important practical issue of peak load access pricing. Finally, let us stress that, to a large extent, we have left aside various dynamic issues : investments (in particular unobservable specific ones) for network development (see however Remark 4 in Section 3), consequences of regulators' limited well as regulation of entry. Inefficient commitment power, as developement of the network, more costly rents of asymmetric information, and inadequate entry behavior are then to be expected. These questions can be addressed within our general regulatory framework (see ch. in Laffont-Tirole (1993) for the basic principles). 29 1, 9, and 13 APPENDIX 1 : Optimal regulatory rule (Section Let n(f3) be the co-state variable associated with U. m Using the transversality condition The = fi(0) = 3) From the Pontryagin principle a/09). we have differentiation of the Hamiltonian H= {S(q (po(m + V( qi ( Pl (l3),p 2 (/3),qM{3),p 2 (j3))) +x[po(/3)q (p (P)) +Pi(0)<h(pi(fl),M0)) +P2(P)<h(Pl{0),Pi(0))] -(1 + A)[tf(e (/?) +C l + e x (/?)) +C (f3,e (/3),q (p (f3)) ({3,eM,q1 (p-MP2(0))) + qx{pi{P)MP)) + ft(M0),ft(0))) + cq2(pM,P2(P))] - W(fi)}flfi) - /,(/3)^( e o(/3)+e (^)){^(/3,Co(^ eo (/3),?o(po(/3))+?i(Pi^),p 2 (/?))+?2 (Pi(/?),P2(/?))), 1 qo(po(0)) + <h(pi(0),P*(0)) + <h(Pi(P),P2(0))) +|j(/3, Ci{0, e l (/3), qM/3),p2 (0)), qi(P M,P2(P))) } with respect to p and , px , p 2 gives (24), (25) and (26) and with respect to (31). 30 eo, t\ gives (30) APPENDIX Part (i) results dC k d(3 = is —* k for same the from = Sub-cost observability : To prove that sub-cost observation (37). is useless 1,2 note that, at the optimal regulatory policy, That for all k. is, when d 2 Ek / (—dEk /8Ck , Q(P),qi(0)} under sub-cost unobservability, where C{/3) This incentive scheme announces type is incentive compatible. To show = CQ (0) + this, it suffices to — Ci(0). show that if choose the sub-cost allocation {Co(/3),Ci(/?)} corresponding will it ) given the firm's scheme under sub-cost observability, {t(0),Co (0) ,Ci(/3),Q({3),qi(fi)} consider the associated incentive scheme {t(0),C(0) to its 2 announcement, even though the regulator cannot control sub-costs. Note that from (37), there exists a such that ^(P,C k (f3),Q k ((3)) =a k = Q,l and CQ 0) + C 0) = C0), l where Q =Q and Qi = qx . But the condition d 2 Ek /dCk 80 = implies that ^0,C k 0),Q k (J3)) = a for all Jfe, as well. From our assumption, {C when it announces 3. We (/3) + C*i(/3) = C({3)} is the unique cost minimizer for type can thus implement the same allocation when sub-costs are unobservable as when they are. 31 APPENDIX 3 : Competitor market power Because of the dichotomy assumption, the optimization with respect to the access price can be carried out under complete information ma* {S(q PO, (p )) + : V(?i£pi, p2 ), 72(^1,^2)) -po?o(po) -PiViiPiito) ~ + c)q2 (pi,p2 )} (a (A3.1) s.t. poqo(po) + piqi(pi,P2) + aqiipuPi) > -ci?i(pi,p 2 ) + P2j-^(pi,P2) - ftOi.Pa) From + qi{pi,P2) + ?2(pi,P2)) co(go(po) ^(eo + (A3.2) ei) + c)^-(Pi,P2) = (a (A3.3) 0. (A3. 3) a +c= p2 gafoi.P?) = p 2 (l + g—— ! l/T} 2 (pi,p 2 ))- 222.f stlPi.Pz) Substituting into (A3.1), (A3. 2) max \S(q P0.P1.P2 (po)) we have + V(qi(p lj p 2 ),q2(pi,P2))-Poqo(Po)-Pi<li(PliP2)-P2<h(Pl,P2)+ L ~ 7~ *' x ^iPllPij s.t. Po<7o(Po) +Pl9l(Pl,P2)+P2?2(Pl P2) J -ci?i(pi,P2) - c?2 (pi,P2) - > 2 1 + A(/3) J v , 7 Let + ?l(Pl,P2) Co(?o(Po) + V>( e o + ci). 72lPl ? P2J denote the multipher of the contraint. We substitute A to X(f3) below for convenience. Maximizing with respect dV to (pi,P2) dqi dqx dpi +(1 + A) dV we obtain dcfr __ dq2 dpi d?i : d<h ^92 _ dp x dpx , 3pi 32 +?2(Pl,P2)) <9?2 /dgi 9pi Vdpi , ^<72 dp! d<?i dg 2 dp\ dp x 1 dK dp 2 dqi +(1 dV dq^_ dq 2 77 2 dq 2 dqi dq2 dp 2 2 dp 2 2 dp 2 dq + A) dp 2 5p 2 dpi dq c— dp dqi \dp2 dp 2 J 2 'Ci-z op 2 dgl dpi V dp2 2 \ \ 72 d *9<72 dpi dpi Pi dg2 dqi - p2 — Co - - Cq -?i C! + (Piq-i dpi V + 1 c A -72 dp 2 dp 2 d (p 2 q2 Op 2 \ 772 + p— , -i - Pi Vi ~ Co — — co Ci dpi dpi 71 1 c + + A 1 P2 + A dqi 72 The second term (dqi dq 2 dq2 dq 2 Kdpi'dpi dpi dpi p 2 ( dqi dqi dqi dqi dpi' dpi dpi'dpi 2 El 11 §3x. _ i3X + Piqi + P272 + A. El 72 ^2L tji dpi\ . 772 dq2 d ( dpidpi d dqi ^22. ' dpi dpi 1 ^ dqi dpidp 2 ( dqi 1 P^A '1} 33 ^ + + 11V2 7i272i R7V1 1 12m <fr dpim \Tji/ 7i272i ~ dqi ] d + Pi7i2ai7 (*)] 7i72 1 dpidpiKrjiJl Pi72a|-(^)+P272i4(^ - d f \rji dpidpi\t}iJ 7i72 1 \ d fpiqi dqi dpi dpi . dpi 'dpi r] dpi of the right-rand side equals 11I P: d (p 2 q2 dqx_ dcfr £i 772 ' dqi qi dpiT) 2 . Pi - Co -c A x Pi Pi - Co - A + 1 A ^4(i)+P272l4(^) 7l72 7l A c P2 here + 1 fti 1 1 1 72 72 1 mi - 712 721 ^7iafe(^)+P27i2^(^) 7i72 - 7i272i R k = Pkqk- Since a a = = p2 — c — Ei we have , HP) Coo r 72 1 + A(/?) 1 :P2—" X(fi) 72 1 + P 2 7i^(^)+Pi7i2^fe) A(/3) 7i72 with 72 = 7i72 -712721 72' 27i72 + 72i72 - 72i7i2 - 7i 34 - Vi 712721 APPENDIX The dichotomy 4 Restricted regulation : implies that price formulae can be obtained as in the full information case. Let W = S(q ) + V(qi, q2 ) -p - Piqi - p 2 ?2 q denote the consumer net surplus, and MP = dqir qi + p- \p x - co - dq 2 + p- [p2 . , c, ] dpi co - c] dpi denote the monopohst's marginal profit and let P denote its profit. The planner solves W max {po.Pi.w} subject to P-V= (BB) p q +p x + qx (p 2 - MP = Let (1 + hand side of to pi is A) and v denote the shadow {BB) c x qx - xj) = and 0. prices of the maximized with respect is - CoQ - c)q 2 to two constraints. Because the p x the , left- first-order condition with respect : dW dMP \ dpi f V dpi J Because the net consumer surplus decreases with sumption on profit, we obtain 1/ < 0. The prices, and from our concavity derivative of the Lagrangian with respect to as- p2 yields dW — + - dp 2 The ing. first-two terms are the The , (1 -dP dMP + A)— + !/-— = standard terms obtained in the absence of monopoly pric- correction in the access pricing formula (a d(M P)/dpi, which is equation shows that lowered. But pi is n 0. dp2 dpi = p2 — c) depends on the sign of positive under generalized strategic complementarity. for a given p\ the price p 2 and therefore the access price, should be also increased , and the result ambiguous. However, functions and constant returns to scale both effects cancel, giving the formula as This last in the case of full regulation. 35 for linear same demand access pricing APPENDIX = Let 6" be defined by 6' V(p u p2 +b- - a) 5 V( Pl ,p 2 ). Hence 38' 0^ = 91- ft(«) 89' = ft + (p 2 dp 2 - ft(a) 80' Let AR = (/?! - - ci Co)(qi(a) - qi )) Maximizing (45) with respect to Pi,p2 ,a GOT \ qi (a) (1 + A) [fo "Aft + ( 1 + -(i-c(n) + - Cl - A) [( - yields -(i-G(e-)) \q2 + (1 {pi + +g(6')[q2 <?(*) - Aft(a) - (1 + (p 2 - - Co)ft. c2 : dq -a- ca)^(«)]] 2 + (p 2 - d - Co)p + (p2 - Co - c2 Pl ~ Cl " + dq x co) ir )p]] = A)AiE ^df^ + (pi_Ci - A) - c 2 )ft(a) co)2jL(a) +^*)(ft-?i(a))(l G(P) \q 2 (a) + (l+\) - a (A5.1) {P2 ~ a ~ C2) df {a) }} 2 +(P2 - C2 " co) a|]] -q (a)](l+\)AR=0 A) (pi (.45.2) 2 - Ci -^ (a)+(P2_a " C2) dpi +<7(0*)ft(a)(l+A)Ai2 dp l: (a) 2 =O (-45.3) Hence G(O') Aft(a) + (1 + A)( Pl - dqi Cl - co)^-(a) dpi -(A <?2 + (a) (l + A)(p 1 dft, -,\d<l2/dpi(a) -c - Co )^(a)) 1 / <9p 2 •(l-G(n) Aft +g[9')(l + (1+A) + X)AR q\ (pi - Ci - Co)5pi -qi(a) -92(a) 36 ]] dq2 /dp 2 (a). h [p 2 - Co dq2 /dpi(a) dq 2 /dp 2 {a). - c2 ) — dpi! = 0, and [1 - G{6-) A<7 2 + (1 + A) (Pi ~ Cl - Co) -5^?l + (1 + A)Afl AR, we = (a) fe- 32 (i-g)S 3p2 - C2 - . x 5 <72 Co) dp 2 O the case will consider -^ iw) +(i - G) G (^ (a) A) (p 2 dp 2 +<K0")?2(l Given the ambiguity of the sign of /_ , 1- AR = 0. - - ^- G )Sf (Pi a -<?)§* (P2-Co-C 2 ci Co) ) -Gx{ql (a)-q2 (a)^)-X(l-G)q1 -(l-G)Xq2 or (1+A)A Pi — Ci — CQ p2 - Co - <*1 a2 c2 9« det A= G(l dP7 - G)^-.A(a) + (1 £21 3ps G) 2 A («) with dq 2 dq2 dqi dq x dq2 dqx dq2 dpi dp 2 dpi dp 2 dpi op 2 dp 2 dpi 3<?i A~ = l (1-G)2a 9pj _(1_ (7)^22. 1 det A , > a-ofe ^M + 3ps (1 _ (71^2L ap,' This yields X[G(q1 (a)Pi - ct - c q2 (a)^)+(l-G)q ](l-G)^ + (l-Gy\ q /2 a 1 „ = (1+A) (l-G)2 A + G(l-G)A(a)(§£/i£(a) \[<?( 9l (a) P2 - fc(<0|i£|) + (1 ~ G)ft](l - G)fg- - A(l - G)q2 GAla) dp; — Ct — Cq = (1 + A) (l-G)*A + G(l-G)A(a)(f*/fa(a) 37 :i-g)£ 9pi ^_ Gm^(l + *ll$)+(l-G){V2 + Pi-cx-q, = L Pi -c -co P2 1 V2 + ^ GA(a)^J {^V a 2 + A- + + (l-G)A rii2) ^+i8)+( 1 A g 1 + Afi -^(*+'*>j_ with A= A(a) These formulas reduce From = 7/17/2 - 7/i(a)7/ 2 (a) 7/127/21 - 7/ 12 (a)77 2 i(a). to the classical superelasticity formulas when G = 0. (A5.3) cp-a _ P2 A l + Pi-ct - 1 A7/ 2 (a) l + cp 772(a) pi Al7/ 2 (a) 7/ 33 u (a) Rxja) _ T] 1 7/ 2 (a) (a) _ff 2 /' 7/" 2 A , "1 + A^2 G A(a)^^ + (l-G)A A" _1 p2 - c 2 - Cp p2 APPENDIX The 6 program regulator's then is -H x - {3 e x (/?)) + c)q2 Pricing rule : H (0 - eM) + c) - pMqi(pM,pM - -po(/%o(po(/3)) -( Pl (/3) Component Implications of the Efficient : t (pM, Pl (0) - Hx (/3 - e x {P)) + c) + U({3)]dF((3) s.t. f/(/3) = -V'(e (/5) W) > + e 1 (^)), o, and MJ)qo(M0))+Pi(P)<iM0),Pi(0) - Hi(0 - d(0)) + + - Pl (/3) H x {fi [ -i/ - (/3 e - ei(/3))j? 2 (pi(/?),Pi(^) (/3))(<7o(po(/?)) - - ei (/3)) ?1 ( Pl (/3), Pl (/3) Proceeding as in Section 5 0'(e o -r e x ) ty (e + ei ) = = ff£( 9o - H x {f$ etG9)) c) + c) + c)) eM) + c) - tf(0) - 0( eo co + ?i + ft) - f$~\0)f0)d0 ~ + Sjwmw — w - ci)p— + (Pa A(g) l ( ^ + . + A(/3) 39 + "(go r , ei ) + c - coJo— -(ft+ft) / + A(^)\^22i + A(/3) l ^ ^ lHI = Pi-Cog-Ci„ = Pi - + (/?) + we obtain fl,fli +(Pi - ei(/?)) + ?i(Pi(/3),Pi(/?) - #i(/3 - +?2(pi(/3),Pi(/3)-^i(/3-e 1 (/3)) -fft(/3 - i5Ti(/J c) 1 17"' ^zi + d) A(fl = ft e^/?)) = 0. with *0J (-£-). p2 \qi +92/ —7— •72 721 ><7l+92 M?12- Hence 00 ,_M0)_PL = c 1 with 772 + A(/3) 77" = f^' 40 * A(j9) 1 + P2 A(0)itf MIT LIBRARIES 3 9080 00922 4913 APPENDIX 7 : The The Lagrangian of the H ){qo(po) poqo(po)- (P-e monopoly's optimization problem + q {puPi- H - -0(e o first-order conditions {/3-e l (l3)) + c) l l +( Pl -#!(/?- eOJfafa, Pl The Component Pricing Efficient H {$ - e x x {ff)) po with respect to pQ and p x are ,fdqi , If a ~ qQ and a! ~ qx + =a 90 1- q2 dq2 dq x v — Co Po Pi is \ - Co - l — i/ »7o Ci Pl r/j dq2 , po where + q2 (p u p 1 - H (P-e - o-iPi). apo l-v p 2 ^2 given by (47). 41 Pl under price caps is + c) + q2 (p1 ,p1 - + ei) + v(p - a (po-coH rule 1 H x tf - 1 (/3)) e x (0)) + c)) + c)) REFERENCES BAUMOL, W.J. (1983), "Some Subtle Pricing in Railroad Regulation", In- ternational Journal of Transport Economics, 10, 341-355. BAUMOL, phone W.J. (1993), "Deregulation and Residual Regulation of Local Tele- Service', AEI Studies in Telecommunications Deregulation, Ameri- can Entreprise Institute CAVE, M. (1992), kets in the Public Policy Research, for "The Growth of Competition ENSTT UK", Colloque - in New York University. Telecommunications Mar- Le Management des Entreprises de Reseau. LAFFONT, and J.J. J. TIROLE (1990a), "The Regulation of Multiproduct Firms", Journal of Public Economics, 43, 1-66. LAFFONT, and J.J. J. TIROLE American Economic Review, LAFFONT, J.J. and and Procurement, NOLL, R.G. (1990b), "Bypass and Creamskimming" 80, 1042-1061. TIROLE (1993), A MIT Press. J. Theory of Incentives in Regulation (1989), "Telecommunications Regulation in the 1990s" in Directions in Telecommunications Regulation Policy, Vol. ed. Duke VICKERS, kets", University Press, J. (1989), Newberg, Durham and London. "Competition and Regulation in Vertically Related Mar- mimeo. WILLIG, R.D. bing 1, P. New (1979), "The Theory of Network Access Pricing", (ed.), Issues in Public Utility Regulation, Public Utilities Papers. ii906 6: 42 in H.M. Tre- Michigan State University Date Due n t I 6 W* 1 I Lib-26-67