Sie sind auf Seite 1von 24

The RAND Corporation

Trigger Price Regulation Author(s): David J. Salant and Glenn A. Woroch Reviewed work(s): Source: The RAND Journal of Economics, Vol. 23, No. 1 (Spring, 1992), pp. 29-51 Published by: Blackwell Publishing on behalf of The RAND Corporation Stable URL: http://www.jstor.org/stable/2555431 . Accessed: 14/05/2012 17:17
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Blackwell Publishing and The RAND Corporation are collaborating with JSTOR to digitize, preserve and extend access to The RAND Journal of Economics.

http://www.jstor.org

RAND Journal of Economics Vol. 23, No. 1, Spring 1992

Trigger price regulation


David J. Salant* and Glenn A. Woroch *

We consider the difficulty of achieving efficient prices and investments when returns on a public utility's projects are vulnerable to opportunisticratemaking. We model the long-term relationship between a firm and its regulator as a time-dependent supergame in which the regulator sets price ceilings to maximize surplus and the firm invests to maximize profit. Wefind history-dependent strategiesthat supportself-enforcing, mutually beneficialequilibria. Equilibriumpayoffs are close to the planning solution provided interestrates are small enough and capital depreciatesfast enough. We concentrate on "triggerprice regulation" where, in response to inefficient behavior, the regulator cuts price down to operating cost and thefirm curtails investment. This mechanism performs well even withproduction economies and with restrictions on players' threats.

1. Introduction * A long-term relationship develops between a public utility and its customers when each must commit resources for an extended period. The firm invests in specialized, immobile plant and equipment while utility customers make complementary investments of their own. Several authors observed that during the course of the relationship, returns on such investments are vulnerable to "opportunistic behavior." I The firm may refuse to expand capacity to meet growing demand or even to replace worn-out plant, allowing service quality to deteriorate. Attempting to serve the interests of consumers, a regulatory agency may disallow recovery of "imprudent" investments or engage in regulatory lag and excessively slow depreciation.

* GTE Laboratories Incorporated. We are grateful to Tom Lyon, Stan Reynolds, Bob Rosenthal, and Ingo Vogelsang for detailed comments. Discussionswith Jim Friedman, Roy Radner,and Mike Toman were also valuable.Bill Taylorgave helpful comments on an earlierversion that was presented at the biennial conference of the InternationalTelecommunications Society, June 1988 in Cambridge, Mass. The views expressed in this article are those of the authors and should not be attributedto GTE Corporation or any of its subsidiaries. ' Williamson ( 1975 ) and Klein, Crawford,and Alchian ( 1978 ) study the incentive to appropriatequasi-rents in a world of perfect information. Opportunism also arises when, despite promises to the contrary, a regulator incorporatesthe firm's private information about technology and demand revealed through its operation. For details see Baron (1988).

29

30

THE RAND JOURNAL OF ECONOMICS

Investment incentives might be restored if actions of the firm and the regulator are policed by outside parties. Legislaturesand courts do constrain behavior of the participants, but limited information makes their control highly imperfect. Whatever their range of discretion, all parties will continue to pursue immediate gains at the expense of long-term efficiency. The length of the firm-regulatorrelationship plays a crucial role in this story. With a finite horizon, agreements will invariably unravel because the temptation for opportunism just before the relationship dissolves is too strong. By induction, the same holds for each of the preceding periods. The conclusions change when the relationship lasts indefinitely. In that case, results from the study of repeated games suggest that the firm and its customers can secure the mutual benefits forgone through shortsightedness. The longer the relationship lasts, the greaterthe potential losses and, therefore, the greaterthe rewardsto continued cooperation. If players can credibly threaten to punish each other severely, then most any outcome superior to the static equilibrium could be sustained without outside intervention. Applying the insight of this Folk Theorem to the firm-regulator relationship is the principal goal of this article. We seek price and investment plans that are mutually beneficial for investors and consumers and at the same time self-enforcing. To illustrate how this logic applies to regulation, consider a simple example in which a one-time sunk investment will supply a service at any scale.2 Assume away physical depreciation and demand growth so that no additional capital outlays are necessary. Left to its own, the firm will undertake this venture only as long as monopoly pricing over its lifetime covers its capital cost. The dead-weight loss that arisesjustifies some form of rate regulation. First-best prices should equal marginal cost, which in this case is zero. To recoup the initial outlay, secondbest prices call for a constant mark-up each period. It is unclear whether a regulatory agency will adhere to these prices as time unfolds. Staff turnover and political crises weaken its resolve to fulfill past promises. The agency can please current consumers by cutting price, perhaps to a level that just keeps the firm in business. Exercising a little foresight, a profit-seeking enterprise will never invest in the first place. Prospects for better outcomes improve when this relationship continues indefinitely. At the extreme, a regulator must consider the loss to an infinite generation of consumers resulting from the firm's reaction to a price cut. This sum can be large relative to the efficient level if the weight attached to distant generations is not too small. In the case of a perfectly durable, lump-sum investment, the regulator does not care because the firm will continue to produce at any nonnegative price. Investment need not be so lumpy, however. When capacity can be added smoothly, the firm and regulator could agree on a candidate price-investment path. By accumulating capacity gradually, the firm holds back some investment. Should it ever reach the efficient capacity, it would lose all leverage when capital is perfectly durable. To deter opportunism by the regulator, it must approach a limit that falls short of the efficient level. In this way, the firm and regulator strike a balance between the benefits of continued expansion and the temptation to appropriate returns on existing capacity. When capital depreciates, repeated investments are needed to replace worn-out plant and equipment-even if investment is lumpy. Then, whether the efficient solution could be supported as an equilibrium depends on the speed of depreciation. The faster capital depreciates, the less the regulator can gain from confiscatory rates. Thus, neither the de-

This case is examined in closer detail in Salant and Woroch ( 1991 ).

SALANT AND WOROCH

31

preciation rate nor the discount factor can be too small if the firm and regulator are to voluntarily cooperate. We model this relationship as a time-dependent supergame played by a profit-maximizing enterpriseand a surplus-maximizingregulator.Startingwith an initial stock of capital, the utility invests in durable plant and equipment to provide a single service. Capacity additions incur an adjustment cost that favors gradual accumulation. Furthermore, all capacity costs are assumed to be sunk. As a benchmark, we solve the planning problem of maximizing discounted consumer surplus while ensuring the firm breaks even. The solution is a dynamic version of the Ramsey-Boiteuxpricing rule together with an investment path approachinga limiting capital stock. Much of the article is concerned with how closely this solution can be approximated by a noncooperative equilibrium. We first look at the set of equilibria when the firm and regulator must each commit at the outset to complete investment and price paths. We find that the rigidity imposed by such commitments prevents any benefits from cooperation being attained: at the unique equilibrium the regulator never allows any profit, so no investment ever occurs. We then consider the possibility that both sides can, at any date, alterprice or investment plans. In that case, an equilibrium exists in which the regulator follows the efficient price path as long as the firm invests efficiently. Any deviation "triggers" a punishment: the regulatorcuts price down to operating expense and the firm curtails investment. In general, the more that players can "punish" each other in this way, the greater is their incentive to cooperate. Indeed, the lower the regulator can set price and the more the firm can divest, the larger is the set of equilibrium outcomes that trigger strategies can sustain. In particular, we find that, for a sufficiently small interest rate and depreciation rate, trigger strategies support the planning solution. A linear example illustrates the triggermechanism and partially characterizesthe equilibrium set. The example confirms that the effectiveness of trigger price strategies depends on values of initial capacity in addition to the interest and depreciation rates. Several novel implications follow from our study of the firm-regulatorrelationship that apply to the design of regulatory institutions. A first step toward cooperation is the creation of conditions conducive to free and active bargainingbetween the firm and regulator. Next, by raisingthe damage from defecting from the negotiated outcome, regulatoryreform makes better outcomes possible. This argues for such unorthodox prescriptions as a relaxation of the ban on confiscatory rates and the removal of minimum standards for service quality.

2. The model
* Production and consumption occur at an infinite number of discrete points in time. The firm produces a single, perishable product with a known technology. Except for the capital accumulation equation, demand and cost conditions are unchanging and separable
across periods.3 0 Demand. Let Pt and qt be the actual price and quantity in period t. Price is bounded . above by a ceiling ^t Stationary demand is Q(p), from which we derive the inverse demand P(q). Consumer surplus is

S(p, q)

JP(x)dx-pq.

(1)

I The analysis can accommodate systematic demand growth or technological progress or even small perturbations around the stationary path (see Salant and Woroch ( 1992)). Arbitraryswings in demand alter the nature of the planning solution.

32

THE RAND JOURNAL OF ECONOMICS

When the price charged exactly equals the demand price (i.e., p = P(q)), (1) becomes the usual Marshallian consumer surplus, S(q) S(P(q), q), where S'(q) = -qP'(q) and S"(q) = - [P'(q) + qP"(q)] are both assumed to be nonpositive.
0 Production and costs. Before production can begin, the firm sinks an entry cost of F that is independent of the scale of operation. This figure includes expenditures on planning, site preparation, and any necessary permits and licenses. Production requires variable factors and durable capital. For expositional simplicity we assume a constant unit variable cost of v ? O.' Later we allow average variable costs to fall. Either way, output is constrained by the amount of installed capacity: qt ? qt. The firm begins with initial capacity & ? 0, which depreciates at a geometric rate 5.If gross investment in period t is I,, next period's capacity is just qt = It + q't-1, where y = I - 6 is the rate at which capital survives. Investment is irreversible (i.e., I, 2 0). Construction costs in period t are C(It), where C(0) = 0 and C(* ) is a convex, increasing function. To rule out cases where investment is never worthwhile, we require that, when starting from scratch, the marginal benefit of capacity exceeds its marginal cost: S'(O) > C'(O). Convexity of construction costs is not inconsistent with natural monopoly.5 Although unit costs of building capacity may be rising, its durability reduces cost of production in the future. Such scope economies give an incumbent firm an advantage in production over multiple periods. Furthermore, the sunkness of investment confers a first-mover advantage that may be insurmountable by subsequent entrants. Combined, sunkness and durability of capital create a natural monopoly that warrants rate regulation.

o The objective functions. All parties-the firm, consumers, and the regulator-use a constant market interest rate r to discount the future by the factor p = 1/(1 + r). The firm's owners are solely interested in the stream of investment returns. In the short run, the cash flow is just 7rt= (pt - v)qt - C(It). Over the long run, they evaluate the firm's performance according to discounted profits:
00

ll(p,

q,

A;

r, 8)
-

= z
t=

pt((pt
A

- v)qt -

C( q~t-

yt-i)),

(2)

where we have made the substitution It = t- yqt-. Net benefits to consumers is given by discounted consumer surplus:
00

S(p, q; r)

z ptS(pt, qt).
t= I

(3)

As a representative of consumers' interest, the regulatory agency adopts maximization of this measure as its objective. 3 The planning solution. Here we describe and partiallycharacterizethe second-best priceinvestment path. The planner's problem is to maximize the surplus Q subject to the firm's

Since variable cost is independent of capacity, the model must be modified to accommodate cost-reducing innovation. This could be done by letting v depend on cumulative R&D expenditures. This is part of the subject of the companion article by Salant and Woroch ( 1992). S Allowing for the possibility of increasing returns to scale in construction would complicate the analysis in at least two ways: (i) timing of investment would matter, and (ii) the minimum size of a capacity increment relative to depreciation and discount rates would have to be small enough for triggerpolicies to work. We believe several important classes of natural monopoly are captured with the given assumptions that allow increasing returns in production and unrecoverable set-up costs.

SALANT AND WOROCH

33

participation constraint: II 2 llo. The term llo includes the opportunity cost of the nextbest venture. As written, the planner places no explicit weight on the firm's profit. This formulation sacrificesno generality,however. First of all, a "fair"returnon investment could be imbedded directly into 11o. Second, by varying ll1, the set of second-best optimal investment plans can be derived. As long as the firm earns at least llo, it will voluntarily enter this industryeven with the prospect of regulation. When H1o exceeds unregulated monopoly profits, however, a subsidy must be paid for the firm to embrace regulation. A solution to the planning problem (denoted by *) consists of a sequence of capacities, price caps, and quantities { qt*, p^t*, } that solves q* maximize 4V(p,q; r) subject to ll(p, q, q; r, 6) 2 11o+ F
Pt < P(qt),
qt < O~tt qt

(PP)

Yqt-l

for each t given an initial capacity No. Choice of price Pt has been suppressed because the regulator does not gain from departing from the price ceiling Pt. Furthermore, a simpler version of (PP) in which price exactly clears the market (i.e., 3t = P(qt)) yields the same solution.6 A solution to (PP) occurs at a singularity of the Lagrangian:
00

?C =
t=1

pt{S[P(qt),

t] + X[[P(qt)

v]qt- C(qt-

yqIt-)] + /t(qt-

4t)}.

(4)

A solution to the problem exists and satisfies the usual Kuhn-Tucker conditions (see Luenberger ( 1969)). Necessary conditions for closely related welfare problems have been studied on various occasions.7 When there is no excess capacity (i.e., qt = t), the first-orderconditions reduce to a period-by-period, inverse-elasticity rule:
Pt- Mct Pt q1
Et

(5

where,as usual, Et

-P(qt))/P'(qt)qt

is the price elasticity of demand and


=(X -

1)/XE

(0, 1)

is the "Ramsey number." Note that the value of X, and likewise 7, derives from the profit constraint and, hence, depends on both a and r. Here MCt is a dynamic marginal cost:
MCt = v + C'(It)
-

ypC'(It+1).

(6)

The first two terms in (6) are simply marginal operating and investment costs; the last term accounts for the savings of reduced investment in the following period from an additional unit of current investment, after properly discounting and deducting for depreciation. At the planning solution, capacity approachessome limiting level for any initial capacity. This limit capacity,j(r, 6, 11o, &o),will dependon IIo, r, and 6 as well as the initial level,
6 Had we included the constraint p, < P(q,), the corresponding Kuhn-Tucker conditions would be vP[P(qj)- pt] = 0, where v, is the associated multiplier. If p, < P(q,), then the complementary slackness conditions imply pt(Xqt -qu) = vP'(q,) = 0, and so X = 1. When X = 1, pt drops out of all the first-orderconditions and reduces to those derived from (4). Note too that when X = 1, setting unit price p, differentfrom P(qu), and allowing the planner to make lump-sum transfers in the amount of [p, - P(q,)]q, from investors to customers would not increase the value of the Lagrangian. 7Brock and Dechert ( 1985) establishthe necessity of a nearly identical condition. In their dynamic investment problem, output is always equated to capacity.

34

THE RAND JOURNAL OF ECONOMICS

q0. When initial capacity is small, the planning solution calls for capacity to gradually increaseover time. But when initial capital stock is large,additional investment is unnecessary until capacity depreciates to near its limit value. We derive other properties of the planning solution as needed below.
O Steady states. The system is in steady state when capacity is unchanging, so that investment exactly replaces depreciated capital: It = bqt-I for all t 2 1. Bars over the variables indicate steady-state values: q^t= c, pt = f = P(,j), and so forth. Many initial capacities can be maintained as steady states, provided capacity starts at that level. Of these, many also generate revenue that covers investment, production, and opportunity costs, i.e., any q such that [P(q) - v]q - C(aq) 2 r(Ho + F). The largest such capacity provides an important benchmark, since it maximizes consumer surplus over all feasible steady states:

maximize S(q) subject to [P(q) - v]q - Cbq)


?

r(IIO+ F).
-

(SS)

Let q*(b, r, Hlo+ F) be the solution to (SS). As r -O 0,jqj*(b,r, IIo + F) q*(b) and its associated price, f*(b) = P(q*(a)), satisfy P*(b) = v + C[bq*(b)]/q*(b).

*(a), where
(7)

A consequence of Lemma 1 and the proof of Corollary 2 is that the solution to planning problem (PP) converges to the single steady state (ft*(b), j4(b)) as r -O 0 and t -s 00 for any initial qc and IIo + F. Call the path pt = ft*( b), qt = j*( b) the steady-state planning solution. When H1o Fare both zero, this is the surplus-maximizing, steady-stateoutcome and for any positive r. When the system does not start in a steady state, the planning solution has capacity settle down to the limit capacity qj(b,r, H1o, Its value is closer to the steady-stateplanning do). solution qi*( 6) the closer the interest rate r is to zero. For any fixed r, the limit capacity is increasing in the initial capacity c%. 3. The strategic approach

* The planning solution presumes the regulator has direct control over production and investment. Invariably,a privatelyowned and managed firm prefershigher prices and smaller capacities than the levels dictated by efficiency. Therefore, we inquire whether it is possible for the regulator to sustain the planning solution when it sets price ceilings while the firm controls investment and production.

o Strategies. Assume that, when selecting their actions, the firm and the regulator enjoy perfect recall and perfect foresight. At the outset, the firm can elect not to enter the market by withholding its resources from production. At the beginning of each period, the firm invests in capacity and the regulator sets a price ceiling. Figure 1 diagrams the extensive form of the game. For much of what follows, these decisions are made simultaneously. Simultaneous moves reflect an information lag: neither the firm nor the regulatorcan react instantaneously to unexpected deviations by the other. This lag allows either party to "cheat" by deviating from any anticipated path before it becomes known to the other. In the meantime, the firm can earn a one-time profit by reducing, or perhaps ceasing, investment. The regulator can augment its surplus for a single period by lowering the price ceiling. Given a price ceiling and installed capacity, price-quantity pairs must meet three conditions:(i) pricecannotexceedthe ceiling,(ii) quantitycannotexceed capacity,and (iii)

SALANT AND WOROCH FIGURE1 EXTENSIVE FORMOF THEGAME F Stay out Enter R
'I0

35

Pi, R P2 F

GA
/ S

quantity cannot exceed demand realized at the chosen price. All three constraintsare satisfied in the shaded regions of Figure 2. These constraints still leave a wide range of feasible price-quantity pairs. Each pair yields a different payoff to the two parties. We resolve the indeterminancy by selecting the price-quantity pair that maximizes short-run profit:

[p(

q,q)

qU,)] = argmax {(p - v)q: p

p, q < , q Q(p)}

(8)

We could have used other rules to select a feasible pair, but this one has several attractive features. First, the firm's apparent discretion over short-run production and pricing makes sense when it possesses superior or more timely cost and demand information. This information asymmetry makes direct control impractical for a regulator. Furthermore, we can show that if price and quantity were at the firm's discretion, the firm would nevertheless maximize short-run profit as in (8).'
8 The rule also limits the price-quantity pairs that are realized. Implicit in (8) is the assumption that the firm will operate at full capacity as long as the regulator'schoice of a price cap allows it to cover its operating expenses.

Typically, regulated firmsareguaranteed some minimalrateof return, so the firmwill shutdownwhenprice and

36

THE RAND JOURNAL

OF ECONOMICS

2 FIGURE PRICEANDQUANTITY DETERMINATION FINAL OF

PX
q

iL~~~~~~~~~
D~~~~~~~~~
q

(a)

(b)

Strategies are rules that determine actions at each date as functions of the history up to that time. The history of the game up to period t is ht = (qoP3l, cil
, Pt-,,
qt-1)

(9)

A strategy for the regulator is a sequence of functions 0 each period t as a function of the history At: ht At E

{ /t } specifying a price ceiling in


9
i

1* (10) a capacity

Similarly, a strategy for the firm is a sequence of functions for each t as a function of the history
A1t: ht

{ At} specifying

>t

'y

_It E 9.(1

Notice how the assumption of irreversible investment constrains the firm's actions in each period. 0 Equilibria. A subgame is the game remaining given a history ht. A subgame perfect equilibrium is a pair of strategies (-/*, 1*) in which, given any period t 2 1 and history ht, (i) 4j* (restricted to the subgame) maximizes the firm's discounted profit from period t on when it expects the regulator to use 0*, and (ii) k* (restricted to the subgame) maximizes the discounted surplus from period t on given the firm is using 41* to set capacities. A steady-state equilibrium is a strategy pair, (/*, 4t*), satisfying (i) and (ii) above such that capacities and price ceilings are constant along the equilibrium path: = P(*) for each t, where h* denotes histories in ,*(ht*) = cjo = * and /*(ht*) =
A

fails to cover at least some portion of its capital costs. Additionally, the firm may be able to reduce output below capacity when the price ceilings do not allow adequate return on investment. Such considerations can easily be included by reinterpretingv as the smallest mark-up over operating costs that will be acceptable to the firm and a as the fraction of capacity the firm will use when it can no longer earn an adequate return on its capacity.

SALANT AND WOROCH

37

= which qs = q* and i3s 5* for all s < t. Capacity remains at its initial level do at a steadystate equilibrium. Our formulation of the firm-regulatorrelationship has a couple of features that set it apart from other applications of supergames. First, due to the irreversibilityof capital, the sequence of constituent games is time dependent: each new play of the game depends on the history of play through the current capital stock.9 Also, an inherent asymmetry exists between the two players. Unlike oligopoly models, for instance, the strategiesand the payoffs of the firm and regulator are qualitatively different.'0 4. Extreme equilibria
* This section characterizes equilibrium outcomes of the game, focusing on the one that maximizes surplus. We show how optimal outcomes depend on interest and depreciation rates. We also demonstrate how the difference between the surplus-maximizing equilibrium and the planning solution vanishes as the interest rate goes to zero. The information lag that forces the firm and regulatorto move simultaneously prevents the planning solution from being an equilibrium outcome. Only when the regulator can observe the firm's investment before setting price ceilings is the planning solution an equilibrium. We begin by finding the equilibrium when players are restrictedto open-loop strategies. In that case their actions at each date are independent of history, except for the irreversibility constraint. There are two reasons for considering this game. First, as in Reinganum and Stokey (1985 ), commitment can be measured by the length of time over which players are unable to deviate from planned actions. Open-loop strategies represent an extreme form of commitment, since players are never able to adjust their actions after the initial period. Second, a unique equilibrium in open-loop strategies exists for any subgame. When closed-loop strategies are available-so actions at each date can depend on the history to that date-the open-loop equilibrium continues to be an equilibrium. Furthermore, it is the minmax point of the game. This means that the joint minmax strategies are an openloop equilibrium in any subgame: neither side would ever have any incentive to deviate from these strategies. i3 The open-loop game. An open-loop strategy for the regulator is a sequence of price ceilings {f1t:t = 1, 2, ..., o } with fit2 0 for each t. Likewise, an open-loop strategy for the firm is a sequence of capacities {cit: t = 1, 2,.. ., o } with cit2 ycqt- for all t. For any initial capacity q0, this game has a unique Nash equilibrium: Pt = v and 4t = 'oy for each t. = It is easy to verify that these strategies form an equilibrium. Let fit v for each t. Then the firm will never wish to invest, because it will not earn a return on its investment.
Obversely, if ct = y q , the regulator forgoes some surplus if price exceeds unit cost.

To see that no other equilibria exist, suppose that for some r > 0 we have T> v. Then the regulator can increase its period r payoff S(p37, qj) by simply reducing j3T.On the other hand, when j. = 0, reducing fit to v has no effect on its payoff. More succinctly, A = v is a
9 Gilbert and Newbery ( 1989) construct a supergame model of the firm-regulatorrelationship without time dependence. In Benhabib and Radner ( 1992), time dependence arises from a renewable resource. Our approach is based on a variation of Friedman ( 1988). 10In this respect our setup is similar to Fernandez and Rosenthal ( 1990), who model sovereign-debt renegotiation as a time-dependent supergame played by a borrower nation and a lender bank. l A Nash equilibrium is a pair of price and capacity sequences such that the firm (the regulator) cannot increase discounted profit (surplus) by adjusting its capacity (price) sequence given that the regulator will adhere to its price (capacity) sequence from the outset.

38

THE RAND JOURNAL

OF ECONOMICS

best reply to qt 2 0 for each t. Further, if q > yqT-yfor some r, then the firm can increase its profitsby reducing investment-unless the marginal revenue is positive at the depreciated stock. We conclude that Proposition 1. (i) For any initial capacity q10, there is a unique open-loop equilibrium: At = V and (t = ytlq0 for each t 2 1. (ii) The unique steady-state, open-loop equilibrium is At = v and 1t= 0 for all t. Proposition 1 shows that when the firm must make simultaneous and independent commitments over an indefinite horizon, both owners and ratepayers suffer. The reason for this inefficiency is that the commitments are made unilaterally and noncooperatively. In their widely accepted analysis, Kydland and Prescott (1977) reach what seems to be the opposite conclusion. They argue that in dynamic policy games the outcome improves when the planner commits to an entire future course of action and all other agents follow. Kydland and Prescott then point out that the planner will typically wish to renege on its promises. For this reason, full optimality may require binding, bilateral contracts. In practice, such bilateral commitment is not always possible and, as we show, is generallyunnecessary.When neither the firm nor the regulatormakes any preplaycommitments, efficient equilibrium can arise, although the no-investment-no-cost-recovery outcome of Proposition 1 remains another equilibrium outcome. This latter outcome forms the conflict point of the game when it is expanded to include closed-loop strategies. Corollary 1. Starting with any history, hT for r > 0, the strategy pair qt(ht) = v and 1t(ht) = yq~t-,, for all t ? r and for any continuation game, (i) is a (subgame perfect) Nash equilibrium and (ii) constitutes the minmax threats of the game. Proof. See the Appendix.

o Optimal threats. So-called minmax threats support the maximal cooperation at equilibrium. The minmax threat for a playerin any continuation game is the equilibrium strategy that holds its opponent to the lowest payoff among the set of equilibria strategy pairs. Here we assume that the firm can, without penalty, stop investing and the regulatorcan deny the firm any further capital recovery. These threats are unrealistically severe. In practice, the firm faces an obligation to serve and a regulator can limit but not forever deny all capital recovery. Even the most severe credible threats are not likely to be carried through ad infinitum without the possibility of renegotiation. Corollary 3 below establishes that our approach can accommodate threats that are less severe or of shorter duration. Corollary 1 reveals the simple stationary form that these extreme threats possess: no
investment and no profits. They yield a payoff of zero for the firm and
z

ptS(ytT+1qT)

t=T

for the regulator, where q^T the capacity at the date X- at which the threat is carried out.2 is

o Steady-state triggerprice equilibria.This section demonstratesthat the planning solution is not an equilibrium when the firm and the regulator simultaneously choose capacity and price at each date. Nevertheless, we can show that the planning solution can be closely approximated at an equilibrium.
Farrell and Maskin ( 1989) and Bernheim and Ray ( 1989) have examined the effect that renegotiation has on the equilibrium set in infinitely repeated games. Their work, which does not strictly apply to the time dependent games considered here, suggest that efficient outcomes can still arise at equilibrium. Green and Porter ( 1984), and Abreu, Pearce and Stacchetti ( 1986, 1990) have analyzed equilibrium in repeatedgames with imperfect monitoring, i.e., where players cannot detect with certainty deviations by others. This work suggests that triggerprice regulation can still be effective when the regulator cannot perfectly monitor the firm.
12

SALANT AND WOROCH

39

When the firm's opportunity cost, Ilo, and sunk entry costs, F, are both zero, the interest rate, r, does not affect the solution to (SS). The following proposition establishes = that at = qi*(6) for each t 2 0 and 3tt P(q*(6)) for each t 2 1 is not an equilibrium outcome where
q-*(6) max {q: [P(q) - v]q - C(6q) 2 0} = argmax {S(q): [P(q)
-

v]q - C(6q) 2 0}.

The reason lies in the fact that the firm could earn a one-time profit whereas (SS) has it exactly breakeven. Proposition 2. The steady-state planning solution qt = q-*(3) > 0 for each t is not a (closedloop) equilibrium outcome. Proof See the Appendix. Section 5 presents an example in which the planning solution is an equilibrium outcome < whenever q^o q-*(5) and the interest rate is not too large (see Proposition 7 below). However, the steady-state optimum, qt* = 4i*(6) for all t, cannot be an equilibrium outcome. For fixed r and 6, it is possible to characterize the maximum capacity that can arise at a steady-state equilibrium. This capacity must provide sufficient profit to discourage the firm from curtailing investment. Depreciation must be fast enough and discounting slow enough that the regulator will not wish to deviate. Only in the absence of depreciation will there be no loss in consumer surpluswhen the firm stops investing. The following proposition characterizesthis maximum capacity. Proposition 3. Let q( 6, r) be defined as the largest q for which
[P(q)
- v]q-

C(q)

-yrC(6q)/(

1 - yr)

(12)

and also
00 00

p ts[P(
t=T

2]t=T

P S( v, -y qr)

( 13)

= for each Xr 1. Then (i) q^t q(6, r) and it = P(q^(b,r)), t 2 1, is an equilibrium outcome = provided q^%q(6, r) and (ii) given any qo > q^(6, at = qo for all t, is not an equilibrium r), outcome. Proof See the Appendix. Notice that as r -O 0,q( 6, r) q- qi*( 5). Thus, this proposition says that as the interest rate tends to zero, the largest initial capacity that can be maintained at a steady-state equilibrium converges to the optimal steady-state capacity. When capital is perfectly durable (i.e., 6 = 0), trigger strategies fail to support any steady-stateoutcome with q-> 0. To see this, consider any q+ > 0 such thatp+ = P(q?) > v. = = If the firm's trigger strategy calls for it to set q^ q+ whenever P5s p+ for all s < t, then the will not wish to maintain ^ = p'. This is so because once steady-state capacity regulator is achieved, no future investment is needed when capital is perfectly durable. The regulator can raise consumer surplus in every period without loss of service. For (p', q+) to be a steady-state equilibrium outcome, ( 13) must hold. The nearer 6 is to 0, the smaller the discount rate must be for the regulator to comply with any steady state 5* > v.

5. Other trigger price equilibria


* Proposition 3 indicates that there are steady-state equilibria that closely approximate the surplus-maximizing, steady-state solution among the class of such paths satisfying the

40

THE RAND JOURNAL OF ECONOMICS

firm's break-even constraint. In this section we look at the performance of trigger strategies in supporting non-steady-state, as well as steady-state, price-output paths. First we allow for an arbitraryinitial capacity. Next we relax the assumption of constant unit operating costs and entertain the possibility of increasing returns in production characteristic of natural monopoly. Finally we consider a variation in the order of moves in which the regulator moves after the firm. In each case we see that trigger strategies can improve upon static noncooperative outcomes. These strategies have the firm and regulator respond instantly to a departure from the planned path with their minmax threats. This applies as well to the defecting party, since self-punishment not only issues a best reply to the opponent's punishment but also penalizes an opponent for a delayed response to an infraction. In addition to being extremely severe, the punishments continue forever. If the future matters enough, the present value of forgone surplus will ensure cooperation. Nevertheless, no matter how damaging the consequences of a breakdown in cooperation, the firm and regulator are unable to attain the planning solution as long as there is some discounting.

o Nonstationary equilibria. Given any initial capacity, we construct an equilibrium path with normalized payoffs arbitrarily close to the steady-state optimal payoffs for positive depreciation rates and an interest rate sufficiently small. To describe the equilibrium set, we must normalize the payoffs, since discounted profit and surplus both diverge as the interest rate goes to zero. This is done by multiplying discounted surplus and profits by the interest rate. These equilibria result in capacity paths converging to q*( 5), the optimal steady-state capacity, as r -O 0 for any initial capacity. Furthermore, any per-period payoff pair, S = S(j) and ir = [P() - v]C(6), with j E [qm(, ), 4*(6)], can be approximated arbitrarilyclosely by equilibrium-normalized payoffs for r sufficiently small. Here qm(6) is the steady-state output that provides the firm with the largest per-period profit, that is,
Given a sequence 6 = { at: t = 1, 2, ..., 9o the regulator can be defined as
00

qm(6)

argmax {[P(q)

v]q

C(6q)}.

} of capacities, the normalized payoff for

NS((Q, r) = r and for the firm as


00

z
t=1

ptS(q^t)= rg'(C; r)

NII(CQ, r) = r

z
t=1

pt[[P(q^t)

- v]q^t-

C(qt-

yq^t)]

= rH(CQ; r, 6).

Recall that the planning solution solves (PP): it maximizes discounted surplus (3) subject to the break-even constraint
00

z pt([P(qt)

v]qt - C(qt -

yqt-l))

? I-o

+ F.

(14a)

Now consider a variant of this problem with a different constraint. Let @**( r, 6, I-o, q^o)
=

{*

(r,

6,

HI, do): t = 1, 2, . . ., 00}

maximize (3) subject to


00

2: p't-+'([P(qt)
t=7

-v]qt

-C(qt

- -qt-,))

2 Ho + F

(14b)

for - = 1, 2, ... , 0o .

SALANT AND WOROCH

41

The solution to this second problem satisfies a stricterprofit constraint than the original planning problem. ( 14b) requires that at any date r the value of the firm never falls below lO, while ( 14a) merely requires that the initial market value of the firm exceeds nlo. Equivalently, (1 4b) implies the firm has a nonnegative cash flow. Which constraint is appropriate depends on the circumstances facing the firm. The value of the firm's opportunity cost may change over time. When initial investment represents a sunk cost, the firm's opportunity cost jumps upon deciding to enter the industry. Nevertheless, the two paths converge when there is no discounting. A characterization of the constrained optimum path, Q* *( r, 6, I-o, q^o), a first step is to find how closely the planning solution can be approximatedat an equilibrium. To simplify notation, we include F in I-o for the remainder of this section.
with 11o < max { [P(q) - v]q - C(6q) }, the capacity path Lemma 1. For any (r, 6, I-o, q^o) as q* *(r, 6, IIo, q^o)* q*(r, 6, I-o, q^o) t -0o0 for some q*(r, 6, Ilo, '?o)< 00 .

Proof See the Appendix. The next lemma illustrates an equilibrium strategy pair having a "trigger property" and yielding normalized payoffs for the regulator arbitrarily close to the optimal payoffs from ( 14b). Lemma 2. Given any To? 0, ro > 0, Ho > 0, and e > 0, there is an r(E) > 0 such that for any r < r(E) the following strategy pair is a subgame perfect equilibrium: p[q* *(ro 6, 1O, ?o)] [
=*ht

if

ht = h*

v and 41* (ht)

otherwise

(15)

rq**(ro, t
Iqt-l

6, IIo, o)

if

ht = h(

~~~~~~~~~~(16
=**(r, 6, no, do)

otherwise, P(q**),andq**

where h*= (o, P3 * *, q* ,..., foreachs =1, 2 ..., t-1. Proof. See the Appendix.

pt-, qt-),ps*

A direct consequence of Lemma 2 is the following proposition, which states that if the interest rate is sufficiently small, equilibrium payoffs can approximate the payoffs of the steady-state planning solution as closely as desired. Proposition 4. Given e > 0, there is an r(E) such that the payoffs to the regulator from the capacity path Q *( r) determined by the equilibrium strategy pair ( 15 ) and ( 16) satisfies INS((cQ*(r),r)
whenever r < r( e).
-

S(q*(6))I

<e

Proof: See the Appendix. Immediate from the proof of the above proposition is that, starting from any initial capacity, a range of capacities can be reached and maintained at an equilibrium capacity, provided the discount rate is sufficiently small.
Corollary 2. Let J E [qm(6), j*(6)), where qm(6) = argmax {[P(q)
-

v]q

C(6q)}.

Given
at(ht)
a-nd

> 0 and q^o, there exists an r and a subgame perfect equilibrium (0, A) such that J as t INS )o each t and s, ht = (do, p), Jl . . , pt-1, qt- ) ps = P(qs) where, for
I ASh - Vt-I andn INSTQf/-Y
I efovr anyn r E

Js

(O. PA)

42

THE RAND JOURNAL OF ECONOMICS

Observe that unless qo _ q*( 5), the normalized profit and surplus of the equilibrium outcomes are bounded away from 0 and S( q*( 6)), respectively. This is so because the cost of building up capacity to 4j*( 5) can never be made up, even though its normalized value tends to zero as r -- 0. Thus, S(q-*( 5)) must always exceed any equilibrium payoff when qo <4(6) Further, the normalized value of the firm's profits can never vanish at any date t when qt_1> 0 and pt > v, as the firm can always earn a one-time profit by not investing and then shutting down. However, as r -- 0, the firm's per-period profits along an equilibrium path converging to some qc> 0 can be made to approach zero. The punishments implied by the triggers in (15) and (16) are unrealistically grim. Corollary 3 below follows directly from Lemma 2 and Proposition 4 and relies on less severe or shorter punishments. Note that although less severe punishments imply a smaller equilibrium set for any interest rate, it remainstrue that equilibriumcan approximatethe planning solution for sufficiently small interest rates. Reducing the severity of the punishments requires only minor modifications in the proof of Proposition 4. To see this, suppose the regulator can prevent the firm from abanwhere doning its investment. In particular,suppose that the firm cannot invest less than 6cq^, 6 > 5', so that the firm must replace some but not all of actual depreciated capital. In addition, assume the regulator cannot force price below v'> v. The punishments 3t = v' and qt = qt-l)] qt-l = qtq^t-l + 6'q^tj = [1 - (6 outcomes and can replace the punishments in the right-hand then can support cooperative side of both ( 15) and ( 16 ) for appropriately reduced discount rates. Summarizing, Corollary 3. Given any qo > 0, ro > 0, Ho > 0, and e > 0, there is an r(E) > 0 such that for any r < r(E) the following strategy pair is a subgame perfect equilibrium: {P[q* *(r0, 6, Ho, '1o)]
=*ht

if

ht = h*
(15a)

v' and fq*(ro, 6 Ho, q0) if

otherwise ht = h*

1qt-

otherwise,

> where v + 647*( > v'> v, 6 > 6Y 0, and where h * is any history in which the last defection 5) occurred at date r and then not again for T periods. That is, the firm sets capacity at q, = -yqs-l for each s = r + 1, ... ., r + Tand then qj* = q(r, 6, 110,q,+T) at all other dates
s, while the regulator sets the price ceiling at p3 = v' for each s
= X

+ 1, . . .,

X- +

T and

then p* = P(qs*) thereafter. Proof: See the Appendix. Since punishment occurs for T periods after the last defection, capital does not accumulate at the cooperative rate. Instead, the shortfall must be made up gradually once cooperation resumes.

o Increasing returns in production. Up to this point, production has exhibited constant returns to scale. Yet natural monopoly is often associated with significant scale economies. Proposition 5 generalizes Lemma 2 and Proposition 4 to allow for increasing returns to scale in production up to capacity. In particular, we suppose that operating costs are of the form
w0 h0a
where v(0) = 0, v

V(qf q) =
A --.. O an

v(q)

for for

q?< q> (17)

A'q<O

SALANT AND WOROCH

43

This is equivalent to assuming a generalized Leontief production of the form q = G[Min {aL, j3K}], where L and K are the amounts of variable and fixed inputs and a and 3 are production coefficients. Scale economies are captured by G(O) = 0, G' > 0, and G' > 0.

Proposition Suppose for any rHi0< a, for some II > 0, 5.


q*( , r, Ho) exists. Then the strategies
-

argmax {qP(q)

v(q, q) - C(q)

rHO:q -q

P[40(r,9 , ion, ^o)] v(q, yqyt-,)/'yqt-,

if

ht

(t1

otherwise

and
Ot(ht) =

r4I'*(ro,
IYqt-i

,HIIjO)

if

ht=hh*
(19)

otherwise

form a subgame perfect equilibrium where q-t**(ro,6, I-o, do) maximizes (3) subject to * (14b) and ht* = (qo, qr P(4*), ..., q-t*-, P(q4t*-l)).Also, the associated normalized surplus, NS(&* *( ro, 6, Ho, &o)), converges to steady-state optimal surplus S(q*( 3)) as ru- 0. Proof. See the Appendix.
0 Equilibrium when the firm and regulator alternate moves. Often, regulators observe a utility's investments before they set price ceilings. In some cases, they condition rates on the level of investment the firm undertakes. Such cases demand a sequential move game. To begin, let the firm first set capacity at and the regulator follow with price ceiling A. Whereas the steady-state planning solution could only be approached when moves were simultaneous, we show that it can be achieved when the regulator enjoys a second-mover

advantage. 13

As before, a strategy for the firm is a rule determining the choice of capacity in each period t as a function of the history through period t - 1. Now a strategy for the regulator ^ specifies the choice of as depending on ht and qt. the firm's capacity in period t. Let h= [ht, t]. Consider the following strategy pair, (X*, l*): 0*(h+) Pt
Lv

if

h+ =(h*,') qt t

otherwise

~~~~~~~~(20)
)

qt* At*(ht) =(21 = yqt-l

if

ht=ht

otherwise,

where ht*= (do, qi,IP*, Pt-,), the optimal path through period t -1. Given 6 < 1, this strategypair is a subgame perfect equilibrium that attains the planning solution provided r is not too large. The firm earns zero profit along the equilibrium path, and since it moves first,it cannot earn more by setting off the trigger.But once a deviation has occurred, neither player has an incentive to cooperate.

6. Proposition Given 3 < 1, there is a ro > 0 such that the planning solution is a subgame perfect equilibrium outcome whenever 0 < r < ro.

13

Where pitis set after the firm has committed to qt, then the planning solution will always emerge as a

subgame perfect equilibrium outcomewithlinearinvestment costs.

44

THE RAND JOURNAL OF ECONOMICS

The remaining possibility has the regulator commit to a price ceiling for each period before the firm determines its capacity. In that case we find the same "regulatorylag" as in the simultaneous-move game. As in Proposition 2, this lag implies that the planning solution is not an equilibrium outcome because the firm is able to earn a one-time profit by defecting.

6. A linear example
* A simple example illustrates how triggerprice regulation can sustain win-win outcomes. Consider the case of linear investment cost: C(I) = I, (22)

and let both unit operating cost v and the fixed entry cost F be zero. In this case, dynamic marginal cost in (6) is constant and increasing in both the interest and depreciation rates: MCt = fp(b + r). Substituting (23) into the pricing rule (5) and then rearrangingyields (23)

t - fp( + r) -X
Pt
Et

(24)

This expression characterizes second-best prices when there is no excess capacity, which c holds as long as 'yqO j*. Since it is independent of qj,-,, its solution is a steady state: =ft* for all t ? 1. Thus, the planning solution requires investment qt= qt= 4* and Pt of q* - 'yqoto bring capacity immediately up to its limit value 11*;thereafter, investment maintains this level by covering physical depreciation, &j*. The participation constraint in (PP) in the linear case is
00

-10 c z pt[P(qt)qt
t= 1

f(q't -

yqt-1)]

(p*

#6)4*/r - f3yp(4* - 4o)


S

(25) ..,t 1
(26)

strategies: Given capacitiessatisfying (24) and (25), we constructthe followingtrigger


* =
p

if

qs=qs I

Ps=P*,

vV

otherwise if *vs = *, otherwise.


Ps =P , s= 1, ..,t-

(27)

lzyqt-l

path with payoffsto defectingWe now want to comparepayoffsalong the cooperative A valuesalongthe defecfirstfor the firm,and then for the regulator. tilde(-) distinguishes tion path.
a Defectionby the firm.We firstshow when the firmwill not wish to defectin the first will A period.The firm entersperiod 1 with capacity oyqo. trustingregulator set the price cost:Pt = v, ceilingat Pi = p*, but shouldthe firmdeviate,pricewill be cut to operating

2. In that event, its best reply is to cease all investment: qt = ytq0,t 2 2.

this In period 1, a defectingfirm selectsql to maximizeone-periodprofit.Generally, unlessmarginalrevenueis positiveat that prowill call for no investment(i.e., ql = 'yqo) defectionin period 1 savesthe firm up to fl(q*- qo), the cost ductionlevel. Therefore, in of plannedinvestment.It collectsat least qy0*jO revenuesthat period,so thattotal profit from defection (discounted to period 0) is bounded below: P[Pi*p - l(i q-yo)] ? oypj5q^o. profit Suppose,instead,that the firmadheresto the optimalpath.Then its first-period
equals p (*=*
-

_yo),

which is negative whenever initial capacity is less than the

SALANT AND WOROCH /

45

limit capacity qO< 4*. Cooperation yields profits discounted to period 0 of Ho. The firm will cooperate provided 10 > yppj*qo. (28)

In subsequent periods t ? 2, the firm's payoff along the candidate path (discounted to date t) is (f5* - #6)j*/ r, which from (25) is at least flpy(c* - go) + Ho. Defection yields at most a one-time profit (discounted to date t) of ypy*qj*. 3 Defection by the regulator. We now solve ( 13 ) to calculate the values of r and 6 for which the regulator will not defect. The regulator'sdiscounted payoff from cutting price to Pt = 0 for each t 2 1 is

pS(O, + z2 ptS(0, lyt-lq*) ci*)


2

(29)

In comparison, the surplus along the planning solution is


00

z ptS(ft*,

c*) = S(U*, j*)Ir.

(30)

Cooperation dominates defection whenever (30) exceeds (29):


00

S(f *, 4*)Ir

pS(O,

oa*)+ 2: pts(ol 1Yt-lq*)2

(31)

Note that the right side of (31 ) tends to 0 as r tends to 0. Proposition 7. Suppose that qo < 4* and that I-o is sufficiently large so that (28) and (31) hold. Then, for r sufficiently small, the planning solution is a subgame perfect equilibrium outcome with trigger strategies described in (26) and (27) when the regulator and the firm move simultaneously. When the firm chooses capacity before the regulator sets the price ceiling, the planning solution is once again an equilibrium outcome if the interest rate is not too large. The conditions imposed on (28) and (31 ) in this proposition govern how much profit and surplus the firm and regulator must receive to induce them to cooperate. Inspection of the participation constraint (25) reveals that as r -- 0, smaller values of 5* will still allow (28) to be satisfied. Condition (31) indicates that as the interest rate gets small, average surplus need only be positive to be guaranteed that the regulator will wish to continue cooperating.

o Special case of linear demand.A more complete characterizationof equilibriais possible if we take demand to be linear: P(q)=a-bq
for 0 < q c a/b, so that the price elasticity is
Et

(32)

(a - bqt)/bqt.

(33)
(34)

Consumer surplus then becomes S(p, q) = aq - bq2/2 - pq, which equals bq2/2 whenever p = a - bq. We require sufficient demand to cover average total cost in a steady state, or a - bq > #3 for some q. Solving (24) yields the steady-state capacity:
-*

a - flp(b + r)

46

THE RAND JOURNAL

OF ECONOMICS

Substituting (35) into (25) and rearrangingresults in a quadratic in 1 + 7:


(1 + 7)2bba2/r + (1 + n)[p1'ycxa (a - f35)a/r] + (Ho - pfryqo) = 0.

(36)

The firm prefers to cooperate when (28) holds, which in this case is 2 10? ypco[na + 3p(b+ r)]/(l The regulator prefers to cooperate when
00

).

(37)

z
1=1

p'b4 2/2 = bj*2/2r

c ai*/(1

+ r - Ay) bi*2/[2(1

+ r-

_2)]

(38)

or equivalently, 4* > [2a(1 + r 2)r]/[b( + 2r - 2) (39)

Expressions such as (37) and (39) sweep out an entire range of discounted profit and surplus that can be supported when interest rates are small. On the frontier, each level of surplus is associated with a differentprofit. Which of the many profit-surpluspairs is selected must be decided by the players. Furthermore, triggerstrategiessuch as (26) and (27) specify a particular price-investment path to be used. In fact, other combinations of prices and investments could yield exactly the same payoff levels. Aware of this indeterminacy, the firm and regulator must jointly decide on one such path, or risk a loss of efficiency. 7. Policy implications

* Opportunism constrains the growth and operation of a regulated utility. Efficiencyminded policymakers cannot ignore its effects. Full commitment to a policy is not a solution, however. On the contrary, flexibility of response raises a strong deterrent to inefficient behavior. With this in mind, we find that trigger price regulation produces some unorthodox prescriptionsfor the design of policies toward natural monopoly. These lessons apply equally well to other forms of monopoly regulation,including rate-of-returnand price-capregulation. Although our stylization of price control lacks some basic features of the typical price-cap scheme, it captures the essential spirit of that scheme. 14 Furthermore, modification of our setup to have the regulator set the allowed return on investment instead of a ceiling price is straightforward.Gilbert and Newbery (1989) carry out such an exercise when capital is not durable. How well a particular regulatory program performs relative to another will turn on the characteristics of the particular market. It clearly is a subject that warrants further analysis. Whatever the institutional framework, several novel implications follow from our analysis of trigger price regulation. Facilitate negotiations between the firm and regulator. We have seen that, if they cooperate with one another, investors and consumers can realize huge benefits. Mutual gains require a climate conducive to cooperation and coordination that can be achieved only through free and active negotiation. Restrictions on the exchange of offers, and on the terms of agreement, only prevent the bargaining process from pricing and investing efficiently. This observation is not new. Coase ( 1960 ) pointed out the value of costless negotiation in a general setting of bilateral externalities. We expand on Coase's point by demonstrating that, due to the multiplicity of equilibria, the attainment of a win-win outcome requires a degree of coordination that is possible only if negotiation costs are not too high.
O
14 The first implementation of price caps was for British Telecom and had its roots in Littlechild ( 1983). The U.S. version applied to AT&T is codified in Federal Communications Commission (1988). Linhart, Radner, and Sinden ( 1983) and Vogelsang( 1989) offerearlyspecificationsof the price-capmechanism and analyzeits performance.

SALANT AND WOROCH

47

On the other hand, policies that facilitate bargainingcan yield unwanted consequences. Specifically, face-to-face negotiations raise the temptation for managers of the firm and staff at the commission to strike a deal at customers' expense. Free access to negotiations should eliminate much corruption, but then the administrative process becomes slow and cumbersome when every interest group is allowed to intervene.

O Raise the conflict point. The equilibriawere supportedwith triggerstrategiesthat credibly threaten to halt any capital recovery and all investment. We recognized that in practice, the use of such grim strategies is unlikely. As we show, less damaging threats might still support the same payoffs but will invariably shrink the equilibrium set. The degree to which negotiation can improve on static, noncooperative behavior depends on how damaging both parties find a breakdown of the bargaining process. Paradoxically, measures that make the threat point more distasteful may actually improve and enhance the stability of the regulatory climate. This suggests that each party should be given expanded freedom to respond to opportunistic behavior. By relaxing its obligation to serve, for instance, a firm can react to unrenumerative rates by reducing service quality or by abandoning a service altogether. Symmetrically, a regulator could mete out,a more severe punishment for inadequate investment if the constitutional ban on confiscatory rates were eased. It might also be allowed to cut rates in ancillary markets served by the firm. The method used to finance investment can have important strategic effects as well. Specifically, the larger the fraction of capital financed by debt, the less shareholders stand to lose if the regulator drives the firm into bankruptcy. Limits on the ability of a public utility to use debt financing therefore make cooperative outcomes more likely and more stable. Policies that reduce payoffs associated with the threat point, while desirable from a bargaining perspective, are not without tradeoffs. A more severe threat point enlarges the set of equilibria to include both better and worse outcomes. And the damage done is greater if the trigger is set off by mistake or during a brief lapse of rationality. Outside intervention by courts and legislaturesmay directlylimit opportunisticbehavior, lessening the need for noncooperative punishments. But whatever the powers of these third parties,there invariablyremains some opportunity for improvement. Triggerprice regulation offers an avenue to exploit these mutual gains. o Raise the discount factors. The more weight the regulator and the firm attach to the future, the higher will be the costs of defection. Lengthening the tenure of commissioners and their staffs, or instituting a system of overlapping terms, would encourage them to internalize the impact of their actions. Permitting regulatory staff limited employment opportunity in the industry after leaving the agency is another way of increasing the costs of defection. Similarly, managers will place more weight on the firm's future by using properly designed compensation schemes or by opening up post-retirementemployment opportunities to regulatory officials and staff."5 Allowing such post-retirement opportunities has both benefits and costs. While such positions serve to mitigate opportunisticbehavior, corruption may result if these employment opportunities are used as bribes.
O Restrict the firm's capital structure. Our results establish the connection between equilibrium and the firm's capital structure. In particular, excessively durable capital exposes the firm to expropriation should it adhere to efficient accumulation rules. Durability raises

15 Salant ( 1991 ) illustrates how overlapping terms can result in cooperative outcomes in an infinitely repeated game with finitely lived players.

48

THE RAND JOURNAL OF ECONOMICS

the regulator's temptation to renege on any cooperative plan. Proposition 4 and Corollaries 2 and 3 illustrate the role played by depreciation. To capture the benefits of their relationship, both parties would like to limit the commitment the firm undertakes when it invests. Institutions could prohibit adoption of technologies that require highly sunk expenditures on plant and equipment. Compromising production efficiency may be a small price to pay for strategic stability.

8. Conclusion
* We model the regulation of a public utility as a long-term relationship between a firm and its regulator.After formulating a time-dependent supergame,we extend a Folk Theorem to show that when the participants are sufficiently patient and capital is not too durable, the planning solution can be approximated arbitrarilyclosely at an equilibrium. We propose "trigger price regulation," in which the regulator, responding to any deviation from the efficient path, cuts price down to operating cost and the firm curtails investment. This policy continues to perform well for differentspecificationsof the technology, sequencing of moves, and limitations on noncooperative behavior. Elsewhere we have demonstrated that its attractive properties hold when the firm invests in cost-reducing innovations or product enhancements, and when capital is lumpy. Its effectiveness over a wide range of environments positions trigger price regulation as an attractive option for regulatory reform. APPENDIX
* Proofs of Corollaries 1 and 3, Propositions 2 through 5, and Lemmas 1 and 2 follow.

Proof of Corollary 1. (i) If the regulatoranticipates the firm will set q, = -yd,- for each t ? r, then consumer surplus is maximized by settingfi, = v for each t. This choice of a price ceiling has no effect on future investment. Obversely, if the firm believes the regulatoris going to set fi, = v for t ? r, then it has no incentive to incur additional investment expenses, so c, = -y4q, for t -r. and (ii) To see that these strategiesthreaten minmax penalties, notice that they provide the firm with zero profit for the remainder of the game, and the regulator'spayoff is limited to the maximum surplus that can be amassed utilizing existing capacity. The firm can always choose to exit and so cannot be held down to any lower payoff. And the regulator cannot be held to any lower payoff if, as we assume, it can set rates at operating costs that keep the firm's existing capacity running. Q.E.D. Proof of Proposition 2. Let ' and t be any strategy pair having the equilibrium property that fi, = f*(6) and = q, = j*( 6) for each t. Notice that at this steady-state optimum, p~q, j*( 6)j*( 6) = vq*( 6) + C(bq*( 6)), making profit zero in each period. Suppose defection occurs in period 1. The firm sets capacity at 41 = -yqO= -y,*( 6) provided static monopoly output is less than -y'2o. Assuming the regulator sets price at f*( 6), the firm earns a profit in that period equal to -yC(6q*(6)). Further, by ceasing all subsequent investment, the firm avoids any future losses regardless of the regulator'sactions. Therefore, this defection gives the firm greaterprofits than does A, which is zero. Q.E.D. Proof of Proposition 3. Let q = d(6, r) be as in Proposition 3: -(q) = [P(V)-v]7C(6o) 2 yrC[64]/(1 - yr), (Al)

for each r 2 1 let


00 00

Z, p'S[P(4),

4] 2 Z pIS(V, _y'-7),

(A2)

and let j3(6, r) = P[c2(6, r)] be the corresponding price. Given 6 < 1, both inequalities (Al) and (A2) are eventually satisfied as r approaches zero, since the right side of (Al) goes to zero as r -> 0, and for fixed 6 < 1, (A2) is satisfied for r sufficiently small. (In general, one of (Al) and (A2) will bind while the other is slack.) Thus, - C(64) -> 0 and so d(6, r) -> #*(6), the optimal steady-state capacity. In the following, fix [P() -v] 6 < I and letq* =*(b). Consider the following closed-loop strategieswhich have a triggerproperty: p(b, r) (hi) = 0d* if h, = h* (A3)

oherwise

SALANT AND WOROCH


rq(6,r) ,*(h)=

/ 49

if

h,=hh*
(A4)

b4-_

otherwise,

t - 1. Thenthe strategy pair(A3), 4(6, r) for each r = 1, 2, (A4) form a (subgame perfect) Nash equilibrium if at any point in the game, and for any history up to that point, both the regulator and the firm would wish to continue using those strategies (assuming the other one will do so as well). Now, given no previous defection up through date r, (A3) and (A4) imply that defection by the regulator does not pay, provided
=

whereh * is the sequence = f (6, r) and q, PT

2
t=Tr

ptS[pi(b

r),

q4(b

r)]

2 zptS[v,
t=Tr

t-yTq(b,

r)].

(A5)

But this is guaranteed by (A2). Once a deviation from h* has occurred, Corollary 1 says that the firm will never wish to incur additional investment expenditures. Given h, = h*, the largest one-period profit the firm can obtain by defecting is [fi(r, 6) where, for r sufficiently close to zero,
pf(r, 6)
=

v]-y4(6, r)

yC[64(6, r)]/(l

yr),

(A6)
(A7)

v + C[64(6, r)]/[(l

yr)4(6, r)],

as in (Al). Equivalently, there is a p close to one satisfying (A7), so that the profit from continued cooperation is
co 2:
t=I p t [pi(b r) v] 4(b r) -C[ 64(b r)
]}-

(A8)
(A9)

Then, for p sufficiently near 1, (Al) implies that (A8) can be rewritten as
-yC[64(6, r)] - yr

which is just (A6). Therefore, the firm cannot gain from defection when p is sufficiently large. Observe that for largervalues of q, (A l ) would be violated for sufficientlylarge p. Thus, no largersteady-statecapacity is sustainable as a (subgame perfect) Nash equilibrium outcome, making the strategies (0*, P*) result in the surplus-maximal steady-stateoutcomes. Finally, it is important to note that our trigger strategies call for minmax punishments of defections. As a result, outcomes supported by these strategies cannot be Pareto dominated by an equilibrium supported by any other pair of closed-loop strategies. Summarizing these arguments gives
Proposition A. Given any 6, 0 < 6 < 1, there exists an F> 0 such that for r <
r-, the triggerstrategies(0*, P*) given by (A3) and (A4) form a steady-state (subgame perfect) Nash equilibrium providing the regulatorwith the largest surplus from among the set of steady-state outcomes.

Proof An immediate implication of the above is that for any

q E [qm(6),
-

4(6, r)]

where

qm(b) = argmax{[P(q)

v]q

C(6q)}

is the profit-maximizing steady-statecapacity, strategiesof the form (A3) and (A4) can be used to support a steadystate outcome with qt = qjand pt = P(4c) for all t. This follows because the above arguments in no way depend on the steady-state capacity being 4*; we merely require that [P(j)-v] -C(&q) > 0. Q.E.D. t = 1, 2, . Proof of Lemma 1. We show that for any '2o the sequence { q* *(r, 6, Ho, 0o): and strictly monotone. Such a sequence must then have a unique limit point. For do [P(qj*(r, 6, Ho)) - v]1*(r, 6, HO) C(64*(r, 6, Ho)) = rHo, it mustbe the casethat
4*(r, 6, HO) > q**(.)
<

oo } is both bounded i*( r, 6, Ho) where

> q*i*(.)

for each t, from the assumptions that S'(0) > C'(0) and that S" < 0 and C" > 0. Similarly, concavity of S( *) and whenever qo> 4*(r, 6, Ho). Q.E.D. convexity of C(*) imply that i*(r, 6, HO) q,**(.) <q*-*(*)
_

Proof of Lemma 2. First notice that for the firm, continued play of (0*, P*) yields a normalized profit that remains * - C(q,* - -yq * ) -* roHo as t -* oo, and strictly positiveas r - 0. This is because(i) 7r* = [P(q)* * )-v]q* therefore (ii) NIH = lim,0 r{ z H,/( 1 + r)'} = lim,0 r[(ro/r)Ho] = r0H0> 0. Defection yields a one-period payoff of no more than II = max { [P(q)
-

v]q: q 2 0 } and nothing thereafter, which has a normalized profit

tending zeroas r -> 0. to

50

THE RAND JOURNAL OF ECONOMICS

Further, adherence to 0* results in a payoff for the regulator converging to S[qc*(6)] as r -> 0. In contrast, deviation from 0* at date t generates a normalized surplus of no more than r 2i S1-y-t) (

which tends to zero as r -> 0. Thus, for r sufficientlysmall, deviation from 0* does not pay for the regulator. Q.E.D. Proof ofProposition 4. By Lemma 1, for rHo< HI max { [P(q) path (Q**(r6, HO,co) = {q**(r, 6, Ho, do)} satisfies *(r, 6, HO,qO) q,* for some q*( r, 6, Ho, co). Also notice that q*(r, 6, HO,Jo) 4 q*(b) as r - 0.
-

v]q

C(6q): q 2 0 }, the constrained optimal

q*(r, 6, HO,40) as t --> 0

If not, then it must be the case that q*( r, 6, Ho, co) -> q' as r -> 0 for some q' # j*( 6). And it must also be the case that q' < j*( 6), as 7*( 6) is the maximal capacity level that can be sustained by the firm while still earning revenues in each period exceeding the total of rHoplus operating and depreciation costs. But for r sufficientlysmall, one can always construct a path satisfying the profit constraint ( 14b) that provides the regulator a greater surplus than any path converging to q' < j*( 6). Consider the steady-state capacity qr = Max {q: [P(q)
-

v]q

C(6q) > r[Ho + C(ci*(6))]}.

Observe that as r - 0, qr-- *( 6). Let Q(r) be the "fastest," i.e., shortest number of periods, path to qrsatisfying constraint ( 14b). (Note that if c^o qr, this path reaches qrin one period.) Then as r -> 0, < NS(Q(r), r)
->

S(i*(6)) > S(q').

Thus, q*(r, 6, 11o, go) -> 7*(6) as r -> 0 for any So > 0. Then for ro such that S[ q*( r, 6, Ho, 'o) ] > S[ q*( 6)I - , the strategypair ( 15 ) and ( 16) results in a capacity path with the desired properties. Q.E.D. Proof of Corollary 3. First, for T large, the punishment payoffs are approximately the same as is the case with ( 15) and ( 16). Also, suppose for any 0 < r < ro, Ho in (I14b) satisfies r0i0/r > ro max
{qt}

z
t=I

qtpt[P(qt) - v'] - C(qt - yq,).

Also, notice that continued play of (0*, ip*) yields a normalized payoff for the firm that approaches r0110as r -> 0. And defection by the firm gives it a normalized payoff that the above inequality implies is strictly less than that. So the firm would not wish to defect from ( 16a). Further, the regulatorwill adhere to ( 15) and collect a normalized payoff converging to S[c*( 6)] as r -> 0, whereas defection produces a payoff of S(q') for some 0 c q' < 4*( 6). Note that once a defection occurs, neither side will wish to reward the other for T periods. Thus, after the punishment period ends, the incentives are as if there were no defections. Q.E.D. Proof ofProposition 5. First, notice that an argument identical to the one in Lemma 1 can be used to establish that
-* the path4t(r, 6, Ho, 2o)converges somelimit value,4*(r, 6, Ho, qo).Also,as r -* 0, 4*(r, 6, Ho, qo) q*(6). to

Second, the incentives for the firm to defect from the candidate equilibrium path are the same as in the first part of the proof of Lemma 2, and an identical argument implies that A is always a best reply to k. That the regulator would find 4 a best reply to A for any history follows from the second half of the proof of Lemma 2. The regulator's surplus from a deviation from ' is strictly smaller, given ht when v" < 0, than was the case in Lemma 2, where v" = 0. Given a defection has already occurred, incentives are the same as in Lemma 2. Finally, that 4*( r, 6, Ho, qo)must converge to i*(6) as r -> 0 also follows from the same argument given in Proposition 4. Q.E.D.

References
ABREU, D., PEARCE, D., AND STACCHETTI,E. "Optimal Cartel Equilibria with Imperfect Monitoring." Journal of

Economic Theory, Vol. 39 (1986), pp. 251-269. "Toward a Theory of Discounted Repeated Games with Imperfect Monitoring." Econometrica, Vol. 58 (1990), pp. 1041-1064. BARON, D. "Design of Regulatory Mechanisms and Institutions." In R. Schmalensee and R. Willig, eds., Handbook of Industrial Organization. Amsterdam: North-Holland, 1988. Approach." BENHABIB, J. AND RADNER, R. "The Joint Exploitation of a Productive Asset: A Game-theoretic Economic Theory, Vol. 2 (1992), pp. 155-190.

SALANT AND WOROCH


BERNHEIM,

51

B.D. AND RAY, D. "Collective Dynamic Consistency in Repeated Games." Games and Economic Behavior, Vol. 1 (1989), pp. 295-326. BROCK, W. AND DECHERT, W. "Dynamic Ramsey Pricing." International Economic Review, Vol. 26 (1985), pp. 569-591. COASE, R. "The Problem of Social Cost." Journal of Law and Economics, Vol. 3 (1960), pp. 1-44. FARRELL, J. AND MASKIN, E. "Renegotiation in Repeated Games." Games and Economic Behavior, Vol. 1 (1989), pp. 327-360. FEDERAL COMMUNICATIONS COMMISSION. "Policy and Rules Concerning Rates for Dominant Carriers."Docket No. 87-313, 1988. FERNANDEZ, R. AND ROSENTHAL, R. "StrategicModels of Sovereign-Debt Renegotiations." Review of Economic Studies, Vol. 57 (1990), pp. 331-349. FRIEDMAN, J. "A Modified Folk Theorem for Time-dependent Supergames."Mimeo, University of North Carolina, February 1988. GILBERT, R. AND NEWBERY, D. "The Dynamic Efficiencyof RegulatoryConstitutions."Working paper, University of California, Berkeley, Department of Economics, December 1989. GREEN, E. AND PORTER, R. "Noncooperative Collusion Under Imperfect Price Information." Econometrica, Vol. 52 (1984), pp. 87-100. KLEIN, B., CRAWFORD, R., AND ALCHIAN, A. "Vertical Integration, Appropriable Rents, and the Competitive Contracting Process." Journal of Law and Economics, Vol. 21 (1988), pp. 297-326. KYDLAND, F. AND PRESCOTT, E. "Rules Rather Than Discretion: The Inconsistency of Optimal Plans." Journal of Political Economy, Vol. 85 (1977), pp. 473-491. LUENBERGER, D. Optimization by VectorSpace Methods. New York: John Wiley, 1969. LINHART, P., RADNER, R., AND SINDEN, F. "A Sequential Principal-Agent Approach to Regulation." In O.H. Gandy, P. Espinosa, and J. Ordover, eds., Proceedings from the Tenth Annual Telecommunications Policy Research Conference, Norwood, N.J.: ABLEX Publishing, 1983. LITTLECHILD, S. "Regulation of BritishTelecommunications' Profitability." U.K. Department of Industry,February 1983. REINGANUM, J. AND STOKEY, N. "Oligopoly Extraction of a Common PropertyNatural Resource:The Importance of the Period of Commitment in Dynamic Games." International Economic Review, Vol.26 (1985), pp. 161173. SALANT, D.J. "A Repeated Game with Finitely Lived OverlappingGenerations of Players." Games and Economic Behavior, Vol. 3 (1991), pp. 244-259. AND WOROCH, G.A. "Crossing Dupuit's Bridge Again: A Trigger Policy for Efficient Investment in Infrastructure." ContemporaryPolicy Issues, Vol. 9 (1991), pp. 101-114. . "Promoting Capital Improvements by Public Utilities: A Supergame Approach." In W. AND Neuefeind and R. Riezman, eds., Economic Theory and International Trade: Essays in Honor of John Trout Rader III, New York: Springer-Verlag,1992. VOGELSANG, I. "Price Cap Regulation of Telecommunications Services: A Long-run Approach." In M. Crew, ed., Deregulation and Diversification of Utilities. Boston: Kluwer Academic, 1989. WILLIAMSON, 0. Markets and Hierachies. New York: Free Press, 1975.

Das könnte Ihnen auch gefallen