Sie sind auf Seite 1von 12

OXFORD REVIEW OF ECONOMIC POLICY, VOL. 13, NO.

REGULATION, COMPETITION, AND THE


STRUCTURE OF PRICES

JOHN VICKERS
All Souls College, Oxford1

Many competition policy issues in regulated industries concern the structure of prices charged by multi-product
firmsfor example price discrimination, non-linear pricing, cross-subsidies, and network access pricing. This
article first sets out the (Ramsey) principles of optimal pricing to recover fixed costs. The sometimes conflicting
aims of promoting competition and pursuing social objectives are brought into the analysis. Questions of whether
to allow pricing structure discretion to the firm, and how much, are considered next. With asymmetric informa-
tion, some discretion is often desirable, but its optimal form is hard to characterize. The article then turns to the
controversial network access pricing problemon what terms should an integrated dominant firm be required to
supply inputs required by its rivals? Finally, there is discussion of pricing structure regulation in the transition
from more to less regulation, which, it is to be hoped, is in prospect in parts of the regulated industries as effec-
tive competition develops.

I. INTRODUCTION removal of legal barriers to entry. Conduct regula-


tion, which may take the form of explicit monopoly
Competition policy in regulated industries has three controls and/or competition policy measures, con-
broad aspectsstructure, liberalization, and conduct strains the pricing and other behaviour of dominant
regulation. Structural policies include break-up deci- firms. Most obvious is the question of how to
sions such as those taken (or not taken) at the time of regulate the level of prices charged by firms with
privatization, merger controls, and scope-of-busi- monopoly powerby RPIX price caps or other-
ness restrictions. Liberalization policy is about the wise.

1
This paper was presented at the British Association Annual Festival of Science at Birmingham University, and at the Warwick
Business School conference on Rebalancing Utility Prices in September 1996. The paper was written during a period of sabbatical
spent at London Business School. I am grateful to the Economic and Social Research Council and the Office of Fair Trading for research
funding under the Contracts and Competition programme (grant L114251038). Much of the paper reflects joint work with Mark
Armstrong. Thanks also go to Chris Doyle, John Kay, Robin Nuttall, and Jean Tirole for valuable discussions and comments. The
usual disclaimer applies.

1997 OXFORD UNIVERSITY PRESS AND THE OXFORD REVIEW OF ECONOMIC POLICY LIMITED 15
OXFORD REVIEW OF ECONOMIC POLICY, VOL. 13, NO. 1

The purpose of this paper, however, is to provide tion, the theoretical principles of which are only
some theoretical background for policy questions partly developed, are the subjects of section III.
concerning the structure of prices charged by multi-
product firms in regulated industries.2 Many compe- Section IV brings wholesale prices, and their relation
tition policy issues in regulated industries come to retail prices, into the analysis. It discusses the
under this heading, for example price discrimination, problem of the terms on which an integrated domi-
non-linear pricing (e.g. volume discounts), cross- nant firm should be required to supply inputs
subsidies, predatory pricing, pricing of inputs (e.g. notably access to essential facilities such as natural
network access) used by rivals, universal service monopoly networksrequired by its rivals. This
obligations and geographical averaging. Though the network access pricing problem is perhaps the most
focus is on pricing behaviour, the economic princi- controversial of all pricing structure questions in
ples relating to non-price behaviour (e.g. refusal to regulated industries.
supply) are often similar.
The transition from more to less regulation, which,
A major reason why pricing structure questions are it is to be hoped, is in prospect in parts of the
difficult is the problem of conflicting objectives regulated industries, involves removal of some price
or, perhaps more accurately, of too few instruments controls. The resulting question of how remaining
chasing a number of objectives. Thus efficient re- price controls should be structured so as to avoid
source allocation calls for (marginal) cost-reflective distortions of competition and efficiency is taken up
pricing. But with economies of scale or scope, that in section V.
leaves a problem of cost recovery. Where liberaliza-
tion has occurred, pricing structure policies might
distort effective competition by thwarting the entry of II. THE RAMSEY PRICING BENCH-
efficient firms or by inducing the entry and growth of MARK
inefficient ones. Further tensions might arise if
social objectives (e.g. distributional concerns) or (i) Least-bad Departures from Marginal Cost
political constraints (e.g. uniform nationwide pricing Pricing
requirements) are important. Section II below exam-
ines these issues, and resulting trade-offs, starting Consider the pricing problem of a pure monopolist
from the bench-mark of Ramsey pricing. that supplies n different products (which might be
differentiated by time, place, state of the world, or
Another fundamental source of difficulty has to do even type of buyer). Let ci and pi denote the marginal
with information. Regulatory and competition au- cost and price of product i, and assume for now that
thorities generally know less about cost and demand neither varies with quantity (but see below on non-
conditions, and about firm behaviour, than those in linear pricing). Let xi (p) be the demand for product
the industry. Decentralized information would ide- i, which generally will depend on the whole set of
ally be used in setting prices, but it might be hard to prices p = (p, . . . , pn) and not just pi. If v(p) is
elicit, and indeed firms might seek privately to aggregate consumer surplus, then xi(p) = vi(p),
exploit informational advantages to the detriment of where vi(p) is the partial derivative of v with respect
the public interest. This suggests that firms should to pi.
have some discretion over their pricing structure
choices, but perhaps with constraints on the exercise Assuming that there are no significant externalities
of that discretion. For example, instead of many or other distortions elsewhereso that ci measures
detailed price controls, the firm might be constrained social as well as private marginal costit is clearly
by just one or two broad price caps, with supplemen- optimal to set prices equal to marginal costs: pi = ci.
tary rules on permissible ranges of price variation Then, and only then, will prices reflect the resource
(sometimes called floors and ceilings) or cross- costs caused by individual consumers decisions,
price restrictions (e.g. bans on kinds of price dis- which is a requirement for efficient resource alloca-
crimination). Broad price caps and limits to discre- tion.

2
For more detailed analysis, see Laffont and Tirole (1993, ch. 3) and Armstrong et al. (1994, chs 35).

16
J. S. Vickers

The trouble is that marginal cost pricing will This discussion has so far assumed uniform pric-
obviously not cover all costs if, as in major parts ingi.e. that price per unit does not vary with
of the regulated industries, there are economies of quantity purchased. Non-linear pricing, in particular
scale or scope. Suppose that the monopolist has a quantity discounts, can often do better. For example,
fixed cost F in addition to its variable costs. Then consider two-part pricing, which involves a fixed
its profit (p) = i (pi ci )xi (p) F. Marginal cost charge as well as a price per unit. To be feasible, this
pricing would lead to = F. Unless this deficit requires consumption metering and no resale oppor-
can be covered by a lump-sum payment to the tunities, conditions which are common in the utility
firmwhich would have to be financed somehow, industries, and, indeed, there are fixed quarterly
with resulting social costs caused by higher dis- charges for gas, electricity, and telephone services.
torting taxes elsewhere in the economythe firm The benefit of two-part pricing is that it enables the
will be unable to finance its activities. If the variable price to be closer to marginal cost, while the
firms revenues have to cover all its costs, the fixed charge contributes to the firms fixed costs.
question is how best to do this. Ramsey pricing is Indeed, if no one got excluded from consumption
the answer to this question. 3 altogether, two-part pricing with variable prices equal
to marginal costs could achieve the first best (given
Formally, the problem is to choose prices p to the other current assumptions).
maximize welfare v(p) + (p) subject to (p) > 0.
One might guess that the solution to the problem This accords with the Ramsey principles above. If
would be to have the same price/cost mark-up demand for access to the service is perfectly inelastic,
(p i ci )/pi on each product. This is wrong: equal then all fixed-cost recovery should be done via the
mark-ups are not generally optimal. Much closer to fixed part of the two-part tariff, and there is no
the truth is the proposition that optimal pricing demand distortion. If, however, the number sub-
causes the same proportional quantity reduction for scribing to the service varies with the fixed charge,
each product, relative to quantities demanded at then there should be some mark-ups on the variable
marginal cost pricing. More precisely, Ramsey prices prices. Ramsey principles apply not just to uniform
are such that a small proportional change in tax and two-part tariffs but also generally to the theory of
rates, where (pi ci) is the tax on product i, would optimal non-linear pricingsee Wilson (1993, es-
cause the same proportionate reduction in the com- pecially ch. 5).
pensated demandi.e. substitution effectfor each
product.4 Thus the equi-proportionate quantity re- (ii) Distributional Concerns
duction principle is correct for small increases in
taxes and provided that income effects are ignored. The principle that mark-ups should be higher on prod-
ucts for which demand is less price-sensitive, and
This description of Ramsey pricing is much more especially its implications for the fixed element in
general than the inverse elasticity rule, which says charges, seems to conflict with concerns for income
that price/cost mark-ups should be inversely propor- distribution.5 High fixed charges on essential services,
tional to elasticities of demand. That rule is, of like poll taxes, are regressive. Ramsey principles can
course, implied by the more general principle in the readily be adapted to accommodate distributional con-
commonly assumed case of independent demands cerns and, moreover, the Ramsey tax problem is better
and no income effects. In some special cases, the motivated when they matter, because otherwise a uni-
general principle implies equal mark-ups, but gener- form poll tax would be rather efficient.
ally it does not. It follows that Ramsey pricing
generally entails price discrimination: if ci = cj, the The many-person Ramsey rule involves welfare
Ramsey prices pi and pj are likely to differ. weights. The welfare weight of consumer h is the
3
The classic paper is Ramsey (1927).
4
See Mirrlees (1976). Note that the tax is defined here as the difference between price and marginal cost, whereas the mark-up
is that difference in proportion to price. So taxes are measured in money units whereas mark-ups are dimensionless numbers (e.g.
percentages).
5
Hancock and Waddams Price (1995) analyse distributional implications of pricing structure changes as liberalization happens
in the British gas industry.

17
OXFORD REVIEW OF ECONOMIC POLICY, VOL. 13, NO. 1

increase in social welfare resulting from a small and unaccountability. Regulators, perhaps like cen-
increase in hs income. Depending on ones attitude tral bankers, should have focused objectives.
to income distribution, poor people might have higher
welfare weights than rich people. Let the marginal (iii) Effects of Competition
social value of an income transfer (MSVT) to h be
equal to hs welfare weight minus the costs of How does competition affect this analysis? If compe-
meeting the extra demands resulting from the trans- tition to the dominant firm takes the form of a price-
fer. taking fringe, it is straightforward formally to extend
the Ramsey analysis to the competitive case. Let si(p)
The Ramsey principle about equi-proportionate re- be the supply of the competitor group function for
ductions in compensated demands still holds, but in product i. Then the residual demand curve facing the
relation to MSVT-weighted demands. With optimal dominant firm is given by xi(p) si(p). If competitor
pricing, a small proportional increase in taxes profit and consumer surplus have the same welfare
would cause the total substitution effectsi.e. the weight, then the principles in section II(i) above, with
(unweighted) compensated demand changesto be residual demand in place of total demand, apply
proportional to MSVT-weighted demands. Depend- equally to the competitive case. If not, then welfare
ing on the welfare criterion, and on what has been weightings modify the analysis along the lines of
achieved by other distributional policy instruments, section II(ii).
the many-person Ramsey rule could even entail
subsidies of some products. In any event, Ramsey Thus, in principle, it is quite easy to characterize the
principles are entirely capable of accommodating optimal pricing structure, given the need to cover
distributional concerns. fixed costs and given that the price vector p is the only
instrument available. But unless fixed costs are small
It certainly does not follow, however, that regulators and distributional concerns are modest, this con-
have applied, or should be allowed to apply, anything strained optimum may fall well short of productive
like the principles of the many-person Ramsey rule. efficiency.
In practice, political economy considerations collide
with normative welfare economics. Most of the taxes If, for example, the optimum has a large tax (pi ci)
and subsidies implicit in utility prices historically are on product i, then the fringe will supply units of the
more plausibly the result of political incentives and product that could have been more efficiently sup-
pressures, coupled with inertia, than of rational plied by the dominant firma kind of cream-
normative calculation. Implicit taxes are less subject skimming. (The fringe supplying too much is analo-
to legislative scrutiny and approval than explicit gous to consumers demanding too little of the prod-
taxes. uct in question.) On the other hand, if product j is
subsidized at the marginfor distributional or po-
Should regulators take distributional considerations litical reasons, perhapsthen there is productive
into account in making their decisions? I think not inefficiency in the other direction as more efficient
except perhaps where specific duties with distribu- competitor supply is partially or wholly foreclosed.
tional aspects, together with the financial and other Moreover, a subsidy to product j that is not financed
means of carrying out those duties, have been given independently will generally increase the tax, and
to them by government or parliament. The basic hence productive inefficiency, associated with prod-
reason is that distributional policies are generally uct i.
better pursued by other branches of government,
which is not to say that the latter can achieve a There are three broad approaches to the possible
distributional outcome that could not be improved by conflict between liberalization and the tax/subsidy
the use of regulatory instruments. Rather, the view is patterns that might exist in pricing structures. The
that the advantages of regulators having discretion to first is not to pursue liberalization. Then monopoly
pursue distributional ends are outweighed by disad- profits can be used not only to cover fixed costs but
vantages of capture, influence activities, uncertainty, also to pursue various distributional ends. But this

18
J. S. Vickers

fails to exploit the various benefits of competi- would be better to give some pricing structure discre-
tion, and tends to result in the persistence of some tion to the firm.
subsidies that are both inefficient and unwar-
ranted. The general advantage of giving the firm some
pricing discretion is that decentralized information
The second approach is to liberalize but not worry too can be exploited. Thus, provided that the regulated
much about potential cream-skimming. Indeed, some price index is reasonably well constructed, we might
encouragement to entry via the tax wedges (pi ci) expect relative prices to reflect relative costs or
might be thought legitimate, at least in the short term, elasticities better if the firm has some freedom over
in view of various barriers to entry that may exist, pricing structure. But there are dangers that freedom
incumbent advantages, infant firm considerations, and will be abused, for example anti-competitively. Thus
so on. But, especially as liberalization extends, this is pricing structure freedom is often limited by more or
not a very satisfactory position either. less explicit restrictions on prices in relation to costs
(as with rules against cross-subsidy) or in relation to
The third and most attractive approach is to address each other (as with bans on price discrimination).
both the tax/subsidy issue and the entry barrier The question of how best to influence and constrain
problem directly. Rebalancing of prices, which com- the exercise of pricing discretion is a difficult one.6
petition will sooner or later compel unless it is
hindered, is the way to tackle subsidies that turn out (i) Global Price Caps
not to be desired once their costs are exposed, and
public subsidy or explicit taxes are needed to recon- One proposal, which Laffont and Tirole (1996) have
cile efficient competition with desired subsidies. analysed, is to have a single global price cap on the
(Indeed, liberalization might assist the provision of firms product range. This would cap an index P(p)
explicit subsidies by reducing and revealing their of the firms prices but leave the firm free to choose
costs.) That is, an additional instrument is needed to its pricing structure subject to that cap. The general
remove conflict between objectives. This is entirely advantage of discretion can be illustrated by Figure 1.
practicable, as the fossil fuel levy in electricity and Compared with the non-discretionary price vector
the proposed universal service fund for telecommu- p, it would clearly be better to allow the firm
nications illustrate. freedom to choose any prices such that v(p) > v(p)
because the firm would generally do better and
The tension between competitive pressures and im- consumers (in aggregate) would be no worse off. (In
plicit taxes in pricing structures has been most acute this case the index P(p) is simply v(p).) This will
in the context of network access pricingsee further induce a Ramsey pricing pattern, though probably
section IV below. with non-zero profits, but it does require a lot of
knowledge about consumer preferences.

III. PRICING DISCRETION AND THE Less informationally demanding, and even simpler,
PROBLEM OF INFORMATION is the linear price constraint given by pi xi(p) > pi
xi(p), which says that the firm can set any prices that
Ramsey pricing requires a lot of information about do not increase the cost of buying the product bundle
cost and demand conditions. The regulated firm is demanded at the non-discretionary pricessee Fig-
likely to know much more about that than the regu- ure 1. Again, the firm will generally gain from having
lator. Moreover, if price controls are committed for the discretion, and so too will consumers.
a period of years ahead, industry conditions may well
change before they are reset. Rather than imposing a If the weights in a linear global cap are set correctly
particular price vector upon the regulated firm, it in proportion to the quantities demanded at the
therefore seems likely that in many circumstances it resulting pricesthen Ramsey pricing is induced.

6
The general theory of incentive design with multi-dimensional uncertainty is not well developed. Armstrong (1996) examines
the multi-product monopoly non-linear pricing problem, which is of this type.

19
OXFORD REVIEW OF ECONOMIC POLICY, VOL. 13, NO. 1

Figure 1
Advantage of Pricing Discretion
p2

linear price constraint

p p
2

Isov curve

p
p 1
1

Thus a very simple form of price cap can bring about (ii) Floors and Ceilings to Combat Cross-subsidy
optimal pricing in a decentralized manner that ap- and Predatory Pricing
pears not to be too informationally demanding for the
regulator. It could be said that setting the weights Another, complementary, approach to the issue of
exactly right would call for a great deal of informa- discretion is to bound the freedom of the firm by
tion, but this does not upset the general point that prohibiting cross-subsidies or predatory pricing,
pricing structure discretion is often desirable, and perhaps by setting floors and ceilings between
that it can be implemented in simple ways. which prices are allowed to vary. Thus Baumol and
Willig, among others, have advocated the combined
Armstrong and Vickers (1996) analyse the desirabil- use of incremental cost floors and stand-alone cost
ity of pricing discretion, and the optimal form of ceilings on prices.7
discretion, for a series of examples involving cost or
demand uncertainty. While some discretion is desir- The (average) incremental cost of product i is the
able when there is cost uncertainty, this is not addition to total cost caused by producing i rather
necessarily so for demand uncertainty. Thus zero than not producing it at all, divided by the number of
discretion might be best if the demand elasticity is units of product i, holding constant the production of
higher in markets where the level of demand is other goods. The (average) stand-alone cost of prod-
higher. This is because the firm likes to charge higher uct i is the cost per unit of producing i in isolation.
prices in larger markets, but it is socially desirable Consider the example with constant marginal costs in
(for Ramsey reasons) for prices to be inversely section II(i) above, and assume that the fixed cost F
related to elasticities. When discretion is desirable, it is completely jointi.e. the entire fixed cost F is
is shown to imply Ramsey-like pricing in some entailed by the supply of any product. Then the
simple symmetric examples, but departures from incremental cost of product i is ci , and the stand-
Ramsey pricing are optimal more generally for rea- alone cost is [ci + F/xi ]. These floors and ceilings
sons having to do with asymmetric information. must apply not only to individual products but

7
See, for example, Baumol and Sidak (1994, chs 5 and 6).

20
J. S. Vickers

also to sets of products. Thus, for example, the also discusses alternative approaches. However, there
stand-alone cost of products i and j in combination is is an important point about a possible effect that
[ci + cj + F/(xi+xj)]. Otherwise a lot of profit could be regulation might have on incentives for predatory
made by charging the stand-alone cost for each behaviour. The discussion of pricing discretion in
product. section III(i) above, which was in a static setting,
implicitly assumed that the firm always maximized
These incremental and stand-alone cost tests would profit, which rules out predatory pricing because a
be met in a contestable market thanks to the forces of necessary condition for predatory pricing is (tempo-
potential competition, but why use them in a regu- rary) sacrifice of short-run profit. But incentives for
lated monopoly? One point is that they do not require such behaviour might exist and, indeed, might be
demand information. Another argument is that they strengthened by broad price caps insofar as price cuts
give precisely the notion of subsidy-free prices, in one market allow price increases elsewhere, thereby
because consumers of each product at least cover the reducing the profit sacrifice entailed by predatory pric-
extra cost that their consumption causes, so no ing. This suggests that constraints to guard against
financial burden is placed on others, and no consum- predatory pricingperhaps in the form of floors and
ers pay more than they would pay if they broke away ceilingsmight be a sensible accompaniment to broad
and went to an alternative supplier with access to the price caps (see further section IV below).
same technology. Subsidy-free prices do not always
exist, but under reasonable natural monopoly condi- (iii) Bans on Price Discrimination
tions Ramsey prices are in the non-empty set of
prices that are subsidy-free, as are many other prices. Unlike price floors and ceilings, which constrain
pricing structure by reference to costs, bans on price
Baumol (1996), following the spirit of Areeda and discrimination do so by placing cross-restrictions on
Turner (1975),8 has urged a similar approach to prices, so their enforcement does not require cost
predatory pricing. Baumol argues that pricing above information. Price discrimination can be defined in
average avoidable cost does not threaten to drive out general terms as the sale of different units of a
a more efficient rival firm and so cannot be predatory. product at prices not corresponding to differences in
Avoidable cost is the same as incremental cost, the cost of supplying them.9 This definition covers
except that sunk costs are ignored. As with the non-linear pricing as well as cases where different
subsidy tests described above, this test of predatory terms are offered to consumers in different markets
pricing applies to products in combination as well as despite common cost conditions. It also covers cases
individually. Which costs are avoidable depends on of uniform pricing (e.g. geographically) where costs
the time horizon. In the long run, for example, all differ.
costs are avoidable, and the approach boils down to
average incremental cost (which is simply average As explained above, optimal pricing generally entails
cost in the single product case). Baumol (p. 62) price discrimination when there is a need to finance
argues that the relevant time horizon is that over fixed costs (or subsidies that are not funded other-
which the price in question prevailed or could reason- wise). So a benign social planner should be allowed
ably have been expected to prevail. This might be to engage in price discrimination. But it does not
hard to judge. follow that the same is true for a profit-maximizing
firm. Indeed, analysis for unregulated monopoly
A number of the criticisms that have been made of the shows that the welfare consequences of profit-max-
AreedaTurner rule, for example its possible vulner- imizing price discrimination are ambiguous (see
ability to strategic behaviour, apply also to Baumols Varian, 1989). Unless price discrimination causes
proposal. We shall not rehearse those criticisms here output to rise, it is bad, but it might desirably open up
since they do not concern regulationsee Ordover new markets (e.g. markets for high-demand consum-
and Saloner (1989) for an analytical survey which ers in the case of non-linear pricing).

8
Areeda and Turner advocated a marginal cost criterion in principle, proxied by average variable cost in practice.
9
See Whish (1993, p. 503).

21
OXFORD REVIEW OF ECONOMIC POLICY, VOL. 13, NO. 1

More relevant here are the welfare consequences of of inefficient entry and cream-skimming if access is
price discrimination by a regulated monopolist. Natu- required to be sold too cheaply.
rally, these depend on how the monopolist is regu-
lated.10 With linear price caps that have fixed weights The access pricing problem can properly be viewed
proportional to demands, price discrimination is as an instance of the general pricing problem. Al-
good for consumers as well as the firmsee section though the discussion in the previous two sections
III(i) above. But average revenue regulation, which implicitly supposed that p was a vector of retail
caps [pixi(p)/xi(p)], though it causes output to prices, it applies equally well if p includes wholesale
rise, tends to be detrimental to consumers and can prices. But the controversy that surrounds the access
have very distorting incentives on pricing structure. pricing problem is such that it deserves a section of
its own. The two leading approaches to the problem
Another approach is to allow price discrimination are the efficient component pricing rule (ECPR)
subject to the constraint that no consumer is worse advanced by Baumol, Willig, and others,11 and the
off than with uniform pricingi.e. that all continue Ramsey/global cap analysis of Laffont and Tirole
to have the option of buying at the uniform price. This (1994, 1996).
may have attractions, especially in the case of non-
linear pricing: discounts to large buyers could happen (i) The Efficient Component Pricing Rule
without harming smaller consumers (unless the latter
are firms competing with the former). Consider a vertically integrated firm M that supplies
a retail output and an upstream input (say network
A significant difference between this form of op- access) necessary for producing the output. Let p and
tional tariff and a global cap is that giving a discount a respectively denote Ms retail and access prices. (In
to one consumer group does not relax the price more general analysis these would be vectors and
constraint for others, and so any incentive to engage non-linear pricing might be allowed.) Let b and b+c
in predatory pricing may be diminished. Bans on denote the marginal costs of access and of final
price discrimination, by increasing the cost to the output, so c is the marginal cost of the downstream
firm of predatory pricingsince price has to come activity. Firm M also has fixed cost F, which is
down in all markets, and not just those where entry assumed joint, so b and c are incremental as well as
has occurredmight further reduce the risk of it marginal costs. Firm M faces price-taking
occurring. At the same time, however, such bans competitor(s) downstream whose product might be
might prevent the firm from legitimately and pro- differentiated from Ms. The question is how a
competitively responding to rivalry. should be regulated.

In general terms, the ECPR says that a should be set


IV. THE PRICING OF INPUTS SOLD TO equal to the direct incremental cost of access plus the
COMPETITORS opportunity cost of (i.e. lost profit from) supplying
it. The ECPR is often expressed as the margin rule:
Rivals to BT and British Gas have had to obtain key
inputs from themin particular, access to telecom- a = p c, (1)
munications or pipeline networks. The pricing of
access to essential facilities can be complex even which says that the margin (p a) between Ms retail
when (as now in rail and electricity) their ownership and access prices should equal its incremental cost in
is separate from competitive activities, but it is the competitive activity. If Ms opportunity cost of
especially controversial when there is vertical inte- supplying the input to competitors is [p (b+c)] per
gration. On the one hand, it seems that vertically unit, then (1) corresponds to the opportunity cost
integrated dominant firms might have a strong inter- formulation since it can be expressed as
est in denying rivals access to key inputs on reason-
able terms. But, on the other hand, there are dangers a = b + [p (b+c)]. (1a)

10
See Armstrong et al. (1994, especially section 3.3.2).
11
See, for example, Baumol and Sidak (1994, ch. 7).

22
J. S. Vickers

The reasoning behind the ECPR is that it is necessary prices are the best that can be done given the instru-
for productive efficiency. A lower access price (in ments available and constraints. If there were no
relation to the retail price) would lead to excessive fixed cost recovery problem, then marginal cost
entry by inefficient rivals, and a higher access price pricing would be optimal: a = b and p = b + c. That
might deter more efficient rivals. On the face of it, the satisfies (1), but the ECPR was hardly intended as a
ECPR has the further advantage of being cost- corollary of marginal cost pricing.
based and not requiring information about demand
elasticities and so on. Laffont and Tirole (1996), responding to the com-
mon criticism that Ramsey pricing requires an unre-
The immediate natural objection that the ECPR is a alistic amount of demand information, have argued
recipe for the preservation of monopoly profitsand that an appropriate global price cap that embraces
associated allocative inefficiencyis rebutted by its access and retail prices symmetrically can induce
proponents when they say that its full validity is Ramsey pricing in a decentralized manner (and,
conditional upon the retail price p being regulated so moreover, avoid other competitive and regulatory
that no supernormal returns accrue to the firm. distortions that asymmetric regulation might cause).
The logic is essentially as described in section III(i)
(ii) Ramsey Pricing of Access above. The worry that such a global cap might
encourage a predatory price squeeze against rivals
Laffont and Tirole (1994) analyse the problem of i.e. raising a while lowering pcould be met by
choosing p and a to maximize welfare when the fixed supplementary constraints, possibly of the form p
cost F has to be financed from revenues. Their base a > c, which is similar to the ECPR margin rule (1).
model has price-taking rivals who have constant
returns to scale and whose product is differentiated (iii) A Synthesis?
from that of firm M. This being a Ramsey problem,
it is not surprising that the (constrained) optimal Notwithstanding this last point, the ECPR and
access price has the form Ramsey approaches appear rather different.
Armstrong et al. (1996) attempt to provide a synthe-
a = b + [Ramsey term involving super- sis. They note, first, that optimality of the margin
elasticities of demand],12 (2) rule (1) depends on some key assumptions that have
not always been made clear in expositions of the
which looks rather different from the seemingly ECPR. These include:
cost-based ECPR. Whether (2) gives a higher
access price (in relation to p) than (1) is ambiguous. homogeneous productthe rivals and M supply
It is generally desirable that access revenues should the same retail product;
contribute to the fixed-cost recovery problem, which fixed coefficients technologyone unit of output
the ECPR ignores. This is a reason for raising a requires one unit of input; and
above its ECPR level. On the other hand, Laffont and no bypassonly M supplies that input.
Tirole have product differentiation in their model,
and this is a reason for lowering a (see further When these assumptions are relaxed, and in the
below). absence of a fixed cost recovery problem, the optimal
access price a for a given (optimal or not) retail price
The LaffontTirole model is more general than the p, is given by
simple ECPR model in two major respectsit takes
account of the need to cover fixed costs and it allows a = b + [p (b + c)], (3)
for product differentiation. Its solution does not
satisfy the necessary condition for productive effi- where is the displacement ratio defined as [Change
ciency because of the fixed-cost recovery constraint. in Ms sales of final output as a changes] divided by
As with Ramsey pricing more generally, the optimal [Change in Ms sales of input to rivals as a changes].

12
Super-elasticities of demand take into account cross-price effects as well as own-price effects. Where products are (imperfect)
substitutes, the super-elasticity of demand for each product is lower than its own-price elasticity.

23
OXFORD REVIEW OF ECONOMIC POLICY, VOL. 13, NO. 1

The second term on the right-hand side of (3) is the prices when there are fixed costs to be recovered. (Of
opportunity cost of access, and hence (3) can be course, they imply marginal cost pricing when there
regarded as a more general expression of the ECPR is no fixed-cost recovery problem.) One way of
than the margin rule (1). expressing the Ramsey access price is as the ECPR
price plus a normal elasticity term.
Only if the three assumptions above hold does the
displacement ratio = 1, in which case (3) and (1) are (iv) Multi-product Implications
equivalent. Otherwise < 1, so (3) implies a lower
access price, for a given retail price, than the margin Some of the main implications of the access pricing
rule (1). Corresponding to the relaxation of the three principles discussed above arise in multi-product
assumptions, can be decomposed into terms relat- settings. Suppose, for example, that there is just one
ing to product differentiation, variable coefficients, basic kind of access, but that it can be used to provide
and bypass. Calculation of necessarily involves a a variety of different services.13 Ramsey principles of
good deal of elasticity information (cross-price efficient cost recovery generally imply that access
elasticities of demand, elasticities of substitution, the put to different uses should be priced differently
elasticity of alternative input supply). This is not (according to elasticities, etc.). So if usage-depend-
apparent from (1), because (1) is correct only if ent pricing of access is feasible, it is usually optimal.
assumptions hold which imply that the relevant
elasticities are zero or infinite. If fixed-cost recovery is not a binding constraint, but
Ms retail price vector p is fixed by regulation
If the fixed-cost recovery constraint binds, then the (optimally or not), then (3) suggests that usage-
optimal p and a satisfy an expression of the form: dependent pricing of access is warranted because the
opportunity cost to Mthe product of the displace-
a = b + [p (b + c)] ment ratio and Ms retail margindiffers as between
+ [Ramsey term involving standard own-price uses. Thus access used by rivals to supply direct
elasticity]. (4) substitutes for profitable services supplied by M
might be priced more highly than access used to
The last term in this expression is a standard own- supply services complementary to those offered by
price elasticity, not a super-elasticity as in (2). But M. (In fact, matters are somewhat more complicated
since (4) addresses the same question as that an- in the multi-product case because the overall oppor-
swered by Laffont and Tirole, it should be equivalent tunity cost to M is the sum of the opportunity costs
to (2). Indeed, the super-elasticity in (2) can be on its various offerings.)
expressed as the sum of the second and third terms on
the right-hand side of (4). So if (4) does indeed give Further questions arise when the set of products is
the precise meaning of the ECPR, it follows that the endogenous. There is a case for saying that access
optimal access price derived by Laffont and Tirole can used to provide new services ought to be priced
be viewed as the ECPR price plus a standard Ramsey lowere.g. temporarily exempted from Ramsey
term. taxationin order to encourage innovation. There
is a concern that adoption of the ECPR might
Thus the ECPR and LaffontTirole Ramsey ap- encourage M to introduce products inefficiently in
proaches appear capable of some form of synthesis. order to create and recoup opportunity costs that
The ECPRin its general opportunity cost formu- would not otherwise exist. These examples illustrate
lation, of which the margin rule is just a special the general point that ex-post optimality (of pricing in
casegives the optimal access price for a given retail our context) might distort ex-ante incentivesby
price when there is no fixed-cost recovery problem. failing to encourage desirable behaviour or by en-
Ramsey principles give optimal retail and access couraging undesirable behaviour.

13
Section IV of Armstrong et al. (1996) contains analysis of the more general case with N final products offered by the incumbent,
R final products offered by the competitive fringe, and M types of access.

24
J. S. Vickers

V. PRICE CONTROL UNDER PARTIAL globalin effect the index weights of deregulated
DEREGULATION prices become zero. It certainly does not follow
that the case for discretion over the structure of the
Firm Ms retail pricing was assumed to be compre- prices that remain regulated is undermined, but
hensively regulated in the analysis of access price care is needed to avoid distortions induced by
regulation discussed in the previous section. But regulation.
what if some or all of Ms retail prices are deregulated?
This question is becoming increasingly important as For example, suppose that firm M supplies an input
liberalization policies in the regulated industries (access) and an output to each of two markets. At
offer the prospect that the scope of monopoly price retail level market 1 is monopolized, but market 2 is
controls might be reduced. reasonably competitive. Firm M has a monopoly on
the supply of access to both markets. In natural
In terms of the basic framework above, in the event notation its four prices can be denoted (a1, a2, p1, p2).
of deregulation, the retail price p becomes endog- If p2 is removed from price control and if a single
enous and influenced by the regulated access price a. linear price cap is applied to the other prices with
The way that p(a) varies with a then enters the weights proportional to the quantities of them de-
analysis of optimal a. If firm M still has some market manded, then an obvious distortion would arise.
power, there is reason to reduce a, relative to (3) Firm M would have an excessive incentive to raise a2
above, in order to bring p closer to marginal cost in order to induce a higher p2. (Recall that Ms rivals
(b + c) and hence improve allocative efficiency. On in market 2 require the access input from M. Their
the other hand, as a comes down, the margin rivalry constrains the retail margin rather than the
[p(a) a] might widen, so that productive effi- retail price level.) With appropriate index weightings
ciency worsens as rivals inefficiently take more this incentive could be curbed, but the design of a
business from firm M. In any event, Ms break-even single price cap would appear to be significantly
constraint might limit the extent to which a can be more complex when that cap ceases to have global
reduced. It is not clear that the ECPR gives much coverage.
useful guidance in these circumstances.
Pricing structure issues arising from partial deregu-
In multi-product settings, deregulation of a subset lation are of considerable policy importance, but are
of prices requires amendment to global price cap imperfectly understood. They provide further open
proposals. Caps with partial coverage are not questions in the theory of pricing discretion.

REFERENCES

Areeda, P., and Turner, D. (1975), Predatory Pricing and Related Practices under Section 2 of the Sherman Act, Harvard
Law Review, 88, 637733.
Armstrong, M. (1996), Multiproduct Nonlinear Pricing, Econometrica, 64, 5175.
Vickers, J. (1996), Multiproduct Price Regulation under Asymmetric Information, mimeo, University of
Southampton.
Cowan, S. G. B., and Vickers, J. S. (1994), Regulatory Reform: Economic Analysis and British Experience,
Cambridge, MA, MIT Press.
Doyle, C., and Vickers, J. S. (1996), The Access Pricing Problem: A Synthesis, Journal of Industrial Economics,
44, 13150.
Baumol, W. (1996), Predation and the Logic of the Average Variable Cost Test, Journal of Law and Economics,
39, 4972.
Sidak, J. G. (1994), Toward Competition in Local Telephony, Cambridge, MA, MIT Press.
Hancock, R., and Waddams Price, C. (1995), Competition in the British Domestic Gas Market: Efficiency and Equity,
Fiscal Studies, 16(3), 81105.
Laffont, J.-J., and Tirole, J. (1993), A Theory of Incentives in Procurement and Regulation, Cambridge, MA, MIT Press.
(1994), Access Pricing and Competition, European Economic Review, 38, 1673710.

25
OXFORD REVIEW OF ECONOMIC POLICY, VOL. 13, NO. 1

Laffont, J.-J., and Tirole, J. (1996), Creating Competition through Interconnection: Theory and Practice, Journal of
Regulatory Economics, 10, 22756.
Mirrlees, J. (1976), Optimal Tax Theory: A Synthesis, Journal of Public Economics, 6, 32758.
Ordover, J., and Saloner, G. (1989), Predation, Monopolization and Antitrust, in R. Schmalensee and R. Willig (eds) (1989).
Ramsey, F. (1927), A Contribution to the Theory of Taxation, Economic Journal, 37, 4761.
Schmalensee, R., and Willig, R. (eds) (1989), Handbook of Industrial Organization, Amsterdam, NorthHolland.
Varian, H. (1989), Price Discrimination, in R. Schmalensee and R. Willig (eds) (1989).
Whish, R. (1993), Competition Law, 3rd edn, London, Butterworths.
Wilson, R. (1993), Nonlinear Pricing, Oxford, Oxford University Press.

26