Four Reasons Why High Access Charges May Not Facilitate Collusion


This section provides a number of caveats to the view developed in the previous section that networks can raise each other's costs and facilitate retail collusion through high access charges. For this purpose, it relaxes in sequence several assumptions underlying this view: weak endogenous-marginal-cost effect, absence of second- and third-degree price discrimination, and pricing of calls to the caller only.

Evasion of the Access Tax through the Buildup of Market Share

In order to obtain the monopoly price at the retail level, the networks must agree on an access charge above marginal cost. How much above that cost depends on the substitutability between the networks. Were they not to compete for market share, as in the case of international telecommunications with monopoly national carriers, it would be optimal for the domestic monopolies to set the settlement charge at marginal cost. The companies' perceived marginal costs would then be equal to the true marginal costs, and each would charge the monopoly price in its reserved territory. In general, though, competition for market share erodes markups, and so the enforcement of collusion requires raising the access charge in order to bring the retail price back to its monopoly level. Tougher retail competition, therefore, calls for higher deviations from marginal cost pricing of access.

The question then is, Does a high markup above the marginal cost of access indeed generate high retail prices? To see that this is not a foregone conclusion, let us return to the two effects unveiled in the previous section. First, for given market shares, a higher access charge raises marginal costs and thus retail prices. This is the raise-each-other's-cost effect we have focused on. Second, for a given access charge set above marginal cost, a network's perceived marginal cost per call decreases with its market share; this is the endogenous-marginal-cost effect.

A network's perceived marginal cost of a call depends on the product of the access charge markup (difference between the access charge and the marginal cost of access) and of the rival's market share. Thus, as long as the access charge departs little from marginal cost, a reduction in the rival's market share does not much affect the network's perceived marginal cost. And indeed, LRT shows that for small departures from marginal cost pricing of access, an equilibrium in retail prices exists and is unique, and the retail price is indeed an increasing function of the access charge. The endogenous-marginal-cost effect is small and is dominated by the raising-each-other's-cost effect.

When retail price competition is intense, however, the networks want to impose substantial markups on access in order to make up for the lack of markup at the retail level. But as the access charge grows, the endogenous-marginal-cost effect becomes more important, and a market share buildup starts reducing the networks' perceived marginal cost substantially. This cost reduction creates the temptation to charge low retail prices in order to avoid paying the "access tax."

Indeed, as the access charge grows above marginal cost, there comes a point at which the networks are no longer willing to charge a high retail price and instead want to undercut. Where this point lies depends on network substitutability. If networks are very substitutable, then small price cuts reduce perceived marginal costs substantially.

As an illustration, suppose that the networks succeed in generating the retail monopoly price through a high access charge. In a symmetric equilibrium, the networks share the market and furthermore do not make net access charge payments to each other because the flows in and out of a network are balanced. If networks are very substitutable, then a network can corner the market, that is, double its market share, through only a modest price cut. At the lower price, the undercutting network makes a little bit less than the monopoly profit on each customer, but has twice as many customers; and it still does not incur an access deficit, since the rival network has no subscribers and therefore there are no off-net calls. Ironically, it is precisely when very high access charges are required in order to sustain collusion that only small price cuts are needed for networks to reduce their fraction of off-net calls substantially. And indeed LRT shows that no retail equilibrium exists with high substitutability and high enough access charges.

Last, let us compare the network interconnection problem with the patent pool paradigm of section 5.3. The key difference between the two situations is that in the patent pool paradigm the firms cannot evade the wholesale tax. In particular, if a firm stops sharing and corners the retail market, this firm must pay (at least) twice as much to the patent pool venture and thus to its rival for the use of the patents. In contrast, a telecommunications carrier's net access payment does not increase when it stops sharing the market and corners it. Put differently, the firms' perceived marginal production cost is exogenous in the case of payments to a patent pool joint venture, and endogenous in the case of two-way access to local networks.

Nonlinear Pricing

Under linear retail pricing, a high access charge may facilitate collusion because a network cannot sell more telecommunications services to its subscribers without incurring an access deficit. This reluctance to boost per-customer volume is what gives rise to high retail prices. Under linear pricing, a network is torn beween the desire to expand market share in order to benefit from the fat premium from enlistment of customers and the concern about generating an access deficit, since both are associated with a price cut.

Suppose now that networks offer two-part retail tariffs: a monthly subscriber charge (cum perhaps a connection charge) and a per-minute calling rate or usage price. We previously assumed that the monthly subscriber charge is set equal to zero, or, more generally, that it is exogenously set. Let us, in contrast, assume that the networks' monthly subscriber charges, and not only the per-minute calling rates, are unconstrained. Networks now have two instruments each, and can separate the building of market share from the generation of call volume. While an increasing call volume is still associated with a decrease in the usage price, a network can use the subscriber charge to build market share without inflating its outflow.

Indeed, the higher the per-customer profit generated by high access charges, the tougher the competition for market share. Thus networks compete fiercely on the fixed-fee dimension and offer low, perhaps negative, subscriber charges. (An example of a "negative" subscriber charge is the giveaway by mobile operators of telephone handsets or sharp subsidies to their purchase.) A case in point is the U.K. market in which high termination charges lead to high connection subsidies for mobile subscribers (Armstrong, 1997b).

Considers a simple example of competition in two-part tariffs. The consumers' variable demand is known by the operators; actually it is the same for all consumers. A standard result on two-part tariffs is that with known consumer demand, firms optimally set the usage price at marginal cost and then extract consumer surplus through a fixed fee.

The application of this result to our context requires two remarks. First, the extraction of consumer surplus in oligopoly is limited by the extent of competition. Under tough competition (high substitutability), little surplus can be extracted from the consumer. Second, the usage price is set equal to the relevant marginal cost, namely, the network's perceived marginal cost. An access markup implies that the perceived marginal cost lies above the true marginal cost. That is, under an access markup the networks reduce volume, just as under linear pricing. The raise-each-other's-cost effect still prevails; however, it does not enable firms to collude because they wage a fierce war along the fixed-fee dimension.

The access charge is no longer an instrument of collusion. Indeed, the firms' equilibrium profit is entirely determined by the extent of their differentiation and is independent of the access charge. While this result is extreme, it illustrates well the general logic of the erosion through nonlinear pricing of the fat profits generated by access charge markups.


Termination-Based Price Discrimination

We have assumed that networks do not discriminate on the basis of whether the call terminates on their network or on the rival network. Yet, in the presence of an access markup (or for that matter, discount), a network's marginal cost depends on whether the call is on or off net. With an access markup, say, off-net calls are privately more costly to produce than on-net calls, and so, on purely cost-based grounds, networks would like to charge more for off-net calls. Note incidentally, that there is no technological obstacle to such price discrimination. In some instances there are no institutional constraints either, and it is indeed used. For example, in many countries, calls terminating on mobile networks are more expensive that those terminating on fixed networks. In the United Kingdom, cable companies (which offer local service) partly engage in termination-based price discrimination in that they offer free calls to subscribers belonging to their networks. In contrast, the incumbent BT does not engage in such termination-based price discrimination. 

This section, based on Laffont, Rey, and Tirole (1998b), explores the implications of allowing networks to discriminate in this manner.

An interesting feature of this new form of competition is that termination-based price discrimination reintroduces network externalities among consumers. Interconnection together with uniform (non-termination-based) pricing implies that consumers, when choosing a network, do not take into account the choice of network by the people they will want to call.In contrast, if on-net calls are cheaper than off-net calls, consumers are better off if the people they want to call select the same network. More generally, it is easy to see that an access markup (access charge above marginal cost) or an access discount (access charge below marginal cost) generates, respectively, positive or negative net work externalities among consumers. Such externalities can be labeled tariff-mediated network externalities.

First, high access charges do not facilitate collusion as well as under uniform pricing. The reason is basically the same as for nonlinear pricing. Suppose that the networks set two prices, one for on-net calls and the other for off-net calls; so, we consider a particular form of third-degree price discrimination but retain the linear pricing assumption. (The analysis is not much altered when combining the two forms of price discrimination; see Laffont, Rey, and Tirole, 1998b.) Whereas networks cannot build market share without incurring an access deficit under uniform pricing, they can do so with termination-based price discrimination. The access deficit depends on the volume of off-net calls, and therefore on the off-net price, but not on the on-net price; so a network can build market share by reducing its on-net price without increasing its access deficit. The reason why collusion may fail, as with nonlinear pricing, is that the network can compete along a dimension (on-net price here, monthly subscriber charge under nonlinear pricing) that does not inflate the access tax bill.

The second interesting insight relates to the welfare implications of allowing termination-based price discrimination. At first sight, one might believe that such discrimination is socially wasteful because it is neither cost nor demand based. Or, to be more precise, it is generated by a divergence between the privately perceived and the social costs of calls. The networks have no demand-based incentive to price discriminate according to termination, because a consumer's willingness to call another consumer is independent of the identity of the latter consumer's network. We conclude that termination-based price discrimination introduces a distortion in the consumers' marginal rates of substitution between on- and off-net calls.

We should, however, remember that the industry is in a second-best situation, and so adding a distortion to a set of existing distortions may turn out to increase social welfare. This is actually the case here in some circumstances. First, price discrimination intensifies competition, as we already noted. Second, when the networks are poor substitutes, price discrimination tends to alleviate the second marginalization generated by an access markup. Overall, the welfare implications of termination-based price discrimination are ambiguous.


Box 5.5 Reception Subsidies

Last, let us observe that networks may still compete away high access charges even if they charge linear prices and if they cannot discriminate on the basis of call termination. We have until now assumed that the call receiver neither paid nor received anything for receiving calls. In the presence of access charge premiums or discounts, though, networks may want to let the call receiver internalize the net benefit or cost to the network of terminating the call. In particular, when the access charge exceeds the marginal cost of terminating the call, the network may want to give the call receiver money for receiving the call.

Although the point can be made generally, it is most simply illustrated in the case of a very unbalanced calling pattern. Suppose that there are two types of consumers: those who only call, and those who are called and never call. Suppose further that the networks are undifferentiated. Then, in an equilibrium, both networks charge a linear price equal to the marginal cost of calls plus the access premium for calls, and pay a subsidy equal to the access premium for call reception. That is, both networks tag the full access premium onto the price of calls and subsidize call reception to the level of the access charge premium. 

Global Price Cap and Incentives to Exclude


Recall that a major drawback of LRIC regulation is that it does not allow the operator to make a margin on its bottleneck segment. The operator then has strong incentives to deny access to its rivals through nonprice methods, which in turn call for close monitoring by regulatory agencies. Incentives for exclusion are much attenuated if treatment of accessservices as a normal business segment is allowed and the operator can make money on them.

Under a global price cap (GPC), the operator manages its product lines "symmetrically," as it has no built-in incentive to favor one over another. In particular, excluding buyers of interconnection services amounts to mutilating a potentially quite profitable activity. A global price cap provides the operator with the flexibility to choose which product lines, retail or wholesale, are profitable. The theoretical analysis confirms that behavior that excludes rivals, raises their costs, or limits their demand tends to reduce the operator's profit. This reduction is particularly clear in the case where rivals have no market power and the exclusionary practice consists in raising the rivals' cost. The exclusionary practice then is tantamount to the operator's raising its own cost of providing the retail service through the competitive rivals. In contrast, we saw that even in this simple case the operator has a strong incentive to exclude under fragmented regulation. Therefore, one should not hastily transpose concerns that are legitimate under fragmented regulatory schemes to global price cap regulation.

Global price caps thus enable regulation to be more light-handed, for global price caps reduce perverse incentives and therefore diminish the need for regulatory oversight of the operator's decisions. A global price cap scheme, therefore, is more compatible with deference to the operator's business judgment than existing schemes. A global price cap, however, still involves discretion with respect to the weights in the cap and to their revision process.
Global Price Cap and Predation Global price caps raise the possibility of predation. To the extent that predation is often mingled with exclusion, a more common practice, this fear may be exaggerated, for we have observed that a global price cap tends to eliminate, rather than create, incentives for exclusion.

A price squeeze, however, is easy to carry out under a global price cap: The operator can increase the access charge and reduce its final price while keeping the price cap constraint satisfied. It thereby hurts its rivals on the retail market considerably. One can then conceive of use of this strategy for predatory purposes. That is, the operator might reduce its profit until the next price review, but eliminate rivals who otherwise would have been used by the regulator as benchmarks in the future. The profitability of such predatory behavior unfortunately has not yet been analyzed; one can only presume that the threat of predation is more relevant when competitors do not have a long purse and when their assets cannot easily be purchased and managed by another company in case of bankruptcy.