Armstrong Sappington Regulation Competition

Journal of Economic LiteratureVol. XLIV (June 2006), pp. 325366

Regulation, Competition, andLiberalization

MARK ARMSTRONG AND DAVID E. M. SAPPINGTON

In many countries throughout the world, regulators are struggling to determinewhether and how to introduce competition into regulated industries. This essay exam-ines the complexities involved in the liberalization process. While stressing the impor-tance of case-specific analyses, this essay distinguishes liberalization policies thatgenerally are procompetitive from corresponding anticompetitive liberalization policies.

325

Armstrong: University College London. Sappington:University of Florida. We are grateful to Sanford Berg,Severin Borenstein, Cory Davidson, Roger Gordon, AntonioEstache, Jerry Hausman, Mark Jamison, Paul Joskow,Mircea Marcu, John McMillan, Paul Sotkiewicz, JohnVickers, Catherine Waddams, Helen Weeds, and anonymousreferees for helpful comments and observations.

1. Introduction

Economists have developed an extensiveset of principles for regulating a monop-oly supplier. The benefits of unfettered,pervasive competition are also well docu-mented and well understood. However, ourunderstanding of the precise conditionsunder which regulated monopoly supply ispreferable to unregulated competition islimited. Furthermore, we know relativelylittle about optimal liberalization policiesthe policies that govern the transition tocompetitive market conditionsin caseswhere competition is deemed superior tomonopoly.

The purpose of this essay is to explorethese two issues, both of which are of sub-stantial practical importance throughout the

world. These issues are particularly relevantin key network industries (such as thetelecommunications, natural gas, electricity,transport, and water industries) where scaleeconomies can render production by manyfirms uneconomic, but where some compe-tition may be useful to help disciplineincumbent suppliers of key services. Ouranalysis of the choice between regulatedmonopoly and unregulated competition, likeour analysis of the design of liberalizationpolicies, emphasizes the problems thatimperfect information and imperfect institu-tions pose for the design of industry policy.

This essay has three main parts: (1) sec-tion 2 reviews some recent experience withliberalization policy; (2) sections 3 through 5consider the choice between unregulatedcompetition and regulated monopoly; and(3) sections 6 and 7 consider the design ofliberalization policies in settings where com-petition is preferred to monopoly.

Section 2 summarizes selected experi-ences with liberalization in three networkindustries where liberalization has garneredsignificant attention in recent years:

Jun06_Article1 5/2/06 4:50 PM Page 325

326 Journal of Economic Literature, Vol. XLIV (June 2006)

telecommunications, natural gas, and elec-tricity. The experiences in Chiles telecom-munications industry, the United Kingdomsnatural gas industry, and the state ofCalifornias electricity industry have variedgreatly, ranging from substantial success inChile to substantial failure in California.

Section 3 begins the discussion of thechoice between unregulated competitionand regulated monopoly by analyzing simpleformal models of these two forms of indus-trial organization. The models emphasizethe role of imperfect information. Section 4considers additional factors that affect thechoice between regulation and competition,including the resources available to the reg-ulator, the regulators independence andautonomy, the need to ensure ubiquitousservice at affordable prices, and the impor-tance of investment and innovation. Section5 notes that, even when monopoly supply isthe preferred mode of industry operation,some of the benefits of competition may besecured by allowing potential operators tobid for the right to serve as a regulatedmonopoly supplier.

Section 6 begins the discussion of liberal-ization policies in settings where the benefitsof competition outweigh its costs. The poten-tial merits and risks of direct entry assistanceare considered first. Liberalization policiesthat are likely to reduce the intensity of long-term industry competition (and thereforegenerally are not recommended) arereviewed next. These policies include pro-viding temporary monopolies or oligopolies,excluding foreign investors, specifying mar-ket share targets for industry participants,restricting incumbent suppliers asymmetri-cally, and affording competitors undulyfavorable long-term access to the incum-bents infrastructure. Section 7 discusses lib-eralization policies that typically willenhance the intensity of long-term industrycompetition (and therefore are recommend-ed). These policies include reducing thecosts that customers incur when they switchsuppliers; rebalancing the prices charged by

the incumbent supplier to better reflect itsoperating costs and otherwise redesigningthe regulations imposed on the incumbentsupplier to account explicitly for emergingcompetition; privatizing state-owned enter-prises; establishing appropriate accessprices; and increasing monitoring, datareporting, and antitrust scrutiny, at least dur-ing the early stages of liberalization.

Two central themes emerge in this essay.First, even the comparatively simple choicebetween regulated monopoly and unregulat-ed competition can be intricate and complexin practice. Second, the decision to intro-duce competition into an industry is only thebeginning of a journey down a long andwinding road that can present many obsta-cles and detours. Furthermore, the bestroute from monopoly to competition can dif-fer substantially in different settings.Therefore, there is no single set of directionsthat can guide the challenging journey frommonopoly to competition in all settings.

Even though detailed, comprehensivedirections typically are not available, somebroad conclusions (summarized in section 8)that can serve as useful guide posts can bedrawn. These broad conclusions include thefollowing five. First, the greatest potentialgains from competition tend to arise when(1) industry scale economies are limited rel-ative to consumer demand; (2) the industryregulator has limited information, limitedresources, and limited instruments withwhich to craft policy; (3) the regulators com-mitment powers are limited; and (4) subsi-dization of the consumption of some of thedominant suppliers services either is notcritical or can be achieved by means otherthan through distortions in the suppliersprice structure. Second, there are a widevariety of liberalization policies, and themerits of the different policies vary consid-erably. Therefore, it generally is more appro-priate to inquire about the benefits and costsof specific liberalization policies than to askwhether liberalization per se is desirable orundesirable. Third, liberalization policies


Armstrong and Sappington: Regulation, Competition, and Liberalization 327

1 The following discussion is drawn primarily fromAhmed Galal (1996), Pablo T. Spiller and Carlo G. Cardilli(1997), David M. Newbery (1999, pp. 12325), CecileAubert and Jean-Jacques Laffont (2002, pp. 3941),Laffont (2005, pp. 21416), and Ricardo D. Paredes(2005).

that primarily aid some competitors andhandicap others on an ongoing basis can hin-der the development of vigorous long-termcompetition. Fourth, liberalization policiesthat remove entry barriers and empowerconsumers to discipline industry supplierstypically are the best means by which to fos-ter vigorous long-term competition. Fifth,well-designed liberalization policies canfacilitate a transition from stringent, detailedregulatory control to less intrusive antitrustoversight. However, during the transitionprocess, heightened regulatory and antitrustscrutiny may be required.

2. Recent Experience with Liberalization

To help ground the ensuing discussion ofthe principles of liberalization policy, we beginby reviewing briefly the recent experiencewith liberalization in Chiles telecommunica-tions industry, the United Kingdoms naturalgas industry, and Californias electricityindustry.

2.1 Liberalization in ChilesTelecommunications Industry1

The General Law of Telecommunicationsopened Chiles telecommunications industryto competition in 1982. This law generallydid not restrict the number of licenses thatwould be granted to deliver telecommunica-tions services in Chile. Importantly, thelicenses were for nonexclusive provision ofwireline and wireless services, both local andlong distance. The primary obligationimposed on all telecommunications opera-tors in Chile was to connect their networks incompliance with specified technical require-ments. Other terms and conditions of inter-connection typically were left to negotiationamong the operators.

In 1982, local telecommunications servic-es were supplied almost exclusively byCompena de Telfonos Chile (CTC), whilelong distance telecommunications serviceswere supplied almost exclusively by Entel.Not surprisingly, these two state-ownedenterprises were not anxious to connecttheir networks with the networks of emerg-ing rivals. Few interconnection agreementswere signed and competition was limiteduntil Chile implemented a dispute resolu-tion process that ensured the timely execu-tion of interconnection agreements. Longdistance competition also was limited untilan equal access requirement was imposed in1994. The requirement stipulated that everytime a customer placed a long distance call,she had to explicitly designate (via dialing aspecial code) a long distance company tocarry the call. The special code for each longdistance carrierincumbent and entrantalikehad the same number of digits. Mostimportantly, a call was not automatically car-ried by Entel if the caller did not specify analternative carrier.

Once new suppliers were afforded sub-stantial opportunity to provide telecommu-nications services on terms comparable tothose faced by incumbent suppliers, compe-tition began to flourish in Chile. The num-ber of fixed lines more than tripled (fromroughly one million to more than three mil-lion) between 1992 and 2000, and the exten-sive waiting list for (fixed) wirelinetelephone service that had inconveniencedChiles citizens for so many years quickly dis-appeared. The number of mobile telephonesubscribers in Chile increased nearly ten-fold (from roughly 36,000 to more than 3.4million) during the 19912000 period andnearly doubled again (to more than 6.7 mil-lion) between 2000 and 2003. Prices formost telecommunications services in Chilealso have declined substantially since 1990.

By 2003, nearly one-fourth of the fixedtelecommunications lines in Chile were sup-plied by CTCs competitors rather than bythe incumbent supplier. Competitors



2 See Paredes (2005), for example.3 The material is drawn from Mark Armstrong, Simon

Cowan, and John Vickers (1994, chapter 8) and Newbery(1999, chapter 8).

secured especially large market shares inChiles urban regions. This pattern of entrylikely reflects in part the lower unit costs ofserving urban areas given their relativelyhigh population densities. This entry patternalso may reflect regulated prices of basictelephone service that are further abovecosts in urban regions than in rural regions.The regulated prices of local telephone serv-ices in rural regions often do not reflect fullythe relatively high unit costs of serving theseregions. Instead, prices in rural regions tendto be set closer to (and sometimes below)cost while prices in urban regions often areset substantially in excess of cost. Such pric-ing patterns are designed to ensure thatbasic telephone service is affordable to allcitizens. These patterns may be less pro-nounced in Chile than in some countriesbecause a separate fund has been estab-lished to subsidize the purchase of basictelephone service in rural regions and low-income urban regions. Nevertheless, injurisdictions where CTC is deemed to be adominant supplier of local telephone servic-es, it is required to set the same price forbasic services across substantial geographicregions. Some suggest that this restriction onCTCs ability to target price reductions tothose regions in which competition is mostintense may be one cause of the substantialmarket share that competitive suppliers oflocal telecommunications services have beenable to secure in Chile.2

2.2 Liberalization in the United KingdomsNatural Gas Industry3

Natural gas was one of the principal net-work industries in the United Kingdom tar-geted for privatization under MargaretThatchers program of industrial reform.Natural gas was first extracted from the U.K.Continental Shelf in 1965 and British Gas(BG) was formed as a state-owned company

4 Alternatively, BG could have been separated into anational pipeline business and several regional suppliersbefore privatization, an approach subsequently pursued inthe U.K. electricity industry.

5 This Y factor enabled BG to pass through its costs ofpurchasing gas to consumers, and thereby limited BGsincentive to secure low gas prices. Subsequent regulationpermitted BG to pass through only a more exogenousindex of gas costs.

6 The 2 percent value for X was widely believed tounderestimate the realistic potential for productivity gains.Of course, lenient price regulations enhance revenuesfrom privatization.

in 1972. BG held a legal monopoly over thesale of gas to consumers and a legal monop-sony over the purchase of gas from produc-ers in the U.K. fields. Since gas explorationand production involve substantial sunkcosts, BGs monopsony power necessitatedthe use of long-term (e.g., twenty-five year)purchase contracts to limit expropriation ofthe substantial investments made by gasproducers.

An initial attempt to liberalize the gas sec-tor occurred in 1982 when entrants wereauthorized to employ BGs pipelines (atterms negotiated with BG) to supply gas tofinal customers. In addition, BGs legalmonopsony was removed. Although thesesteps were intended to facilitate competitionin the gas industry, competition did notemerge immediately.

BG was privatized as a vertically integrat-ed monopoly in 1986.4 At this time, con-sumers were divided into two categories forregulatory purposes: (1) the tariff market,consisting of those consumers who pur-chased less than 25,000 therms of gas peryear, and (2) the contract market, consistingof the remaining higher-volume consumers.BG continued to enjoy a legal monopoly inthe tariff market after privatization. BGsprices in this sector were regulated: theannual increase in the average price pertherm was limited to RPI + Y X, whereRPI is the inflation rate, Y is a measure ofthe increase in the price BG pays for gas,and X is a productivity improvement factor.5

X was set at 2 percent in 1986, but raised to5 percent in 1992.6



7 These access charges were distance related. BecauseBGs retail prices did not vary across geographic regions,this pricing policy enabled entrants to secure a higherprofit margin when serving customers located close to thegas landing point.

Competition in the contract marketwas potentially open to competition andBGs prices in this market were not regu-lated. Although it was legal, competitiondid not emerge in this market until 1990.The delay reflected in part limited regu-lation of the prices BG charged retailcompetitors for access to its pipelines.BG also practiced price discrimination inthe contract market, charging higherprices for gas to those industrial userswho had no reasonable alternative energysource (e.g., electricity).

To combat these perceived problems,new regulations were imposed in 1988.The regulations required BG to (1) pub-lish a tariff on which any customer in thecontract market could purchase gas (inan attempt to limit price discrimination);(2) publish a corresponding tariff of net-work access charges (although thesecharges remained otherwise unregulat-ed); and (3) purchase no more than 90percent of the gas in any newly discov-ered gas field. This final requirementessentially forced some limited entry intothe contract market. By 1990, though,BG supplied 93 percent of the gas sold inthis market.

Limited competition in the contractmarket promoted three further regulato-ry reforms in the early 1990s. First, themonopoly threshold in the tariff sectorwas reduced from 25,000 therms to 2,500therms, thereby permitting smaller cus-tomers to purchase gas from competitors.Second, the prices charged for access toBGs pipelines came under regulatorycontrol.7 Third, BG was required toreduce to 40 percent (by volume) itsshare of sales to the contract market by1995. At the same time, new gas fieldswere being opened, and entrants were

8 As a result, BG was left with long-term obligations topurchase gas that it no longer needed to serve its own con-sumers. When BG was privatized, investors were promisedthat BG would be able to supply approximately two-thirdsof the demand for natural gas by volume until 2009. Asregulation and liberalization developed, this promise wasnot honored.

9 The following discussion is based primarily on SeverinBorenstein (2002).

able to buy supplies from these newfields.8

Although regulation did not force BG toseparate its pipeline and gas supply opera-tions, intense regulatory scrutiny and oner-ous separate accounting requirements ledBG to undertake such vertical divestiturevoluntarily. Furthermore, the monopolyfranchise threshold of 2,500 therms wasremoved entirely in 1998, so all consumerswere permitted to purchase gas from suppli-ers other than BG. Competition intensified,and all retail price controls were removed in2002. Only access charges continue to beregulated. Many consumers continue to pur-chase natural gas from BG even though BGcharges more than some competitors forwhat is largely a homogeneous product. Thisfact may suggest that customer switchingcosts or other causes of customer inertia areimportant in this industry.

2.3 Liberalization in Californias ElectricityIndustry9

Although the state of Californias experi-ence with liberalization in its electricityindustry is quite recent, the experience isalready legendary. In 1996, California enact-ed legislation that introduced five primarychanges in the states electricity sector. First,electricity generation and wholesale pricesfor electricity were deregulated. Second, asthey were required to do, the three incum-bent (vertically integrated) suppliers of elec-tricity sold a sizable portion of theirgeneration capacity, focusing instead on thetransmission and delivery of electricity. By1999, the incumbent suppliers had sold tofive independent suppliers generationcapacity that produced roughly one-third of



10 It has also been suggested that some of the few largesuppliers of electricity may have exacerbated the problemby intentionally reducing their supply of electricity, there-by raising the market-clearing price of electricity and,hence, the profits of electricity suppliers. (See Paul L.Joskow and Edward Kahn 2002, for example.)

the states electricity consumption. Third,retail competition was introduced as retailcustomers were permitted to purchase elec-tricity from firms other than the incumbentsuppliers. Fourth, a ceiling was imposed onthe retail price that incumbent supplierscharged for electricity. Fifth, the threeincumbent suppliers were afforded strongfinancial incentives to buy and sell electrici-ty through the California Power Exchange(Cal PX) during its first four years of opera-tion. Cal PX was established in 1998 to runthe day-ahead market for electricity in thestate of California. Although the incumbentsuppliers participation in Cal PX helped toensure substantial short-term supply of anddemand for electricity, the incumbents cor-responding limited use of long-term con-tracts for the purchase or sale of electricityultimately proved to be detrimental.

In the summer of 2000, unusually hightemperatures and very dry conditions pro-duced both a substantial increase in thedemand for electricity and a significantreduction in the available supply of electric-ity from hydroelectric generating units in thestate. When less efficient generating unitswere dispatched to meet the high demandfor electricity, the wholesale price of elec-tricity soared well above the ceiling that hadbeen imposed on retail electricity prices.Consequently, incumbent suppliers incurredsevere financial losses. The higher wholesaleprices were not sufficient to augment thesupply of electricity greatly due to the longlead time necessary to construct new gener-ating capacity and due to the particularlyhigh costs faced by established producers ofpeak load capacity. These high costs reflect-ed less efficient operating technologies, par-ticularly high wholesale prices for naturalgas, and expensive pollution permits.10

In 2001, the state of California adopteddrastic measures to quell the crisis that haddeveloped in its electricity industry. Thestate established itself as the wholesalebuyer of power for the incumbent suppliersin response to the reluctance of other parties (including wholesale suppliers ofelectricity) to deal with the financially dis-tressed incumbents. The state also raisedsubstantially the prices incumbent suppli-ers could charge for electricity, especially tolarge retail customers. In addition, to pre-vent these customers from securing elec-tricity from alternative suppliers at lowerprices, the state terminated retail competi-tion. In essence, Californias experimentwith deregulation and liberalization hadended.

3. Regulated Monopoly and UnregulatedCompetition

In an economic paradise, where a regula-tor is omniscient, benevolent, and able tofulfill any promise he makes, competitioncannot improve upon regulated monopoly.In such a paradise, the regulator will ensurethe firm produces the ideal range of servicesat the lowest possible cost and will set wel-fare-maximizing prices for these services.Consequently, industry performance wouldnot improve if an additional firm operated inthis setting.

Of course, the real world differs markedlyfrom this paradise. In practice, regulatorsinvariably lack important information aboutthe markets they oversee and so will not beable to direct and control perfectly the activ-ities of a monopoly producer. Because of itsdaily operation in the industry and its directcontact with consumers, the regulated firmwill be better informed than the regulatorabout the demand for the regulated servicesit supplies, the minimum possible currentcost of delivering the services, and thepotential for less costly future provision.This information asymmetry generally givesrise to an unavoidable trade-off between



11 Martin Loeb and Wesley A. Magat (1979) provide aformal statement and proof of this conclusion. William W.Sharkey (1979) critiques Loeb and Magats analysis. JorgFinsinger and Ingo Vogelsang (1981, 1982) and David E.M. Sappington and David S. Sibley (1988) suggest dynam-ic modifications of the Loeb and Magat policy that affordless rent to the regulated firm and/or can be implementedwith less precise information about the surplus generatedby the firms activities.

12 This model is drawn from the large literature thatexamines the design of regulatory policy when the regulat-ed firm is better informed about its environment than theregulator. Reviews of this literature include BernardCaillaud, Roger Guesnerie, Patrick Rey, and Jean Tirole(1988), David P. Baron (1989), Laffont and Tirole (1993),Laffont (1994), and Armstrong and Sappington (forthcom-ing). Joskow and Richard Schmalensee (1986), Armstrong,Cowan, and Vickers (1994), Glenn Blackmon (1994),Sappington (1994), Robert Mansell and Jeffrey Church(1995), Sappington and Dennis L. Weisman (1996), andJoskow (2005), among others, provide less technical dis-cussions of incentive regulation. Michael A. Crew and PaulR. Kleindorfer (2002) and Vogelsang (2002), for example,provide critiques of this literature.

rent and efficiency: the firm can be motivat-ed to operate efficiently but only if it isawarded substantial rent for doing so. Inparticular, the firm will operate at minimumcost and attempt to satisfy the needs anddesires of customers only if it is awarded thefull surplus that its activities generate.11

However, such a generous award to the reg-ulated firm typically will provide it with sig-nificant rent and thereby reduce the netbenefits enjoyed by consumers. To limit therent that accrues to the regulated firm, someinefficiency typically is tolerated.

To examine the optimal resolution of thistrade-off and to examine the impact of lim-ited information on the choice betweenregulated monopoly and unregulated com-petition, consider the following simplemodel.12

3.1 A Simple Model of Regulated Monopoly

Suppose that when regulated monopoly isimplemented the regulator faithfully pur-sues the social goal of maximizing theexpected value of V + U, where V denotesthe surplus enjoyed by consumers, U is thefirms rent, and [0,1] is a parameter.Because 1, society values consumer wel-fare at least as highly as the welfare of share-

13 Laffont (2005, pp. 12) reports that for each dollar oftax revenue the government collects, citizens in a devel-oped country bear a cost of approximately $1.30 (so =0.3). The corresponding cost typically is substantiallygreater in developing countries.

holders, perhaps because consumers are lesswealthy than shareholders or because manyshareholders reside in other jurisdictions.

A transfer payment T from consumers tothe firm entails a reduction of [1 + ]T in thesurplus enjoyed by consumers. The parame-ter 0 represents the social cost of publicfunds. This cost arises from the distortionscreated by the taxes imposed on con-sumers/taxpayers to raise the funds.13 Noticethat because public funds entail unit cost1 + , each dollar of public funds secured bytaxing the profit of the regulated firm can beemployed to increase consumer/taxpayerwelfare by 1 + dollars.

The monopoly supplies a single product atregulated unit price p 0. The demandcurve for this product, q(p), is commonknowledge. The regulator sets both the unitprice, p, for the regulated product and atransfer payment, T, from consumers to theregulated firm. The firm is obligated to serveall customer demand at the establishedprice.

The firm incurs a fixed cost of operation,F, and a constant marginal cost of produc-tion, c. For simplicity, the firms marginalcost can take on one of two values, cL or cH.Let cH cL > 0 denote the differencebetween the high and the low marginal cost.The firm knows from the outset of its inter-action with the regulator whether its mar-ginal cost is high or low. The regulator doesnot share this information and neverobserves the firms marginal cost directly.The regulator perceives the two possiblecost realizations to be equally likely. For sim-plicity, the firms fixed cost of operation, F, isassumed to be common knowledge. Thefirm seeks to maximize its rent, U, which isthe sum of its profit, q(p)[p c] F, andthe transfer payment, T, it receives from theregulator.



14 See Roger B. Myerson (1979), for example.

Suppose the regulator announces that hewill set unit price pi and deliver transfer pay-ment Ti to the firm when the firm reports itsmarginal cost to be ci, for i L,H. When thefirm with cost ci chooses the (pi,Ti) option, itsrent will be Ui q(pi)[pi ci] F + Ti. Therevelation principle ensures there is no lossof generality in examining regulatory policiesthat induce the firm to report its marginalcost truthfully.14 Therefore, social welfarewhen the firms marginal cost is ci is:v(pi) [1 + ]Ti + [q(pi)[pi ci] F + Ti]

= wi(pi) [1 + ]Ui,where wi(pi) v(pi) + [1 + ][q(pi)[pi ci] F], and where v(p) denotes consumer sur-plus when price p is established. (v(.) is aconvex function of p.)

If the firms realized marginal cost wereobserved publicly, the regulator wouldimplement the (Ramsey) price that maxi-mizes wi(.) when cost ci is realized, fori L,H. These full-information prices will bemarginal-cost prices (pi = ci) when is zerobecause transfers from consumers to thefirm entail no direct social costs in this case.In contrast, when is large, payments toconsumers financed by the firms profit arehighly valued and the full-information priceswill approximate the prices chosen by anunregulated profit-maximizing monopolist.

When the firms marginal cost is notobserved publicly, it must be the case thatq(pL)[pL cL] F + TL q(pH)[pH cL] F + TH or, equivalently,

(1) UL UH + q(pH),

to ensure the firm truthfully reports its lowmarginal cost of production. To ensure thefirm finds it profitable to operate when itsmarginal cost is high, it must be the case thatUH 0. Because social welfare declines asthe firms equilibrium rent increases (when < 1 or > 0), UH is optimally held to zero.To limit the firms equilibrium rent, con-straint (1) also will hold as an equality under

15 G might also include the surplus lost when regulationretards industry innovation. Charles Jackson, Tracey Kelly,and Jeffrey Rohlfs (1991) and Jerry A. Hausman (1997)estimate that delays in licensing wireless telecommunica-tions providers in the United States reduced consumersurplus by billions of dollars.

16 This fact is evident from expression (2).

the optimal regulatory policy, soUL = q(pH). Therefore, total expected wel-fare if price pi is set when marginal cost ci isrealized (for i L,H) will be:

(2) 12[wL(pL) [1 + ]q(pH)]

+ 12 wH(pH) [1 + ]G,

where G is a (fixed) cost of regulation thatis financed with public funds. This costmight include the salary of the regulatorand his staff, for example, and all othercosts associated with acquiring essentialinformation about the regulated industry.15

Differentiating expression (2) with respectto pH reveals that, when the regulator can-not observe the firms marginal cost, he isable to achieve the level of expected wel-fare that he could achieve if the firms costswere observable, but the high marginal costwas cH, where:

(3) cH cH + [1 1+] > cH.

Therefore, the optimal price when the highmarginal cost is realized is the full-informa-tion (Ramsey) price corresponding to theinflated cost cH. The inflated price when cHis realized reduces the number of units ofoutput on which the firm can exercise itscost advantage when cL is realized.Therefore, the increase in pH and the associ-ated reduction in TH limit the rent thataccrues to the firm when its marginal cost iscL. Because the firm has no incentive tounderstate its production cost, there is novalue to distorting the firms activities whenit reports its marginal cost to be cL.Consequently, the optimal price when thelow marginal cost is realized will be the full-information price corresponding to thefirms actual cost (cL).

16



17 Section 6 considers the possibility that both firmsmay not find it profitable to operate in the industry. SeeEmmanuelle Auriol and Laffont (1992) and Michael H.Riordan (1996), for example, for related models in whichthe duopoly is regulated.

Three features of the optimal regulatorypolicy in this simple setting characterizeoptimal regulatory policy in more generalsettings where the monopoly supplier is bet-ter informed than the regulator about keyfeatures of the regulated industry. First, thefirm generally commands rent from its supe-rior information. Second, to limit this rentand thereby generate greater surplus forconsumers, the regulator will design a menuof options from which the firm can make abinding choice. A well-structured menu ofoptions can induce the firm to employ itssuperior knowledge to secure outcomes thatare better for both the firm and consumersin more favorable environments (e.g., whenthe firm has lower operating costs). Third,the optimal regulatory policy generallyinduces inefficient performance to limit thefirms rent.

3.2 A Simple Model of UnregulatedCompetition

Now consider the following simple modelof unregulated competition. Suppose twofirms produce a homogeneous product andengage in Bertrand price competition. Eachfirm knows its own constant marginal cost ofproduction and its rivals marginal cost,c {cL,cH}, when it sets its price. No transferpayments to or from the firms in the indus-try are possible in this setting. In particular,the firms profits cannot be appropriated bythe government to reduce the general taxburden elsewhere. Therefore, the social costof public funds plays no role in this settingand social welfare is v(p) + , where isindustry profit. Both firms find it profitableto operate in the industry.17

Each firm has the low marginal cost, cL,with probability 1/2. The firms costs may becorrelated. Let [1/2,1] represent theprobability that the two firms have the same 18 The probability that both firms have high cost is /2,

the probability that both firms have low cost is /2, and theprobability that a given firm has low cost while its rival hashigh cost is [1 ]/2.

19 The profit-maximizing price for a firm with the lowmarginal cost is assumed to exceed cH. Consequently, whenonly one firm has the low marginal cost, it will serve theentire market demand at price cH in equilibrium.

cost.18 The firms costs are perfectly corre-lated when = 1. Their costs are uncorrelat-ed when = 1/2. Bertrand competitionensures the equilibrium price will be cHexcept when both firms have low cost, whichoccurs with probability /2. A firms operat-ing profit is zero unless it has the low mar-ginal cost while its rival has the highmarginal cost, in which case the firms profitis q(cH).19 A firm realizes this positive prof-it with probability [1 ] /2. Consequently,expected industry profit in this duopoly set-ting is [1 ]q(cH), which declines as thefirms costs become more highly correlated.Notice that the probability that the industrysupplier has low marginal cost is (1 /2).This probability decreases as increases. Incontrast, the probability that the industryprice will be cL is /2, which increases as increases.

Ignoring the firms fixed costs of production(F) for now, social welfare in this unregulatedduopoly setting is:

(4) 2 v(cL) + [1 2 ]v(cH) + [1 ]q(cH).

3.3 Comparing Regulated Monopoly andUnregulated Competition

Regulated monopoly offers four potentialadvantages over unregulated competition inthis simple setting: (1) industry prices canbe controlled directly; (2) transfer pay-ments can be made to the firm to providedesired incentives; (3) the firms profit canbe taxed to generate public funds, therebyreducing the deadweight losses associatedwith other sources of public funds; and (4)duplicative fixed costs of production can beavoided because there is only one industrysupplier.



20 If = 1, regulated monopoly and unregulated duop-oly provide the same level of expected social welfare.

Unregulated competition has three corre-sponding potential advantages: (1) the likeli-hood that the industry producer has the lowmarginal cost is higher than under monopolybecause even if one firm fails to secure thelow cost, its rival may do so; (2) the presenceof a rival with correlated costs reduces theinformation advantage of the industry pro-ducer; and (3) any direct, operational costs ofregulation (e.g., the salaries of regulators andtheir staff) are avoided. The first of thesepotential advantages of unregulated duopolyis referred to as the sampling benefit of com-petition. The second potential advantage willbe referred to as the rent-reducing benefit ofcompetition.

To compare the performance of regulatedmonopoly and unregulated competition inthis setting, initially suppose = 0, F = 0,and G = 0, so there is no social cost of pub-lic funds, no fixed cost of production, and nodirect cost of regulation. Expressions (2)and (3) reveal that the maximum expectedwelfare under monopoly regulation in thiscase is:

(5) 12v(cL) +12 v(cH + [1 ]).

A comparison of expressions (4) and (5)provides four conclusions regarding the rel-ative performance of regulated monopolyand unregulated duopoly.

First, unregulated duopoly delivers ahigher level of expected social welfare thandoes regulated monopoly when the duopo-lists costs are perfectly correlated (so = 1).20 When costs are perfectly correlat-ed, the industry producer never has a costadvantage over its rival and so commands norent in the duopoly setting. Furthermore,competition drives the industry price to thelevel of realized marginal cost. Therefore,the ideal (full-information) outcome isachieved under duopoly but not undermonopoly, where regulated prices diverge

21 When demand is perfectly inelastic (so q(p) 1, forexample), expression (4) is weakly greater than expression(5) whenever /2 [ 12], which is always the case.22 The convexity of v(.) implies the expression in (5) isat least [v(cL) + v(cH) [1 ]q(cH)]/2. Therefore, the dif-ference between expressions (5) and (4) is at least [1 ][v(cL) v(cH)]/2 q(cH)[(1 )/2 + (1 )]. Because

this expression is nondecreasing in , it is at least [1 ][v(cL) v(cH)]/2 q(cH)/2. This term is positive when

demand is sufficiently elastic.

from marginal cost in order to limit the rentthe monopolist commands from its privi-leged knowledge of costs. In this case, then,unregulated duopoly is preferred to regulat-ed monopoly even though the former offersno sampling benefit. The benefits of compe-tition arise entirely from rent reduction inthis case.

Second, when demand is perfectly inelas-tic, unregulated duopoly produces a higherlevel of expected welfare than does regulatedmonopoly.21 When demand is perfectlyinelastic, price distortions do not affect out-put levels and, therefore, do not serve to limitrent. Consequently, only the probability ofobtaining a low-cost supplier affects expectedwelfare, and this probability is higher underduopoly than under monopoly because of thesampling benefit of competition.

Third, regulated monopoly will generatea higher level of expected welfare thanunregulated duopoly when demand is suffi-ciently elastic (and < 1). When demand isvery elastic, prices that do not track costsclosely entail substantial losses in surplus.Prices track costs more closely under regu-lated monopoly than under unregulatedduopoly.22

Fourth, unregulated duopoly outperformsregulated monopoly when the differencebetween the high and the low marginal cost() is sufficiently close to zero. Monopolyrent and duopoly profit are both negligiblein this case, and so the choice betweenmonopoly and duopoly depends upon whichregime produces the low marginal cost morefrequently. The sampling benefit of compe-tition ensures the duopoly regime does so



23 The benefits of unregulated competition may be lesspronounced if industry competition is less pronounced.For example, Cournot competition may better describeindustry interaction than Bertrand competition.Alternatively, industry producers might collude in settingprices. (Recall that limited industry competition may havecontributed to the substantial increase in the wholesaleprice of electricity in California in 2000.)

24 In contrast, if the fixed cost of regulation, G, is suffi-ciently large, unregulated duopoly will outperform regu-lated monopoly because the former avoids this cost, byassumption.

25 In developing countries where consumers have limit-ed income, the corresponding limited demand for keyservices can result in higher average production costswhen scale economies are present.

whenever costs are not perfectly correlated(so < 1).23

More generally, if the fixed costs of opera-tion (F) are sufficiently large, regulatedmonopoly will outperform unregulatedduopoly in the simple model analyzed herebecause monopoly avoids the duplication offixed costs.24 This conclusion is a corollary ofthe more general observation that monopolysupply will minimize industry costs if pre-vailing scale economies are pronounced rel-ative to industry demand.25 Scale economiesoften are pronounced in network industries,where substantial physical infrastructure(e.g., a gas, water, or electricity distributionsystem or a telecommunications network)must be deployed in order to deliver serviceto customers.

When the social cost of funds () is con-sidered, regulated monopoly offers an addi-tional advantage over unregulated duopoly.The monopolists rent can be taxed to funddesirable social projects, thereby reducingthe need to employ other (potentially morecostly) means to raise revenue.

4. Additional Considerations

The simple models considered in section 3abstracted from several important institu-tional factors that can affect the optimalchoice between regulated monopoly andunregulated competition. These institutionalfactors include (1) the resource constraintsthe regulator faces; (2) the potential role of

26 For further discussion of the effects of institutions onregulatory policy, see Brian Levy and Spiller (1994, 1996),Newbery (1999), Roger G. Noll (2000), J. Luis Guasch(2004), Ioannis N. Kessides (2004), Mark A. Jamison,Lynne Holt, and Sanford V. Berg (2005), and Laffont(2005), among others. We focus on the effects, rather thanthe origins, of a countrys institutions. Daron Acemoglu,Simon Johnson, and James Robinson (2005) trace a coun-trys institutions to its colonial origins. In countries wherepoor living conditions rendered settlement unattractive,for example, little effort was devoted to developing thecountrys institutions. Such effort was more prominent incountries where long-term settlement was more attractiveand more widespread.

27 The theory of optimal regulation when the firm isprivately informed about several aspects of its operationawaits further development. Jean-Charles Rochet and LarsStole (2003) provide a survey of the theory of multidimen-sional screening.

28 Jon Stern (2000) documents the limited regulatoryresources that are available in many countries. Stern sug-gests that resource sharing among regulatory agencies canhelp to mitigate partially the effects of severe shortages ofcritical regulatory resources in some settings. Noll (2000)also emphasizes the merits of sharing regulatory resources.

regulation in pursuing distributional objec-tives; (3) the instruments available to theregulator; (4) the prevailing degree of regu-latory independence and accountability; (5)the ownership structure of the incumbentindustry producers; and (6) the importanceof industry investment and innovation.These factors are now considered in turn.26

4.1 Resource Constraints

Although the regulator was not omniscientin the models of section 3, he had consider-able knowledge of the regulated industry. Inpractice, a regulators information can be farmore limited.27 A regulators difficult task ofoverseeing and directing the activities of amonopoly supplier can become nearlyimpossible when the regulators informationand expertise are severely limited and whenhe lacks the physical and financial resourcesto overcome these limitations. Consequently,allowing competition to replace regulatoryoversight as the primary means of motivatingand disciplining the incumbent supplier canbe advantageous when the efficacy of regula-tory oversight is severely compromised bylimited regulatory resources.28



29 The public finance literature (e.g., Anthony B.Atkinson and Joseph E. Stiglitz 1976) notes that when anonlinear income tax is feasible, prices of goods and serv-ices often should reflect production costs. Laffont andTirole (2000, section 6.2) consider the implications of thisconclusion for pricing policy in regulated industries. RobinBurgess and Stern (1993) review the theory and the prac-tice of tax policy in countries with limited wealth.

The resources of other oversight agenciesalso influence the relative merits of monop-oly supply and liberalization. For example,competition may be less likely to imposemeaningful discipline on a dominant incum-bent supplier if the antitrust agency thatoversees industry competition has limitedexpertise and meager physical and financialresources. Consequently, both the relativeand absolute level of resources available tooversight agencies influence the relativemerits of monopoly and liberalization.

4.2 Income Redistribution and UniversalService Goals

The analysis in section 3 abstracted fromany differences in the cost of serving differ-ent customers and did not model explicitlydifferences in the wealth levels of differentconsumers. Such differences can raise con-cerns about universal servicethe ubiqui-tous delivery of essential services ataffordable rates. When a government haslimited ability to redistribute income direct-ly (perhaps because income taxes are widelyevaded, for example), the regulated prices ofessential services can constitute an impor-tant means to promote universal service.29

To illustrate, suppose a country wishes toensure that all its citizens have access toclean water or to basic communications serv-ices at low prices. Because prices tend toreflect costs under unfettered competition,individuals that are particularly costly toserve (because of their geographic location,for example) may face unduly high prices forkey services under unregulated competition.In contrast, a regulated monopolist willagree to serve high-cost (e.g., rural) cus-tomers at relatively low prices if it is permit-ted to offset the associated financial losses by

30 The difference between expressions (4) and (6) is[v(cL) v(cH)]/2 + q(cH)[12 ]. The convexity of v(.)implies this expression is at least [/2 ( 12)]q(cH).This expression is nonnegative because it is a decreasingfunction of and is nonnegative at = 1.

charging prices sufficiently above the costsof serving low-cost (e.g., urban) customers.Consequently, universal service (and otherdistributional) concerns can cause regulatedmonopoly to be preferred to unregulatedcompetition.

4.3 Available Instruments

The policy instruments available to over-sight agencies also influence the relative mer-its of monopoly and competition. Toillustrate, return to the simple setting consid-ered in section 3, where the regulator wasable to set prices and deliver transfer pay-ments. In practice, regulators are not alwaysable to deliver transfers to the firms they reg-ulate. If the regulator lacked this ability in thesetting analyzed in section 3.1 and wished toensure the monopolist never terminated itsoperations, the regulator could do no betterthan to set a single price equal to the highmarginal cost, p = cH (assuming fixed costs arezero). This policy would generate expectedwelfare

(6) v(cH) + 12q(cH).

It is apparent that the level of expectedwelfare in expression (6) is less than the cor-responding level in expression (4).30

Consequently, unregulated duopoly isalways preferred to this restricted form ofmonopoly regulation in the setting consideredin section 3.1.

More generally, if a regulator has limitedability to reward a monopoly supplier forsuperior performance and penalize the firmfor inferior performance, the regulator maybe unable to motivate the firm to serve con-sumers well. This is the case regardless ofhow well the regulator understands thefirms capabilities and consumers prefer-ences. Similarly, if the regulator is not



31 See Nicolas Curien, Bruno Jullien, and Rey (1998)and Laffont and Tirole (1990a), for example.

32 Of course, a country with limited ability to collecttaxes also may have limited ability to enforce desiredprices. The relative strengths of the countrys institutionsare important to consider in such cases. A country may bebetter able to measure electricity consumption andenforce associated charges than to prevent citizens fromhiding wealth in order to evade income taxes, for example.

authorized to compel the firm to report dataon its operations, the regulator will find it dif-ficult, if not impossible, to make informed pol-icy decisions, even if the regulator has ampleresources and ability to analyze and interpretdata. Consequently, limited powers to rewardor penalize the firm or to compel data report-ing can render monopoly regulation ineffec-tive, and so can increase the relative merits ofunregulated competition.

A regulators powers to control the activi-ties of new suppliers also can affect the rel-ative merits of monopoly and competition.To illustrate, suppose the regulator cannotimpose any regulations on new competitors.In particular, suppose a regulator cannotlimit the range of services that competitorsoffer or tax any of these services. In thiscase, even though liberalization may help tomotivate the incumbent supplier to reduceits operating costs, it may allow competitorsto engage in cream-skimming. Cream-skim-ming is the act of serving the most profitable(e.g., urban, business telecommunications)customers and leaving the incumbent sup-plier to serve the less profitable (e.g., rural,residential telecommunications) customers.Cream-skimming can limit the ability of theincumbent supplier to finance particularlylow prices on some services with substantialprofit earned on other services and therebyundermine socially desirable pricing struc-tures.31 Therefore, even though managedcompetition (which entails the regulation ofboth incumbent suppliers and new entrants)might be preferable to monopoly, monopolymay be preferable to unfettered competi-tion when an ineffective tax system requiresthat universal service be pursued throughindustry prices.32

33 For instance, Michael I. Cragg and I. AlexanderDyck (2003) find that the financial compensation of man-agers is not as closely linked to the performance of the firmin state-owned enterprises as it is in privately owned firms.

34 See Jnos Kornai, Eric Maskin, and Gerald Roland(2003).

35 William L. Megginson, Robert C. Nash, andMatthias van Randenburgh (1994), Juliet DSouza andMegginson (1999), Rafael La Porta and Florencio Lpez-de-Silanes (1999), Kathryn L. Dewenter and Paul H.Malatesta (2001), Megginson and Jeffry M. Netter (2001),Scott J. Wallsten (2001), and Simeon Djankov and PeterMurrell (2002), among others, present (sometimes mixed)evidence that the efficiency of SOEs increases after priva-tization, particularly in the presence of substantial industrycompetition and independent regulators.

36 Of course, in practice, this greater potential need notalways translate into more pronounced reductions inindustry costs. Fumitoshi Mizutani and Shuji Uranishi(2003), for example, report that liberalization in Japanspostal industry did not substantially increase the produc-tivity of the SOE.

4.4 Private versus State Ownership

The extent of government ownership ofthe dominant incumbent supplier also canaffect both the merits and the most appro-priate form of liberalization. A firm that islargely owned by the government can be lessresponsive to the oversight and demands ofshareholders than are privately ownedfirms.33 Furthermore, state-owned enter-prises (SOEs) can face softer budget con-straints than their privately ownedcounterparts in the sense that the govern-ment may be more willing to tolerate lossesby the firm and to finance the firms contin-ued operation despite poor historic financialperformance. Because of the correspondingdiminished incentive to minimize produc-tion costs,34 an SOE may operate with high-er costs than a privately owned monopoly.35

Therefore, liberalization may have greaterpotential to reduce industry costs when theincumbent monopoly supplier is owned pri-marily by the government than when it isowned primarily by private investors.36

4.5 Regulatory Independence andAccountability

The simple models considered in section 3presumed the regulator faithfully pursuedthe social objective. In practice, a captured



37 See George J. Stigler (1971), Richard A. Posner(1974), Gary S. Becker (1983, 1985), Laffont and Tirole(1993, chapter 11), and James F. Dewey (2000), for exam-ple, for models of capture and the strategic choice ofexpenditures to influence regulators.

38 See Yeon-Koo Che (1995) and David J. Salant(1995), for example.

39 Bengt Holmstrom and Paul Milgrom (1991),Holmstrom (1999), and Mathias Dewatripont, Ian Jewett,and Tirole (1999) examine the effects of long-term careerconcerns on short-term behavior. Such concerns can, forexample, induce a regulator to focus primarily on per-formance dimensions that are readily observed (e.g., short-term price reductions) and less on performancedimensions that are more difficult for potential employersto observe in a timely fashion (e.g., prospects for viablelong-term industry competition).

regulator may not do so. A regulator is said tobe captured by the firm he regulates whenthe regulator generally implements policiesthat further the interests of the firm at theexpense of the broader social interest.37

Many factors increase the likelihood of regu-latory capture. For example, a regulatoryagency with limited expertise and resourcesmay be forced to rely heavily on the adviceand information supplied by the firm whenformulating policy. Alternatively, the firmmay routinely offer attractive employmentopportunities to regulators who have provedto be cooperative, and the countrys laws maynot preclude such offers.38 Also, the firmmay provide a sizable portion of the regula-tors ongoing budget and may have some dis-cretion over the timing and magnitude of itscontributions to the regulators budget.When factors like these lead to a high likeli-hood of regulatory capture, the entry of addi-tional competitors may be the best way toimpose meaningful discipline on the incum-bent supplier and otherwise ensure long-term gains for consumers.39

The importance of regulatory independ-ence from short-term popular opinion also isapparent. Producers in network industriestypically incur substantial sunk (nonrecover-able) costs. These producers also often deliverservices to a large portion of the population,and so the prices of these services are of sub-stantial public concern. Together, thesetwo elements create substantial risk of

40 See Oliver Williamson (1975) for a pioneering treat-ment of the problem and Newbery (1999) for an extensivediscussion of the problem of regulatory commitment.Some authors (e.g., Witold J. Henisz 2000) have devel-oped indices to measure expropriation risk in differentcountries. Jamison, Holt, and Berg (2005) review theseindices.

41 Wallsten (2001) finds that the privatization of state-owned telecommunications providers in Latin Americaand Africa is associated with improved industry perform-ance when the industry regulator is independent (in thesense of not being directly under the control of a govern-ment ministry), but not otherwise. Geoff Edwards andLeonard Waverman (2006) find that regulatory independ-ence is associated with lower charges for access to the net-work infrastructure of state-owned incumbent suppliers oftelecommunications services.

expropriation by well-meaning but short-sighted regulators. In response to public pres-sure, regulators may reduce prices as far aspossible toward variable production costs. Aslong as regulated prices allow the firm torecover its variable production costs, the firmwill prefer to continue to produce than to ter-minate its operations. Thus, in the short run,the regulator gains by securing low prices andongoing production of key services. Althoughsuch a policy may provide short-term gains, itcan have substantial long-term costs. Potentialand actual producers will realize they areunlikely to recover any sunk costs they incur.Consequently, they will be reluctant to incursuch costs, and so existing network infrastruc-ture will be permitted to decay and new net-work infrastructure will not be built. Thus, ifregulators are to design and implement poli-cies that best serve the long-term interests ofconsumers, they must be able to develop pol-icy credibility by resisting short-term pressuresto renege on long-term promises.40

A regulators commitment powers can beenhanced by a variety of factors, includingstrong legal institutions and a long tenure.Strong legal institutions can thwart attemptsby other government agencies to intervene inthe day-to-day operations of the regulatoryagency and can thereby enhance a regulatorscommitment powers by reducing the likeli-hood that the terms of announced regulatorypolicies will be changed.41 In particular, stronglegal institutions can enforce long-term



42 Kessides (2004, p. 102) describes settings where gov-ernments have changed the terms of legal contracts.

43 See Newbery (1999), Salant and Glenn A. Woroch(1992), and Paul Levine, Stern, and Francesc Trillas (2005)for formal analyses of this effect. Levine, Stern, and Trillas(2005) also emphasize the potential value of delegatingauthority to an independent regulator with substantial con-cern for the welfare of the firms he regulates.

44 Timothy Besley and Stephen Coate (2003) employdata from the United States to examine the effects of themethod by which state regulators are selected on perform-ance in the states electric power industry. The authors findlower prices and reduced investment (as proxied by thefrequency of power outages) in states where regulators areelected than in states where regulators are appointed.

45 See Maskin and Tirole (2004) for a formal analysis ofthis trade-off.

contracts between the regulator and the firm,and can prevent a regulator from changing theterms of announced policies (including prom-ised returns on investment) in response topressure from other government agencies orthe citizenry at large. When legal institutionsare weak, pressure groups may anticipate sub-stantial benefits from convincing the regulatorto renege on the promises he has made to thefirm.42 Thus, the independence of a countrysjudicial system and the ability of a country toenforce the terms of legal contracts can affectthe optimal design of industry policy.

A regulator may also be better able to pur-sue policies (such as delivering promisedrewards to the regulated firm) that promotethe long-term, rather than the short-term,interests of consumers if his tenure as regu-lator is relatively long. Long-lived regulatorswith sufficient concern about future industryoutcomes will realize that short-term expro-priation of an incumbent producers invest-ment will discourage future investment bythe same firm or its successor.43 In addition,a regulator may feel less pressure to panderto popular opinion if he is not elected bydirect vote of the citizenry.44 Thus, a settingwhere the regulator serves a fairly long termand faces reappointment by the governmentmay provide an appropriate trade-offbetween independence and accountability.45

There is an additional benefit of a reason-able degree of regulatory independence fromdirect intervention by other government

46 In this respect, regulatory independence can deliverbenefits similar to those provided by the privatization of astate-owned enterprise. Privatization increases the cost tothe government of intervening in the day-to-day opera-tions of the firm, and thereby renders more credible thegovernments promise to refrain from such intervention(Sappington and Stiglitz 1987). Regulatory independenceenhances the commitment powers of regulators in similarfashion, once the mission and goals of the regulatoryagency are stated clearly.

47 Spiller and Vogelsang (1997) and Tonci Bakovic,Bernard Tenenbaum, and Fiona Woolf (2003) emphasizethe value of coupling substantial regulatory independencewith a clearly specified regulatory contract (e.g., a regula-tory license). Under such coupling, regulators will have theindependence required to implement impartially thedetails of a politically popular contract, but will not beempowered to implement any policy of their choosing.Using a large dataset of monopoly franchise auctions,Guasch (2004, table 1.16) reports that the original terms ofthe franchise contract were renegotiated 61 percent of thetime when there was no separate regulatory body respon-sible for contract administration. The corresponding per-centage was only 17 percent when the contracts wereadministered by a separate regulatory body.

48 The independence granted the telecommunicationsregulator in Jamaica is believed to have led to a decline inindustry investment as investors feared the regulatorwould use his autonomy to expropriate investments (Levyand Spiller 1994).

49 See Laffont and David Martimort (1999) andMartimort (1999b).

agencies. A government that is unable tointervene in the short-term operations of aregulatory agency is compelled to state asfully and clearly as possible the mission of theregulatory agency. Failure to do so will affordthe regulatory agency the opportunity to pur-sue its own mission and goals, rather thanthose of the government more broadly.46 Acoherent and transparent statement of theprinciples that will guide industry policy pro-vides greater certainty for industry partici-pants, which can encourage investment andfacilitate long-term planning.47

When regulatory capture is likely, explicitrestrictions on the regulators autonomy andcommitment powers may be desirable inorder to limit the regulators ability to pursueinterests other than the long-term socialinterest.48 For example, key powers mightbe dispersed among multiple regulators ordivided between a regulator and other gov-ernment agencies.49 Alternatively or in addi-tion, the regulators discretion in formulating



50 See Laffont and Tirole (1990b).51 Martimort (1999a) shows that it can be optimal to

impose more severe restrictions on a regulatory agencyover time as the potential for regulatory capture increases.

52 If competition undermines a regulators (otherwisesubstantial) powers to promise attractive returns to theincumbent monopoly supplier, it is possible that competi-tion may be inadvisable when the incumbent monopolist isinvesting and otherwise satisfying the long-term needs ofconsumers reasonably well. In extreme cases, unregulatedmonopoly might be the most effective way to deliverstrong investment incentives if the absence of regulation,itself, constitutes a meaningful commitment not to expro-priate a monopolists earnings. Notice that, although a reg-ulator may have particularly pronounced ability toexpropriate investors because of his direct control over abroad range of industry policies, other government entitiesalso may have substantial ability to expropriate investors byimposing excessive profit taxes or indirect taxes, for exam-ple.

53 Wesley M. Cohen and Richard C. Levin (1989)review empirical studies of the relationship between mar-ket structure and innovation.

policy might be substantially restricted.50

The tenure of individual regulators mightalso be limited.51

Competition can play a particularly valu-able role in disciplining and motivating theincumbent supplier to pursue the best inter-ests of consumers when the regulators pow-ers to do so are (inherently or intentionally)limited. However, competition is not neces-sarily a panacea. Weak commitment powersmay limit nonfungible investment underboth unregulated competition and regulatedmonopoly. Competitors, like the incumbentsupplier, will recognize that their invest-ments may be expropriated by a regulatorwith substantial expropriation powers andlimited commitment powers, and so may bereluctant to undertake the investmentrequired to improve industry perform-ance.52 Thus, there is seldom an effectivesubstitute for strong commitment powers.

4.6 Investment and Innovation

The link between industry structure andthe incentive to innovate and to reduce costsis complex even in the absence of regula-tion.53 Competing firms may have greaterincentive than an unregulated monopolyprovider to reduce operating costs in part

54 See Kenneth J. Arrow (1962). Of course, this argu-ment presumes that intellectual property protection pre-cludes competitors from immediately copying theinnovators discovery and thereby eliminating the financialgain from innovation.

55 This argument presumes that capital markets areimperfect. This is a reasonable assumption in the contextof innovation because innovators often are unable to con-vince investors of the merits of their potential innovationwithout revealing the details of the innovation (and there-by forfeiting some of the potential financial gains from theinnovation). See, for example, James J. Anton and DennisA. Yao (1994, 2002).

56 See Joseph Schumpeter (1950) and Glenn C. Loury(1979), for example.

57 Guthrie (forthcoming) provides a more completereview of the literature that examines the impact of regu-lation on infrastructure investment.

58 A regulatory policy that delivers no extra profit to afirm as its realized production costs decline effectivelyexpropriates any investment the firm might make in anattempt to secure lower production costs.

because industry output (and thus thepotential cost savings from a reduction inmarginal production cost) is greater undercompetition than under unregulatedmonopoly.54 In contrast, substantial marketconcentration can encourage innovation forat least two reasons. First, the profit amonopolist generates can serve as a valuablesource of research and development (R&D)funding.55 Second, the prospect of substan-tial monopoly profit can be a compelling rea-son to undertake R&D investment.56

Regulatory policy can affect infrastructureinvestment differently than it affects innova-tive effort and investment designed toreduce operating costs.57 To illustrate thispoint, first consider rate of return regulation,which promises a fair return on prudentlyincurred investment. When expropriationcan be avoided, such a promise can deliverstrong incentives for infrastructure invest-ment. In contrast, because it requires rev-enues to track costs closely, rate of returnregulation (like other forms of cost-plusregulation) typically provides limited incen-tive for innovation and cost reduction.58

Now consider price cap regulation, whichtypically permits revenues to diverge fromrealized costs for a specified period of time(e.g., four years) but does not promise



59 The regulatory policy implemented for British Gas in1986 (which permitted the average price of natural gas toincrease at the rate RPI + Y X) is an example of price capregulation. Sappington (2002) provides an overview ofprice cap regulation plans. Also see Jan Paul Acton andVogelsang (1989), Jordan Jay Hillman and Ronald R.Braeutigam (1989), Armstrong, Cowan, and Vickers(1994), and Jeffrey I. Bernstein and Sappington (1999), forexample. Notice that a promise to allow revenues todiverge from realized costs can play much the same role asintellectual property rights protection that allows an inno-vator to derive substantial financial gain from her innova-tion.

60 A. Michael Spence (1975) observes that, when qual-ity cannot be regulated directly, rate of return regulationmay provide stronger incentives to supply quality thanprice cap regulation. Sappington (2005a) reviews the liter-ature on service quality regulation in utility industries.

61 Regulatory policies that deliver a consistent rate ofreturn on investment also can reduce the variance ininvestors financial returns, and thereby reduce the regu-lated firms cost of capital.

62 Alternatives to these two forms of regulation alsomerit consideration. As Schmalensee (1989) notes, earn-ings sharing plans (which specify explicit rules for sharingthe firms realized earnings with consumers) can help toavoid extreme distributions of rent that can prove difficultto enforce in practice. Sappington (2002) provides a dis-cussion of earnings sharing plans and relevant referencesto the literature. Vogelsang (2005) explains how earningssharing policies might be coupled with other regulatoryregimes to provide appropriate short-term incentives forcost reduction and long-term incentives for investment.

specific long-term returns on investment.Although such a policy can provide substan-tial incentive for short-term innovation andcost reduction,59 it may provide limitedincentive for long-term infrastructure invest-ment.60 Therefore, the choice between rateof return regulation and price cap regulationwill depend in part on the type of investmentthat is most important to secure. In settingswhere the top priority is to induce the regu-lated firm to employ its existing infrastruc-ture more efficiently, price cap regulationmay be preferable. In contrast, in settingswhere it is important to reverse a history ofchronic underinvestment in key infrastruc-ture, rate of return regulation may bepreferable.61

The appropriate choice between rate ofreturn regulation and price cap regulationalso is influenced by industry volatility andregulatory commitment powers.62 As costsand demands change over an extended time

63 This is the case even if the price cap plan makesexplicit adjustments for changes in economywide outputprices and changes in key input prices.

64 Of course, if a regulator has considerable discretionover which of the firms investments qualify for the prom-ised rate of return, serious problems of opportunism canarise under rate of return regulation, just as they can ariseunder price cap regulation.

65 Guasch (2004, table 1.16) reports more widespreadrenegotiation of franchise contracts under price cap regu-lation than under rate of return regulation.

66 Harold Demsetz (1968) provides the seminal discus-sion of this observation. Laffont and Tirole (1987), R.Preston McAfee and John McMillan (1987), and Riordanand Sappington (1987), among others, provide formalanalyses of the optimal design of monopoly franchises.

period, prices and costs will invariablydiverge under price cap regulation, possiblyleading either to financial distress or to par-ticularly large profit for the regulated firm.63

Neither of these outcomes is likely to becredibly sustained in settings where thecountrys institutions are weak. Therefore, insuch settings, rate of return regulation maybe preferable to price cap regulation in thepresence of considerable industry volatility,particularly if infrastructure investment isdesirable.64 The cost-plus nature of rate ofreturn regulation, which ensures profits areneither excessive nor insufficient, can renderits implicit commitment to set prices thattrack costs closely more credible than theprice commitments encompassed in a pricecap plan.65

5. Franchise Bidding

Even when pronounced scale economiesrender direct competition in the market pro-hibitively costly, competition for the right toserve as a regulated monopoly supplier (inthe form of franchise bidding) can sometimescapture for consumers much of the surplusthat strong competition in the market wouldgenerate.66 To illustrate the potential value offranchise bidding, return to the simple modelconsidered in section 3 and suppose twofirms bid for the right to serve as the soleprovider of the product in question underterms specified by the regulator. Each firm is



67 The regulator can achieve the ideal (full-information)outcome in this setting if the firms costs are correlated.See, for example, Joel S. Demski and Sappington (1984)and Jacques Crmer and Richard P. McLean (1985).

privately informed about its own marginalcost of production (c {cL,cH}). The twopossible cost realizations again are equallylikely for each firm. Furthermore, for sim-plicity, the firms cost realizations areuncorrelated.67

The optimal franchise bidding policy in thissetting takes the following form. The twofirms report their cost realizations simultane-ously. If only one firm reports the low costrealization (cL), that firm is selected to be themonopoly supplier, sell the product at unitprice pL, and receive transfer payment TL. Ifboth firms report low costs, one of the firms isselected at random to operate the (pL,TL) con-tract. If both firms report high costs (cH), oneof the firms is selected at random to operateunder the (pH,TH) contract.

The contracts are optimally designed toensure the firms report their costs truthfully.Consequently, a firm with the low marginalcost will be selected to serve as the monop-oly supplier with probability 3/4. (Thus, fran-chise bidding secures the sampling benefitof competition.) A firm with the high mar-ginal cost will operate with probability 1/4.Truthful reporting by the low cost firm willbe a best response to truthful reporting bythe rival potential producer if

(7) 34 UL 14 [UH + q(pH)],

where Ui [pi ci] q(pi) F + Ti for i = L,H.The binding participation constraint will bethat of the firm with the high marginal cost,so UH is optimally held to zero. Inequality (7)will optimally be satisfied as an equality tolimit the rent afforded a firm with the lowmarginal cost. Consequently, expected wel-fare under the optimal pricing policy is:

(8) 34 [wL(pL) 13[1 + ]q(pH)]

+ 14 wH(pH).

68 This is the case because v(cH + [1 ]) v(cH) [1 ]q(cH). Therefore, expression (9) will exceed expression(4) if 2[v(cL) v(cH)] q(cH)[1 + ]. This inequality holdsfor all values of 1.

69 In practice, franchise auctions often take differentforms. For example, firms often are invited to specify aprice at which they are willing to supply the product inquestion to consumers, and the firm that bids the lowestprice might be awarded exclusive production rights. Thisauction, which involves no transfer payments from firms tothe government (or vice versa), typically will result in themost efficient firm serving consumers at a price thatreflects its average cost of production. Such an auctionresembles normal price competition but has the addedadvantage that it can permit profitable operation even inthe presence of scale economies. Alternatively, the supplyprice might be determined in advance and firms could bidon the amount they will pay the government (or the pay-ment they will require from the government) for the right(and obligation) to provide the service at the stipulatedprice.

Expression (8) reveals that pH is optimallychosen to maximize wH(.) [1 + ]q(.),while pL is optimally chosen to maximizewL(.). Therefore, prices in this franchise bid-ding setting are the same prices that areimplemented in the regulated monopoly set-ting. (Recall expression (2).) The transferpayment (TL) to the low-cost supplier isreduced in the franchise bidding setting,though. The firm will reveal its superior capa-bilities despite being promised a smallertransfer payment because cost exaggerationentails substantial risk of being excluded fromthe industry. (Thus, franchise bidding securesthe rent-reducing benefit of competition.)

Because franchise bidding and monopolyregulation implement the same price for agiven cost realization, total expected welfareis the same under the two regimes except thatthe likelihood of a low-cost supplier increasesfrom 1/2 to 3/4. In the case where there is nocost of social funds (so = 0), expected wel-fare in the franchise bidding setting is:

(9) 34 v(cL) +14 v(cH + [1 ]).

It is apparent that expression (9) exceedsexpression (5). Expression (9) also exceedsexpression (4) when = 1/2.68

In summary, franchise bidding outper-forms both monopoly regulation and duop-oly competition in this simple setting.69 It



70 This possibility is known as the winners curse. SeeVijay Krishna (2002, pp. 8485), for example.

71 The terms of the franchise contract can be altered tocounteract the problems that arise when bidders have lim-ited financial resources. For example, rather than bid on alump-sum franchise fee, potential operators might bid onthe fraction of realized profit (or revenue) that they arewilling to share with their customers. See Tracy R. Lewisand Sappington (2000) for details. Che and Ian Gale(1998, 2000) provide related analyses.

does so by employing both transfer pay-ments and price regulation to pursue socialgoals by ensuring the benefit of scaleeconomies through the selection of a singlesupplier and by securing the sampling andrent-reducing benefits of competition.

More generally, franchise bidding canhave its drawbacks. If potential operatorsdo not have comparable skills, accurateinformation, and substantial financialresources, bidding for the right to serve asthe monopoly supplier may not be intense.Firms will have little interest in bidding fora contract they expect to lose to a morecapable operator. In addition, a firm will bereluctant to bid aggressively on a contractif it may win the contract not because it isthe most capable producer but because itunderestimates most severely the true(common) cost of operating the fran-chise.70 A firm may be similarly reluctantto bid for a contract when it suspects otherbidders (e.g., an incumbent provider of theservice in question or a firm that providessimilar services in another geographicregion) have better information regardingthe financial returns the contract is likelyto provide. Furthermore, a firm cannot bid more for a contract than its financialresources allow.71 When pronounced dif-ferences in skills or information and/or adearth of qualified bidders with substantialfinancial resources limit the intensity offranchise bidding, such bidding will not capture for consumers the surplus they would enjoy if multiple, well-informed firms with similar capabilities

72 A decision to exclude foreign investors from the bid-ding process (to promote nationalism or to further nation-al security, for example) can limit substantially theintensity of franchise bidding. Such exclusion also canreduce the value of the franchise since the skills andexpertise of the successful bidder affect the profit theenterprise can generate. Roman Frydman, Cheryl Gray,Marek Hessel, and Andrzej Rapaczynski (1999) report thatenterprise revenues increase substantially when an SOE issold to outside managers, but not when it is sold to indi-viduals who managed the SOE before its privatization.

73 Guasch (2004) analyzes more than 1,000 franchiseauctions in Latin America and the Caribbean between1985 and 2000. He finds that more than 50 percent ofelectricity franchise contracts and 75 percent of waterfranchise contracts were renegotiated. The average timebetween franchise award and renegotiation was approxi-mately two years. Renegotiation was initiated by the cho-sen operator more often than it was initiated by thegovernment.

74 When all relevant dimensions of performance are notspecified clearly in a regulatory contract, a potential sup-plier may be able to bid the highest franchise fee, notbecause it is the least-cost supplier of the services in ques-tion, but because it will deliver the least on all of the per-formance dimensions that are not specified in theregulatory contract. Alejandro M. Manelli and Daniel R.Vincent (1995) identify conditions under which considera-tions of this sort render it optimal for a regulator to nego-tiate a contract with a single selected supplier rather thanto award the franchise to the firm that bids the most tooperate under a contract that does not specify fully all rel-evant dimensions of performance.

competed against each other in the marketplace.72

Franchise bidding also may provideopportunities for the selected producer tohold up the government. After it is award-ed the monopoly franchise, the chosen pro-ducer may attempt to renegotiate the termsof the original contract in order to securemore favorable terms.73 The governmentmay be susceptible to renegotiationdemands in order to avoid the appearance offailure in the procurement process or toavoid substantial transaction costs associatedwith reauctioning the franchise.74

Franchise bidding also can fail to provideideal incentives for investment. If the dura-tion of the franchise contract is short relativeto the useful life of a desirable sunk invest-ment, the chosen supplier may be reluctantto undertake the investment if the firmstenure as the monopoly supplier is likely to



75 See Williamson (1976), for example.76 Robin A. Prager (1989), Mark A. Zupan (1989a,

1989b), and Yasuji Otsuka (1997) present evidence whichsuggests that these difficulties with franchise competitionare not always insurmountable in practice.

77 See Laffont and Tirole (1988). Investment also mightbe encouraged by adopting technologies that employ morefungible physical assets (e.g., wireless rather than wirelinefacilities in the telecommunications industry).

end before it can fully recover the cost of theinvestment. Consequently, the chosen sup-plier may adopt an inefficient productiontechnology (one that employs an undulysmall level of sunk costs) and all potentialsuppliers may reduce their bids for the rightto serve as the monopolist.

Although long-term contracts can, inprinciple, help to overcome this problem,long-term contracts seldom are a panacea.In practice, it typically is impossible todelineate all relevant contingencies in acontract.75 Furthermore, rigid contractscan preclude valuable adaptations to chang-ing industry conditions. In addition, eventhe most carefully crafted long-term con-tracts may not be enforced if the prevailinglegal institutions are weak. For all thesereasons, even long-term contracts typicallyare unable to deliver ideal investmentincentives. In some settings, it may be pos-sible to adjust franchise bidding policies tocounteract some of these problems.76 Forexample, auction rules that favor theincumbent supplier can enhance theincumbents incentive to undertake sunkinvestments in the presence of short-termcontracts.77

In summary, auctions for the right to bethe sole supplier can help to limit monopolyrents and to select the most efficient indus-try supplier. However, franchise auctionsseldom eliminate the need for regulation.Moreover, the same strong regulatory insti-tutions that are necessary for effectivemonopoly regulation are required for thesuccess of franchise bidding contracts.When these institutions are present,

78 Regulators can sometimes benefit by ensuring thatalternative suppliers are available to replace the incum-bent supplier as needed. Although it can be costly to main-tain a second source of production, the ability to readilyshift some or all production to the second source canimpose useful discipline on a monopoly supplier. See, forexample, Rafael Rob (1986), Anton and Yao (1987),Demski, Sappington, and Speller (1987), Riordan (1996),Riordan and Sappington (1989), James D. Dana, Jr. andKathryn E. Spier (1994), and Anton and Paul J. Gertler(2004).

though, franchise bidding can constitute auseful additional instrument that regulatorscan employ to select and discipline soleproviders in network industries.78

In concluding this section, we note thatyardstick competition can secure many ofthe same benefits as franchise bidding insettings where different monopolists oper-ate in distinct geographic markets. (Forexample, different water distribution com-panies often serve different regions of acountry.) In such settings, each monopolistmight be compensated on the basis of howits performance compares to the perform-ance of other monopolists. For example,the compensation delivered to each firmmight be set equal to an index of the real-ized costs of the other firms, rather than itsown costs. This and other forms of yardstickcompetition can provide strong incentivesfor efficient performance by all monopolistswhen they are known to operate in similarsettings. When the firms operate in envi-ronments that differ substantially (in geo-graphic area, terrain, weather, orpopulation density, for example), explicitcorrections for relevant differences can beimportant. Such corrections typically willbe necessary to avoid compensation that isunduly generous for some firms and undu-ly meager for others. Appropriate handi-capping can be difficult in the presence oflimited information about the precisenature of the variation in the firms operat-ing conditions. However, some relativeperformance comparisons generally can



79 See Andrei Shleifer (1985), Joel Sobel (1999), RogerCarrington, Tim Coelli, and Eric Groom (2002), andMehdi Farsi, Massimo Filippino, and William Greene(2005), for example. Armstrong and Sappington (forth-coming, section 4.1) provide further discussion and refer-ences. Notice that the entire benefit of yardstickcompetition arises from the rent-reducing impact of oper-ation by multiple firms. Yardstick competition does notadmit a sampling benefit because the firm that supplies agiven market is fixed exogenously.

80 This discussion draws in part from Armstrong,Cowan, and Vickers (1994, chapter 4).

help to discipline and motivate monopolysuppliers.79

6. Entry Assistance and Anti-CompetitiveLiberalization Policies

The foregoing discussion suggests thegreatest potential gains from competitionwill tend to arise when: (1) industry scaleeconomies are limited relative to consumerdemand; (2) the industry regulator has limit-ed information, limited resources, and limit-ed instruments with which to craft policy; (3)the regulators commitment powers are lim-ited; and (4) subsidization of the consump-tion of some of the dominant suppliersservices is either not critical or can beachieved by means other than regulating thesuppliers price structure.

The design of liberalization policy is ofparamount importance in settings wherecompetition is deemed to be a superioralternative to a prevailing monopoly regime.The purpose of this third segment of thepresent essay is to discuss the principles thatunderlie the design of sound liberalizationpolicy.80 After considering the subtle issue ofexplicit entry assistance, this section focuseson liberalization policies that can hinder(rather than promote) vibrant, long-termindustry competition. Section 7 reviewspreferable competition-enhancing liberal-ization policies.

It is important to emphasize at the outsetthat the discussion in both this section andthe next necessarily entails some subjective

judgments. Furthermore, even though thepolicies considered in this section generallyare not recommended, some of the policiescan, in theory, enhance welfare in certainsettings if the regulator is particularly wellinformed. Therefore, definitive, unequivocalconclusions about liberalization policies aredifficult to draw. The ensuing discussion isintended to provide general guiding princi-ples (to the extent possible) rather than pre-cise, comprehensive prescriptions anddefinitive conclusions.

Successful liberalization is seldom as sim-ple as removing all legal restrictions on entryinto the regulated industry. Entrants facemyriad economic barriers to

Armstrong Sappington Regulation Competition

Documents

Transcript of Armstrong Sappington Regulation Competition