1

Network Economics, Dynamic Competition,

and Regulation

Howard ShelanskiCargèse, Corsica

May 3, 2005

2

What is a “Network Industry?”

Broad definition: An industry that connects users to each other or to producers through a system (or systems) of compatible, interoperable physical infrastructure.

Economic definition: An industry whose good or service increases in value to each individual user as the number of other users of the same good or service increases.

Examples:

Software platforms

Communications systems

3

What is a Network Externality?

The consumer’s perspective: A “network externality” is the added benefit a consumer gets as additional consumers join the same network or use the same product.

Telephone service benefits a consumer if she can call 10 other people, but it becomes increasingly valuable to her as she can call 100, 1000, or millions of other people.

The competitive firm’s perspective: A firm’s “network externality” raises rivals’ costs and creates a barrier to entry by new firms.

Once a firm has the market lead and provides the greatest network benefit to subscribers, rivals must not only beat the price and technology of the leading firm, but must do so by enough to compensate consumers for the network benefit they lose in switching to the smaller rival.

4

Are Network Effects Common or Rare?

Network effects might at first appear common. We benefit in our consumption of many goods as more people demand them. I have easier access to pizza, and pay less for it, because many others also like pizza and many suppliers have entered the market. But such price and output effects are not true network effects.

In a true network externality, the increased benefit comes directly from other people’s consumption, price effects held constant.

When defined in this way, network effects become less common, although they tend to appear in important industries.

Key: Ask whether, when price is held constant, the value of the service increases with the number of consumers.

5

Network Externalities Distinguished from Economies of Scale

Scale economies arise when a producer’s per-unit costs fall as output increases.

In contrast, network externalities arise when a consumer’s benefit grows as total consumption increases.

Scale economies and network externalities need not go together. A telephone network might raise its per-subscriber costs by expanding into a rural area thus reducing scale economies; but that same expansion increases the network externality the system provides by increasing the number of people each subscriber can call.

6

Network Externalities and Lock-In Effects

Lock-in effects arise when consumers find it more economically rational to stick with an existing product/service rather than to switch to a competing one.

Network externalities can create lock-in by making the alternative provider less attractive in terms of the benefit it will provide to the consumer. Even if a competing provider offered payment (or a very low price) to switch, a consumer might decline because of the network benefit she would lose.

Not all lock-in arises from network externalities; lock-in can occur if other switching costs of moving to the competing good or service are too high. Service termination penalties, incompatibility with already-purchased complementary goods, and sunk costs are factors that might keep consumers from switching even to otherwise better choices.

7

Network Externalities Pose Regulatory Challenges

Network markets may tip toward monopoly: As one network gains a market lead, it offers a greater benefit to consumers. New customers receive a greater network benefit from joining the leader and existing customers have less incentive to switch from the leader to a competitor. These forces grow in a self-reinforcing cycle that can lead to network monopoly.

But network monopoly is not, as in conventional markets, clearly bad for consumers, because they receive a large benefit from the network externality that can compensate for monopoly pricing.

A key question for regulators is therefore whether competition can be introduced without reducing the network externality.

If the network externality cannot be had without the monopoly, rules designed to end the monopoly may harm consumers. In such cases regulators may therefore have to make trade-offs between conventional price/output objectives and network externalities.

8

Regulation when Competition and Network Externalities can Coexist

When interconnection or inter-operability of networks is possible, then multiple networks can compete and jointly create the network externality for each network’s consumers.

But this does not mean regulators should mandate interconnection whenever possible.

Interconnection has costs:

Short-run costs of putting in place the facilities through which interconnection occurs—often a minor concern.

Long-run costs of deterring innovation and network investment: A firm may invest less in its network if the benefits of that investment will be shared by interconnecting rivals.

9

Basic Lessons

Network regulation will always have both costs and benefits

As markets and technologies change, marginal costs and benefits may change

Regulation of network industries is therefore likely to evolve over time as the comparative costs and benefits of regulation change.

10

A Simple Example—the Horizontal Case

Suppose 2 competing telephone networks have entered a market and the larger refuses to interconnect with the smaller.

Regulatory question: Should the government require the networks to exchange calls?

Benefits of regulation:

(1) Price competition: Consumers get lower prices than monopoly and get an even higher network benefit as the market grows through price effects.

(2) Non-price competition: May induce the rivals to introduce vertical features (like voice mail or caller ID) to attract customers.

11

(Simple example, cont’d)

Costs of regulation:

(1) Reduction of beneficial investment: network owners may have less incentive to invest in network improvements that rivals can share.

(2) Increase in wasteful investment: network owners may have incentive to invest in sorting and degrading the quality of traffic from the rival’s customers.

12

(Simple Example, cont’d)

How regulators decide what tradeoffs to make, and whether they should mandate interconnection, can shift over time.

Consider a 3-period scenario:

13

Period 1

Early period of network development: If there is not yet full deployment of the network, investment in infrastructure may be more valuable than competition among the developing networks. One € invested in the network has greater social value than one € saved on price. Long-run network benefit due to deployment is greater than short-run network benefit from interconnection. No mandated interconnection.

14

Period 2

Mature period of network deployment: Networks are fully deployed and the market structure might be monopolistic or divided among a few firms. Now, one € saved on price might have higher value than one € invested in the network. Short-run network benefit from interconnection is greater than the long-run network benefit from marginal deployment. If the firms have not voluntarily interconnected, mandate interconnection.

15

Period 3

Period of technological change: The original network becomes outdated and needs renewed investment. As the value of the services the old network can support declines compared to the services an updated network could support, the marginal value of one € saved on price declines relative to the marginal value of one € invested in the network. Long-run network benefits of new service deployment are higher than short-run benefits of new service interconnection. Deregulate interconnection, at least for new services and infrastructure.

16

Simple Example—Vertical Dimensions

Imagine that one of the two competing telephone companies in our example introduces a vertical service, voice mail perhaps.

Two regulatory questions arise:

(1) Must the firm allow other telephone companies access to its voice-mail service?

A possible tradeoff here is between investment in vertical services and competition in basic telephone service.

(2) Must the firm allow other voice-mail providers to serve its basic telephone subscribers?

A possible tradeoff here is between investment in vertical services by outside firms and investment by the network owner in the underlying network if such investment would benefit unaffiliated vertical competitors.

17

(Cont’d)

Of the two sets of tradeoffs above, (2) is probably the more important. As a general matter of competition policy we generally let firms benefit from their proprietary innovations and prefer to let the market force rivals to come up with their own competing innovations, so (1) is of less concern.

With respect to (2), network access for rival providers of vertical services, we have greater concern because of the practical impossibility and social undesirability of requiring a small creator of a vertical service to also integrate into ownership of a network over which to provide that service.

18

(Cont’d)

The benefits of competition and diffuse innovation in vertical services may be very high, but they will not always be higher than deployment and development of the underlying network. As in the horizontal case, the relative values of the activities will change, and hence so should the regulatory decision regarding vertical access. Again, if network regulation is to enhance welfare, it cannot be static.

19

Lessons from the Simple Example

Regulators intervening in network markets must be sensitive 2 things:

(1) Whether market participants will provide efficient levels of interconnection without regulatory intervention;

(2) What the welfare tradeoffs are from regulated interconnection at a given time in the development of the relevant market.

In sum, regulators must decide whether regulation is necessary to achieve interconnection and whether interconnection will on the whole be beneficial.

20

A Real Example: The Case of “Internet Neutrality.”

In general, “net neutrality” is the idea that the Internet should be open in all dimensions for the free flow of traffic and services.

Horizontal dimension: Networks should interconnect to exchange traffic on a non-discriminatory basis: customers should detect no difference in traffic receive from within their network or from outside.

Vertical dimension: Network owners should not be able to discriminate among service providers to which consumers seek access over the network.

21

Motivations for “Net Neutrality”

The argument, typified by Lessig and Lemley (2000), is that the internet is becoming the closed province of a few, powerful network owners who discriminate against rival networks and in favor of particular content and services. They argue that this discrimination deters innovation at the edge of

the network that has been so beneficial to society.

Advocates seek regulation that prohibits discrimination between or within networks, so that content and applications developed at the periphery can flow in a neutral, unimpeded manner between internet end users.

The argument advocates a strong form of interconnection and interoperability in access to internet transport networks.

22

What are the Limiting Principles?

Unless (1) unregulated networks will arrive at inefficient voluntary interconnection rules and (2) welfare increases monotonically with competition and interconnection, horizontal net neutrality cannot always be the correct policy.

In the vertical case, unless (1) non-discrimination always increases vertical innovation, and either (2) innovation in vertical services and applications is the source of all welfare gains, or (3) vertical non-discrimination rules never deter innovation in the network itself, then vertical non-discrimination requirements will not always be welfare-enhancing.

23

Horizontal Case: Is Regulation Necessary to Achieve Good Interconnection between networks?

Even this threshold question is difficult to answer and sensitive to assumptions about specific market facts.

24

Model 1 (adapted from Cremer, Rey, & Tirole (2000))

Two firms compete only for new customers;

The firms independently set interconnection quality Ө to a value between 0 and 1. If Ө = 0 the networks do not exchange traffic. If Ө = 1, subscribers to different networks reach each other as if they were on the same network.

Consumers’ willingness to subscribe to either network is measured by service quality si:

si = v[(ßi + qi) + Ө(ßj + qj)]

where v is the consumer valuation parameter, q stands for the number of new users attracted to network i or j, and ß is the number of existing users of network i or j.

25

(cont’d)

The Game: Each firm sets its value for Ө, and then sets the quantity of service it will sell. New subscribers then choose whether to subscribe, and which network to subscribe to.

Effects of Ө: (1) A higher value of Ө raises the value of both networks, and

attracts more new customers to both providers.

(2) But, the increased Ө will disproportionately benefit the smaller network by increasing the number of calls its subscribers can make with high quality by more than the expansion of high-quality calls for subscribers to the larger network.

The result will be that with a high Ө the larger network will get higher profit per new subscriber, but fewer new subscribers.

26

(cont’d)

Result: If existing customers have switching costs and demand is not saturated, then a firm’s incentive to degrade interconnection will increase with its market lead. Because the lowest value of Ө will determine si, the larger network controls the behavior of new consumers.

Policy implication: Regulatory intervention may be warranted in unsaturated markets with a clear market leader But, must still make sure interconnection will not have

offsetting costs through deterred investment in network deployment.

27

An Alternative Model

What if the specific conditions of the model just discussed do not hold? What if firms are not just competing for new customers but are competing for each other’s existing customers? This describes the case in which a market is mature and growth of the customer base is low (period 2 in the simple example).

Foros and Hansen (2001) show conditions under which increased interconnection quality will, under such circumstances, be in the interests of both the dominant network and the smaller rival.

28

(cont’d)

The intuition: As growth in quantity of service ceases to be an option for the firms, they engage in Bertrand price competition instead of a Cournot game (the firms cannot simply choose either to expand or reduce supply.)

The higher the interconnection quality, the higher the price each firm can set.

Both firms increase profits-per-customer from higher Ө.

The smaller network becomes relatively more attractive to consumers and takes market share from the larger one,

but not by enough to offset the increased profits from the rising price. So, the firms voluntarily interconnect with high quality.

29

Lessons for Horizontal Network Regulation

The under-provision of interconnection may occur at precisely the time that mandating interconnection will be most costly: i.e. when network deployment is growing and the service market is developing (period 1 in the simple model).

Before intervening, regulators should therefore have good evidence that (1) competition cannot survive without regulated interconnection and (2) that investment deterrence will not be so great as to cancel the benefits of network competition.

Several empirical studies (Crandall, et al. (2004); Hausman (2002)) suggest the tradeoffs are significant.

30

(Lessons, cont’d)

Whether firms will voluntarily reach high levels of interconnection quality cannot be determined solely from market structure. One cannot say that regulatory intervention is needed just because the market is a duopoly. Instead, specific market facts matter and the market may under-produce interconnection when demand is growing, but may sufficiently produce (or overproduce) interconnection when demand is static.

Even with the above question resolved, the welfare benefits of interconnection versus lost network investment needs to be determined.

Regulated horizontal network interconnection may not always be the correct policy.

31

Vertical Net Neutrality

Basic intent is to ensure that an applications provider at the edge of the network is never foreclosed from reaching consumers.

But who should be barred from discrimination, and what are the tradeoffs?

Unintended consequences: Hogendorn (2005) has shown that mandating non-discrimination by networks against ISPs could induce ISPs to compete among themselves by discriminating further upstream against content and applications providers.

Further barring vertical discrimination by ISPs will limit product differentiation and the alternatives available to consumers.

So, a simple mandate of vertical non-discrimination throughout the internet may be counterproductive.

32

Tradeoffs

Value of vertical net neutrality also depends on complex welfare calculations

Comparative economic values to society of different kinds of innovation affect net welfare benefits of strict net neutrality. So it is important to recognize that emphasis on innovation at the edge may have costs in the form of lost innovation in the platform.

Consumer choice may come at a cost, and some consumers may prefer less choice at a lower price (walled gardens) to more choice at a higher price. So it is important to recognize that emphasis on non-discriminatory accessibility may be good for some producers but less good for at least some consumers. Here there is also a hard balance between social welfare effects and consumer welfare effects.

33

Vertical Policy Implications

As digital networks compete, end-to-end neutrality should be a goal but not an unconstrained mandate:

One constraint should be ensuring sufficient incentives and ability to upgrade the platform even if doing so might, in the short run, affect access for some applications/content providers.

Regulators should not reflexively bar differences in access quality or exclusive deals between networks and applications providers:

Limited packages might benefit consumers. But a network owner should not be able to bar services that allow competitive package offerings. For example, preferential caching by the network for an affiliated content provider might be o.k., but barring access for a 3rd party provider of caching services should raise concerns.