Does Practice Follow Principle? Applying Real Options Principles To Proxy Costs in US Telecommunications

Transcription

1 Does Practce Follow Prncle? Alyng Real Otons Prncles To Proxy osts n U Telecommuncatons Mark A. Jamson Drector of Telecommuncatons tudes Publc Utlty Research enter And Assocate Drector for Economc and Busness tudes enter for nternatonal Busness Educaton and Research Warrngton ollege of Busness Admn. Unversty of Florda For the Proceedngs of the Worksho The New nvestment Theory of Real Otons and ts mlcatons for the ost Models n Telecommuncatons October haro Audtorum olumba Unversty would lke to thank Davd angton Tracy Lews James Alleman Wllam harkey teve lutsky and Wllam Baumol for ther helful comments and suggestons. Any errors are my own resonsblty. 1

2 ecton. ntroducton n ths aer analyze whether current U ractces n estmatng ncremental costs for telecommuncatons comanes rovde effcent nvestment ncentves to regulated ncumbent local exchange comanes (LEs). The mlementaton of the Telecommuncatons Act of 1996 (Act) has rased ths ssue. 1 n mlementng the Act the Federal ommuncatons ommsson (F) and many state Publc Utlty ommssons (PUs) are usng ncremental cost studes wth some mark-u for shared costs for establshng rces for nterconnecton recrocal comensaton unbundled network elements (UNEs) 2 and unversal servce. 3 Regulators use these ncremental cost estmates whch come from roxy cost models 4 to set celngs on these rces. Ths s a sgnfcant change n regulatory ractces. Before the Act regulators used ncremental cost studes to set rce floors for comettve or otentally comettve servces. Ths change n the use of ncremental cost studes has affected stakeholder nterests n the studes. Prevously LEs had an ncentve to kee the cost estmates low n order to obtan greater flexblty to lower rces n comettve markets. Now these comanes' nterests are to kee the estmates hgh to obtan greater flexblty to rotect revenue streams and affect comettors' costs. Ths does not mean that LEs would always choose to rce at the celng f the celng were hgh. LEs may rce below the celng to dscourage the develoment of comettve networks or to smly meet cometton for UNEs. Even f the rce celng were hgher than the rce the LE actually charged the hgher celng stll has value to the LE because t gves the LE the oton to rase rces should the market stuaton change. Also when ncremental cost studes formed the bass for rce floors some comettors to LEs wanted hgh cost estmates to secure rce umbrellas n comettve markets. Now comettors want to kee the estmates low to decrease the ayments that they have to make to ther comettors the LEs. Ths new stakeholder dynamc has romted new nvestgatons nto the arorateness of how regulators erform and aly ncremental cost studes. The current debate focuses on 1 Telecommuncatons Act of 1996 P.L. No tat. 56 (1996). 2 nterconnecton recrocal comensaton and UNEs all nvolve a comettor connectng to the LE's network. n ths aer address UNE rcng. Models for recrocal comensaton are more comlex than the model use n ths aer because both LEs and new entrants ay and receve recrocal comensaton. 3 By "rce" for unversal servce mean the comensaton that the regulator allows the servce rovder to receve n exchange for chargng subcomettve rces. ecton exlans ths n more detal. 4 A roxy cost model s a cost model that comutes cost for a non-exstent reresentatve comany rather than for a secfc comany whch used to be the ractce. 2

3 Total Element Long Run ncremental ost (TELR) studes because the F and many PUs are adotng them (Jamson 1998b; alnger 1998). One element of ths debate has concerned the alcablty of Real Otons nvestment analyss. 5 Ths debate has rased concerns that TELR-based rces do not nduce effcent nvestments on the art of LEs because TELR studes do not reflect economc derecaton of LE nvestments uncertanty n demand and nter-temoral oortunty costs. n ths aer analyze these ssues by consderng a model n whch the regulator affects LE nvestment decsons and market outcomes through rce controls. By comarng ths model to TELR ractce show that rces based on exstng TELR models could dscourage effcent LE nvestment by understatng ncremental costs. These models understate LEs' ncremental costs by narorately udatng nut rces wthout adjustng derecaton and by overstatng exected demand. also show that the neffcency s not as great as some clam. For examle Hausman (1996) states that TELR models omt derecaton demand uncertanty and Real Otons. show that the frst two clams are ncorrect and that the effect of omttng Real Otons s ambguous. have organzed the remander of the aer as follows. ecton descrbes the economc model. By "model" mean the stylzed assumtons make about how regulators regulate LEs how LEs make nvestment decsons and how LE comettors buy UNEs. Ths s a non-techncal exlanaton that non-economsts should fnd readable. The Aendx contans the techncal exlanaton. ecton examnes how varous roertes of TELR models affect the effcency of LEs' nvestment decsons. ecton V s the concluson. ecton. The Model Ths secton descrbes the model and ales t to dentfy effcent rces. A. Descrton of the Model My model n ths aer consders an LE that roduces multle roducts and that makes nvestment decsons subject to regulatory rce controls and uncertan demand. 6 assume 5 For examles of ths debate see Hausman (1996) and Hubbard and Lehr (1996). The ssues address n ths aer are based uon the ssues rased n these documents. 6 n ths aer use the term roduct to nfer a roduct beng sold n a artcular market. o for examle a UNE n one market would be consdered a searate roduct from the same UNE sold n another 3

4 that the LE enjoys economes of jont roducton from roducng multle roducts but make no assumtons about whether the LE s a natural monooly;.e. do not assume that a sngle LE could serve the entre market demand more effcently than two or more other frms (Baumol 1977; Jamson 1998a) although do not reclude t. The LE roduces both retal roducts and UNEs. ome of the LE's retal roducts are subject to unversal servce oblgatons (UOs). defne a roduct subject to UOs as a roduct where the regulator reures the LE to charge a rce that s lower than what would be found n a erfectly comettve market. n other words the LE cannot charge the regulator's mandated rce wthout an external source of fundng or wthout chargng suercomettve rces for other roducts. f demands are ndeendent then by defnton the er unt fundng needed to allow the LE to charge the mandated rce and earn ts cost of catal s the dfference between the comettve market rce and the mandated rce. assume that the LE s rces and nvestment affect demand that demand s stochastc and that buyers buy less at hgher rces than they do at lower rces all other thngs beng eual. LE nvestment affects demand because hgher levels of nvestment mrove the ualty of the UNEs and retal roducts makng customers of ether tye of roduct more wllng to buy from the LE. further assume that uantty roduced and LE nvestment determne the LE's roducton costs. 7 ncreases n uanttes roduced ncrease costs. nvestment s also a cost but one that lowers other roducton costs. n effect ths assumton means that the LE can trade sunk nvestments for varable costs and vce versa although the trade may not be one-for-one. The LE's decson to roduce or not roduce a roduct even f the regulator makes the choce for the LE changes the LE's costs and revenues. call the change n costs the ncremental cost of the roduct. The ncremental cost covers the LE's entre roducton of ths roduct. For examle f the LE roduces ten thousand unbundled loos n a market and ths causes the LE's total cost for all roducts to ncrease from $900 mllon er month to $900.2 mllon er month then the ncremental cost of rovdng unbundled loos n ths market s $ er month. Lkewse call the LE's change n revenues from rovdng a roduct the ncremental revenues for the roduct. As n the case of ncremental cost t s the change n the market. By defnng roducts ths way segregate roducts along the dmensons of techncal characterstcs geograhc market customer tye and dstrbuton channel. 7 For smlcty assume that nvestment both mroves ualty and decreases oeratng costs. An examle of such an nvestment mght be the urchase of a dgtal swtch whch offers a hgher ualty sgnal than older technologes and also has lower rces for sare arts. Not all nvestments do both. 4

5 comany's total revenue that s mortant. o the ncremental revenue from the roduct s the change that t causes n the LE's total revenues. To smlfy dscusson generally assume that addng or drong a roduct does not affect demand for other LE roducts. Ths assumton allows me to refer to rces rather than ncremental revenues. dro ths assumton when consder Real Otons. My model has the regulator the LE and the customers make decsons n seuence. The regulator makes the frst decson. he selects the rces for UNEs and UOs usng TELR models. Because the regulator uses these models rces are ted to nether the LE's actual economc costs nor ts earnngs as n rate of return regulaton. assume that roerly aled t s techncally feasble for the TELR model to relably estmate the LE's ncremental cost. 8 Also the regulator enforces her other rce reurements by settng mum rces and not by rate of return regulaton. The LE makes the second decson. t chooses whether and how much to nvest and executes ts nvestment choce. The LE knows the regulators' rce celngs but does not know uanttes that UNE and UO customers wll buy. assume that the LE and the regulator know the mnmum and mum amounts that these buyers could urchase and the robabltes of them buyng any artcular amount between the mnmum and mum. The regulator reures the LE to suly all that customers want to buy from the LE. further assume that the LE wants to mze exected rofts and s rsk neutral. ustomers make the last decsons. They make ther buyng decsons based uon the rces the LE charges the nvestment-nduced ualty of the LE's roducts and other factors. These other factors were unknown at the tme that the LE made ts nvestment decson. Ths gves demand ts stochastc roertes. Once these buyers make ther urchasng decsons the LE ncurs the remander of ts costs for rovdng the roducts and receves ts revenues. To smlfy the analyss assume that the demand and suly effects of nvestment are such that the LE sells no roducts and ncurs no ncremental costs f t chooses to make no nvestment. 8 Wthout ths assumton the effects of mscalculatng ncremental cost would ambguous. Regulated rces cause over nvestment or under nvestment n secfc regons. Mscalculatng ncremental cost rovdes ncentves for neffcency f and only f the mscalculaton romts the regulator to move regulated rces from one regon to another. 5

6 Ths model s a reasonable aroxmaton of what s haenng n telecommuncatons n the U.. Regulators are settng UNE and UO rces usng TELR models generally wth contrbutons to shared costs. Even though these rces aly to LEs' exstng networks and not just new nvestments the rces sgnal the LE how much roft t can exect from new nvestments that may be used for UNEs and UO roducts. Also nvestment can ncrease ualty and lower other roducton costs. Furthermore LEs do not know demand wth certanty when they nvest. LEs nvest to relace facltes servng current demand and to serve rojected new demand. n the case of exstng demand LEs can have a reasonable amount of certanty that demand n the near future wll be smlar to what they are exerencng today. However t s always ossble that demand wll decrease f comettors lace facltes that eventually serve exstng demand or f customers move to other locatons. 9 Projected new demand s also uncertan. Poulaton shfts housng develoments that do not lve u to forecasts and slowdowns n economc growth can all result n realzed demand beng less than forecasted demand. Also new demand may not last for the entre average economc lfe 10 of the facltes for the same reasons that current demand may declne. B. Alcaton of the Model Assume that the regulator wants the LE to choose an nvestment level that mzes the total beneft that LE nvestment can brng to the economy. The benefts of are the decrease n LE unt roducton costs and the ncrease n the value of LE roducts brought about by the ncrease n LE roduct ualty. The cost of s smly the cost of the nvestment. f the regulator had comlete and erfect nformaton and f the regulator had comlete control over the LE and customers the regulator would choose and the otmal uantty by euatng margnal benefts wth margnal costs and reurng the LE to make the nvestment and customers to buy. But n the real world the regulator does not have comlete and erfect nformaton and does not have comlete control over the other layers. 9 Telecommuncatons lant has only lmted otental for servng demand n more than one geograhc locaton. The most fungble eument s crcut eument. Techncans can remove ths eument from one locaton and lace t n any other locaton that uses the same technology. ome swtchng eument can also be moved to another locaton that uses the same technology. Feeder cable can be used to serve any demand that occurs n the feeder lannng area. As a result f demand decreases along art of the feeder cable's route the dled orton of the feeder cable becomes avalable to serve demand n another art of the route. However feeder cable n one route cannot be moved to serve demand n another feeder route. Dstrbuton cable has lmted fungblty. 10 Regulatory accountng ales one derecaton lfe to all telecommuncatons of a artcular tye. Even f ths derecaton lfe s correct on average t may be too long for some locatons and too short for others. 6

7 nstead the regulator must use ncentve mechansms to nduce the LE to choose and customers to buy or at least to nduce them to choose amounts close to and. 11 To determne whch rce celngs wll nduce the most effcent nvestment the regulator uses backwards nducton; n other words she starts at the end of the seuence of decsons that wll take lace n resonse to her rce controls and works backwards through the decsons to determne whch rce control wll gve her the most desrable outcome. he begns by consderng the LE's customers stuatons. These customers wll choose uanttes of LE roducts based uon the rces the LE charges and. However because there s a stochastc element to demand the regulator does not know wth certanty how much customers wll urchase at artcular rce and nvestment levels. nstead lke the LE she knows how rces and nvestment affect demand and has exectatons about the stochastc effects. he bases her exectaton on her knowledge of the range of ossble uanttes demanded and the lkelhood of each otental level of demand. Havng formed her exectatons about how customers wll resond to the LE s rces and nvestments and knowng that the LE shares these exectatons 12 the regulator consders how the LE wll resond to her rce controls and ther shared exectatons about customer demand. From the LE s ersectve once the regulator establshes her rce celngs whch call M where M reresents all of the LE s roducts the LE chooses ts nvestment level to mze ts rofts. call the rces that relate to UNEs and UOs where reresents the UNEs and UO roducts. The LE s actual rofts wll be the dfference between ts realzed revenues and the sum of ts realzed roducton costs and nvestment. However at the tme the LE makes ts nvestment decson t does not know what ts actual rofts wll be because t does not know how much customers wll buy. o the LE bases ts nvestment decson on ts exected rofts whch are eual to the rces the LE wll charge tmes exected demand mnus exected roducton costs and mnus nvestment. Next the regulator attemts to choose the rce that nduces and. Unfortunately there s no rce that does ths because the regulator has only one tool and s tryng to determne 11 There may be constrants that kee the regulator from nducng LE to choose. For examle the LE may osses rvate nformaton about how much lowers oeratng costs. Also the Act reures regulators to base rces on cost. n certan stuatons such as when cost-based rces rovde LE customers wth surlus at the margn ths restrcton may cause the LE to under nvest. 12 n ractce the regulator may have less nformaton about customers than does the LE. n such a case the regulator wll have to allow the LE to earn extra rofts n order to nduce an effcent nvestment decsons. But even wth ths the LE wll under nvest. 7

8 two outcomes. The tool s the rce er unt sold. 13 The two outcomes are the nvestment made and uantty sold. f the regulator had a two-dmensonal rce -- one dmenson that reflected uantty sold and another that reflected ualty -- then she mght be able to acheve her effcent outcome. do not model ths ossblty because have found no regulators dong ths n ractce. (Jamson 1998b) What the regulator does nstead s choose a rce that mzes total socal surlus subject to the LE's and customers' decson-makng rocesses. Total socal surlus s the dfference between the value customers lace on UNEs and UO roducts and the cost of rovdng them. The amounts they choose are and. As the next subsecton exlans n more detal the regulator chooses a rce that covers the LE's exected ncremental cost. The effects that rce changes have on nvestment determne whether the rce s above or below the LE's margnal cost. f rce ncreases ncrease nvestment then the otmal rce s above margnal cost. The reverse s true f the rce ncreases decrease nvestment. The Aendx descrbes ths n more detal.. ost-based Prces Now consder how the regulator s effcent rce control comares wth the LE s costs. Recall that the LE chooses by euatng the nvestment s margnal effect on exected revenues wth ts margnal effect on exected cost. The margnal effect on revenues s smly tmes the exected margnal effect on demand. The margnal effect on cost s smly the margnal effect on exected roducton costs (ncludng effects on demand) lus the margnal nvestment. Ths means that must reflect the exected margnal cost. Because the LE cannot charge drectly for ualty the LE chooses nvestment levels that cause margnal costs to be ether above or below. The drecton and magntude of the devaton from margnal cost deends on how nvestments affect demand and costs. The nteractons among demand costs and nvestment are comlex but n general large nvestment-nduced changes n demand cause the LE to kee margnal costs below esecally f nvestment causes the demand curve to become steeer. 14 Both of these condtons mght occur f ualty s mortant to customers 13 Accordng to my nternatonal survey results (Jamson 1998b) rces for nterconnecton and network elements are generally lnear but not always. assume lnear rces for smlcty. 14 The Aendx rovdes suffcent condtons for when the LE would choose to kee margnal cost above rce and for when the LE would choose to kee margnal cost below rce. 8

9 esecally at hgher rces. call ths necessary relatonsh between rces and margnal costs the otmzaton constrant. The otmzaton constrant nduces the LE to make an effcent nvestment only f the LE s wllng to nvest. The regulator ensures that the LE s wllng to nvest by alyng the same rocess that she used to develo her otmzaton crtera but wth two slght dfferences. The frst dfference s that she must consder the total effect of the nvestment and not just the margnal effect. n other words she must consder the nvestment s total effect on the LE s exected revenues and the nvestment s total effect on the LE s costs to nduce the LE to nvest. The second dfference s that she does not need to concern herself wth euatng the effects. he only needs to ensure that the revenue effect s at least as great as the cost effect. As long as ths s true the LE s wllng to artcate n the regulator s mechansm. call ths the artcaton constrant. Wth a small amount of algebrac manulaton whch show n the Aendx the artcaton constrant can be exressed as beng greater than or eual to the exected ncremental cost of the LE choosng = over = 0 dvded by the exected demand. n other words the regulator s rce celng must be greater than or eual to the LE s exected TELR (whch may be dfferent the regulator s estmated TELR) dvded by exected demand. alnger (1998) ales a dynamc model to rovde a useful exlanaton of the LE s ncremental costs. He exlans that the ncremental cost s a current rce that makes the current value of the ncremental revenues just eual to the current value of the ncremental oeratng costs lus the ncremental nvestment gven that these values wll contnue to be eual n all future erods. n other words the otmzaton constrant and the artcaton constrant should be understood to mly resent values of rces to be charged exected uanttes to be sold and exected roducton costs over the lfe of the nvestment. 15 He also shows that the assumton that the exected lfe s known whch s the tycal assumton n TELR studes understates ncremental cost. He further shows that the ossblty that technology mrovements wll ncrease the caacty of assets decreases ncremental cost. 15 Ths does not mean that rces have to be constant. Rather t means that the regulator and the comany vew the regulator s rce control decson as establshng current rces and rces for each erod over the lfe of the asset allowng that the rces may change from erod to erod. 9

10 D. Real Otons Now consder the effect of Real Otons on the otmzaton constrant and the artcaton constrant. Real Oton theory states that f nvestng at the current tme creates or destroys future oortuntes the foregone or ganed value of these oortuntes should be consdered n estmatng the value of the nvestment decson. (Trgeorgs 1996) To examne ths effect label as ψ the roducts assocated wth the LE s alternatve nvestment whch the LE cannot make because t s nvestng for. ψ could be another set of roducts or smly an nter-temoral change n when an nvestment to rovde could be made. That s to say ψ could reresent the same roducts as but rovded on a dfferent tme schedule or a dfferent way. ψ mght also nclude retal roducts that the comettors sell n cometton wth the LE. Regardless of whether ψ reresents a temoral or nter-temoral dfference from or both t s ossble to re-exress the otmzaton constrant and the artcaton constrant to exlctly nclude Real Otons. Dong so changes the otmzaton constrant to the reurement that must eual the exected margnal cost lus the adjustment for the LE not beng able to charge drectly for ualty lus the exected margnal unt value of the foregone otons. The artcaton constrant becomes the reurement that be greater than or eual to the exected ncremental cost of the LE choosng = over = 0 lus the value of the otons that are foregone by ncreasng nvestment from 0 to dvded by the exected demand. n other words the regulator s rce celng must be greater than or eual to the LE s exected TELR lus the change n oton values dvded by exected demand. ncororatng Real Otons mroves LE ncentves to nvest effcently but the effect on the otmzaton and artcaton constrants s ambguous. Hausman (1996) argues that Real Otons values are ostve meanng that nvestments always foreclose otons (alnger 1998). f ths s the case TELR-based rces nduce LEs to nvest too lttle. Other wrters (Abel et. al 1996; Trgeorgs 1996) ont out that Real Oton values can be ostve or negatve. n the context of rcng UNEs Real Otons mght be negatve f the nvestment creates oortuntes. For examle the nvestment mght be necessary to mantan a market resence and avod costs of re-enterng a market. Because the drecton of the effect on s ambguous regulators would need to assess the effects of Real Otons on a case-by-case bass f they choose to consder Real Otons n regulatng UNE and UO rces. 10

11 E. Real Otons as EPR ncororatng Real Otons nto UNE and UO rcng s effectvely an alcaton of the effcent comonent rcng rule (EPR). The EPR whch s also called the Baumol-Wllg rule recommends that comettors ay LEs ther oortunty costs. n other words the rces an LE would charge to comettors would ensure that the LE would make the same amount of roft regardless of whether t succeed n the comettve orton of the market. The EPR formula for settng UNE rces (called wholesale rces n the formula) s (Baumol and dak 1994): Wholesale rce = Retal rce - [Retal Wholesale ] Or alternatvely Wholesale rce = Retal marku Wholesale (1) Where s the acronym for ncremental cost and Retal marku = Retal rce Retal. omarng Euaton 1 to the artcaton constrant wth Real Otons Wholesale rce reresents (settng asde UO roducts for the moment) Wholesale reresents the ncremental cost of and Retal marku reresents the Real Oton value of ψ. Because Real Otons s an alcaton of the EPR several conclusons from the EPR lterature are alcable to Real Otons; namely: The underlyng model assumes that comettors are frnge comettors that can offer only some subset of what the LE roduces. (Wllg 1979) f ths assumton does not hold then Retal marku s too large of a marku above Wholesale to romote effcency. (Jamson 1998c) Retal marku should contan no monooly rofts. (Tye 1994) The retal market should be homogeneous or the Retal marku should be adjusted for the dfferences n value to customers. (Wllg 1979; Armstrong and Doyle undated) The LE s less lkely to try to rotect markets from cometton and dscrmnate aganst comettors than wth a lower Wholesale rce. (Ordover ykes and Wllg 1985) 11

12 ecton. How TELR Prces Affect nvestment ncentves Ths secton descrbes how varous roertes of TELR studes affect LE nvestment ncentves. descrbe the relevant features of TELR studes frst. Then dscuss the concerns wth economc derecaton demand uncertanty and Real Otons. A. TELR tudes The basc TELR formula begns by estmatng the ncremental catal exense (APEX) that causes. The formula then multles an annual carryng charge by the ncremental APEX. Ths carryng charge conssts of the cost of catal nvestment related taxes and derecaton exense. The result s an annual carryng cost of the APEX. The formula then adds annual oeratng exenses to ths annual carryng cost and exresses the result on a relevant unt bass; for examle er unt er month. Tradtonally APEX has been the man drver n ncremental cost studes. Ths was because analysts assumed that APEX drove almost all carryng costs and exenses or at least that there was a strong ostve correlaton. rtcal assumtons for APEX calculatons have been: 1. Technology By necessty a TELR study assumes a artcular technology s used to rovde the network elements. Assumtons can vary but my understandng s that the F refers to assume that studes assume the most effcent or least-cost technology that s generally avalable. 2. Network archtecture -- TELR studes tend to assume a scorched node;.e. exstng central offces and cable routes are assumed to be fxed but technologes can change. 3. Utlzaton or Fll factor -- Utlzaton or fll factor refers to the ercent of the caacty of a faclty that the study assumes wll be used. For examle an assumton of 50% utlzaton of a 200-ar cable means that t s assumed that 100 ars are used. The F aears to favor average fll. However calculatng average fll s very dffcult because n ractce every node and lnk has a dfferent fll and the flls change on a regular bass. The last tme examned TELR models the Hatfeld TELR model nterreted average fll to mean that the scorched-node network would be otmzed wth resect to fll but wth the constrant that network 12

13 facltes must be urchased n standard szes. o for examle f a cable route needed 202 coer ars the model would use the next sze cable above 200 ar for estmatng TELR Derecaton -- The models use whatever derecaton rates the regulator sets. 5. ost of catal -- The models use whatever cost of catal the regulator sets. 6. Oeratng exenses -- Tradtonally ncremental cost studes estmated oeratng exenses by multlyng APEX by an exense/asset rato calculated from LEs' accountng records. beleve ths s the method used by the TELR models that the F s consderng. Recently some LEs have begun usng actvty based costng. B. omarson of Economc Prncles wth TELR Practce: Derecaton Hausman (1996) argues that TELR studes do not nclude derecaton or at least nclude too lttle derecaton. The frst argument s ncorrect. The revous subsecton descrbes how TELR studes ncororate derecaton on assets. Wth resect to the amount of derecaton Hausman (1996) s correct that TELR studes should nclude economc derecaton and the effects of decreases n nut rces. These effects are art of ncremental cost. t s unlkely that the studes do ths because even f the studes use arorate derecaton lves and derecaton methods the studes' technology assumtons contnually udate the nvestment amounts accordng to technology and rce mrovements. As technology becomes more effcent and unt rces declne udatng the technology assumtons lowers derecaton exenses all other thngs beng eual. Unless ths effect s ncororated nto the derecaton ths udatng causes TELR models to understate derecaton. n my exerence regulatory derecaton ractces do not consder ths dynamc of usng derecaton n TELR models. As a result there s a rsk that rces wll be too low.. omarson of Economc Prncles wth TELR Practce: Demand Uncertanty Hausman (1996) also argues that TELR studes do not adeuately reflect the effects of demand uncertanty on LEs' rreversble nvestments. Ths aears to be at least artally true but not to the extent clamed. 16 n a dynamc sense ths creates a smultanety roblem because rce affects uantty demanded. 13

14 My model ncororates demand uncertanty and rreversble nvestments. n the model the LE makes the nvestment before the LE knows demand and the LE s unable to reverse the nvestment. The artcaton constrant states that the rce must be greater than or eual to exected average ncremental cost. The numerator for the average ncremental cost ncludes all of the costs the LE ncurs to rovde ncludng costs ncurred for demand that does not materalze or that does not reman for the entre economc lfe of the lant. The denomnator s the exected sales. As a result average ncremental cost reresents an average over rojects whch lve u to ther demand exectatons rojects whch exceed ther demand exectatons and rojects whch fal to lve u to ther demand exectatons. All rreversble nvestments are covered n the artcaton constrant. 17 Lkewse the otmzaton constrant ncororates exected demand and costs over all rojects ncororatng the UNE or UO. Unfortunately t aears that the TELR models that the F s consderng do not lve u to the otmzaton and artcaton constrants wth resect to uncertan demand. As subsecton A above exlans the F has determned that TELR studes should assume an average utlzaton amount. Ths results n er unt TELRs eual to average ncremental costs. However the mlementaton of ths decson n the Hatfeld model does not take nto account all factors that cause utlzaton to be less than otmal. 18 ecfcally by otmzng the network based on current demand and future growth the studes fal to consder rojects that fal to lve u to the nvestors' demand exectatons. As a result the alcaton of the TELR models understates average ncremental cost and margnal cost. Hausman (1996) also states n the context of uncertan demand that TELR-based rces truncate the amount of rofts that LEs can exect from nnovaton thus dscouragng nnovaton-related nvestments. Hs asserton s correct. n fact any regulaton that lmts the amount of roft that an LE receves from an nvestment has ths effect. TELR-based rces are more onerous than for examle regular rce ca constrants n that normal rce cas allow comanes to adjust rces and so earn extra rofts on a artcular servce should market condtons ermt. However the roblem s wth rce lmts on nnovatons not wth demand uncertanty. At least n theory the roer remedy for ths roblem would be to have few f any TELR models are statc and so gnore ths roblem. 17 Even though usng average fll would remedy the rreversblty of nvestments for ndvdual rojects t does not remedy rreversblty for the LE's nvestment as a whole. Ths rreversblty should be reflected n the cost of catal. Henry Ergas (1998) argues that tradtonal cost of catal tools such as APM do not adeuately reflect ths rreversblty. Hubbard and Lehr (1996) argue that they do. Not beng an exert on cost of catal wll leave ths debate to others. 18 Accordng to a F staff reort the BPM2 model uses smlar fll factors. 14

15 rce constrants on nnovatons. n ractce t s hard to determne when somethng s an actual nnovaton and not just oortunstc use of regulatory rules. D. omarson of Economc Prncles wth TELR Practce: Real Otons Hausman (1996) also asserts that TELR studes understate ncremental cost because they fal to reflect the value of Real Otons -- nter-temoral nvestment otons foreclosed when the LE chooses. Ths may be true n some nstances but t s dffcult to magne that the effect s large. The EPR euatons n ecton llustrate how an LE s choce to make UNE and UO roduct nvestments whch call must rovde greater roft than the LE s best alternatve nvestment whch call Ψ that forecloses for the LE to choose. The EPR euatons also llustrate how Real Otons created by lower the roft needed from to nduce the LE to make the nvestment. To foreclose nvestments must occuy some sace that Ψ reures. Ths sace could be: 1. Physcal. Ths would be sace n for examle condut buldngs or rado sectrum that Ψ could occuy n the future. f ths s the only avalable sace for Ψ then the full value of Ψ s n the Real Oton. By "value of Ψ mean the net resent value of cash and otons from Ψ. f there are other saces that Ψ could occuy then only the ncremental loss f Ψ 's full value would be ncluded. 2. atal. Ths would nclude the consumton of catal that could not be relaced n the future excet at a hgher cost 3. Demand. The nvestment mght suly some consumer demand that could be served wth hgher valued nvestment sometme n the future 4. ost. The nvestment mght suly consumer demand that could be served wth lower cost nvestment sometme n the future. 5. Rghts. The nvestment mght cause the LE to lose some legal rght that has value. For examle f a rural LE urchased other exchanges n ts state t mght become suffcently large to lose ts rural LE status. As a non-rural LE the LE would be subject to the unbundlng and other local exchange cometton reurements of the Act. 15

16 To create an oton an nvestment must create an oortunty that dd not exst revously. For examle an nvestment n oenng local exchange markets to cometton creates a long dstance oton for Regonal Bell Oeratng omanes who are currently rohbted from rovdng ths servce. f does foreclose or oen a roftable nvestment ψ t may be necessary to aly somethng lke the EPR to ncent the LE to make the nvestment. 19 f t s necessary to consder the oortunty cost of a foreclosed nvestment n rcng nterconnecton or UOs t s also necessary to ncororate all of the cost and revenue effects of takng the foreclosed nvestment. For examle f the alternatve nvestment s a delayed nvestment n cable used to serve a artcular market then the cost of re-enterng the market should be ncororated as should the revenue loss from helng comettors ncrease ther market enetraton. Also because the nter-temoral oortunty cost ales only certan condtons and can be ostve or negatve t should be modeled searately from the TELR cost. A general adjustment to all TELR estmates would not necessarly mrove nvestment ncentves. Also makng the nter-temoral oortunty cost a searate model makes t easer for regulators to assess the valdty of the cost and the assumtons made when estmatng t. E. omarson of Economc Prncles wth TELR Practce: Other Factors There are other factors that affect the arorateness of TELR ractces. These may also decrease nvestment ncentves below an otmal level. One factor s the alcaton of the mark-u above TELR. ome regulators omt the mark-us. Others rovde mark-us only to cover shared costs. The frst ractce rovdes an neffcent ncentve to decrease nvestment and the second ractce mght. Recall that TELR s the sum of all margnal costs. The common assumton n telecommuncatons s that margnal costs are constant as roducton ncreases. n local exchanges roducton economes come from densty and scoe rather than scale. f hgher rces nduce greater nvestment whch s also a common assumton then the effcent rce s above margnal cost and therefore above TELR exressed on a er unt bass. ontrbutons to shared costs mght rovde the arorate mark-u but t would be only by accdent because the formulas for sreadng shared costs do not consder nvestment's effects on ualty. 19 However as has been shown n the EPR lterature t s narorate to consder any orton of these 16

17 Another factor s the omsson of the effects of rvalry on arorate rces. f LEs are subject to general multlateral rvalry (Jamson 1996) the rvalry forces LEs to rce below stand-alone cost for ndvdual roducts and for grous of roducts. Ths n turn means that LE rces for ndvdual roducts and for grous of roducts must exceed TELR for the LE to reman fnancally vable. Also the lack of oortunty to charge for ualty may lower the ualty of servces customers ultmately receve because LEs beneft from nvestng n ualty only through ncreased demand. Ths ncreased demand asses along to the LE only a orton of the value that the nvestment creates. Ths may rovde comanes wth an uneconomc ncentve to merge n areas wth large amounts of network nterconnecton because the merger allows the merged comany to nternalze some of the benefts of nvestment n ualty. The last factor s the regulatory rocess for usng TELR models affects rsk. assume n ths aer that the regulator commts to a rce schedule that the LE s certan wll hold over the entre lfe of the nvestment. Ths s unlkely to be true. U and nternatonal exerence wth such models ndcate that regulators can cause wde swngs n cost study results by changng a few crtcal assumtons. n the days of rate of return regulaton assumtons n fully dstrbuted cost studes and n rate cases rmarly affected when costs would be recovered and from what servce. Whle these ssues were mortant to market effcency they dd not rse to the level of today s use of cost studes where changes n assumtons determne total revenues wthout a clear remedy for nter-temoral errors excet to seek stranded cost recovery. f LEs beleve that regulators wll act arbtrarly or oortunstcally wth the TELR models then the LEs wll under nvest. ecton V. oncluson n ths aer consder ssues that the Real Otons debate has rased regardng the estmaton and alcaton of ncremental costs. show that current TELR models underestmate ncremental costs but not to the extent that some clam. also show that concerns wth nter-temoral oortunty costs are effectvely an alcaton of the EPR. Ths means that much of the lterature about the EPR should be alcable to ths ssue. There are other ssues that have not dscussed or modeled that also nfluence the effects of usng current TELR models for rcng UNEs and UO roducts. One such ssue s the hgher rofts that reresent monooly rofts. 17

18 effect of stranded cost remedes on LE nvestment decsons. Economc lterature on contract breach remedes aears to mly that some stranded cost remedes would allevate under nvestment concerns and may actually encourage over nvestment. ustomer contrbutons for lne extensons and contrbutons by real estate develoers may also have these effects. My model does not consder rvalry n the LE's markets. As menton n the revous secton multlateral rvalry would rovde alternatves for both the LEs and the customers. Ths would generally create addtonal rcng constrants. Also customer's oortuntes to selfrovde UNEs affect outcomes. Recrocal comensaton also affects the economcs of buyng and sellng UNEs. Also my assumton about the nature of nvestment s ute secfc. assume that nvestment roduces both ualty and reduces costs. Ths may be true for some nvestments but may not be true for all. A more thorough study s needed that consders both secfc nvestments and jont nvestments such as the nvestment assume. Lastly have not addressed whether the regulator should try to estmate TELR accurately or try to over estmate or under estmate TELR. have assumed that the regulator can come farly close. t may be that the regulator knowng that she has a robablty of error should seek to overstate TELR or understate TELR because one has less of a negatve effect on effcency. 18

19 Aendx My model consders a multroduct LE that makes nvestment decsons subject to regulatory rcng constrants and uncertan demand. The LE roduces roducts M N where N s the set of all roducts n the economy. The LE s roducts M are the roducts that fall ether nto the category of UNEs or nto the category of roducts subject to UOs. A roduct s subject to an UO f the regulator reures the LE to charge a rce < φ where φ s the rce that LE would charge n a erfectly comettve market. Demand s gven by ( ) where nature determnes and reresents nvestment that the LE can undertake. nvestment mroves ualty so ( ) > 0 where subscrts denote frst dervatves. Also ( )/ < 0 M. ( M ) s the LE s cost functon and ( ) = ( M ) ( M\ ) < ( ) s the ncremental cost of roducng. assume that < 0 < 0 ( M )/ > 0 and 2 ( M )/ < 0 M. The ncremental revenue effect of roducng s s: R ( M ) = M ( M ) M M\ ( M ) M\ assume there are no demand cross-elastc effects between and M\ so R ( M ) = ( ). assume that f the LE makes no nvestment t roduces no UNEs and UO roducts because customers would not buy them. n other words ( 0) = 0 and (0 0) = 0. suress comettors' rofts by assumng that they oerate n erfectly comettve markets. also assume that the regulator wants to mze weghted surlus Z mn LE LE [ w V ( ) w ( ( ) ) ( w w ) ( ) ] 0 where w LE and w are the weghts gven to LE rofts and customer surlus resectvely and V( ) s the customer s gross surlus. For smlcty assume that these weghts are eual. Ths smlfes the regulator s roblem to mzng 19

20 Z ( V ( ) ( ) ) 0 (2) mn There are three tme erods. n the frst erod the regulator selects the rce vector for UNEs and UOs usng TELR models. The regulator does not know or. However the regulator knows that [ mn ] and s dstrbuted accordng to the cumulatve densty functon F(). The regulator also knows the LE s roft mzng and rsk neutral so the regulator s able to accurately estmate the LE's best resonse functon to customer demand and the regulator's rce controls. The regulator also selects the rce vector M\ usng some rce cang mechansm and not rate of return regulaton. The urose of ths assumton s to remove oortuntes for cost shftng and ncentves for addng the rate base. n the second erod the LE chooses. The LE knows. The LE does not know but knows the range and densty functon just as the regulator does. n the thrd erod nature chooses customers buy ( M ) and the LE receves ncremental revenues of ( ). To solve Euaton 2 the regulator chooses the otmal rce vector by backwards nducton. For smlcty reresent the grou of customers as a sngle customer. The regulator calculates that n the last stage of the game ths customer mzes utlty accordng to { V ( ) ( )} 0 [ 0 ] The customer s frst order condtons are V ( ) = (3) assume that second order condtons are satsfed. Euaton 3 mles an otmal uantty choce ( ). The regulator then calculates the LE s best resonse functon to the customer s choce. The LE mzes rofts accordng to 20

21 21 solatng gves because of the assumton that demands are ndeendent. Ths gves the frst order condtons or where ((arg) s the exected value of arg. Euatons 4a and 4b are the otmzaton constrants and mly an otmal nvestment ( ). The LE s wllng to make the otmal nvestment as long as these hold and whch s the artcaton constrant. f the regulator were able to choose nvestment drectly the regulator's choce would be to otmze Euaton 2 wth resect to uantty and nvestment. Ths would gve the frst order condtons [ ] 0 mn ] [0 M M M M M M π (4a) 1 0 mn [ ] 0 mn ] [0 π (4b) 1 ( ( ( (5a) ( (

22 0 = 0 = [ V ( ) ( )] mn [ V ( ) ( )] 1 (7) mn (6) The regulator s unable to satsfy these condtons n ths model. show ths by combnng the regulator's frst order condtons wth the customer's and LE's frst order condtons. ombnng Euatons 3 and 4a gves 0 ( ) ( V ( ) ( ( ) ) ( ( ) ) mn 1 (8) From Euaton (6) V ( ) ( ( ) ) = 0 so Euaton (8) becomes ( ( ) ) = ( 1 whch means Euaton (7) becomes V = 0. o the regulator can satsfy her frst order condtons only n the secal case where the customer s margnal value of nvestment s zero and nvestment s margnal effect on the LE s to decrease cost dollar for dollar. Because the regulator cannot dctate uantty and nvestment the best the regulator can do s mze the followng [ V ( ( ( ) ( ) ( ( ( ) ( )] ( ) 0 mn whch gves the followng frst order condtons 22

23 23 assume that second order condtons are satsfed. To solate rce use the customer's frst order condtons from Euaton 3 to obtan From the LE's frst order condtons n Euaton 4a the value nsde the {} s zero so the regulator's frst order condtons become or Euaton 10a shows that the regulator's otmal rce celng wll eual margnal cost only f the customer's margnal value of nvestment s zero. By assumton V > 0 and < 0 so the sgn of determnes whether the otmal rce celng s above or below margnal cost. The regulator's otmal rce celng s above margnal cost f an ncrease n rce ncreases nvestment and the regulator's otmal rce celng s below the margnal cost of an ncrease n rce decreases nvestment. To determne the sgn of dvde the total dervatve of the LE's frst order condtons (Euaton 4a) wth resect to rce by the total dervatve of ts frst order condtons wth resect to nvestment. n other words [ ] ) (9a 0 mn V V V = V = 0 1 mn mn ) (10a 0 mn V = (10b) V = ( ( (

24 24 From second order condtons π < 0 so the sgn of π determnes the sgn of. π s Rearrangng terms and reversng sgns to get rd of the negatve sgn n front of the uotent gves The sgn of the frst exresson deends uon the sgn of the exresson nsde the arentheses because < 0. The second exresson's sgn deends uon the sgn of because the sgn of the exresson nsde the arentheses deends uon the sgn of. f > 0 then the regulator's rce s above margnal cost and the exresson s negatve. f < 0 the reverse s true. s ostve by assumton. o suffcent condtons for > 0 are [ ] mn PD[ PLQ (13) 0 (12) 0 (11) 0 < < > < > < > > π π =

25 25 ondtons 11 hold f the extra margnal costs caused by the nvestment-nduced demand growth domnate the nvestment-nduced decrease n margnal costs and f nvestment causes the nverse demand curve to be steeer. ondtons 12 hold f the nvestment-nduced decrease n margnal costs domnates the extra margnal costs caused by the nvestment-nduced demand growth f nvestment causes the nverse demand curve to be steeer and f the combned effects of the steeer nverse demand curve and rce exceedng margnal cost domnate the other effects. ondtons 13 hold f the costs caused by the nvestment-nduced demand growth domnate the nvestment-nduced decrease n margnal costs f nvestment causes the nverse demand curve to flatten and f the combned effects of the margnal cost decreases and demand changng wth rce ncreases and nvestment ncreases domnate the other effects. uffcent condtons for < 0 are ondtons 14 hold f the nvestment-nduced decrease n margnal costs domnate the extra margnal costs caused by the nvestment-nduced demand growth nvestment causes the nverse demand curve to be steeer and nvestment's effect on demand s domnated by all other effects. ondtons 15 hold f the extra margnal costs caused by the nvestment-nduced demand (16) 0 (15) 0 (14) 0 < < < < > > > > <

26 26 growth domnates the nvestment-nduced decrease n margnal costs nvestment causes the nverse demand curve to be steeer and the combned effects of the steeenng nverse demand curve and rce exceedng margnal cost domnate the other effects. ondtons 16 hold f the costs caused by the nvestment-nduced demand growth are domnated by the nvestment-nduced decrease n margnal costs nvestment causes the nverse demand curve to flatten and the combned effects of the margnal cost decreases and demand changng wth rce ncreases and nvestment ncreases domnate the other effects. Now consder the effects of Real Otons. The customer s frst order condtons n Euaton 3 stll hold. However the LE s mzaton roblem and frst order condtons become where ρ s the net absolute value of the Real Otons foreclosed or oened by. gve ρ a lus sgn f the net value of the Real Otons s ostve and gve ρ a negatve sgn f the net value of the Real Otons s negatve. assume that some orton α [0 1] of ρ s a socal beneft or cost so 1-α of ρ s rvate to the LE. The regulator s mzaton roblem and frst order condtons are now where the frst order condton solves to [ ] 0 mn ] [0 ± ρ π (4c) 1 0 mn ± ρ [ ] 0 mn ± ρ α V (9b) 0 mn V V V = ± ρ α

27 mn ( ( ) ( V ± ( α 1) ρ ) mn ( ) 1± ρ = 0 whch when combned wth the LE's frst order condtons becomes ( ( ) ( V ± ( α 1) ρ ) = 0 mn or ( ) ( V ± ( α 1) = ( ( ρ Because ρ > 0 and α - 1 < 0 the effect of Real Otons s to decrease the mark-u above margnal cost f the Real Oton s oened by nvestment and to ncrease the mark-u above margnal cost f the Real Oton s foreclosed by nvestment. Real Otons do not affect the sgn of. The artcaton constrant becomes ( ( ) ) ( ( ) ( ± ρ (5b) whch s the EPR. 27

28 Bblograhy Armstrong M. and. Doyle. (no date). Access Prcng Entry and the Baumol-Wllg Rule Dscusson Paer No Unversty of outhamton. Baumol Wllam J. (1977). On the Proer ost Tests for Natural Monooly n a Multroduct ndustry Amercan Economc Revew 67: Baumol Wllam J. and J. Gregory dak. (1994). Toward ometton n Local Telehony. ambrdge Massachusetts. Dxt A. K. (1990). Otmzaton n Economc Theory. Oxford. Ergas Henry. (1998). "Valuaton and ostng ssues n Access Prcng wth ecfc Alcatons to Telecommuncatons" nfrastructure Regulaton and Market Reform: Prncles and Practce M. Arblaster and M. Jamson eds. Melbourne Australa Hausman Jerry. (Arl ). Testmony before the alforna Publc Utltes ommsson. Hubbard R. Glenn and Wllam H. Lehr. (July ). "atal Recover ssues n TLR Prcng: Resonse to Professor Jerry A. Hausman" n the Matter of mlementaton of Local ometton Provsons of the Telecommuncatons Act of 1996 Docket No Jamson Mark A. (1998a). "A Further Look at Proer ost Tests for Natural Monooly." (unublshed). Jamson Mark A. (1998b). "nternatonal urvey of nterconnecton Polces: Fnal Reort." (unublshed). Jamson Mark A. (1998c). "Regulatory Technues for Addressng nterconnecton Access and ross-subsdy n Telecommuncatons" nfrastructure Regulaton and Market Reform: Prncles and Practce M. Arblaster and M. Jamson eds. Melbourne Australa Kahn Alfred and Wllam Taylor. (1994). The Prcng of nuts old to omettors: A omment Yale Journal on Regulaton 11: Mtchell Mtchell Werner Neu et. al. (1995). The Regulaton of Prcng of nterconnecton ervces (unublshed). Ordover J. A. A. O. ykes and R. D. Wllg. (1985). "Nonrce Antcomettve Behavor by Domnant Frms toward the Producers of omlementary Products" Anttrust and Regulaton: Essays n Memory of John J. McGowan F. M. Fsher ed. ambrdge Massachusetts. alnger Mchael A. (1998). Regulatng Prces to Eual Forward-Lookng osts: ost-based Prces or Prce-Based osts? Journal of Regulatory Economcs 14: Trgeorgs Lenos. (1996). Real Otons: Manageral Flexblty and trategy n Resource Allocaton. ambrdge Massachusetts. Tye Wllam. (1994). The Prcng of nuts old to omettors: A Resonse Yale Journal on Regulaton 11: Wllg Robert. (1979). The Theory of Network Access Prcng ssues n Publc Utlty Regulaton H. Trebng ed. East Lansng Mchgan