Citation
Restructuring in network industries : competition and mergers in telecommunications

Material Information

Title:
Restructuring in network industries : competition and mergers in telecommunications
Creator:
Jamison, Mark A
Publication Date:
Language:
English
Physical Description:
ix, 126 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Customers ( jstor )
Economic competition ( jstor )
Long distance telephone services ( jstor )
Marginal revenue ( jstor )
Market prices ( jstor )
Monopoly ( jstor )
Prices ( jstor )
Production costs ( jstor )
Telecommunications ( jstor )
Textual collocation ( jstor )
Dissertations, Academic -- Economics -- UF ( lcsh )
Economics thesis, Ph.D ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 2001.
Bibliography:
Includes bibliographical references (leaves 119-124).
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Mark A. Jamison.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
026987146 ( ALEPH )
47291843 ( OCLC )

Downloads

This item has the following downloads:


Full Text










RESTRUCTURING IN NETWORK INDUSTRIES: COMPETITION AND
MERGERS IN TELECOMMUNICATIONS








By

MARK A. JAMISON


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2001

























Copyright 2001

By

Mark A. Jamison





















To my wife (Patti), my children (John and Laura),
and my parents (John and Judae)















ACNOWLEDGMENTS


I must first thank my supervisory committee members. Dr. David E. M.

Sappington, the chair of the committee, took numerous hours to give direction and advice and to correct errors. His knowledge and insight were invaluable. Dr. Steven Slutsky, a member, provided important guidance throughout my graduate studies and was a quick study of my many efforts. Dr. Chunrong Ai, a member, provided expert advice on econometric problems and the presentation of econometric studies. Dr. Heather Elms, the external member, gave good direction and insights beyond the normal range of economics. I hereby thank them all again.

Outside of my supervisory committee, I am grateful to Dr. Sanford Berg, who

made it possible for me to come to the university, and whose opinions and leadership are respected worldwide. I am also indebted to Dr. W. Andrew McCollough for his foresight in allowing Dr. Berg and me to embark on the enterprise that we have built at the university. I am also grateful for advice from Dr. Walter Beckert and Dr. David Figlio. A special thank you is due to my colleague, Janice Hauge, without whom graduate studies would have been impossible.

I would also like to thank Eric Chiang, Josh Silverboard, Jason Mills, Eileen Pun, Verizon, Sprint, SBC, and BellSouth for their assistance with data. Thanks are also due to James Sichter of Sprint and Mark Sievers, Ed Lowery, and Link Hoewing of Verizon for their willingness to discuss issues and provide insights.








Lastly, I would like to thank all of the Public Utility Research Center (PURC) and departmental staff that have made the world work during my tenure as a student. Cynthia Stehouwer and the rest of the PURC staff kept my project finances and tasks in order; Misty Swain kept my graduate program in order; Dian Studstill kept all administrative matters flowing; and Dorthy Padgett provided important assistance in producing my book.















TABLE OF CONTENTS
Page


ACKNOW LEDG M EN TS .......................................................................................... iv

A B ST R A C T ............................................................................................................... viii

CHAPTERS

1 RESTRUCTURING IN NETWORK INDUSTRIES: COMPETITION
AND MERGERS IN TELECOMMUNICATIONS............................. 1

2 MARKET CONCENTRATION AND INCENTIVES TO
DISCRIMINATE AGAINST RIVALS IN NETWORK
IN D U STR IES .......................................................................................5

2.1 Introduction........................................................................................ 5
2.2 The M odel........................................................................................ 10
2.3 Symmetric, Single Market Case ......................................................17
2.4 Monopolist Entry into a Competitive Market..................................18
2.5 New Competitive Entry into a Formerly Monopoly Market ............20
2.6 C onclusion .......................................................................................24

3 NETWORK INTERCONNECTION AND TELECOMMUNIATIONS
COMPETITION: THE CASE IN THE U.S......................................26

3.1 Introduction......................................................................................26
3.2 Literature R eview .............................................................................31
3.3 H istorical Background .....................................................................33
3.4 T he M odels ......................................................................................39
3.5 M odel R esults ..................................................................................54
3.6 C onclusion ....................................................................................... 63

4 NETWORK EXTERNALITIES AND CROSS-BORDER MERGERS
IN NETWORK INDUSTRIES...........................................................64

4.1 Introduction......................................................................................64
4.2 The M odel........................................................................................69
4.3 Q uality C hoices................................................................................75









4.4 Analysis of Pure Cross-Border Mergers..........................................79
4.5 M ixed M ergers.......................................................................... ...... 88
4.5 C onclusion .......................................................................................93

5 C O N C LU SIO N .........................................................................................95

APPENDICES

A PROOFS AND LEMMAS FOR CHAPTER 2.........................................99

B APPENDIX FOR CHAPTER 3..............................................................107

C PROOFS AND LEMMAS FOR CHAPTER 4.......................................110

R E FER EN C E S ......................................................................................................... 119

BIOGRAPHICAL SKETCH .................................................................................... 125













Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

RESTRUCTURING IN NETWORK INDUSTRIES: COMPETITION
AND MERGERS IN TELECOMMUNICATIONS By

Mark A. Jamison

May 2001

Chairperson: David E. M. Sappington Major Department: Economics

This dissertation addresses three issues. The first is whether large service providers in network industries, such as large Internet backbone providers, have an incentive to discriminate against their smaller rivals. Using Cournot interactions, each network provider simultaneously chooses quantity, quality for communications within the provider's own network (internal quality) and quality for communications between the provider's network and other networks (external quality). Larger networks choose higher internal quality than do smaller networks. All networks choose lower external quality when connecting with smaller networks. Large networks and small networks choose identical external quality when interconnecting with each other. Lastly, incumbent providers are sometimes willing to raise rivals' per customer costs, but never their interconnection costs.

The second issue is how regulatory policies relating to network interconnection affect entrants' incentives and abilities to enter local telecommunications markets and








expand supply. I find that low prices for reselling incumbents' services, leasing facilities, and exchanging traffic increase entry, but time for entry, access to incumbents' buildings, and reform of universal service subsidies do not. However, if prices for leasing facilities are low relative to incumbents' retail prices, then less entry occurs, presumably because incumbents exert effort to protect their retail profits. Higher prices for exchanging traffic increase entrants' market share, presumably because they target customers (such as Internet Service Providers) who receive more calls than they make. Low prices for reselling services do not cause entrants to choose reselling over other supply methods.

The third issue is how mergers affect market performance in a network industry. Network providers choose quality for communications within the provider's own network, quality for communications between the provider's network and other networks, and output. The dissertation shows that cross-border mergers provide firms with an incentive to increase industry output because mergers allow firms to internalize positive network externalities. The dissertation further shows that cross-border mergers improve the welfare effects of horizontal mergers in certain circumstances. Lastly, mergers that are both cross-border and in-market increase welfare if the aggregate size of the markets that the pre-merger firms had in common before the merger is large relative to the aggregate size of the markets that they individually served before the merger, and if the value customers place on network externalities is high.















CHAPTER 1
RESTRUCTURING IN NETWORK INDUSTRIES: COMPETITION
AND MERGERS IN TELECOMMUNICATIONS


This dissertation consists of three studies of the restructuring of communications industries.

The first study examines market concentration in the Internet. This is one of the issues that caused the collapse of the WorldCom-Sprint merger. I examine an oligopoly model in which firms simultaneously choose quantities, network quality for internal communications (internal quality), and interconnection quality for communications between networks (external quality). I show that firms base their quality choices on network size, customer value of interconnection, and the cost of interconnection. Networks choose lower external quality when connecting with smaller networks than when connecting with larger networks. Larger networks and smaller networks choose the same external quality when interconnecting with each other. I also find that larger networks will not optimally choose to sabotage rivals by raising the rivals' costs of interconnection.

I develop and explain my results by applying a static oligopoly model. Firms

have identical cost functions of providing quality, but may have different cost functions for production. Cost asymmetries arise from differences in network sizes. An incumbent can raise its rivals' costs, for example, by delaying access to essential facilities. Firms








compete for customers and seek to maximize their individual profits by simultaneously choosing quantity, internal quality, and external quality.

I examine a case in which two firms serve a single market and compete in this market against a third firm that has a monopoly in another market. I show that a Nash equilibrium exists in this setting wherein the firms choose identical external qualities. Furthermore, each smaller firm chooses an internal quality that is lower than its external quality for interconnecting with the larger firm. Smaller firms make these choices because the number of customers reached through a network affects firms' quality choices and the smaller firms' smaller quantities of customers are less valuable than the larger firm's larger number of customers. Also, because I assume that smaller firms are symmetric in size, the smaller firms choose external qualities for interconnecting with each other that are equal to their internal quality choices. I then examine a case in which the first two firms can enter the former monopoly market, but are subject to a cost disadvantage because of the incumbent's established network. I show that firms agree upon external quality and that the incumbent's quantity choice is larger than its rivals' quantity choices. Finally, I examine whether the incumbent would sabotage its rivals by raising their costs. I show that the incumbent may raise its rivals' costs for quantity in certain situations, but the incumbent never raises its rivals' costs for interconnection, even when it can do so costlessly. The incumbent is unwilling to raise rivals' interconnection costs because the higher costs would cause the rivals to choose an external quality that is below the incumbent's optimal choice.

In my third chapter, I find that regulators' basing interconnection prices on incremental cost has given incumbents an incentive to hinder competitive entry.








However, these prices have not led incumbents to hinder entrants' abilities to gain market share. I further find that the opportunity for entrants to place facilities in incumbents' buildings (called collocation) is important to encouraging facilities-based entry and causes entrants to substitute facilities and unbundled network elements for resale-based entry. Higher symmetric reciprocal compensation prices discourage entry and cause entrants to select customers that receive more calls than they send. High wholesale discounts for resold services encourage entry, but do not encourage entrants to substitute resale-based entry for facilities-based supply.

In my fourth chapter, I analyze mergers in network industries. I develop a model in which up to three firms serve three markets, which share network externalities. I first examine pure cross-border mergers, mergers between firms that do not serve any of the same markets prior to the merger. In the initial setting, each firm serves a single market, two of the three markets are served by monopolies and the remaining market is served by a duopoly. I examine the Cournot-Nash equilibrium quantities as a useful benchmark. I then impose an exogenous merger between one of the duopolists and one of the monopolies. The merger allows the new firm to internalize network externalities between the two markets that it serves. With quality constant, his increases quantities and improves welfare unless the merger increases costs by an amount greater than the marginal value of the network externalities that the merged firm internalizes. The merged firm increases its market share in the duopoly market. I show that a merger between larger firms improves welfare more than a merger between smaller firms. I also show that quantities and welfare increase with the number of markets that are added together by the cross-border merger.






4

This dissertation proceeds as follows. Chapter 2 examines network

interconnection and incentives to discriminate and raise rivals' costs. Chapter 3 is an empirical study of the effects of regulatory interconnection policies on the development of competition. Chapter 4 studies the effects of mergers. Chapter 5 is the conclusion. All proofs are in the appendices.














CHAPTER 2
MARKET CONCENTRATION AND INCENTIVES TO DISCRIMINATE
AGAINST RIVALS IN NETWORK INDUSTRIES


2.1. Introduction


The liberalization of telecommunications markets has been marked by a number of mergers and alliances, some of which have been record setting. In 1995, US telecommunications businesses were involved in acquisitions worth $39.1 billion. This increased to $154.8 billion by the first half of 1998. Global mergers and acquisitions in the information technology, communications, and media industries jumped 87 percent between 1997 and 1998 to $488.8 billion (Capron and Mitchell, 1997; Broadview, 1998). Telecommunications accounted for 20% of all merger and acquisition activity worldwide in 1999. Recent examples of mergers and alliances include the merger of Bell Atlantic with NYNEX in 1997, and then with GTE in 2000; SBC's acquisition of Pacific Telesis in 1997, Southern New England Telephone in 1998, and Ameritech in 1999; AT&T and British Telecom's formation of a global joint venture in 1999; WorldCom's purchase of MCI in 1998; and Vodafone's acquisition of AirTouch in 1999 and Mannesmann in 2000.

Competition regulators, utility regulators, and others often raise concerns that telecommunications mergers might decrease competition. Such concerns prompted the European Union (EU) to place restrictions on Global One and on the now defunct British Telecom and MCI alliance (McDavid, 1997). The EU also required MCI to divest a






6

portion of its Internet business as a condition of approving WorldCom's purchase of MCI (Cr6mer, Rey, and Tirole, 2000). The EU halted WorldCom's planned purchase of Sprint because the EU's competition commissioner believed that the merged company would dominate transmission of information over the Internet. Olbeter and Robison (1999) argue that market concentration in the Internet backbone in the US results in some rural states having little access to the Internet. In contrast, in a joint study, the US Department of Commerce and the US Department of Agriculture (2000) found that the high cost of broadband local lines, not access to the Internet, causes rural areas to have less broadband access than urban areas. Cr6mer et al. (2000) argue that larger Internet firms have an incentive to lower the quality of their interconnection with smaller rivals.

In contrast, some research indicates that telecommunications mergers can improve market efficiency. Jamison (1999a) explains how regulation in the US and government ownership of service providers elsewhere have held the industry to an antiquated structure for decades. Market liberalization is forcing incumbent companies to restructure to meet the new economic realities or risk failure. Weisman (1999) shows how mergers among incumbent local exchange companies make it more likely for the merged company to compete against other incumbent local exchange companies. Galbi and Keating (1996), Jasinski (1997), and Jamison (1998; 1999a) explain how network providers become global to attract large multinational communications customers who want their international communications to be provided by an integrated multinational network. Firms such as AT&T and WorldCom meet this demand by establishing local networks in multiple countries and integrating these local networks with their








international networks. Jamison (1999b) shows how this process can increase competition and improve welfare.

This dissertation extends this research by examining market concentration in the Internet. This is one of the issues that caused the collapse of the WorldCom-Sprint merger, the other issue being the US Department of Justice's concern that the combined company would have 27% of the US consumer long distance market. I examine an oligopoly model in which firms choose network quality for internal communications (which I call internal quality) and interconnection quality for communications between networks (which I call external quality). I show that firms base their quality choices on network size, customer value of interconnection, and the cost of interconnection. Networks choose lower external quality when connecting with a smaller network than when connecting with larger networks. I find that larger networks and smaller networks choose the same external quality when interconnecting with each other. I also find that incumbent firms will raise entrants' quantity costs, but will never sabotage rivals by raising the rivals' costs of interconnection.

I develop and explain my results by applying a static oligopoly model in which firms have identical cost structures for quality, but not necessarily for quantity. When a cost asymmetry exists, it results from an incumbent firm having a network that is larger than other firms' networks at the start of the game. Also, an incumbent can raise its rivals' costs, for example, by delaying access to essential facilities. I apply a two-stage game in which an incumbent chooses whether to sabotage rivals in the first stage. Firms compete for customers and seek to maximize their individual profits in the second stage by simultaneously choosing quantity, internal quality, and external quality. In contrast








with Cr6mer et al. (2000), who assumed that internal quality was fixed, I allow firms to choose internal quality. This generalization allows an expanded examination of firms' quality choices.

As a benchmark, I first examine a case in which three identical firms compete in a single market. I find that the firms make identical quality choices. I next examine a case in which two firms serve a single market and compete in this market against a third firm that has a monopoly in another market. I show that a Nash equilibrium exists in this setting wherein the firms choose identical external qualities. Furthermore, each smaller firm chooses an internal quality that is lower than its external quality for interconnecting with the larger firm. Smaller firms make these choices because the number of customers reached through a network affects firms' quality choices and the smaller firms' smaller quantities of customers are less valuable than the larger firm's larger number of customers. Also, because I assume that smaller firms are symmetric in size, the smaller firms choose external qualities for interconnecting with each other that are equal to their internal quality choices. I then examine a case in which the first two firms can enter the former monopoly market, but are subject to a cost disadvantage because of the incumbent's established network. I show that, as in the first situation, firms agree upon external quality and that the incumbent's quantity choice is larger than its rivals' quantity choices. Finally, I examine whether the incumbent would sabotage its rivals by raising their costs. I show that the incumbent may raise its rivals' costs for quantity in certain situations, but the incumbent never raises its rivals' costs for interconnection, even when it can do so costlessly. The incumbent is unwilling to raise rivals' interconnection costs








because the higher costs would cause the rivals to choose an external quality that is below the incumbent's optimal choice.

My results conflict with the findings of Cr6mer et al. (2000). Their model

assumes that differences in firm size result from differences in installed customer bases, which they call attached customers. Attached customers are customers who cannot change service providers nor their quantities purchased for the duration of the game. Firms choose external quality based in part on the value of allowing their own customers with connectivity to rivals' attached customers. Cr6mer et al. (2000) assume that firms must meet the quality demands of new customers, but incur no penalty for ignoring the preferences of their own attached customers. As a result, when two firms interconnect, the firm with the larger number of attached customers chooses a lower external quality than does its rival, which has the lower number of attached customers. Cr6mer et al. (2000) support their assumptions regarding attached customers by explaining that customers sign long-term contracts and quality is difficult to observe. However, contracts that allow firms to ignore customer preferences should not be part of a Nash equilibrium because the customers would be better off never signing such contracts. Furthermore, if quality is hard to observe, it is hard to observe for both new and existing customers, so it is unclear why firms must serve the quality preferences of new customers while ignoring the quality preferences of existing customers. To avoid these difficulties, I assume that asymmetries in firm size result from access to monopoly markets and cost asymmetries. Cr6mer et al. (2000) also assume that firms choose quality before choosing output. This assumption may apply in some network industries, but it does not apply to all. For example, if quality is determined by software features in a physical network,






10

changing quality may be as simple as activating or deactivating a software feature, which can be done in a matter of hours or less. Also, if quality is determined by choosing poorly working circuits versus well-functioning circuits for interconnecting physical networks, then changing quality may be as simple as redirecting circuits, which can be done by entering commands at a computer terminal. To allow for these situations, I assume that firms simultaneously choose qualities and quantities. I find that asymmetry in firm size in my model does not result in the interconnection problems identified by Cr6mer et al. (2000).

The analysis proceeds as follows. Section 2.2 describes the model. Section 2.3 presents the benchmark case in which three identical firms compete in a single market. Section 2.4 presents the case in which one of the firms has a monopoly in another market. Section 2.5 considers the case of entry into a formerly monopoly market and the incumbent's incentives to sabotage rivals. Section 2.6 is the conclusion. Proofs are in the Appendix A.


2.2. The Model


I consider an extension of a model developed by Cr6mer et al. (2000). There are two markets, A and B, for network communications and three firms. Markets are distinct because they are separated by geography. There are q., customers in market E {A, B}. Network providers compete for customers in a single period. qi,i > 0 will denote the number of customers that firm i serves in the market. I assume that q,, is sufficiently large in each market to ensure that there are unserved customers in equilibrium; i.e., in equilibrium q, > qI,,, + q2,,, + q3,p.





11


Customers are identical in each market and a customer of type r e [0, r]

(where r is the same in both markets) obtains a net surplus from buying from firm i at price Pi,, equal to r + s,, - p,,,, where sj,, denotes the value that the customer places on i's network. I assume that ris uniformly distributed.' Customers desire to communicate with customers in both markets, so si., is given by

3 B
Si, = v - Oi,jqjp. (2-1)
j=1 p=A

I let v e (0, 1/4) represent a parameter that reflects the constant marginal value that customers place on network communications of a given quality. This linearity assumption of value follows Cr6mer et al. (2000) and implies that, except for the firms' quality choices, each customer of type ris indifferent with respect to which customers the r-type customer communicates. Because customers always place positive value on being able to communicate with additional customers, si,,, is increasing in quantity. Because firms connect networks across markets, s.i, for customers in one market is increasing in quantity for both markets. Let O. e [0, 1] be firm i's internal quality choice and Oj E [0, 1] be the quality of external interconnection withj's network forj # i. For simplicity, I assume that each firm chooses a single internal quality and, for each network with which it interconnects, the firm chooses a single external quality. In other words, if firm i serves both markets, it provides a single quality interconnection Mj for firm j, as opposed to providing different quality interconnections for each market. I further assume that the value firm i's customers place on communicating with customers of firmj is independent of the quality offered for communicating with customers of firm k, k #j. I


'This assumption results in a linear demand curve.


w








assume that networks are homogeneous except for the providers' quality choices. One implication of these assumptions is that, if networks have the same number of customers and offer the same quality levels, then customers are indifferent between firms.2

To focus the analysis on how network providers can affect the quality of each

other's service offerings, I restrict the quality choices to such things as technical features and reliability that firms can affect when interconnecting their networks. For example, America Online offers proprietary content and functions for its own users and not for users of other Internet Service Providers. I omit other aspects of quality, such as bandwidth for customer network access, which are unaffected by interconnection.

Because only one quality choice can prevail for each network interconnection, the firm with the lowest external quality preference determines the interconnection quality; i.e., the quality of the interconnection between i andj is min { 6j, 0,i}. For example, if one firm chose a capacity of 45 megabits per second and the other chose a capacity of 30 megabits per second, only 30 megabits per second of information could be passed between the networks. The proof to Lemma Al in Appendix A shows that this results in an infinite number of Nash equilibria. In all instances in this chapter, the interconnecting firms make identical choices when optimizing external quality, so I use this as a focal point for the equilibrium choice.

Customers make their purchasing decisions after firms have made their quality and quantity choices. I assume no price discrimination and further assume that prices





2 Technically, the term qi,f in (1) should be qi.-I because customers do not obtain value from accessing themselves. I suppress the -1 and assume that q is sufficiently large that it does not affect the results. If the -1 were included, its effect would be to lower each firm's internal quality relative to external quality.








adjust to firms' and customers' equilibrium choices. Lemma 1 describes customers' network preferences in equilibrium.

Lemma 1. Given the assumptions of the model, each customer of type zris

indifferent between networks at equilibrium; i.e., r + si, - pi, = r + s , - p,, for

every i,j = 1, 2, 3 and i j.

It follows from Lemma 1 that if firm i attracts customers, it has a quality-adjusted price

p - (2-2) Pi,#l - Si,At"", t 22


I define the marginal customer to be the customer that, in equilibrium, is indifferent between buying and not buying network service. Such a customer exists because I assume that T e [0, r], costs are strictly positive, and v, the constant marginal value of connectivity, is sufficiently large relative to firms' costs to ensure that qi,, > 0 for every firm. At equilibrium, the marginal customer will receive zero net surplus and so will have a value of r = pp. Because the distribution of customers is uniform, the quantity of customers that firms choose to serve in equilibrium is simply the total number of customers that lie between the upper bound of customer preferences and p,. Normalizing the density of customers to 1 in each market, the quantity of customers served in market , is

3
.q,,, = r- p,. (2-3)
i=I






14

Combining (2-1), (2-2), and (2-3) gives the customers' inverse demand curve for firm i in market p

3 (3 3
Pi.,=- qJ,+ v ,jq jA +iiqi.A +6i,jqj,B +O,iqiB (2-4) j=J jjpi

Existing (incumbent) firms have identical innate cost functions. New entrants have a cost penalty because they must build networks. Actual costs for some firms may exceed innate costs because incumbents in formerly monopoly markets can raise rivals' costs. For example, incumbents may provide rivals with inferior access to essential facilities or withhold critical network information (Economides, 1998). Entrants cannot raise incumbents' costs in this way because they do not have essential facilities nor do they have network information that the incumbent does not have. Also, incumbents might deny access to rights of way by either refusing access or by using all the capacity of the rights of way. Incumbents might also require points of interconnection to be in locations that are far from its rivals' customers, causing the rivals to incur extra costs to reach their customers. The US Federal Communications Commission recently fined GTE $2.7 million for allegations that GTE denied rivals access to GTE facilities for locating equipment. Incumbents might also delay rivals' market entry by raising legal objections.

Because the quality choices are restricted to quality that can be affected by

interconnection, the costs of adding customers and of providing quality are separable. Incumbents incur a constant marginal cost c > 0 of serving a customer. Entrants' marginal costs are c + co > 0, where co - 0. For simplicity, I assume a symmetric, continuous cost function K(Oj) _ 0 of providing quality Oj e { .1, 4.2, 03} for i= 1, 2,

3. Quality costs are separable from quantity costs and among connections. The








assumption that quality costs are separable is reasonable because it is common for telecommunications firms to have dedicated network equipment for each network interconnection and to have separate equipment for connections for its own customers. I further assume that K(O j) > 0 for all Oj > 0, K(0) = 0, K9> 0, Koo> 0, and Ka o<0.

Extending Economides' (1998) model of raising rivals' costs and Mandy and Sappington's (2000) model of sabotage, I assume that an incumbent and former monopolist, which I call firm 1, in market A incurs a cost (rq) > 0 to raise firm 2's and firm 3's costs by rqq2,A and rqq3.A, respectively.3 I assume (rq) is continuous, 2(rq) > 0 if rq > 0, 20) = 0, r > 0, Xrr > 0, and rq > 0. Furthermore, firm 1 incurs a cost #(ro) > 0 to raise rivals' costs of external quality for interconnecting with firm 1 by K(,ro) - K(0) > 0, where #(ro) is continuous, (ro) > 0 if ro> 0, 0(0) = 0, A > 0, Orr > 0, K(O,ro) > K(0) if ro > 0, K(Qj,0)= K(Oj), Kr > 0, Krr > 0, and r> 0 for all i,j = 1, 2, 3. As a result, firm l's cost of serving ql,A + qlB customers with quality choices 01 = { 6,1, 6.2, 6.3} is

3
C(q,A,q ,,O,) = c(q1A + qB )+ 3K(O,j )+K(,,,)+ (rq)+ 9(r), (2-5)
j=2

and rival j's cost of serving qjA + qjs customers, j 1, is


C(qjA ,qj, 0 )= (C + CO + rq kj,A + cqj,B +K(0j,,, ro)+ EK(, ).
k=2

I assume a two-stage game in which the incumbent chooses its cost-raising

activities and then firms compete in quantity and quality. Presuming Nash behavior, each firm takes its rivals' quantity and quality choices as given when it chooses its quantity and


3 Economides (1998) assumed that raising rivals' costs was costless for the incumbent. This can lead to the incumbent preventing all entry, so I impose a cost on the incumbent.








quality levels. Therefore, (2-4) and (2-5) imply that firm l's profit maximization problem can be written as:


max K/= T -c- ( - Oj qj,A -(I -V1o, ,qA + VJ 1,jqj,B ]qLA
q1,,,Oi ,rq r0 j=2 j=1

+ -C - - VoIj)'j,B - - v9o, k, +V ,jqj,A q1,B j=2 j=1
- K(,j )- (rq )- (r) (2-6) j=l
subject to 0,,j e [0,1] for j = 1,...,3 rq , r - 0
q1., 2 0 for/ = A,B.

Assuming for simplicity that co = 0 in market B, firm j's profit maximization problem, j 1, can be written as:


maxj = -c-co-r - -vjk k,A - Voj, ,A Oj,kqk,B j
q1*J kwj k=1

+ - c - - vOjk )k,B j-(1 j,),B + Oj,kqk,A qj,B kj k=1
3
- EK(Oj,k)- K(0,j,r) (2-7)
k.i
subject to 0,k, e [0,1] for k = 1,2,3
qj. , 2 0 for# = A, B.

To ensure internal solutions that satisfy second order conditions, I assume

K 2v2(r-c)2 Ko(1) > v )2 Kao(1)> (1- ,and - c - co >0. Also, throughout the
4(1-2 (1 - 2v)

dissertation, I assume that identical firms choose identical levels of quantity.








2.3. Symmetric. Single Market Case


In this section I consider the case where three identical firms compete in a single market, which I call market B. I designate the firms as 2, 3, and 4 and, because no firm is an incumbent market B, the firms have symmetric cost functions and no firm can raise its rivals' costs. I consider the equilibrium in which these identical firms choose identical levels of quantity. Proposition 1 provides this section's primary result.

Proposition 1. In the symmetric, single market setting, each firm sets all of

its external quality levels equal to its selected internal quality level; i.e., 0, = 9 for

j i, for all i,j = 2, 3, 4.

In choosing external quality, each firm considers its quantity choice, the quantity choice of the network with which it is interconnecting, and v, the value a customer places on communicating with another customer; i.e., 0. = K;'(vq* BqB ). Firms consider their own quantity choice because this determines the number of customers that are willing to pay prices that reflect the value of the external quality. Firms consider the other firm's quantity choice because more customers on other networks increase the value of the interconnection. Because firms have symmetric quantity choices, they have symmetric external quality choices. Furthermore, firms determine internal quality based on v and their quantity choice squared; i.e., ,i = K0' (vq q )= K (v(qiB )2 ). Their quantity choice is squared because more customers on the firm's own network increase the value of the network, and each customer represents someone who will pay a price that reflects that value. Because in a symmetric equilibrium, all quantity choices are equal, internal quality equals external quality.








From Proposition 1, the symmetric equilibrium quantity for a representative firm i is

s T-c
qiB 4 -v where


Os =G(v(qSB ).

Firms that serve more customers also choose higher quality.


2.4. Monopolist Entry into a Competitive Market


In this section I consider the case where firm 1 has a monopoly in market A and firms 1, 2, and 3 compete in market B. This might represent a situation where the monopoly has merged with firm 4 to enter market B. Costs are symmetric in market B and, because no firm is an incumbent in market B, no firm can raise its rivals' costs. As in the symmetric, single market case, I consider the equilibrium in which identical firms' choices are symmetric. Proposition 2 provides this section's primary result.

Proposition 2. In the setting in which a monopolist from one market enters a

second market which is competitive, the monopolist and its rivals choose the same

levels of external quality for a given interconnection; i.e., O = 6 9 forj # i, for all i,j

=1,2, 3.

Each firm in the monopolist entry setting considers its quantity choice and the quantity choice of the network provider with which it is interconnecting when choosing external quality. Because value increases with the number of customers reached through an interconnection, the firms make symmetric external quality choices even though their quantity choices may be asymmetric. The monopoly firm does not strategically degrade








the quality of its interconnection with smaller rivals. Corollary 1 further describes firms' quality choices. Lemma 2 is useful for Corollary 1.

Lemma 2. In the monopolist entry setting, the monopolist in market A

chooses a greater quantity in market B than does its rivals; i.e., q , > q* for allj # 1.

Firm 1 chooses a higher quantity in market B than does either of its rivals because it internalizes some network externalities. In other words, its quantity choice in market B has synergistic effects with its quantity choice in market A--higher output in market B increases the value of the monopolist's network in market A. Furthermore, higher output in market A increases the value of firm l's network in market B.

Corollary 1 describes the firms' internal and external quality choices.

Corollary 1. In the monopolist entry setting:

a. The monopolist's internal quality choice exceeds its external quality

choices;

b. The rivals implement an internal quality below the external quality they

implement for interconnecting with the monopolist;

c. The rivals implement an internal quality that is equal to the external

quality they implement for interconnecting with each other; and

d. The rivals implement an external quality for interconnecting with each

other that is lower than the external quality they implement for interconnecting

with the monopolist.
That is to say, 0,* > 0* =0, > 0 =0, forj # k, for all j, k= 2, 3.


The monopolist provides the highest quality because its optimal quantity choices are higher than its rivals' optimal quantity choices. These higher quantity choices make








the monopolists competitive network more valuable than its rivals' networks. Furthermore, for connection to a network of a given size, the monopolists' higher quantity choice makes quality more profitable for it than for its rivals. The rivals choose internal qualities that are lower than the quality of their interconnections with the monopolist because connection with the monopolists' network provides more value to their customers than do their own networks. For the same reason, the rivals choose higher interconnection qualities with the monopolist than with each other.



2.5. New Competitive Entry into a Formerly Monopoly Market


In this section, I examine a new entry setting, a situation where firms 2 and 3 enter market A. Because firm 1 is an incumbent, it has a cost advantage of co in A and has the ability to raise its rivals' costs.

Lemma 3 shows that interconnecting firms choose the same external quality if the incumbent does not raise its rivals' interconnection costs.

Lemma 3. If ro= 0 in the new entry setting, then all firms implement the

same external quality for the same interconnection; i.e., O, = 0. forj # i, for all i,j =

1,2,3.

Proposition 3 provides this sections main findings.

Proposition 3. In the new entry setting, the incumbent will raise the entrants'

quantity costs. However, it will not raise the entrants' interconnection costs. That is

to say, rq* > 0 and ro* = 0.

The incumbent chooses to raise its rivals' quantity costs because raising these

costs causes the rivals to reduce their quantities, which allows the incumbent to increase








its output and its profits. However, the incumbent will never raise its rivals' interconnection costs because raising these costs would cause the rivals to implement lower external quality than the level preferred by the incumbent.

In some countries, incumbents have restricted interconnection capacity for rivals. This restriction causes calls or messages between the rival networks to be delayed, dropped, or not completed. This would appear to be in conflict with the conclusion that an incumbent would not raise its rivals' interconnection costs nor strategically degrade interconnection quality. But field interviews with customers revealed that customers generally did not understand network interconnection and believed that the service problem was caused by the rivals' failing to provide adequate internal quality. This perception lowered the demand for the rivals' services, which caused rivals to incur extra costs to obtain and keep customers. Therefore, even though the incumbents' choices related to interconnection, the effect was to raise its rivals' quantity costs.

Corollary 2 further describes the firms' quantity choices and quality choices.

Corollary 2. In the new entry setting:

a. The incumbent chooses higher quantities than does each of its rivals;

b. The incumbent implements an internal quality that exceeds its external

quality;

c. The rivals implement the same external quality when connecting with the

incumbent as the incumbent chooses;

d. The rivals implement internal quality that is equal to the external quality

that they implement for interconnecting with each other; and









e. The rivals' implement internal quality that is lower than the external

quality that they implement for interconnecting with the incumbent.


That is to say, (q.,qqjA,q,B)- qj,jq,A,q,B))> 0 pu=A

and6 ,, >9, =0,1 >0 = ,, forjk, forallj,k=2, 3.

The incumbent chooses higher quantities because it has a cost advantage, and its higher quantity choices drive its higher quality choices. As in previous cases, firms that are interconnecting make equal external quality choices when interconnecting, so no firm prefers a higher or lower interconnection quality than the other firm prefers.

Corollary 3. In the new entry setting, the entrants' quantity choice in market i

is a strategic substitute for the incumbent's quantity choice in market j, for all i #j

and i,j = A, B. However, the incumbent's quantity choice market i is a strategic

complement to the entrants' quantity choice in marketj.

The incumbent and entrants have opposite responses to each other's other-market production because the incumbent's higher internal quality causes it to have a greater response to its internalized network externalities. A firm internalizes more network externalities when it serves two markets than when it serves only one market. Furthermore, higher internal quality internalizes more network externalities than does a lower internal quality. As a result, when a larger firm connects with a smaller firm and both firms serve two markets, the larger firm quantity choice in the first market affects its quantity choice in the second market more than does the smaller rival's quantity choice in the first market. Conversely, the larger firm's quantity choice in the first market affects the smaller firm's quantity choice in the second market more than does the smaller firm's








own quantity choice in the first market. These opposite reactions result from the firms' quality choices. The larger firm implements a higher internal quality than it does external quality for interconnecting with the smaller firm. Therefore, the larger firm's quantity choice has a greater feedback effect than does the smaller firm's quantity choice. Conversely, the smaller firm implements an internal quality that is lower than its external quality with the larger firm. Therefore, the larger firm's quantity choice has a greater affect on the smaller firm's quantity choice.

For example, assume the entrants' increase their quantity in market A. This

higher quality causes the incumbent to choose a lower quantity in market A because the entrants' quantity choice is a strategic substitute for the incumbent's market-A quantity. With respect to the incumbent's quantity choice in market B, the entrants' higher quantity in market A has a direct positive effect on the incumbent's quantity choice in market B. However, this direct effect is weaker than the indirect effect, which is caused by the incumbent lowering its market-B quantity because of its lower market-A quantity. Just the opposite happens for the entrants. Assume that the incumbent unilaterally chooses a higher quantity in market A. In response, the entrants' lower their quantity choice in market A. Their lower quantity choice in market A lowers the positive network externalities that result from their serving both markets. However, their internal quality is lower than their external quality with the incumbent. Therefore, the positive network externality caused by the incumbent's higher quantity choice in market A is greater than the effect of the entrants' lower quantity choice in market A. As a result, they have a higher quantity choice in market B.








2.6. Conclusion


In this dissertation, I examine incentives for discrimination in network

interconnection. I find that a large firm implements the same external quality for a given network interconnection as does its smaller rival. Furthermore, I find that smaller firms offer their own customers an internal network quality that is lower than the interconnection quality that the smaller firms implement with the larger firm.

My results conflict with the findings of Cr6mer et al. (2000). Their model assumes that differences in firm size result from differences in numbers of attached customers, and find that these differences in attached customers cause firms with a higher number of attached customers to prefer a lower quality interconnection than a firm with a lower number of attached customers. They obtain this result because firms ignore the preferences of their own attached customers when choosing interconnection quality, but take into account unattached customers' preferences. The assumption that attached customers have entered into contracts that do not reflect the firm's quality choices, but that unattached customers' contracts do reflect the firm's quality choices, seems difficult to support. If it were optimal for firms to ignore preferences of captive customers, such as when the customers were in a monopoly market, then the monopoly in my model would have done so. That the monopoly in my model did not ignore its captive customers' preferences indicates that it is not optimal for firms to enter into contracts in which customers believe the firms will ignore their preferences when making quality choices. Therefore, these customers are not attached in the sense of Cr6mer et al. (2000). Because it is difficult to construct a scenario in which customers and firms choose contracts that result in attached customers in the sense of Cr6mer et al. (2000), I assume








that asymmetry in firm size results from access to monopoly markets and cost asymmetries. I further assume that firms simultaneously choose qualities and outputs.

This dissertation indicates that the EU may not have been justified in its

conclusion that a larger Internet firm resulting from the MCI-WorldCom merger or the WorldCom-Sprint merger would discriminate against European firms for access to the Internet backbone. It appears true that the larger firm would provide its own customers with a higher quality service than it would provide its competitors, but the larger firm's interconnection quality choice would be no different than the smaller firms' interconnection quality choice for connecting with the larger firm. Furthermore, the interconnection quality the larger firm would choose for connecting with the smaller firms would be higher than the quality the smaller firms would choose for connecting with each other.














CHAPTER 3
NETWORK INTERCONNECTION AND TELECOMMUNICATIONS COMPETITION: THE CASE IN THE U.S.


3.1. Introduction


Opening telecommunications markets to competition has sparked controversy

over what role, if any, the government should play in facilitating entry, regulating prices,

and reforming traditional methods of subsidizing services. Difficult issues in writing the

US Telecommunications Act of 1996 (Act) included how Bell Operating Company

(BOC) long distance restrictions would be lifted,' how and when local telephone

markets2 would be opened to competition, and how state and federal regulators would

share authority. A primary economic issue has been standards for regulating prices that

incumbent local exchange carriers (incumbents) charge new entrants for exchanging calls





SThe AT&T divestiture agreement of 1982 restricted the BOCs from providing long distance service, except in limited areas (called Local Access Transport Areas, or LATAs).
2 The prohibitions on local telephone competition restricted customers and nonlocal exchange carriers from using telecommunications facilities that were owned by someone other than the franchised local exchange carrier, to complete voice and data communications within a local exchange. Customers and nonlocal exchange carriers were allowed to place telephone facilities in local exchanges before the Act, but in most cases, they could use them only for nonlocal communications or for private communications, such as communications within a company. This use restriction made constructing local facilities less economical than it would have been without the restriction. Throughout this dissertation, I use the terms local competition and local telephone service competition to refer to the right to use nonincumbent local exchange company facilities to complete these formerly prohibited local communications.









and for using portions of the incumbents' networks to provide services.3 Call traffic exchange between competing carriers is necessary if customers of one company are to be able to call customers of another company. Payment for exchanging calls is called reciprocal compensation in the US. "Reciprocal" means that both companies involved in an interconnection are obligated to make payments. "Symmetric" reciprocal compensation means the companies charge the same prices to each other. Reciprocal compensation prices are generally symmetric in the US, so I assume symmetry in my models.

The Act provides three methods of market entry, illustrated in Figure 3-1. Some entrants use more than one method. Entrants can build their own facility-based network, lease portions of an incumbent's network, or buy an incumbent's services and resell them. Leasing portions of an incumbent's network is called purchasing unbundled network elements or UNEs. Figure 3-1 shows how an entrant would lease a local telephone line from the incumbent. The line would connect to the incumbent's building. It could then connect to the incumbent's switch or the entrant's switch, depending on how the entrant wishes to use the incumbent's facilities. Entrants that have their own switches must interconnect their switches with those of an incumbent and pay reciprocal compensation for terminating telephone calls on the incumbent's network. Figure 3-1 shows how lines, called trunks, would connect the incumbent and entrant central offices. Likewise, an



3 Traditional voice telecommunications networks consist of lines and switches. Lines either connect customers to the network or connect switches in the network. Switches route calls between customers. Switches are of two types: local switches (also called central offices) that customers connect to and that switch local calls, and long distance switches (also called tandem or toll offices) that route long distance calls from one local switch to another.









incumbent must pay reciprocal compensation for terminating calls on entrants' network. Regarding resale, Figure 3-1 shows an entrant's customer using a resold service. The facility arrangement is just as if it was an incumbent's customer. In a sense, reselling is little more than rebranding the incumbent's service.








CY

custrnes kee ntecrmnnetio Legend
Inambent Facilities
ntntrant Facilities
- - - Leased Facilities Owneratup Vanes
Q incumnbent ut
O Entrant Cstmes





Figure 3-1. Interconnection Arrangements



Economists disagree on how prices for UNEs and reciprocal compensation should be set. Supporting the idea that incumbents should be allowed to charge "high" prices to competitors, Baumol and Sidak (1994a, 1994b, 1995), Kahn and Taylor (1994), Hausman and Tardiff (1995), Larson and Parsons (1994), and Larson (1997) argue that incumbents' prices should be based on the Efficient Component Pricing Rule (ECPR). The ECPR sets the price of an upstream input equal to its incremental cost plus the retail profits that the








incumbent loses to downstream competitors. The ECPR has two purposes: First, to ensure that regulated prices do not provide entrants with an artificial price umbrella that encourages inefficient entry. The second purpose is to avoid giving incumbents price signals that encourage them to hinder efficient competitors.

Arguing for "low" prices, Mitchell et al. (1995), Albon (1994), Economides and White (1995), Tye and Lapuerta (1996), and Tye (1994) point out numerous flaws with the ECPR theory and argue for prices based only on incremental cost. Laffont and Tirole (1994, 1996) and Laffont, Rey, and Tirole (1998a, 1998b) show that the ECPR can be efficient in some circumstances, but that incumbent prices below incremental cost could be efficient in other circumstances.

Despite the voluminous economic debate on these issues, empirical research has been lacking. This is understandable given how recently the Act was passed and given that telecommunications liberalization in other countries is still recent. In this dissertation, I take an initial step in filling this void in the literature by examining how the Act and regulators' interpretation of the Act have affected the early development of local telephone service competition. I find that regulators' adoption of prices based on incremental cost have given incumbents an incentive to hinder competitive entry, as the ECPR theory would predict. However, these prices have not led incumbents to hinder entrants' in their efforts to gain market share. The data do not reveal whether this is the result of a lack of incentive for incumbents, a lack of ability, or a lack of opportunity. I further find that the opportunity for entrants to place facilities in incumbents' buildings (called collocation) is important to encouraging facilities-based entry and causes entrants to substitute facilities and unbundled network elements for resale-based entry. Higher








symmetric reciprocal compensation prices discourage entry and cause entrants to select customers that receive more calls than they send. High wholesale discounts for resold services encourage entry, but do not encourage entrants to substitute resale-based entry for facilities-based supply.

This study provides guidance for US policy makers as they review the progress made under the Act and consider further steps in its implementation. Congress has held several hearings on the implementation and effects of the Act. The Act requires the Federal Communications Commission's (FCC) to review its rules every two years to determine whether the rules are necessary. The Act also requires the FCC to forbear from applying any regulation or any provision of the Act if the FCC finds that the regulation or provision is unnecessary or inconsistent with the public interest. The Act also prohibits state public utility commissions from applying regulations or provisions of the Act that the FCC has decided to forbear from applying. In addition to providing information for US policy makers, this study suggests strategies that incumbents and entrants might choose in advocating policy positions before policy makers, and provides insights that other countries might use in establishing their own telecommunications policies. The World Trade Organization Agreement on Basic Telecommunications is prompting numerous countries to open telecommunications markets to competition.

In this dissertation, I contribute to the literature on competition in

telecommunications by providing an empirical analysis of how US regulatory policies for interconnection affect the development of local competition. Section 3-2 is a review of the literature. In Section 3-3, I provide historical background on telecommunications in the US, including an explanation of portions of the Act. In Section 3-4, I describe my








models and their underlying theory. In Section 3-5, I discuss the estimation results. Lastly, I conclude with policy implications and recommendations for further research.


3.2. Literature Review


Before presenting my results, I examine and summarize the empirical research on the development of competition in telecommunications. One of the earliest contributions in this area is Crandall (1991). He finds that the difference between total telecommunications equipment sales and telephone company equipment purchases increased from 1984 through 1988, indicating an increase in private network investment. Tomlinson (1995) shows that the development of competition from competitive access providers, who provide fiber optic networks for large customers primarily in metropolitan areas, prompts incumbents to build advanced fiber optic rings. Ros (1999) performs a cross-country analysis of how privatization and liberalization affect teledensity (telephone lines per 1000 population) and telecommunications investment. Without distinguishing between different forms of liberalization,4 he finds that competition increases teledensity and decreases investment in the sample countries with liberalization. Dekimpe et al. (1998) find that increased numbers of cellular competitors increases the penetration rate and the rate of diffusion. Blank et al. (1998) show that entry by long distance companies, such as AT&T, into BOC intraLATA long distance markets lowers BOC intraLATA long distance prices.


SThe markets open to competition varied across countries. Countries in the sample liberalized various combinations of local telephone service, domestic long distance, and international long distance.






32

In an examination of the effects of incentive regulation on incumbent investment, Greenstein et al. (1995) find that entry by competitive access providers has no measurable affect on incumbents' investment levels, but that removal of regulatory restrictions on entry encourages incumbents' investment in fiber optics. They also show that removal of restrictions on long distance company provision of intraLATA long distance and on resale of local services decreases incumbent fiber optic investment. Woroch (2000) finds that competitive access provider entry into markets for digital fiber optics in metropolitan areas prompts incumbents to make competing investments. Ros and McDermott (2000) find that repricing local telephone service to remove traditional subsidies from business to residential customers encourages entry by new competitors. In a study on the effects of incentive regulation, Ai and Sappington (1998) find that higher levels of competitive access provider investment in fiber optics is associated with higher incumbent investment in ISDN switches5 and lines and with longer delays by incumbents in resolving customer trouble reports. They also find that intraLATA long distance competition encourages incumbents to invest more in fiber optics and is also associated with long delays by incumbents in resolving customer trouble reports. In a cross-country comparison, Spiller and Cardilli (1997) use a case study approach to conclude that the absence of clear rules on interconnection and parity for long distance competitors in terms of the how customers can use their services, causes delays in entry


5 ISDN stands for Integrated Services Digital Network and was a technology associated with advanced telecommunications.








and disadvantages new competitors. They also find that limiting rights for entrants to use incumbents' networks encourages investment by entrants. Using event analysis, MacAvoy (1997) finds that the California Public Utility Commission's decisions in 1995 to allow local competition caused a 15.73% decline in Pacific Telesis' stock price. Using behavioral tests, MacAvoy (1998) concludes that long distance companies' price-cost margins increased during the late 1980s and through most of the 1990s, even though market concentration declined.


3.3. Historical Background


Today's new era of telecommunications competition has its roots in the

introduction of competition for telephone equipment in the 1960s and in long distance in the 1970s and 1980s. AT&T's primary tool for protecting its monopoly against each new competitive threat was the company's refusal to allow would-be competitors to connect to the BOCs' networks. Because AT&T owned the BOCs, AT&T had end-to-end control of the telephone network and was able to use this control to deny rivals access to the telecommunications market. This tactic led the US Department of Justice in the early 1970s to renew its earlier antitrust suit against AT&T. In 1982, AT&T and the Department of Justice settled the suit, with AT&T agreeing to divest its ownership of the BOCs. The divestiture was completed on January 1, 1984.

The underlying theory of the divestiture was that local telephone service markets are natural monopolies while other market segments are at least potentially competitive. Therefore, the thinking went, dividing the industry into monopoly and competitive companies would facilitate the orderly development of competition. To ensure that the








BOCs did not use their local monopolies to hinder competition in other markets, the divestiture agreement restricted the BOCs from manufacturing telephone equipment and from providing information services and interLATA long distance. In a separate antitrust settlement, the government imposed interLATA long distance restrictions on GTE, too.

But the divestiture's underlying theory was fatally flawed. The boundaries

between local telecommunications and all other telecommunications are artificial and arbitrary. The division between long distance and local telecommunications made long distance companies heavily dependent on incumbent local exchange companies, who were natural competitors to long distance companies. Long distance companies' payments to incumbent local exchange companies for originating and terminating long distance calls constituted nearly 50% of a long distance company's costs. Furthermore, almost 99% of long distance companies' long distance calls passed through the local exchange companies' networks (Jamison, 1995). In part to relieve the resulting strategic tension, long distance companies sometimes bypass portions of the BOCs' local telephone networks by placing long distance network facilities in high traffic areas and connecting directly to large customers.

BOCs would find it in their interest to enter long distance markets. The interLATA restrictions prevent the BOCs from competing for large customers by providing seamless end-to-end services that cross LATA boundaries. Also, BOC provision of long distance would eliminate the double marginalization that occurs when long distance companies take BOC access prices as given when they (the long distance companies) make their profit maximizing decisions (Weisman 1995). Furthermore, BOCs need to enter the interLATA markets if they are to be able to offer the








combinations of local and long distance products that are comparable to the combinations the long distance companies' provide when they enter the local exchange markets. Lastly, the BOCs' traditional monopoly position in local telephone service is being eroded by the convergence of telephone, computers, and media.

The arbitrariness of the distinction between local telephone and all other

communications finally spelled the end of the local telephone monopoly. In the early 1990s, some states began allowing competition for local telephone service. By 1996, the pressure for change was overwhelming and Congress passed the Act. Among other things, the Act removes legal restrictions on local telephone service competition for all geographic areas except those served by small rural telephone companies. In order to facilitate local competition, the Act requires all incumbent local exchange companies to provide entrants nondiscriminatory interconnection and UNEs at any technically feasible point within incumbents' networks at cost-based prices. Creating UNEs allows entrants to enter markets more quickly than they could if they had to build all of their own facilities.

The Act further requires incumbents to offer for resale at wholesale rates any

telecommunications service that they provide to retail customers. Wholesale prices must be based upon retail prices minus the portion attributable to marketing, billing, collection, and other costs avoided by the incumbent when it does not provide the retail service.








This is essentially the ECPR discussed earlier6 (Larson, 1997).

The essential trade-off in the Act is that the BOCs and GTE are permitted to offer interLATA long distance service in exchange for giving up their local monopolies. GTE was permitted to offer long distance immediately upon passage of the Act, as were the BOCs with respect to long distance outside their regions. With respect to long distance within their regions, the key market for BOCs because of their traditional customer base, the BOCs are to be permitted into the market once they have satisfied certain preconditions. Tomlinson (2000) provides details on these preconditions.

The FCC began implementation of the local competition provisions of the Act by adopting rules in August 1996 that incumbents and entrants are to follow to be in compliance with the Act.7 As of 1999, the FCC had issued 231 orders or similar actions related to its implementation of the Act.8

The FCC determined that incumbents' prices should be based upon a measure of incremental cost, which the FCC created and calls Total Element Long Run Incremental Cost (TELRIC). TELRIC incorporates specific assumptions about the incumbents' technology and input prices, which cause TELRIC to understate incremental cost


6 Consider a firm that produces and sells and input, called the upstream product, that is used to produce a final product, that is called the downstream product. This firm also produces the downstream product in competition with its customers for the upstream input. Under the ECPR, the difference between the firm's price for the upstream input and its price for the downstream product is simply the incremental cost of producing the downstream portion of the product.
' FCC 96-325, The First Report & Order In the Matter of Implementation of the Local Competition Provisions in the Telecommunications Act of 1996, CC Docket No. 96-98 and CC Docket No. 95-185, August 8, 1996.
8 It became clear during these proceedings that some items are more controversial than are others. Consensus quickly emerged on items such as white pages, dialing parity, and access to signaling and databases. On other issues, such as prices that incumbents would charge to entrants and collocation there is disagreement among stakeholders. (Harris and Kraft, 1997; Kennard, 1998)








(Jamison, 1999; Weisman, 2000; Mandy, 2000). This decision, and the FCC's extensive unbundling requirements, caused incumbents and state public utility commissions to object to the FCC's local competition rules. Incumbents objected to the FCC's decisions because the decisions gave entrants the right to make extensive use of incumbents' networks at prices below incremental cost. The state commissions objected to the FCC's rules because states believed that the Act leaves these decisions to the states.

The incumbent local exchange companies and the state commissions appealed the FCC's rules to the United States Court of Appeals for the Eighth Circuit. On July 18, 1997, the court released its decision generally agreeing with the states and the incumbents and vacating, among other things, most of the FCC's pricing and unbundling rules. The FCC, MCI, and AT&T petitioned the United States Supreme Court to review the Eighth Circuit's decision. On January 25, 1999, the Supreme Court largely reversed the lower court, holding that the FCC has general jurisdiction to implement the Act's local competition provisions and upholding almost all of the FCC's unbundling requirements. On July 18, 2000, the Eighth Circuit Court of Appeals vacated the FCC's TELRIC methodology and remanded it back to the FCC for revision.

Prior to the Supreme Court's ruling, states adopted an array of regulatory policies for implementing the Act. On relatively noncontroversial issues, such as white page listings and access to signaling and databases, states were reasonably uniform. On controversial issues, such as incumbents' prices, states differed from each other and from the FCC. Based on a review of state commission interconnection decisions located on the National Regulatory Research Institute's web site (NRRI, 1998), 16% of the states chose an incremental cost measure called total service long run incremental cost as their








basis for pricing reciprocal compensation. Total service long run incremental cost generally gives higher estimates of incremental cost than does TELRIC (Jamison, 1999b). Thirteen percent of the state commissions chose TELRIC and the rest chose bill and keep.9 With bill and keep, incumbents and entrants do not pay each other for exchanging minutes. Some states that adopt bill and keep apply the policy only as long as the traffic exchange is relatively balanced. Similarly, 77% of the states chose total service long run incremental cost as their cost standard for UNEs and 19% chose TELRIC. One state chose an accounting cost allocation approach called fully distributed cost as its method.

The Act also mandates collocation, the process by which entrants locate their equipment in incumbents' buildings. There are two types of collocation, physical collocation and virtual collocation. With physical collocation, entrants physically place their equipment in incumbents' buildings where incumbents also have telephone equipment. With virtual collocation, incumbents place equipment in their buildings and dedicate the equipment to use by entrants. This equipment is owned and maintained by the incumbents. Collocation is valuable to entrants because it decreases their costs of interconnection and of accessing UNEs, relative to having to place their equipment some distance away from the incumbents' facilities. Entrants prefer physical collocation because it gives them greater operational control than virtual collocation over their equipment and services.


9 The FCC's policies allow for bill and keep, but do not mandate it.









3.4. The Models


I analyze how regulatory policies affect entrants' entry decisions and the

quantities they sell. I summarize the entry decisions by extending Bresnahan and Reiss's (1991) and Berry's (1992) ordered probit models for entry, which apply a zero-profit equilibrium for entrants. They show that, as additional firms enter a market, quantity produced increases, which causes per-unit profits to decrease. At the equilibrium number of firms, the addition of one more firm makes per-unit profits negative, which makes additional entry unprofitable. I use a standard regression model to summarize entrant system expansion and the resulting effects on entrant market share.

Because of data restrictions, I consider each incumbent's traditional local

exchange areas in a state to be a market. This causes distortions because for a given incumbent, an entrant may choose to enter some of the incumbent's local exchanges and not enter others, and may choose to supply only some areas of a local exchange. Because regulators generally require incumbents to average retail prices across exchanges and to charge lower retail prices in rural areas than in urban areas, and because per customer costs are generally lower in high density, urban areas than in rural areas, I expect entrants to serve urban areas first. Therefore, regions with higher than average proportions of urban areas should have more entry than areas that are below average. To control for this effect, I include in my models a measure of customer density. Higher customer density indicates markets with higher than average proportions of urban areas.

I consider three types of models. The first describes entry. The second describes how entrants expand their market share by constructing their own facilities and by using









UNEs. The third describes how entrants expand their market share by reselling incumbents' services. I first describe the market entry models.

3.4.1. Entry Models

More entry is expected in larger markets, in markets where regulatory policies are conducive to entry and in markets where incumbents have created fewer barriers to entry. Bresnahan and Reiss (1991) explain how market size and the effects of minimum efficient scale affect the number of competitors that can profitably enter a market. They show that in 'small' markets, only a monopolist can profitably serve the market. Then as market size increases, the monopolist's profits increase. At some level of market size, it becomes profitable for a new entrant to enter the market. This new entrant enters when its expected expost profits are strictly positive; i.e., E(c(n= 1)) > 0, where E is the expected value and f'(n=l) is an entrant's profits if n = 1, where n is the number of entrants. Only one entrant enters the market ifE('c(n= 1)) 2 0 and E(sc(n=2)) < 0. At even larger market sizes, more entrants will find it profitable to enter the market. In general, entry occurs up to the point where E(grc(n)) 0 and E(rc(n+1)) < 0. For a given market, price-cost margins and profits decrease as the number of firms increases, the quantity supplied increases, or both increase.

Regulatory policies and incumbent responses to entry affect entry by affecting the profitability of entrant activity in a given market. Policies favorable to entrants and that provide incumbents with little or no incentives to hinder entry increase the probability of entry, while unfavorable policies and aggressive incumbent responses have the opposite effect. In general, low UNE prices favor entrants because entrants can substitute UNEs for their own facilities when facilities are more costly. Readily available collocation








opportunities decrease entrants' costs for using their own facilities and for using UNEs. Likewise, high resale discounts result in low wholesale prices, which make it more economical for entrants to resell incumbents' services, particularly when entrants' facility costs and UNE prices are relatively high.

Reciprocal compensation prices are more complex. Entrants receive more

reciprocal compensation payments than they pay (i.e., are net receivers) if they obtain customers who are net receivers of minutes of calling (which I simply refer to as minutes). An Internet Service Provider (ISP) is an example of a customer that receives more minutes than it sends. A residential customer who uses telephone lines to reach an ISP, and does so for several hours per month, is an example of a customer that makes more minutes than it receives. High reciprocal compensation price-cost margins encourage entry if entrants are better able than incumbents to attract customers who are net receivers of minutes. This effect of high reciprocal compensation price-cost margins decreases as additional entrants enter the market because the number of available netreceiver customers decreases. Low symmetric reciprocal compensation price-cost margins encourage entry when entrants' are more successful than incumbents in attracting customers who are net senders of minutes.

My dependent variable measures the total number of entrants in a market, which includes those that use their own facilities, those that use UNEs, those that are pure resellers, and those that use some combination of the three entry methods. Because entry is new, it is unlikely that the data represent a long-run equilibrium. I adjust for this disequilibrium by including as an explanatory variable the amount of time that has elapsed since entrants were allowed to enter each market.








I use an ordered probit to estimate how regulatory policies affect entry decisions. Ordered probit models are used when dependent variables are discrete and represent ordered outcomes. Underlying the model is a 'virtual' model in which there is an unobserved, continuous dependent variable y* whose conditional mean is assumed to be a linear function of the independent variables (Hausman et al., 1992). In my model, y* measures entrants' abilities and propensities to enter, and incumbents' abilities and propensities to limit entry.

Although y* is unobserved, it is related to my observed discrete random variable for entry. For example, no entry is observed if y* < 0. Likewise, one entrant is observed if0 _y* < ,, two entrants are observed if al __y*
Extending Bresnahan and Reiss (1991) and Berry (1992), I construct likelihood functions for ordered probits by calculating the probability of observing each level of entry. The probability of observing a market with no entry equals: Pr(Ec (n = 1)< 0)= 1- (3fc (n = 1)).

D(.) is the cumulative normal distribution function. jc (n = 1) is the predicted entrant duopoly profits, eis the normally distributed error term, and rc (n = 1)= jc (n = 1)+ 6 equals the actual entrant duopoly profits, where the duopoly is one incumbent and one entrant. Assuming that the profitability of entry decreases as the number of entrants increases, the probability of observing n entrants in equilibrium can be represented as

Pr(rc (n)> 0 and ,rc(n + 1)< 0)= 0(7c (n))- O(c (n + 1)). The residual probability of observing n or more firms is









Pr( c(n) 0)=D(hc(n)).

My entry models explain the number of entrants be examining parameters that affect entrants' expected profits.

3.4.2. Quantity Models

Now consider the second and third types of models, those that describe how

entrants expand their market share. These models follow closely the models for entry. Larger entrant market share is expected in markets where regulatory policies are conducive to entrants and in markets where incumbents have created fewer barriers to entrants serving the market demand. Let represent a customer's type; i.e., the degree to which the customer is a net receiver or a net sender of minutes.10 Assuming cost minimizing behavior, an entrant i provides positive output for customers of type Owhen its expected ex post profits from increasing supply beyond zero are strictly positive; i.e.,


E -l.rq > 0, where q,. represents the quantity that entrant i provides to


customers of type 0 Such an entrant would choose its methods of supply based on its expected ex post profits. For example, an entrant in markets with lower prices for UNEs and higher prices for wholesale services, relative to other markets, should use higher proportions of UNEs to provide services. Greater availability of collocation increases





10 Customers can be viewed as being distributed on a two-dimensional plane that represents their calling volumes. One axis represents the total number of minutes for calls a customer receives and the other axis represents the total number of minutes for calls a customer sends. 0 represents a customer location on this two-dimensional plane.









entrants' market penetration and causes entrants to use less resale when collocation lowers entrants' costs of UNEs or facilities. Also, if the regulated prices for UNEs or wholesale services provide the incumbent with a lower price-cost margin than the incumbent retail services that are displaced by entrants, incumbents could be expected to take steps to hinder entrants' expansion (Ordover, Sykes, and Willig, 1985).

As in the entry models, the effects of reciprocal compensation prices on entrant market share are complex. High price-cost margins for reciprocal compensation encourage entrants to attract customers who are net receivers. Entrants can do this if, for example, their marketing costs for obtaining these customers are low relative to the incumbents' costs of retaining these customers. On the other hand, low price-cost margins for reciprocal compensation prices encourage entrants to increase their use of resale and to increase their supply for customers who are net senders. Entrants can do this if their marketing costs for obtaining these types of customers are low relative to the incumbents' costs of retaining these customers.

I use ordinary least squares regression to examine entrants' market share. I consider two measures of entrant output. The first measure is the number of entrant interconnections to the incumbent. These interconnections, called trunks, are necessary for exchanging minutes between entrant and incumbent customers. Figure 3-1 illustrates how these trunks interconnect central offices. Higher numbers of trunks indicate higher amounts of entrant output for customers served by entrant facilities and, to a certain extent, UNEs. Entrants need trunks for customers served by UNEs only when the entrant is using its own switch; for example, when the entrant has a switch and uses an incumbent's local line UNEs to connect customers to the entrant switch. Trunks









understate entrant supply if the entrants are large because, as an entrant grows, the number of minutes that stay on the entrant's network generally increase. This happens because the probability of a call originating on an entrant's network also terminating on that entrant's network generally increases as the entrant attracts more customers. Also, trunks overstate entrant supply if there are many small entrants, or entrants whose customers are dispersed geographically. This happens because these entrants' low traffic volumes keep them from making efficient use of their local interconnection trunks. Figure 3-1 also illustrates how entrants use a local loop UNE. The local loop UNE connects the entrant customer to the incumbent's building that houses the incumbent's central office. At the central office, the entrant can either connect the customer to its own central office or use switching UNEs to route customer traffic.

My second measure of entrant supply is the number of business resold lines; i.e., the number of units of local telephone service that entrants resell to businesses. Figure 31 shows how entrant customers of resold services use incumbent facilities in the same way that the incumbents' own customers do.

3.4.3. Estimation Models and Data

Table 3-1 describes the data I use for my dependent variables, which are from the United States Telephone Association's (USTA) report to Congressman Thomas Bliley on December 9, 1998 (USTA, 1998). " This sample omits Sprint's local exchange




" Observations include Ameritech (all states), Bell Atlantic (all states), BellSouth (all states), GTE (California, Florida, Hawaii, Illinois, Indiana, Kentucky, Michigan, North Carolina, Ohio, Oregon, Texas, Virginia, Washington, and Wisconsin), SBC (Arkansas, California, Oklahoma, and Texas), and US West (all states) for 1998.








operations because Sprint was not part of USTA at the time the data were reported. COMPT is the number of entrants that are in each market in 1998. TRUNKS is the number of interconnection trunks between the entrants and the incumbent. RESOLDB is the number of the incumbent's business telephone lines that were being resold in 1998.

For the entry models, I create an ordinal variable that represents levels of entry. The Appendix explains the development of this variable.




Table 3-1. Descriptions of Dependent Variables
Name in Standard ObserVariable Model Mean Minimum Maximum Deviation vations

Number ofEntrants COMPT 26.15 0 164 30.10 59
Number of
Interconnection TRUNKS 30,018.81 0 289,299 50,611.44 59
Trunks
Number ofBusiness
Resold Lines RESOLDB 27,009.02 0 183,594 42,696.88 59


For models of entrant output, I estimate entrant market share by dividing the dependent variables and certain explanatory variables by the number of incumbent telephone lines in the market. This indexes these variables according to market size. I explain below which explanatory variables are divided by the number of lines. TRNKPLN represents TRUNKS per line. RSLDBPLN represents RESOLDB per line. RSLDRPLN represents RESOLDR per line.

Table 3-2 describes the data for the explanatory variables. I use the price per

month for a 2-wire local line in urban areas, UNEPRC, as the representative UNE price.12



12A two-wire local line has two wires twisted together that are used to connect the customer to the telephone company central office.










Incumbents have many UNEs. I choose local line prices to represent UNE prices because the sunk nature of line investment and the need for right-of-way and conduit space make lines the most difficult facilities for entrants to construct themselves. Two-wire lines are


Table 3-2. Descriptions of Explanatory Variables Name in Standard ObserVariable Model Mean Minimum Maximum Deviation vations

UNE Prices UNEPRC $16.82 $3.72 $32.00 $6.16 59 Prices for Reciprocal
Compensation RCP $0.0057 $0 $0.0283 $0.0067 59 Business Resale Discount
BUSRSL 0.1794 0.0800 0.2601 0.0374 59 Residential Resale
Discount RESRSL 0.1788 0.0700 0.2500 0.0405 59 Incumbent Local Service
Revenues (000) REVL $749,773 $71,863 $4,731,829 $902,587 59 Incumbent Line Cost
Level USFCOST $265.72 $65.68 $408.81 $57.71 59 Collocation Availability
for Voice Lines COLCV98 23.66% 0% 72.62% 17.54% 59 Quarters entrants in the
market before 1996 NUMQRT 10.75 0 18 15.44 59 1997 Incumbent Service SCMPAMQuality Complaints 97 274.29 6 2,637 466.18 59 1998 Incumbent Service SCMPAMQuality Complaints 98 339.08 8 2,473 506.33 59 Incumbent Total Revenue
1998 (000) REVT $1,511,887 $170,463 $8,460,236 $1,721,406 59 Incumbent Total
Assets (000) TPIS $4,347,098 $493,351 $27,585,598 $4,929,606 59 Incumbent Total Billable
Lines LINES 2,286,606 235,862 16,071,707 2,785,059 59 Number ofIncumbent
Central Offices COS98 230.1 29 752 174.2 59 Local Calling Minutes
for Incumbent (000) MOU98 7,056,890 820,771 36,441,427 7,740,706 59 Incumbent Central Office
Investment (000) COETPIS $760,976 $59,521 $4,581,901 $864,958 59 Presence of Universal
Service Fund Reform USFFUND 0.2881 0 1 0.4568 59

GTE GTE 0.2373 0 1 0.4291 59

Bell Atlantic BA 0.2203 0 1 0.4180 59

BellSouth BS 0.1525 0 1 0.3626 59

SBC SBC 0. 0678 0 1 0.2536 59

US West USW 0.2373 0 1 0.4291 59

Ameritech AM 0.0847 0 1 0.2809 59











the most common technology used for local telephone service. I choose urban prices because I expect most entrants to serve urban areas. Data on these prices are from state commissions (NRRI, 1998; Alabama, 1998), X-Change (1998-1999), and interviews with incumbents. In cases where there are discrepancies, I employ the data provided in state commissions' decisions (NRRI, 1998) when available, and other public data when the commission data is not available.

RCP is the price per minute for symmetric reciprocal compensation. Data on these prices are from the same sources as UNEPRC. Some states have multipart prices for reciprocal compensation. There may be separate prices for terminating minutes at a central office or a tandem office and for different times of the day. To express these prices as a single price, I follow the convention of assuming 6.25% of the minutes terminate at local central offices and the remainder terminate in a tandem. Entrants often connect to the incumbent's network at a tandem office because it gives them access to numerous central offices.

BUSRSL and RESRSL represent the discounts that regulators give entrants for buying wholesale services for business services and residential services respectively. In states where discounts vary by service, I use the smallest percentage discount. In states where residential service discounts are different from business service discounts, I use the residential discount for the models explaining entry and output using UNEs and entrantowned facilities. BUSRSL and RESRSL are highly correlated (R2 = 0.88). In tests of various models, business and residential discounts have similar results. To avoid multicollinearity, I use only the residential discount for these models. Sources for









percent discounts are the same as the sources for reciprocal compensation prices and UNE prices.

I use the ratio of total revenues for basic local telephone services (REVL) and total billable lines (LINES) to indicate incumbents' price levels for local exchange services. I call this ratio LCLPLN. Firms charge many prices, so using a single price is inappropriate.

I use incumbents' Universal Service Fund costs per line, USFCOST, to represent incumbents' costs for telephone lines. Universal Service Fund costs are the local line costs that incumbents report to the FCC for estimating subsidies that are given to small incumbents and rural incumbents for 1998. USF costs are based on incumbents' regulatory accounting records and provide an average cost for all of an incumbent's operations in a state.13

I use the percent of incumbent voice telephone lines accessible by entrants

through collocation in 1998, COLCV98, to represent the ease of collocation. Data are from Tables 3.6 and 3.7 of the FCC's 1998 Local Competition report. (FCC, 1998) There is a risk of endogeneity because higher entrant interest in a market should increase entrant demand for collocation. Higher demand for collocation should increase the incidence of collocation, which could cause a higher percentage incumbent lines to be in central offices with collocation. However, differences between markets should also reflect the ease of obtaining collocation.


13 Regulatory costs are an imperfect measure of economic costs. The accounting processes (Gabel, 1967) and distortive efficiency incentives caused by methods of regulation (Sappington and Weisman, 1996) cause these regulatory costs to deviate from economic costs.








I use the number of quarters in a state from the time the first entrant was given telephone numbers until the Act took affect, NUMQRT, to represent the amount of time that entrants have been operating in a state. Data are from Table 4.8 of the FCC's December 1998 Local Competition report (FCC, 1998).

I use total numbers of customer complaints to state and federal regulators for 1997 and 1998 in metropolitan statistical areas, SCMPAM97 and SCMPAM98, as my measures of incumbent service quality. Low incumbent service quality should encourage customers to buy from facilities-based entrants. Data are from the FCC's ARMIS reports.14 I use SCMPAM97 to examine entry because it represents ex ante entry information. I use SCMPAM98 per line (which I call SCMP98PL) to examine entrant output. I choose 1998 service complaints for my supply models because this represents the quality that customers and entrants experienced at the time supply and purchasing decisions were put into effect.

I use REVT, the incumbent's total operating revenues for 1998 for the market, to indicate market size. For output models, REVTPL represents incumbent total operating revenues per line. REVT data are from the FCC ARMIS reports. Also for supply models, I use several measures of incumbent costs and quantities supplied to analyze incumbent responses to price-cost margins on inputs sold to competitors. TPIS represents the incumbent's total plant in service for 1998, a measure of the incumbent's total investment in facilities. I express TPIS per line as TPISPL. My measure of the number of incumbent lines is LINES, the number of billable telecommunications lines in 14 All ARMIS data are from http://fcc.gov and were downloaded between March 1999 and August 2000.








1998. COS98 is the number of incumbent central offices in 1998. MOU98 is the number of incumbent local telephone minutes in 1998. COETPIS is the amount of incumbent investment in central office switches in 1998. TPIS, LINES, COS98, MOU98, and COETPIS are from FCC ARMIS reports. COS98PL, MOUPL, and COTPISPL are the per-line expressions of COS98, MOU98, and COETPIS respectively.

For the entry models, I include two price-cost ratios, PRCSTUNE and

PRCSTRCP, as explanatory variables. PRCSTUNE is the ratio of the incumbent's UNE price-cost ratio and the incumbent's average retail price-cost ratio. The UNE price-cost ratio is the ratio of UNEPRC to USFCOST. The incumbent's average retail price-cost ratio is the ratio of the incumbent's 1998 total operating revenues and total plant in service. PRCSTRCP is the ratio of the incumbent's RCP price-cost ratio and the incumbent's average retail price cost ratio. The RCP price-cost ratio is the ratio of RCP and the incumbent's total investment in central office switches in 1998 divided by the total number of local exchange minutes in 1998. These ratios reflect the relationships between the price-cost ratios for inputs incumbents sell to entrants and the price-cost ratios for the incumbent's retail services that the entrants displace. Low values of PRCSTUNE and PRCSTRCP indicate that UNE and RCP price-cost ratios are low relative to incumbents' retail price-cost ratios. Ordover et al. (1985) explain that incumbents have an incentive to hinder competitors' success if competition lowers incumbents' profits."5 Examples of incumbents' efforts to hinder entrants might be providing poor service quality to entrants, delaying collocation, and delaying 15 BOCs have a countervailing incentive to cooperate with entrants so that the interLATA restrictions are lifted sooner.








interconnection negotiations. Reciprocal compensation provides a special case for this incentive. If PRCSTRCP is low, incumbents would like to be net payers of reciprocal compensation. In the next section, I examine the coefficients of these ratios to test whether incumbents or entrants have the stronger propensities and abilities to respond to the profitability of using UNEs and paying or receiving reciprocal compensation.

Including the variables PRCSTUNE and PRCSTRCP creates multicollinearity because the variables are constructed from other variables. Therefore, I exclude these variables from the entrant output models and test hypotheses of nonlinear combinations of UNE prices, RCP prices, and incumbent revenues, costs, and quantities supplied. That is to say, I test incumbent responses to UNE price-cost margins by testing the significance of the combination UNEPRC/REVTPL, and I test incumbent responses to RCP price-cost margins by testing the significance of the combination RCP/(REVTPL/MOUPLN). 16

Following Ros and McDermott (2000), I examine the effects of telephone subsidy reform. The Act requires regulators to develop subsidies that are competitively neutral. Traditionally, telephone subsidies, called universal service subsidies, were embedded in telephone company prices and only incumbents' prices were affected by subsidies (Jamison, 1995). Following the passage of the Act, regulators began taking steps to






l6 1 use UNEPRC/REVTPL to test ULINE price-cost margins because the actual ratio of interest, (UNEPRC/USF)/(REVTPL/USF) solves to UNEPRC/REVTPL. Similarly, the reciprocal compensation ratio of interest, (RCP/(COTPISPL/MOUPL))/(REVTPL/COTPISPL) solves to RCP/(REVTPL/MOUPLN).








develop mechanisms for collecting funds for subsidies from all telecommunications service providers and making the subsidies available to all qualified service providers. According to a National Regulatory Research Institute survey (Rosenberg and Wilhelm, 1998), fourteen states had revised or were revising their subsidy policies in 1998. USFUND is a dummy variable that indicates whether the market is in one of these fourteen states.

Lastly, I include dummy variables to identify incumbents. I have a dummy variable for each incumbent, but omit the Ameritech dummy from models to avoid multicollinearity. Incumbent dummy variables may reveal differences in how incumbents processed entrant requests interconnection from 1996 through 1998. Soon after the passage of the Act, incumbents differed in how they processed these entrant requests. One incumbent required entrants to fax their requests for telephone numbers or local lines. Another required entrants to call with their requests, but assigned only one employee to the task of taking entrant orders. Other incumbents worked on electronic methods of taking entrant orders. These differences, as well as differences in regulatory scrutiny and enforcement, could cause incumbent dummy variables to be significant.

To examine potential multicollinearity, I regress all explanatory variables on each other. I also examine the linear correlation of each pair of explanatory variables. Multicollinearity problems occur between the reciprocal compensation variables RCP and PRCSTRCP (R2 = 0.97), between the service complaint variables SCMPAM97 and SCMPAM98 (R2 = 0.92), between the resale discounts for business and residential services (BUSRSL and RESRSL, R2 = 0.88), among variables that indicate market size (for example, REVT and TPIS), and between PRCSTUNE and the variables that are








included in it. To avoid multicollinearity in the entry models, I include no more than one variable from each of the collinear groups, with the exception of collinear groups involving PRCSTUNE and PRCSTRCP. I need these variables in some entry analyses to perform likelihood ratio tests of entrant and incumbent incentives. Regarding multicollinearity in the entrant output models, dividing market size indicators such as REVT by LINES resolves much of the multicollinearity. Otherwise, I include in each model no more than one variable from each collinear group.


3.5. Model Results


In this section I examine the results of my models. I examine the entry models first. I then investigate the models for entrant facilities and UNEs. Lastly, I examine models for resale. Tests of log linear models did not improve the overall fit, so I report only the linear results.

Table 3-3 provides the coefficients and t-statistics for the entry models. One asterisk (*) indicates significance at the 0.10 level. Two asterisks (**) indicate significance at the 0.05 level. Three asterisks (***) indicate significance at the 0.01 level. Table 3-4 shows the marginal effects of explanatory variables. The first three rows show how markets are grouped into nine categories. Rows 2 and 3 show the range of entrants and the number of categories. The remaining rows show how a marginal change in each explanatory variable affects the probability of a market being in one of the nine categories.

I estimate three models. Model 1 examines how UNE prices, reciprocal

compensation prices, resale discounts, customer density, local service price-cost margins,









service quality, market size, and UNE price-cost margins affect entry. The coefficients for UNE and RCP prices are negative, the coefficient for resale discounts is positive, and all are statistically significant. These results indicate that higher UNE, reciprocal


Table 3-3. Regression Results for Entry (COMPT) Explanatory Variable Model 1 Model 2 Model 3
**-0.1523 **-0.1900 **-0.1964 UNEPRC (-2.000) (-2.322) (-2.307)
*-56.038 -70.7291 -51.4000 RCP (-1.879) (-0.733) (-0.501)
**10.0708 **11.1879 **10.7780 RESRSL (2.230) (2.404) (2.288)
***0.0001 *0.0001 0.0001 LNSPCO (2.794) (1.934) (1.409)
***-1.9855 ***-2.0848 **-1.9342 LCLPUSF (-2.939) (-2.940) (-2.509) 0.0088 0.0126
COLCV98 (0.653) (0.901)
-0.0622 -0.0643
NUMQRT (-1.234) (-1.108)
*-0.0010 -0.0010 -0.0010
SCMPAM97 (-1.697) (-1.615) (-1.509)
*** 1.77e-06 *** 1.92e-06 ***2.00e-06 REVT (5.038) (5.108) (5.039)
***14.4117 ***17.435 **16.7705 PRCSTUNE (2.287) (2.589) (2.360) 0.3575 -0.0792
PRCSTRCP (0.099) (-0.021) 0.2729 0.2489
USFUND (0.668) (0.605) 0.6403
GTE (0.730)
**1.1397 0.9940 1.447 BA (2.065) (1.665) (1.867)
***2.2862 ***2.255 ***2.845 BS (3.761) (3.477) (3.121)
***5.0165 ***4.8481 ***5.5051 SBC (4.769) (4.554) (4.183) 0.8116
USW (0.939) 113.70 116.07 116.97
x2 d.f.=11 d.f.= 15 d.f. = 17 Log Likelihood -66.2809 -65.0953 -64.6450

compensation, and wholesale prices discourage entry. The marginal effects in Table 3-4,

which are based on Model 1, show that a marginal increase in UNE prices or reciprocal











compensation prices, or a marginal decrease in wholesale discounts, causes the probability of a market having more than 25 entrants to decrease, the amount of which varies by level of entry. No particular method of entry appears to dominate the others in importance.


Table 3-4. Marginal Effects for Entry (COMPT)
Markets
Category
Number 0 1 2 3 4 5 6 7 8 Entrants
in Market 0-2 3-4 5-7 8-14 15-25 26-38 39-44 45-60 >60 Number
Markets 8 8 5 4 12 11 3 4 4 Marginal Effects

UNEPRC 0.00013 0.00051 0.00687 0.01454 0.01889 -0.00169 -0.03841 -0.00075 -8.20e-05

RCP 0.04838 0.18599 2.52867 5.34936 6.94956 -0.62294 -14.13170 -0.27727 -0.03001

RESRSL -0.00870 -0.03342 -0.45444 -0.96135 -1.24893 0.11195 2.53967 0.04983 0.00539

LNSPCO -8.63e-08 -3.30e-07 -4.50e-06 -9.50e-06 -1.20e-05 1.1 1e-06 2.52e-05 4.95e-07 5.36e-08
LCL
PUSF 0.00171 0.00659 0.08959 0.18954 0.24623 -0.02207 -0.50071 -0.00982 -0.00106
SCM
PAM97 8.63e-07 3.32e-06 4.51 e-05 9.55e-05 0.00012 - 1.10Oe-05 -0.00025 -4.90e-06 -5.40e-07

REVT -1.53e-09 -5.90e-09 -8.00e-08 -1.70e-07 -2.20e-07 1.97e-08 4.46e-07 8.76e-09 9.48e-10
PRC
STUNE -0.01244 -0.04783 -0.65032 -1.37573 -1.78727 0.16021 3.63436 0.07131 0.00772

BA -0.00048 -0.00208 -0.05143 -0.10880 -0.14134 0.01267 0.28741 0.00564 0.00061

BS -7.13e-06 -5.90e-05 -0.10316 -0.21824 -0.28352 0.02541 0.57654 0.01131 0.00122

SBC -6.67e-13 -2.30e-1 1 -0.22637 -0.47887 -0.62212 0.05577 1.26507 0.02482 0.00269



The coefficient for the variable for customer density, LNSPCO, is positive and highly significant, indicating that more entry occurs in densely populated markets. The coefficient for LCLPUSF is negative and highly significant, indicating that higher pricecost margins for local telephone services decrease the number of entrants. This is contrary to the conventional wisdom that economic pricing for local telecommunications services is important for local network competition and indicates that such pricing may actually hinder entry by encouraging incumbents to hinder entry to protect profits. The








coefficient for the variable indicating incumbent service quality in the previous year (SCMPAM97) is negative, but only significant at the 0.10 level. This indicates that poor incumbent service quality hinders entry. This may be because of the importance of resale to entry. If an incumbent's retail services have poor quality, then it should be true that the wholesale versions of these same services will have just as poor, or poorer, service quality.

The UNE-retail price-cost margin ratio, PRCSTUNE, has a positive and highly significant coefficient. This indicates that more entry occurs when incumbents' UNE price-cost margins are high relative to their retail price-cost margins. This is consistent with the theory of Ordover et al. (1985); i.e., incumbents are more likely to hinder entry if the inputs sold to entrants have low price-cost margins relative to the retail services that the entrants replace.

The remaining variables in Model 1, those being for market size (REVT) and

selected incumbents (Bell Atlantic, BellSouth, and SBC), are all highly significant. The importance of market size confirms that there are some economies of scale for entrants. The coefficients for the incumbents are positive, possibly indicating that these incumbents have responded less aggressively to entry than have other incumbents. Consistent with this result, Bell Atlantic and SBC are the only BOCs at the time of this writing to receive permission to enter interLATA long distance markets. The marginal effects in Table 3-4 indicate that SBC has a more positive effect on entry in larger markets than do Bell Atlantic and BellSouth, reflecting the large number of entrants in SBC's Texas and California markets.








Model 2 in Table 3-3 provides the results of adding the other regulatory policy variables to Model 1, namely collocation (COLCV98), time (NUMQRT), price-cost margins for reciprocal compensation (PRSTRCP), and universal service reform (USFUND). Using a likelihood-ratio test, I fail to reject at the 0.10 level the joint hypothesis that the coefficients for these variables are all zero (72(4) = 2.37). Greater access to customers through collocation does not appear to prompt entry. Time does not appear to have been important, perhaps indicating that the passage of the Act was the watershed event triggering entry. Lower price-cost margins for reciprocal compensation have not resulted in incumbents hindering entry. This is probably because, as the market share models indicate, incumbents are sometimes net payers of reciprocal compensation. Lastly, I do not find that subsidy reform has encouraged entry.

Model 3 in Table 3-3 shows entry model results after adding other variables for factors that shift demand or supply. Using a likelihood-ratio test on the hypothesis that the coefficients for all of these added variables and the added policy variables from Model 2 are zero, I am unable to reject the hypothesis at the 0.10 level (72(6) = 3.27). I conclude that Model 1 is the most appropriate model for examining how regulatory policies affect entry decisions.

Tables 3-5 and 3-6 provide the results for the entrant market share models. Model

4 in Table 3-5 shows the results of including all of the explanatory variables in a model for entrants using UNEs and entrant-owned facilities. The primary purpose of this model is to form the basis for testing the signs and significance of nonlinear combinations of explanatory variables. Using an F-test, I fail to reject the hypothesis at the 0.10 level that coefficients for price-cost margins for UNEs, reciprocal compensation, and local








telephone services, resale discounts, subsidy reform, and for the incumbent dummy variables for GTE, BellSouth, and US West are equal to zero (F(8, 40) = 1.19). This indicates that incumbents either do not hinder this form of entrant output, or are unsuccessful in doing so. This also indicates that neither higher margins on local telecommunications services nor subsidy reform prompt incumbents to increase their supply using UNEs or facilities. Indeed, the positive and significant coefficient for reciprocal compensation prices indicates that entrants are increasing UNE and facilitybased supply primarily to customers such as ISPs, which are net receivers of minutes.

Model 5 in Table 3-5 provides results for the TRNKPLN after dropping variables that prove to be insignificant; namely, the resale discount, universal service reform, and the dummy variables for GTE, BellSouth, and US West. These results are comparable to the results of Model 4. Higher UNE prices limit entrant output, as well as entry. Higher local telecommunications services prices encourage customers to buy more from entrants, even though the incumbent's price-cost margin for these services is unimportant. The coefficient for ease of collocation (COLCV98) is positive and highly significant, indicating that collocation is important for entrants that want to build their own facilities that want to use UNEs. The coefficient for NUMQRT is positive and highly significant, showing that that building facilities and assembling UNEs takes time, which gives entrants a cost disadvantage relative to incumbents. Incumbent service quality has a positive and significant coefficient in Model 5. This implies that entrants choose to use their own facilities and UNEs, and that customers choose entrant services, when incumbent service quality is poor. The positive and highly significant coefficient for number of minutes per line is consistent with the conclusion that entrants use their own









facilities and UNEs to target customers based on calling patterns. The negative and highly significant coefficient for Bell Atlantic may indicate that this incumbent has the most aggressive response to entrants targeting ISPs as customers.


Table 3-5. Regression Results for Output Using UNEs and Facilities (TRNKPLN)
Explanatory Variable Model 4 Model 5
**-0.0004 **-0.0003 UNEPRC (-2.254) (-2.026)
*0.2213 *0.1900 RCP (1.801) (1.724)
0.0119
RESRSL (0.680)
*0.00004 **0.00003 LCLPRLN (1.927) (2.244)
***0.0002 ***0.0002 COLCV98 (3.536) (3.546)
**0.0005 ***0.0005
NUMQRT (2.535) (3.020) 6.9692 **9.3746 SCMP98PL (1.534) (2.360) 0.0243 0.0183 REVTPL (1.248) (1.199)
-0.0077 -0.0054 TPISPL (-1.605) (-1.375) 0.0022 **0.0034 MOUPLN (1.213) (2.119) 0.0192 *0.0250 COTPISPL (1.085) (1.872)
0.0014
USFUND (0.797) 0.0047
GTE (0.961)
-0.0015 **-0.0053 BA (-0.450) (-2.634)
0.0043
BS (1.146)
**0.0077 *0.0053 SBC (2.190) (1.986)
0.0063
USW (1.540)
***-0.0316 ***-0.0301 Constant (-3.559) (-3.665) 4.58 6.39 F d.f.= 17,40 d.f. = 12,45 R 0.6608 0.6302








Table 3-6 shows the results for entrant market share using resold business

services. Model 6 shows the results using all explanatory variables. This model forms the basis for a joint hypothesis test of the significance of price-cost margins of UNEs and local telecommunications services, and of incumbent investment, minutes, central office investment, reciprocal compensation prices, and dummy variables for GTE, BellSouth, Bell Atlantic, and SBC. Using an F-statistic, I fail to reject at the 0.10 level the null hypothesis that the coefficients for all of these variables and combinations of variables are zero (F(10, 40) = 1.72). Consequently, I show the results for Model 7, which omits the insignificant variables. The variable for reciprocal compensation prices is in Model 7 because I fail to reject the joint hypothesis when the price-cost margin for reciprocal compensation is included.

Comparing the results of Models 6 and 7, it is apparent that higher UNE prices

decrease resold business lines. Comparing this result with the coefficients for collocation and subsidy reform (which are negative and significant), it appears that entrants primarily use resale of business lines as part of an entry strategy whose main intent is to use UNEs and facilities, and that availability of collocation and universal service subsidies causes entrants to move away from reselling business lines more quickly than they otherwise would. The business resale discount itself is insignificant, indicating that entrant market share using a pure resell strategy is either minor or effectively nonexistent. Consistent with Models 4 and 5, more time for entrants to expand their systems increases the amount of business line resale. Consistent with the discussion of Model 1, poor incumbent service quality decreases entrant use of resale, for reasons stated above. The positive and highly significant coefficient for US West's dummy variable indicates that this









incumbent may be more cooperative with resellers than with facilities and UNE-based entrants, relative to other incumbents.


Table 3-6. Regression Results for Resold Business Lines (RSLDBPLN)
Explanatory Variable Model 6 Model 7
**-0.0010 ***-0.0011
UNEPRC (-2.204) (-2.818) 0.4136 0.4809 RCP (1.173) (1.498) 0.0319 0.0275 BUSRSL (0.591) (0.543) 0.00002 *0.00008 LCLPRLN (0.382) (1.707)
***-0.0003 **-0.0003 COLCV98 (-2.353) (-2.115)
**0.0013 *0.0009 NUMQRT (2.206) (1.736)
-21.0764 *-21.9800 SCMP98PL (-1.612) (-1.723)
-0.0774 *-0.0640 REVTPL (-1.410) (-1.751)
0.0067
TPISPL (0.493)
0.0011 -0.0022 MOUPLN (0.213) (-0.545) 0.0598 0.0079 COTPISPL (1.181) (0.217)
**-0.0110 *-0.0084 USFUND (-2.175) (-1.988)
-0.0220
GTE (-1.578)
-0.0140
BA (-1.462)
-0.0078
BS (-0.736)
-0.0015
SBC (-0.151)
0.0169 ***0.0312
USW (1.442) (4.752) 0.0311 0.0387
Constant (1.231) (1.623) 2.88 3.71 F d.f. = 17,40 d.f. = 12,45 R 0.5500 0.4974








3.6. Conclusion

This dissertation indicates that entrants use a mixture of entry and supply strategies, but that most entrant output is based on using UNEs and entrant-owned facilities. Reselling incumbent services appears to be important for entry, but not for growth in entrant market share. Prices for UNEs and reciprocal compensation are important determinants of entry, marketing, and modes of supply. Higher UNE and reciprocal compensation prices discourage entry, but higher reciprocal compensation prices also cause entrants to market to customers who are net receivers of minutes. Low price-cost margins for UNEs encourage incumbents to hinder entry, but incumbents have little effect on entrant market share. Also, the ease of using UNEs and placing facilities appears to be more important than relative prices for UNEs and wholesale in influencing entrants' choices of methods of supply.

My findings indicate that, if regulators want to increase market penetration for

facilities-based entrants, they should focus on unbundling networks, prices for UNEs, and collocation. Resale discounts for services, while important for entry, have little effect on entrant market share.

Additional work is also needed on this topic. I was unable to test whether the interLATA restriction, which is to serve as a carrot for BOCs to open their markets to competition, is having its intended effect. Also, I was unable to test whether the policies that encourage entrant supply also improve social welfare. A more complete model of how policies affect industry costs and prices is needed to determine welfare effects. Lastly, international comparisons are lacking, as are more in depth analyses of entrant strategies for data and voice services.















CHAPTER 4
NETWORK EXTERNALITIES, MERGERS, AND INDUSTRY CONCENTRATION


4.1. Introduction


The welfare effects of mergers and market concentration have been analyzed

extensively. (See, for example, Salant et al. (1983), Farrell and Shapiro (1990a, 1990b), and Gaudet and Salant (1991).) However, this literature does not examine how network externalities affect incentives to merge and the welfare effects of mergers and market concentration. These issues have taken center stage in recent merger and antitrust cases in network industries. Concerned with market dominance in the Internet backbone, the US and European Union (EU) regulators required MCI and WorldCom to divest Internet assets as a precondition for their merger and later blocked the merger of Sprint and WorldCom. In Silicon Graphics, Inc.'s purchase of software makers Alias and Wavefront, the US was concerned that the horizontal merger of the two software makers would result in higher prices despite the depressing effect that higher prices would have on sales of Silicon Graphics, Inc.'s computer workstations. The US placed restrictions on Silicon Graphics, Inc. to ensure that its products were compatible with competitors' products. (Shapiro, 2000) Similarly, the EU required Vodafone to divest assets in the UK before approving Vodafone's purchase of Mannesman. A US District Court ordered the breakup of Microsoft based in part on the theory that Microsoft attempted to leverage








its dominance in the market for PC operating systems to monopolize the Internet browser market.

I extend the Katz and Shaprio (1985) model for a network industry to examine the welfare implications of mergers in network industries. I examine an oligopoly model in which firms choose output levels, quality within a network (internal quality), and quality between networks (external quality). There are multiple markets and positive network externalities among the markets. Examples of such situations include the following: (1) Telecommunications markets in separate countries or cities (Mergers between SBC and Ameritech, or between Bell Atlantic and GTE, fall into this category.); (2) Internet providers in separate geographic areas (Mergers between US and European Internet providers fall into this category.); and (3) Computer operating software (Windows Me and Windows 2000 fall into this category).' These network externalities imply that an increase in output in one market stimulates demand in other markets.

When a single firm serves multiple markets with network externalities, the firm internalizes network externalities and chooses higher levels of production than separate firms would provide. This fact implies that a merger may increase welfare even if it raises marginal production costs, increases market concentration, or extends market dominance. To analyze this result, I consider mergers that combine firms that operate in different markets, which I call pure cross-border mergers, and mergers that combine firms that operate in some of the same markets, but not all of the same markets. I call these latter mergers mixed mergers.


' Windows Me serves the consumer market for PCs while Windows 2000 is for Internet servers. These two operating systems contain features that allow customers using one to communicate more easily with customers using the other than with customers using a non-Microsoft operating system.








Even if a pure cross-border merger raises marginal production costs for the

merging firms, they will choose higher levels of production after the merger as long as the revenue from internalizing the network externalities exceeds the higher marginal production costs. When a single firm serves multiple markets with network externalities, its additional revenue from an increase in output in market A reflects not only the effects on price and quantity sold in market A, but also the effects that higher market demands have on prices in other markets in which the firm operates. The higher prices in these other markets provide additional revenue for every level of output in market A. I call the effect of a change in the output in A on this additional revenue the marginal extra-market revenue. Positive marginal extra-market revenue causes the merged firms to produce more even if the merger increases marginal production costs.

Also, a pure cross-border merger between large firms may increase welfare more than does a pure cross-border merger between small firms. Recall that a cross-border merger internalizes the positive effects that higher output in one market has on prices in other markets. Large firms profit more from these higher prices than do small firms, and so increase their output more in response to a cross-border merger than small firms do. Although competitors of the merged entity may reduce their outputs (Farrell and Shapiro, 1990b), the output reduction does not offset the merging firms' production increase.2

Pure cross-border mergers cause higher Herfindahl-Hirschman indices even while improving welfare. The indices increase with cross-border mergers because the inside




2 Furthermore, in the network setting, rivals in one market may actually increase their production if the merged firms' higher production in other markets stimulates demand.








firms produce more output after the merger than they did before the merger. This gives the inside firms higher market shares when the outside firms reduce their production.

My findings combine the existing literature on mergers and on network effects. In their seminal article on mergers, Salant et al. (1983) describe how exogenous mergers may decrease joint profits of the merging firms. Farrell and Shapiro (1990a) identify industry conditions and asset transactions among firms that increase industry concentration and worsen industry performance. Farrell and Shapiro (1990b) develop sufficient conditions for profitable mergers that raise welfare and show that the Herfindahl-Hirschman indices can be misleading. Gaudet and Salant (1991) provide rules for determining when a merger increases or decreases welfare.

In their seminal article on network economics, Katz and Shapiro (1985) show

how consumption externalities cause demand-side economies of scale and that firms may choose less product compatibility than is socially optimal. Katz and Shapiro (1994) analyze competition between systems, a collection of two or more components that work together. They describe how customer expectations about the future popularity of systems can result in multiple equilibria or the absence of equilibria. They also show how systems markets make coordination among firms difficult. Liebowitz and Margolis (1994) describe the limits of network externalities. Cr6mer et al. (2000) show how large networks may discriminate against small networks in terms of the quality of network interconnection.

I develop and explain my results as follows. I develop a model in which firms serve three markets, which have network externalities between the markets. In the first stage of the game, consumers form expectations about the size of each firm's network and








each firm's quality choices. In the second stage, firms choose qualities simultaneously. Firms play a quantity game in the third stage and these quantities generate marketclearing prices. Lastly, each customer chooses its network provider by comparing for each provider the value the customer places on the provider's service with its price.

I examine pure cross-border mergers in two settings - a monopoly setting and an oligopoly setting. In the monopoly setting, all three markets are initially served by separate monopolies. I then consider a merger between two of the three monopolies. If quality is unchanged, the merger improves welfare as long as marginal production costs do not increase, or, if they do increase, the increase is less than the marginal extra-market revenue. I show that quantities and welfare would increase even more if all three monopolies merged. In the oligopoly setting, output and net consumer surplus increase as long as marginal production costs do not increase. Quality and welfare unambiguously increase if the additional output stimulates demand sufficiently to cause the merged firms' rivals to increase their production. Mergers between larger firms have greater welfare effects than mergers between smaller firms because larger firms profit more from extramarket revenue than do smaller firms.

I then examine a mixed merger. I consider a model in which there are initially two firms. Each firm is a monopoly in one market and the firms compete in a third market. Using a simulation, I examine the welfare effects of a merger that results in all three markets being served by a single monopolist. I find that welfare increases as long as the effects of internalizing additional network externalities plus the effect of quality changes dominate the effect of losing a competitor in the duopoly market, a situation that








may occur if the duopoly market is small relative to the other markets or if customers place high value on network externalities.

The analysis proceeds as follows. Section 4.2 describes the model. Section 4.3 describes how quality is chosen. Section 4.4 presents the results for pure cross-border mergers. Section 4.5 provides the results for mixed mergers. Section 4.6 is the conclusion. All proofs are in Appendix C. A detailed description of the market demand is also in Appendix C.


4.2. The Model


4.2.1. Demand and Supply

Extending Katz and Shapiro (1985), I consider an oligopoly model in which there are three markets for the network service, which I designate as markets A, B, and C. For purposes of this model, a market is a customer group, such as customers located in a specific geographic region. To keep the markets distinct, I assume that customers cannot migrate between markets for purposes of buying the service; for example, a bank in New York cannot purchase telecommunications service from a provider that does not operate in New York. As many as four firms may operate in the model. The firms are labeled 1, 2, 3, and 4. A firm may serve more than one market.

Let qi'm > 0 denote the number of customers that firm i serves in market m, q' will denote the vector of all q,.m for a single firm i, and q will denote the vector of all outputs of all firms in all markets. For simplicity, I assume that each customer buys at most a single unit of output.









I assume that firms "interconnect" their networks.3 In the setting of physical communications networks, this interconnection would be the lines and technical arrangements that allow customers to communicate across the networks. In the setting of virtual networks, such as computer software, this interconnection could be interpreted as a feature that allows customers of one network to benefit from customers of other networks. For example, software providers may create a feature that allows spreadsheet users to exchange data with database users.

Extending Crimer et al. (2000), let 9i"' [0, 0] represent firm i's quality choice for communications between its customers, for all i = 1, 2, 3, and let 0"' e [0,0] represent firm i's quality choice for external interconnection between its network andj's network, forj # i.4 For simplicity, I assume that each firm chooses a single internal quality and a single external quality choice for interconnecting with each other firm. Let 0' represent the vector of i's quality choices. Quality includes such things as capacity for customers of physical networks to exchange messages, and features, such as instant messaging and technical protocols. A choice of zero represents a refusal to interconnect. I assume that network quality is perfectly observable to firms and customers alike. Furthermore, only one quality choice can prevail for each network interconnection. This is because only a single physical capacity can exist at a single point of interconnection,



3 In some industries, regulators require firms to interconnect their networks. Regulated telecommunications is an example of a network industry where regulators require interconnection. The Internet is generally unregulated. Refusal to interconnect is rare in the Internet, but controversy over type of interconnection is common. See Kende (2000) for an excellent overview.
4 Note that the market designations are irrelevant if each firm has a single interconnection with each other firm.








and if a technical feature cannot be used by customers of network i to communicate with customers of network j, then neither can customers of networkj use the feature to communicate with customers of network i. Section 4.3 describes how differences in quality choices are resolved.

The inverse demand curve for firm i in market m can be expressed as

pi'm pi'm(q, O'). (4-1) I consider a market-clearing price, given output and qualities, and I assume that price decreases continuously with quantity sold in the market. This will be the setting, for example, when there is a continuum of customer types, and higher type customers place more value on communications than lower type customers so that p ,'. < 0 for i,j = 1, 2, 3, where the subscript represents the first partial derivative. Price increases with quality because customers value higher quality network service more than lower quality network service, all other things being equal, i.e., p"" > 0 for i,j = 1, 2, 3. I assume that the marginal value of quality decreases as quality increases. Marginal willingness to pay in one market also increases with quantity sold in other markets because of positive network externalities, i.e., p'. > 0 for i,j = 1, ..., 4, and m h . Furthermore, the marginal value of these positive network externalities decreases with the quantity sold in other markets. Quality and network externalities interact in that the marginal value of the size of the system increases with quality, i.e., p',", > 0 for all i = 1, ..., 4, and m, h = A,
q
B, C. To ensure that an internal solution exists for output choices, I assume that each firm's marginal revenue in a market declines as its rivals increase their output in the market, that customers who value the network service the least value it at zero, and that








each firm's residual demand curve intersects its marginal cost curve from above (Dixit, 1986).'

I assume a customer makes purchasing decisions after firms have made their

quality and output choices, which are based are customer expectations. Each customer chooses to purchase from the firm that provides the greatest net consumer surplus. The marginal customer is the customer that is indifferent between buying and not buying the network service in equilibrium when quantity sold is strictly positive (Katz and Shapiro, 1985).6 At equilibrium, the marginal customer will receive zero net surplus and, for each firm that sells a positive amount, price is equal to the value of the network service to the marginal customer. Prices vary among firms if quality and network size vary among firms. All other things being equal, marginal customers value larger networks and networks with higher internal and external qualities.

I assume that costs for production are separable from costs for quality. Firm i incurs fixed costs K"'m 0 for each market m and a constant marginal cost com > 0 of production. All fixed costs are assumed to be sunk costs. For simplicity, I assume that there are no economies of scope across markets. Let G' (I, 0) represent firm i's cost function for interconnection, where I is a vector such that the (i,j) element of I is the number of interconnections that firm i has with firmj or has the value 1. In the case of physical networks, firm i establishes a costly physical connection with each firm or with each firm in each market. In such cases, the (i,j) element of I is the number of

5 Underlying this inverse demand curve is a customer value function that decreases as customer types increase, increases as quality increases, and increases as the system grows.
6 The assumptions that price is zero for the lowest customer type and that marginal costs are strictly positive ensure that there are unserved customers in equilibrium.








interconnections that firm i has with firmj. In the case of software, there may be no additional costs for a technical feature, such as exporting pure text files, to be available for interfacing with additional software packages. In such cases, the (i,j) element of I is

1. I assume that the cost of interconnection is weakly increasing in quality and is quasiconvex and weakly increases with the number of interconnections, i.e., G,j > 0, G. 0, and I' G, > 0 for i,j= 1, ..., 4, where f is the (i,j) element of I. For simplicity, I assume that interconnection costs are independent of the number of customers. The assumption simplifies notation and does not affect results because the effects of customers on quality costs could be included in cm.

Each firm takes its rivals' quantity choices as given when it chooses its own

quantity levels. Firm i's production in market m is zero if firm i does not operate in the market. Firm i's profit maximization problem can be written as:

C
max T' = ((p',m - ci, i, Kim )- G'
4\0O m=A
subjectto j e [0,0] for j = 1,...,3 (4-2) q1,m > 0 form = A,B,C.

When firm i chooses its profit maximizing outputs, its output choice for each market reflects the effects of internalizing network externalities. For example, if firm i operates in both markets A and B, its marginal benefit from increasing qiA includes the marginal extra-market revenue, pA qI, which is strictly positive and represents the portion of network externalities between markets A and B that firm i internalizes when it operates in both markets. If firm i did not operate in both markets, its output in market A would create network externalities and higher profits for market B, but these profits








would not directly benefit firm i. Therefore, firm i does not consider these profits when choosing its output for market A.



4.2.2 Welfare

The surplus a customer receives from purchasing depends on the innate value the customer places on the network service, on the internal and external quality choices of the customer's network supplier, and the total number of customers who purchase the network services. In each market, a customer only purchases if he values the service at least as much as the marginal customer does. Integrating over all customers who purchase and summing over all firms and all markets gives the net consumer surplus:

C 4 q.'
vnet = O (pim(q'i pi.m(qOi) j i'm -T:, (4-3) m= A i=1 0

where l is the vector of all q', and the weighted social welfare:
4
Z aVel + (1T- a)'( + T), (4-4)
i=1

where a= [1/2, 1 ] is the weight given to net consumer surplus, assuming that the social planner may give preference to consumers, and T; is a transfer payment from consumers to firm i. If a social planner chooses q and 6 to maximize weighted social welfare subject to a non-negative profit constraint for firms, she would ensure that the marginal consumer surplus, plus the positive network externality, are equal to marginal production cost and that the marginal consumer surplus from quality is equal to the marginal cost of quality. Formally, using pointwise optimization:








4 C 4
pi.' + p~;M' + p ;. =cr (4-5)
ji th*m j=1

for all i= 1, ..., 3 and m = A, B, C, and:

pl. = G., (4-6) for all i,j= 1, ..., 3 and m = A, B, C.


4.3. Quality Choices


Choosing quality allows a firm to increase or decrease the consumer value of

network externalities for its service and for its rivals' services. Consider oligopoly setting

1 pre-merger described in Table 1. Firm I operates in markets A and B and firm 2 operates in markets B and C. There are no other firms. Firm 1 can increase network externalities for itself by increasing internal quality or increasing external quality. First consider the situation where firm 1 increases network externalities by increasing internal quality. Increasing internal quality raises demand for firm l's output. Firm 1 in turn increases output for two reasons. First, the higher demand increases marginal revenue for each level of output, which also causes firm I to increase output. Second, firm 1 adds customers in one market to increase the value of its network in its other market because the higher internal quality increases the marginal value from network externalities of adding a customer. The cost of adding these network externalities is the cost of the higher internal quality plus the marginal production costs of increasing output.

Compare this result to the result if firm I increases network externalities by increasing external quality with firm 2. As in the situation of raising internal quality, increasing external quality increases the demand for l's service, which causes 1 to









increase output. However, in contrast to the situation where 1 increases internal quality,

1 does not increase its output to add network externalities. Firm 2's output does this for

firm 1. This leads to Lemma 1.



Table 1. Arrangements of Firms Before and After Merger Between Firms 1 and 2 Firms in Markets before and after Mergers Settings Market A Market B Market C Oligopoly Pre-merger Firm 1 Firms 1 and 2 Firm 2 Setting 1 Post-merger Firm 1 Firm I Firm Oligopoly Pre-merger Firm 1 Firms 2 and 3 NA Setting 2 Post-merger Firm 1 Firms 1 and 3 NA All-monopoly Pre-merger Firm 1 Firm 2 NA Setting Post-merger Firm 1 Firm I NA


Lemma 1. When firms 1 and 2 increase the external quality between their networks,

firm 2's higher output increases the value of firm l's network, which saves firm 1 marginal production costs relative to the situation where 1 raises internal quality.

Because of the saving in marginal production costs, firm 1 prefers higher external quality with non-rivals than it does internal quality. In addition to adding network externalities for firm 1, firm 2's higher output competes with firm l's output in markets where 1 and 2 are rivals. This decreases the profitability to firm 1 of increasing external quality. Therefore, a firm prefers higher external quality with a rival than it does internal quality if the proportion by which the firms' markets overlap is small relative to the value customers place on network externalities. Proposition 1 summarizes this result for Oligopoly Setting 1 Pre-merger. Assumption 1 describes how a change in external quality might affect firm l's revenue.








Assumption 1: The effect of an increase (respectively, decrease) in firm l's

external quality on firm l's revenue is greater than the effect of an increase

(respectively, decrease) in l's internal quality. That is to say, I *aq *q21
q1,. ' p - a+' + pq'2, > ,q P,' * 1*. , (4-7) m=A,B ) m=A,B

where q * represent firm l's optimal output in market m and p.* is the

resulting price, given the firms' quality choices.
, aq' c �X A q2,B
In (4-5), qI,,,, l.m aq2_ __AI, _In (4-5), q 2.C 2 and qA pq ,2 represent the increase in revenue ,B* ,B * aq2,B
for firm I that results from firm 2's greater output and q1 p', 0012 represents the decrease in firm l's revenue.

Proposition 1. The relationship between firm l's preferred internal quality and its

preferred external quality with firm 2 depends upon the aggregate sizes of their

individual markets relative to the aggregate size of the market that they both serve.

More specifically:

a. Firm l's preferred internal quality is lower than its preferred external quality with

firm 2 if and only if Assumption I holds; and

b. Firm l's preferred internal quality is greater than its preferred external quality

with firm 2 if and only if Assumption 1 does not apply and the inequality sign in

(4-7) is reversed.

The relationship between firm l's preferred internal and external qualities depend on the relative sizes of the markets and the value customers place on exchanging








information. For a given value of exchanging information, Assumption 1 will hold if market A or C, or both, is sufficiently large relative to market B. Furthermore, a sufficient condition for Assumption 1 to apply is for firm 2's higher output in B to stimulate more revenue in absolute value for firm 1 in market A than it decreases firm l's
IA * IB
revenue in market B, i.e., q ,A p.1 > q B p ' . This leads to the following corollaries.

Corollary lA. Firm 1 prefers a higher external quality than internal quality if the ratio of

its outputs in A and B is greater than the negative ratio of the slope of its demand in B

1,A IB
to the marginal value of firm 2's output in market A, i.e., q ," pl>q,,.
q p 2

Corollary 1 A applies regardless of the size of firm 2, so even if firm 1 is larger than firm 2, firm 1 would prefer a higher external quality than internal quality. Corollary lB. In the special setting where markets and firms are symmetric, a sufficient

condition for firms to prefer higher external quality than internal quality is for the

marginal value of network externalities to be greater in absolute value than the slope curve, i~e,A l'* >pB *1

of the demand curve, i.e., p 2 > p1

I assume that two firms engage in efficient bargaining to determine external

quality between their networks if the firms prefer different interconnection qualities. The bargaining maximizes the profits of the two interconnecting firms given the quality choices of all other firms. Also, it may be difficult in some industries for firms to adopt an external qualities that are higher than internal qualities. An example would be telecommunications networks if quality represented capacity for communications. In








these industries, if Assumption 1 applied to one or more of the firms such that the bargaining would result in a firm having an external quality that was higher than its internal quality, the firm would choose a single quality for internal and external communications.



4.4. Analysis of Pure Cross-Border Mergers In this section I consider pure cross-border mergers, mergers between firms that are not rivals. After the merger, customers form new expectations about the sizes of the firms' networks and all firms select their outputs and qualities to maximize their profits. I first consider the all-monopoly setting described in Table 1, the situation in which each market is served by a monopoly. The merger that I analyze is a merger of firms 1 and 2. I then consider oligopoly setting 2, the situation in which at least one of the markets has more than one firm.

Consider the all-monopoly setting in which markets A and B are initially served by separate monopolies. Now suppose the monopolies merge. Before the merger firm 1 operates in market A and firm 2 operates in market B. To maximize profits in this permerger situation, each firm equates its marginal revenue from expanding output with



7 From first order conditions, the firms choose a negotiated external quality defined by

e*,(0i*)~ (o' ) z'j (&i To~~" -(I - C~z.i, (o"0 ~:) z'~) ~

which maximizes joint profits. Also from first order conditions, the transfer payment fromj to i is equal to the weighted average of the effect of negotiated quality on their individual profits; i.e.,

TO * "0i ^A " "ji "U * '^P









marginal cost, given quality levels. Now assume that firms I and 2 merge and let 1 represent the merged firm. Assumption 2 describes a possible effect of the merger on marginal production costs.

Assumption 2: The change in marginal production costs in market A

(respectively, market B) is less than the effect that positive network

externalities have on the combined firm's marginal extra-market revenue in A (respectively, B) from firm 2's (respectively, firm l's) original market plus the

effect that quality changes have on marginal revenue, i.e.,

p1, ;A . +1A) 1," + P(" ,' iA ,A 3 +
+ ,A .q , + _ .A0 1 1 q P >( 4 - 8 )
,A2 + . 'A1,A* TT _ + l2 B 1,'B '1,A I,A ol2 + 'A,91.2 +1, 4,A *

where A6"' is the merger's effect on firm l's internal quality. The pre-merger marginal revenue for firm 1 is p1," + p 1',A, q1,A
kp1,A* I,'A * ,*A ,'IJ
o. + P , . qA } ' is the effect of a change in internal quality on marginal J I,A* I, I [I,A* I T_ 1,2)
revenue, o, + p.A,.2 � qA " 912) and is the effect of changing the external quality between firms 1 and 2 to internal quality for the merged firm. Proposition 2 provides this section's initial result. Proposition 2. Suppose Assumption 2 holds in the all-monopoly setting. Then a pure

cross-border merger results in higher output and welfare. If Assumption 2 does not

hold such that inequality in (4-8) is reversed, then a pure cross-border merger reduces

output and welfare.








If Assumption 2 holds, a pure cross-border merger increases output because the

merged firm internalizes positive network externalities. Before the merger, if firm 1 were to increase its output in market A, firm 2's profit in market B would increase because demand in market B would increase. However, because firm 2's profits do not directly benefit firm 1, firm 1 does not take this increase in firm 2's profits into consideration when choosing quantity in market A. After the merger, firm i internalizes the effect that the quantity sold in market A has on demand in market B, and visa versa. Consequently, firm i increases output in both markets A and B above the levels delivered by firms 1 and 2 before the merger, all other things being equal.

Welfare increases (respectively, decreases) when a pure cross-border merger increases (respectively, decreases) quantity because both customers and producers are better off (respectively, worse off) when quantity increases (respectively, decreases). To illustrate, consider a pure cross-border merger that increases output of all firms. All new customers enjoy increased net surplus, because these customers received zero net surplus before the merger and now receive a positive net surplus.8 Also, because the marginal value of network externalities is decreasing in quantity, the difference between consumer value and price increases for existing customers when quantity increases, causing an increase in net surplus for existing customers. Profits are also higher after the merger because both inside and outside firms can always choose the same outputs after the merger as they did before the merger, but both types of firms voluntarily increase output.




8 All customers who purchase receive a positive net surplus, except for the marginal customer, who receives zero net surplus.








Because the pure cross-border merger increases both net consumer surplus and profits, the merger increases welfare.

Because of these external effects, pure cross-border mergers can increase the merged firm's output even if the merger raises marginal production costs. If the crossborder merger for firms 1 and 2 increases marginal production costs in market A by an amount less than the marginal increase in demand in market B that firm I internalizes, then firm I increases output in market A above the level delivered by firm 1. If the increase in marginal production costs exceeds the marginal increase in demand that the merged firm internalizes, then output decreases.

A merger changes how the merging firms choose quality because the connection between markets A and B changes from an external connection to an internal connection. This change may cause quality to decrease or increase depending on how the cost of quality varies with the number of interconnections, the numbers of customers in each market, and the value that customers place on the network service. The following corollaries provide rules of thumb for the effects of pure cross-border mergers on quality. Recall that P is equal to either one or the number of interconnections between firms i and j.

Corollary 2A. A pure cross-border merger in an all-monopoly situation increases

(respectively, decreases) internal quality for a merging firm if the marginal effect on

revenue of increasing internal quality for the merged firm is greater than

(respectively, less than) the ratio of the post-merger fJ to the pre-merger P', times the








pre-merger marginal effect on revenue of increasing internal quality, i.e., internal f iA1, l , B Im 1"m
quality increases for firm i if p; q + p jq > - Pl qm


If the cost of quality is independent of the number of interconnections, it is

sufficient for internal quality to increase for the output in market m to be unchanged by the merger. If the cost of quality is linear with respect to the number of interconnections, then internal quality increases only if the marginal effect on revenue of increasing quality after the merger is three times what it was before the merger. In the special case where the value of network externalities is linear with respect to quality and output, internal quality for customers in market m increases only if the post-merger output in markets A and B is three times the pre-merger output in market m.

In some situations, quality may be fixed by industry standards or by a regulatory authority, or the cost of quality may be sufficiently low so that firms choose the maximum quality. The following corollaries provide two additional rules of thumb for pure cross-border mergers in the all-monopoly situation when quality is not equal to zero and is unaffected by the merger:

Corollary 2B. A pure cross-border merger in an all-monopoly situation increases output

and improves welfare as long as it does not increase marginal production costs if

quality is unchanged.

Corollary 2C. In an all-monopoly situation where quality is unchanged by the merger,

increasing the number of markets involved in a pure cross-border merger increases

output and welfare, all other things being equal.









Corollary 2C is simply a special setting of Proposition 2. Corollary 2D follows directly from Proposition 2. Increasing the number of markets involved in the merger increases the amount of positive network externality that the monopoly is able to internalize. As a result, the monopoly chooses higher output, which improves welfare.

Now consider the oligopoly setting 2 in which, initially, firms 2 and 3 serve

market B and firm 1 is a monopoly in market A. Now let firms 1 and 2 merge and let I represent the merged firm. Proposition 3 and the following corollaries provide the initial results for a pure cross-border merger in this oligopoly setting. Proposition 3. Suppose Assumption 2 holds in oligopoly setting 2. Then a pure crossborder merger results in higher output and welfare. If Assumption 2 does not hold

such that inequality in (4-8) is reversed, then a pure cross-border merger reduces

output and welfare.

When Assumption 2 holds, the merging firms increase their output because they are internalizing network externalities. Depending on the value customers place on network externalities and the relative sizes of markets, rivals to the merged firm either increase or decrease their output in markets. Corollaries 3A and 3B describe sufficient conditions for the rivals to increase their output. Corollary 3A. If Assumption 2 holds in oligopoly setting 2, a pure cross-border merger

causes the merging firm's rival in market B to increase its output if the effect of

positive network externalities from greater output in market A, plus the effect of the

change in external quality, is greater in absolute value than the effect of greater output








in market B, i.e.,

3,B B 33,B iA 1.A* 3B B* B
, +Pq.,. q3. q l + a1.3 +1qBO,3 q ]
_________,.*,.* .'."I (o ,, * *"
1,3 9 1,A* 1,3* +q 2 ,2 3 B 3,B ~3, B - 2,B*
0> +P~. ~ ~ q q q I
-q +q qq The term , Pq;. ., q-1, - q ,A represents the positive effect of


3, * 3 ~B* ). (0 ,3* , q " + q2,B '2,3 an increase in output in market A, 38+ * 3,., *3 q +qq2,B OT+ pqO. "3 q . qIA* +q,8 q' +q'

is the effect of the change in external quality between firm 3's customers and the customers of the merged firm, and -3, ", , i -q2 peSents the negative effect of an increase in the merged firm's output in market B.

Corollary 3A describes the situation where network externalities are sufficiently great to cause profits to increase for the merged firms' rival. Welfare increases because net consumer surplus and all firms' profits increase as a result of the merger. Even if the merged firms' rival decreases its output in response to the merger, output increases in market B because the rival would decrease its output by an amount less than the merging firms' increase. (Farrell and Shapiro, 1990b)

If follows from Corollary 3A that the merging firms' rival increases its output more (respectively, decreases it less) if the number of markets that provide the network service increases, all other things being equal. Likewise, increasing the number of markets where the rival and the merging firms compete causes the rival to increase its








output less (respectively, decrease it more) in response to an increase in the merging firms' output.

These results from Proposition 3 provide additional rules of thumb, which are summarized in the following corollaries.

Corollary 3B. A pure cross-border merger in oligopoly setting 2 improves net consumer

surplus as long as it does not increase marginal production costs and quality remains

unchanged.

Corollary 3C. Increasing the number of markets involved in a pure cross-border merger

improves the welfare affects of the merger if marginal production costs and quality

remain unchanged.

In oligopoly setting 2, cross-border mergers between larger firms are more

beneficial for merging firms and customers than mergers between smaller firms. Recall that before the merger, firm 1, which operates in market A, can choose whether to merge with firm 2 or firm 3, which compete in market B. Further assume that firm 2 is larger than firm 3 before the merger because firm 2 has lower marginal production costs than firm 3 (Farrell and Shapiro, 1990b) and that the merger has the same effect on marginal production costs regardless of which firm merges with firm 1. If firm 1 merges with firm 2, the larger firm in market B, it internalizes more network externalities than if it merges with the smaller firm because the network externalities that it internalizes are determined by the output that the merged firm produces in market B. This quantity is greater if firm

1 merges with the larger firm in market B. The following corollary summarizes these results.






87

Corollary 3D. A large-firm pure cross-border merger in oligopoly setting 2 increases net

consumer surplus more than a small-firm merger.

A large-firm merger reduces industry costs more than does a small-firm merger. The firms in the cross-border merger increase their production relative to the outside firms. Because the larger firm in market B has lower marginal production costs than the smaller firm, increasing the larger firm's market share provides lower industry costs than increasing the smaller firm's market share. This decrease in industry production costs occurs even though the sum of the marginal costs and the variance of marginal costs are both constant. This modifies the results of Bergstrom and Varian (1985) and Salant and Shaffer (1999), who respectively found that a market's output is determined by the sum of the marginal costs of the firms in the market and that there is a negative correlation between aggregate production costs and the variance of marginal costs.

Others have found that higher Herfindahl-Hirschman indices can be associated with higher welfare in Cournot models (Farrell and Shapiro, 1990b; and Salant and Shaffer, 1999). This finding also holds for pure cross-border mergers in network industries. Specifically, when any increase in marginal production cost for the merging firms is more than outweighed by positive network externality effects of the merger, the market share for the merging firms increases. This causes the Herfindahl-Hirschman index and net consumer surplus to increase. Per Corollary 3A, if the merging firm's rivals operate in markets where the merging firms do not and if the rival's output in these markets is large relative to its output in the market where it competes with the merging firm, then the merger unambiguously increases welfare even when the HerfindahlHirschman index has increased.











4.5. Mixed Mergers


In this section I consider an oligopoly setting where a merger decreases the

number of firms in one market and the number of markets served by the merged firm is greater than the number of markets served by individual merging firms before the merger. Mixed mergers, such as the one in this model, raise issues of market concentration and extending monopoly or dominant market position for regulators.

Consider oligopoly setting 1 where firm 1 is a monopolist in market A and a

duopolist in market B. Also suppose firm 2 is a monopolist in market C and serves as the second duopolist in market B. Now assume that firms 1 and 2 merge and let Y represent the merged firm. After the merger, customers form new expectations and the merged firm selects its quality and output to maximize its profits. I call market B the common market because both merging firms operated in it before the merger, and I call the markets A and C the non-common markets because only one merging firm operates in each of these markets before the merger.

The mixed merger has two effects on output. First, the merger removes a

competitor in the common market. All other things being equal, this causes output to decrease in this market. Second, the merger increases the network externalities that merging firms internalize in non-common markets. To illustrate, consider how the merging firms choose output in market A. Before the merger, when firm 1 chooses its output, it internalizes network externalities from its output in market B. After the merger, when the merged firm chooses its output in market A, it internalizes network externalities









from its output in both markets B and C. Not only does the merged firm internalize network externalities from an additional market, but it also internalizes network externalities from a larger output in the common market, market B, because the merged firm serves the entire market. As a result, unless the post-merger output in markets C and B are less than firm l's pre-merger output in market B, the merged firm internalizes more network externalities than the pre-merger firm when choosing output in market A. The corresponding analysis for market C would provide analogous conclusions.

The mixed merger does not necessarily increase network externalities that the merged firm considers when it chooses its output in the common market. Before the merger, the merging firms jointly internalize network externalities from markets A and C when choosing their output for market B. The merger does not change the total number of markets served by merging firms. As a result, unless output in non-common markets increases as a result of the merger, the network externalities internalized for output in market B are no greater after the merger than before the merger. However, the merger does increase the number of markets whose network externalities the merged firm internalizes when choosing output in non-common markets.

Now consider the effect of the mixed merger on quality. Proposition 4 describes how the mixed merger affects quality. Assumptions 3 and 4 provide conditions for Proposition 4.


Assumption 3: The difference between , the ratio of the post-merger and

pre-merger correlation between quality costs and internal connections, times

the pre-merger marginal revenue from internal quality and the post-merger








marginal revenue from internal quality in firm i's pre-merger markets is less

than the post-merger marginal revenue from internal quality in the market that

i did not serve before the merger, i.e.,


Z. . . . -im l,, (4-9)
meM' mem

where Mi is the set of markets served by i before the merger and h M'.

Assumption 4: The difference between the ratio of the post-merger and premerger correlation between quality costs and internal connections, times the

pre-merger marginal revenue from external quality in firm i's pre-merger

markets, is less than the post-merger marginal revenue from internal quality in

all markets, i.e.,

"uqm m * rnaqjim""*. I'm mpe+ m M eM (4-10)
ai~j qnM

Proposition 4. A mixed merger in oligopoly case 1 increases internal quality if

Assumption 3 holds and results in an internal quality that is higher than the premerger external quality if Assumption 4 holds. Alternatively, the merger decreases

internal quality if Assumption 3 does not hold and the inequality (4-9) is reversed and

results in an internal quality that is lower than the pre-merger external quality if

Assumption 4 does not hold and the inequality in (4-10) is reversed.

The merger increases the number of markets that a single firm considers when

choosing internal quality. As a result, if the cost of quality is independent of the number of interconnections, the merger increases internal quality for the markets served by firm i unless the merger causes a decrease in output sufficient to lower marginal revenue from









internal quality in all markets to less than what it was for i's pre-merger markets. The merger also changes from external to internal the communications between markets A and C, between market C and firm l's customers in B, between market A and firm 3's customers in B, and between firm l's customers in B and firm 3's customers in B, which changes how quality is chosen for these interconnections. It follows from Proposition 1 that changing an interconnection between markets from external to internal may lower quality.

I use a simulation to illustrate the mixed merger's effect on output and welfare. For the simulation, I adopt the supply and demand functions explained in the Appendix and assume that the merger has no effects on marginal production costs.9 Figure 1 shows the effect the merger has on net consumer surplus for various sizes of the common market relative to the non-common markets and for various values of network externalities. The curved plane shows the region where the effect of the merger is to increase net consumer surplus. The region where no plane is shown represents a decrease in net consumer surplus. The light colored portion of the plane represents the area where quality is less than one, but more than zero. The merger always decreases net consumer surplus when quality is zero because consumers and firms receive no benefits from network externalities. The merger increases industry profits for all regions, so the curved plane also shows the region for which the merger unambiguously increases welfare.

The change in net consumer surplus increases with the value customers place on network externalities because higher values increase the firm's marginal revenue from


9 The Appendix provides further details of the simulation.




Full Text

PAGE 1

RESTRUCTURING IN NETWORK INDUSTRIES: COMPETITION AND MERGERS IN TELECOMMUNICATIONS By MARK A. JAMISON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2001

PAGE 2

Copyright 2001 By Mark A. Jamison

PAGE 3

To my wife (Patti), my children (John and Laura), and my parents (John and Judae)

PAGE 4

ACNOWLEDGMENTS I must first thank my supervisory committee members. Dr. David E. M. Sappington, the chair of the committee, took numerous hours to give direction and advice and to correct errors. His knowledge and insight were invaluable. Dr. Steven Slutsky, a member, provided important guidance throughout my graduate studies and was a quick study of my many efforts. Dr. Chunrong Ai, a member, provided expert advice on econometric problems and the presentation of econometric studies. Dr. Heather Elms, the external member, gave good direction and insights beyond the normal range of economics. I hereby thank them all again. Outside of my supervisory committee, I am grateful to Dr. Sanford Berg, who made it possible for me to come to the university, and whose opinions and leadership are respected worldwide. I am also indebted to Dr. W. Andrew McCollough for his foresight in allowing Dr. Berg and me to embark on the enterprise that we have built at the university. I am also grateful for advice from Dr. Walter Beckert and Dr. David Figlio. A special thank you is due to my colleague, Janice Hauge, without whom graduate studies would have been impossible. I would also like to thank Eric Chiang, Josh Silverboard, Jason Mills, Eileen Pun, Verizon, Sprint, SBC, and BellSouth for their assistance with data. Thanks are also due to James Sichter of Sprint and Mark Sievers, Ed Lowery, and Link Hoewing of Verizon for their willingness to discuss issues and provide insights. IV

PAGE 5

Lastly, I would like to thank all of the Public Utility Research Center (PURC) and departmental staff that have made the world work during my tenure as a student. Cynthia Stehouwer and the rest of the PURC staff kept my project finances and tasks in order; Misty Swain kept my graduate program in order; Dian Studstill kept all administrative matters flowing; and Dorthy Padgett provided important assistance in producing my book.

PAGE 6

TABLE OF CONTENTS Page ACKNOWLEDGMENTS iv ABSTRACT viii CHAPTERS 1 RESTRUCTURING IN NETWORK INDUSTRIES : COMPETITION AND MERGERS IN TELECOMMUNICATIONS 1 2 MARKET CONCENTRATION AND INCENTIVES TO DISCRIMINATE AGAINST RIVALS IN NETWORK INDUSTRIES 5 2.1 Introduction 5 2.2 The Model 10 2.3 Symmetric, Single Market Case 17 2.4 Monopolist Entry into a Competitive Market 18 2.5 New Competitive Entry into a Formerly Monopoly Market 20 2.6 Conclusion 24 3 NETWORK INTERCONNECTION AND TELECOMMUNIATIONS COMPETITION: THE CASE IN THE U.S 26 3.1 Introduction 26 3.2 Literature Review 31 3.3 Historical Background 33 3.4 The Models 39 3.5 Model Results 54 3.6 Conclusion 63 4 NETWORK EXTERNALITIES AND CROSS-BORDER MERGERS IN NETWORK INDUSTRIES 64 4.1 Introduction 64 4.2 The Model 69 4.3 Quality Choices 75 VI

PAGE 7

4.4 Analysis of Pure Cross-Border Mergers 79 4.5 Mixed Mergers 88 4.5 Conclusion 93 5 CONCLUSION 95 APPENDICES A PROOFS AND LEMMAS FOR CHAPTER 2 99 B APPENDIX FOR CHAPTER 3 107 C PROOFS AND LEMMAS FOR CHAPTER 4 HO REFERENCES 119 BIOGRAPHICAL SKETCH 125 « VII

PAGE 8

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RESTRUCTURING IN NETWORK INDUSTRIES: COMPETITION AND MERGERS IN TELECOMMUNICATIONS By Mark A. Jamison May 2001 Chairperson: David E. M. Sappington Major Department: Economics This dissertation addresses three issues. The first is whether large service providers in network industries, such as large Internet backbone providers, have an incentive to discriminate against their smaller rivals. Using Coumot interactions, each network provider simultaneously chooses quantity, quality for communications within the provider's own network (internal quality) and quality for communications between the provider's network and other networks (external quality). Larger networks choose higher internal quality than do smaller networks. All networks choose lower external quality when connecting with smaller networks. Large networks and small networks choose identical external quality when interconnecting with each other. Lastly, incumbent providers are sometimes willing to raise rivals' per customer costs, but never their interconnection costs. The second issue is how regulatory policies relating to network interconnection affect entrants' incentives and abilities to enter local telecommunications markets and vm

PAGE 9

expand supply. I find that low prices for reselling incumbents' services, leasing facilities, and exchanging traffic increase entry, but time for entry, access to incumbents' buildings, and reform of universal service subsidies do not. However, if prices for leasing facilities are low relative to incumbents' retail prices, then less entry occurs, presumably because incumbents exert effort to protect their retail profits. Higher prices for exchanging traffic increase entrants' market share, presumably because they target customers (such as Internet Service Providers) who receive more calls than they make. Low prices for reselling services do not cause entrants to choose reselling over other supply methods. The third issue is how mergers affect market performance in a network industry. Network providers choose quality for communications within the provider's own network, quality for communications between the provider's network and other networks, and output. The dissertation shows that cross-border mergers provide firms with an incentive to increase industry output because mergers allow firms to internalize positive network externalities. The dissertation further shows that cross-border mergers improve the welfare effects of horizontal mergers in certain circumstances. Lastly, mergers that are both cross-border and in-market increase welfare if the aggregate size of the markets that the pre-merger firms had in common before the merger is large relative to the aggregate size of the markets that they individually served before the merger, and if the value customers place on network externalities is high. ix

PAGE 10

CHAPTER 1 RESTRUCTURING IN NETWORK INDUSTRIES: COMPETITION AND MERGERS IN TELECOMMUNICATIONS This dissertation consists of three studies of the restructuring of communications industries. The first study examines market concentration in the Internet. This is one of the issues that caused the collapse of the WorldCom-Sprint merger. I examine an oligopoly model in which firms simultaneously choose quantities, network quality for internal communications (internal quality), and interconnection quality for communications between networks (external quality). I show that firms base their quality choices on network size, customer value of interconnection, and the cost of interconnection. Networks choose lower external quality when connecting with smaller networks than when connecting with larger networks. Larger networks and smaller networks choose the same external quality when intercormecting with each other. I also find that larger networks will not optimally choose to sabotage rivals by raising the rivals' costs of interconnection. I develop and explain my results by applying a static oligopoly model. Firms have identical cost functions of providing quality, but may have different cost functions for production. Cost asymmetries arise from differences in network sizes. An incumbent can raise its rivals' costs, for example, by delaying access to essential facilities. Firms 1

PAGE 11

compete for customers and seek to maximize their individual profits by simultaneously choosing quantity, internal quality, and external quality. I examine a case in which two firms serve a single market and compete in this market against a third firm that has a monopoly in another market. I show that a Nash equilibrium exists in this setting wherein the firms choose identical external qualities. Furthermore, each smaller firm chooses an internal quality that is lower than its external quality for interconnecting with the larger firm. Smaller firms make these choices because the number of customers reached through a network affects firms' quality choices and the smaller firms' smaller quantities of customers are less valuable than the larger firm's larger number of customers. Also, because I assume that smaller firms are symmetric in size, the smaller firms choose external qualities for interconnecting with each other that are equal to their internal quality choices. I then examine a case in which the first two firms can enter the former monopoly market, but are subject to a cost disadvantage because of the incumbent's established network. I show that firms agree upon external quality and that the incumbent's quantity choice is larger than its rivals' quantity choices. Finally, I examine whether the incumbent would sabotage its rivals by raising their costs. I show that the incumbent may raise its rivals' costs for quantity in certain situations, but the incumbent never raises its rivals' costs for intercormection, even when it can do so costlessly. The incumbent is unwilling to raise rivals' interconnection costs because the higher costs would cause the rivals to choose an external quality that is below the incumbent's optimal choice. In my third chapter, I find that regulators' basing interconnection prices on incremental cost has given incumbents an incentive to hinder competitive entry.

PAGE 12

However, these prices have not led incumbents to hinder entrants' abihties to gain market share. I further find that the opportunity for entrants to place facilities in incumbents' buildings (called collocation) is important to encouraging facihties-based entry and causes entrants to substitute facilities and unbundled network elements for resale-based entry. Higher symmetric reciprocal compensation prices discourage entry and cause entrants to select customers that receive more calls than they send. High wholesale discounts for resold services encourage entry, but do not encourage entrants to substitute resale-based entry for facilities-based supply. In my fourth chapter, I analyze mergers in network industries. I develop a model in which up to three firms serve three markets, which share network externalities. I first examine pure cross-border mergers, mergers between firms that do not serve any of the same markets prior to the merger. In the initial setting, each firm serves a single market, two of the three markets are served by monopolies and the remaining market is served by a duopoly. I examine the Coumot-Nash equilibrium quantities as a usefial benchmark. I then impose an exogenous merger between one of the duopolists and one of the monopolies. The merger allows the new firm to internalize network externalities between the two markets that it serves. With quality constant, his increases quantities and improves welfare unless the merger increases costs by an amount greater than the marginal value of the network externalities that the merged firm internalizes. The merged firm increases its market share in the duopoly market. I show that a merger between larger firms improves welfare more than a merger between smaller firms. I also show that quantities and welfare increase with the number of markets that are added together by the cross-border merger.

PAGE 13

4 This dissertation proceeds as follows. Chapter 2 examines network interconnection and incentives to discriminate and raise rivals' costs. Chapter 3 is an empirical study of the effects of regulatory interconnection policies on the development of competition. Chapter 4 studies the effects of mergers. Chapter 5 is the conclusion. All proofs are in the appendices.

PAGE 14

CHAPTER 2 MARKET CONCENTRATION AND INCENTIVES TO DISCRIMINATE AGAINST RIVALS IN NETWORK INDUSTRIES 2.1. Introduction The liberalization of telecommunications markets has been marked by a number of mergers and alliances, some of which have been record setting. In 1995, US telecommunications businesses were involved in acquisitions worth $39.1 billion. This increased to $154.8 billion by the first half of 1998. Global mergers and acquisitions in the information technology, communications, and media industries jumped 87 percent between 1997 and 1998 to $488.8 billion (Capron and Mitchell, 1997; Broadview, 1998). Telecommunications accounted for 20% of all merger and acquisition activity worldwide in 1999. Recent examples of mergers and alliances include the merger of Bell Atlantic with NYNEX in 1997, and then with GTE in 2000; SBC's acquisition of Pacific Telesis in 1997, Southern New England Telephone in 1998, and Ameritech in 1999; AT&T and British Telecom's formation of a global joint venture in 1999; WorldCom's purchase of MCI in 1998; and Vodafone's acquisition of AirTouch in 1999 and Mannesmann in 2000. Competition regulators, utility regulators, and others often raise concerns that telecommunicadons mergers might decrease competition. Such concerns prompted the European Union (EU) to place restrictions on Global One and on the now defunct British Telecom and MCI alliance (McDavid, 1997). The EU also required MCI to divest a 5

PAGE 15

I 6 portion of its Internet business as a condition of approving WorldCom's purchase of MCI (Cremer, Rey, and Tirole, 2000). The EU halted WorldCom's planned purchase of Sprint because the EU's competition commissioner believed that the merged company would dominate transmission of information over the Internet. Olbeter and Robison (1999) argue that market concentration in the Internet backbone in the US results in some rural states having little access to the Internet. In contrast, in a joint study, the US Department of Commerce and the US Department of Agriculmre (2000) found that the high cost of broadband local lines, not access to the Internet, causes rural areas to have less broadband access than urban areas. Cremer et al. (2000) argue that larger Internet firms have an incentive to lower the quality of their interconnection with smaller rivals. In contrast, some research indicates that telecommunications mergers can improve market efficiency. Jamison (1999a) explains how regulation in the US and government ownership of service providers elsewhere have held the industry to an antiquated structure for decades. Market liberalization is forcing incumbent companies to restructure to meet the new economic realities or risk failure. Weisman (1999) shows how mergers among incumbent local exchange companies make it more likely for the merged company to compete against other incumbent local exchange companies. Galbi and Keating (1996), Jasinski (1997), and Jamison (1998; 1999a) explain how network providers become global to attract large multinational communications customers who want their international communications to be provided by an integrated multinational network. Firms such as AT&T and WorldCom meet this demand by establishing local networks in multiple countries and integrating these local networks with their

PAGE 16

7 international networks. Jamison (1999b) shows how this process can increase competition and improve welfare. This dissertation extends this research by examining market concentration in the Internet. This is one of the issues that caused the collapse of the WorldComSprint merger, the other issue being the US Department of Justice's concern that the combined company would have 27% of the US consumer long distance market. I examine an oligopoly model in which firms choose network quality for internal communications (which I call internal quality) and interconnection quality for communications between networks (which I call external quality). I show that firms base their quality choices on network size, customer value of interconnection, and the cost of interconnection. Networks choose lower external quality when connecting with a smaller network than when connecting with larger networks. I find that larger networks and smaller networks choose the same external quality when interconnecting with each other. I also find that incumbent firms will raise entrants' quantity costs, but will never sabotage rivals by raising the rivals' costs of interconnection. I develop and explain my results by applying a static oligopoly model in which firms have identical cost structures for quality, but not necessarily for quantity. When a cost asymmetry exists, it results from an incumbent firm having a network that is larger than other firms' networks at the start of the game. Also, an incumbent can raise its rivals' costs, for example, by delaying access to essential facilities. I apply a two-stage game in which an incumbent chooses whether to sabotage rivals in the first stage. Firms compete for customers and seek to maximize their individual profits in the second stage by simultaneously choosing quantity, internal quality, and external quality. In contrast

PAGE 17

8 with Cremer et al. (2000), who assumed that internal quality was fixed, I allow firms to choose internal quality. This generalization allows an expanded examination of firms' quality choices. As a benchmark, I first examine a case in which three identical firms compete in a single market. I find that the firms make idenfical quality choices. I next examine a case in which two firms serve a single market and compete in this market against a third firm that has a monopoly in another market. I show that a Nash equilibrium exists in this setting wherein the firms choose identical external qualities. Furthermore, each smaller firm chooses an internal quality that is lower than its external quality for interconnecting with the larger firm. Smaller firms make these choices because the number of customers reached through a network affects firms' quality choices and the smaller firms' smaller quantities of customers are less valuable than the larger firm's larger number of customers. Also, because I assume that smaller firms are symmetric in size, the smaller firms choose external qualities for interconnecting with each other that are equal to their internal quality choices. I then examine a case in which the first two firms can enter the former monopoly market, but are subject to a cost disadvantage because of the incumbent's established network. I show that, as in the first situation, firms agree upon external quality and that the incumbent's quantity choice is larger than its rivals' quantity choices. Finally, I examine whether the incumbent would sabotage its rivals by raising their costs. I show that the incumbent may raise its rivals' costs for quantity in certain situations, but the incumbent never raises its rivals' costs for interconnection, even when it can do so costlessly. The incumbent is unwilling to raise rivals' interconnection costs

PAGE 18

9 because the higher costs would cause the rivals to choose an external quality that is below the incumbent's optimal choice. My results conflict with the findings of Cremer et al. (2000). Their model assumes that differences in firm size result from differences in installed customer bases, which they call attached customers. Attached customers are customers who cannot change service providers nor their quantities purchased for the duration of the game. Firms choose external quality based in part on the value of allowing their own customers with connectivity to rivals' attached customers. Cremer et al. (2000) assume that firms must meet the quality demands of new customers, but incur no penalty for ignoring the preferences of their own attached customers. As a result, when two firms interconnect, the firm with the larger number of attached customers chooses a lower external quality than does its rival, which has the lower number of attached customers. Cremer et al. (2000) support their assumptions regarding attached customers by explaining that customers sign long-term contracts and quality is difficult to observe. However, contracts that allow firms to ignore customer preferences should not be part of a Nash equilibrium because the customers would be better off never signing such contracts. Furthermore, if quality is hard to observe, it is hard to observe for both new and existing customers, so it is unclear why firms must serve the quality preferences of new customers while ignoring the quality preferences of existing customers. To avoid these difficulfies, I assume that asymmetries in firm size result from access to monopoly markets and cost asymmetries. Cremer et al. (2000) also assume that firms choose quality before choosing output. This assumption may apply in some network industries, but it does not apply to all. For example, if quality is determined by software features in a physical network.

PAGE 19

10 changing quality may be as simple as activating or deactivating a software feature, which can be done in a matter of hours or less. Also, if quality is determined by choosing poorly working circuits versus well-functioning circuits for interconnecting physical networks, then changing quality may be as simple as redirecting circuits, which can be done by entering commands at a computer terminal. To allow for these situations, I assume that firms simultaneously choose qualities and quantities. I find that asymmetry in firm size in my model does not result in the interconnection problems identified by Cremer et al. (2000). The analysis proceeds as follows. Section 2.2 describes the model. Section 2.3 presents the benchmark case in which three identical firms compete in a single market. Section 2.4 presents the case in which one of the firms has a monopoly in another market. Section 2.5 considers the case of entry into a formerly monopoly market and the incumbent's incentives to sabotage rivals. Section 2.6 is the conclusion. Proofs are in the Appendix A. 2.2. The Model I consider an extension of a model developed by Cremer et al. (2000). There are two markets, A and B, for network communications and three firms. Markets are distinct because they are separated by geography. There are customers in market // e {A, B}. Network providers compete for customers in a single period, qi,^ > 0 will denote the number of customers that firm / serves in the market. I assume that q^ is sufficiently large in each market to ensure that there are unserved customers in equilibrium; i.e., in equilibrium q^>q],^+ qi,^ +

PAGE 20

11 Customers are identical in each market and a customer of type r g [0, r] (where r is the same in both markets) obtains a net surplus from buying from firm / at price pi,^ equal to r + s,^ p,. ^ , where Si,^ denotes the value that the customer places on /'s network. I assume that ris uniformly distributed.' Customers desire to communicate with customers in both markets, so 5,,^ is given by I let V E (0, 1/4) represent a parameter that reflects the constant marginal value that customers place on network communications of a given quality. This linearity assumption of value follows Cremer et al. (2000) and implies that, except for the firms' quality choices, each customer of type ris indifferent with respect to which customers the r-type customer communicates. Because customers always place positive value on being able to communicate with additional customers, 5,,^ is increasing in quantity. Because firms connect networks across markets, 5,,^ for customers in one market is increasing in quantity for both markets. Let <^,, g [0, 1] be firm /'s internal quality choice and G [0, 1] be the quality of external interconnection with fs network for j i. For simplicity, I assume that each firm chooses a single internal quality and, for each network with which it interconnects, the firm chooses a single external quality. In other words, if firm / serves both markets, it provides a single quality interconnection for firm j, as opposed to providing different quality interconnections for each market. I fiirther assume that the value firm /'s customers place on communicating with customers of firm j is independent of the quality offered for communicating with customers of firm k,k^ j. I ' This assumption results in a linear demand curve.

PAGE 21

12 assume that networks are homogeneous except for the providers' quality choices. One imphcation of these assumptions is that, if networks have the same number of customers and offer the same quaUty levels, then customers are indifferent between firms. To focus the analysis on how network providers can affect the quality of each other's service offerings, I restrict the quality choices to such things as technical features and reliability that firms can affect when interconnecting their networks. For example, America Online offers proprietary content and fiinctions for its own users and not for users of other Internet Service Providers. I omit other aspects of quality, such as bandwidth for customer network access, which are unaffected by interconnection. Because only one quality choice can prevail for each network interconnection, the firm with the lowest external quality preference determines the interconnection quality; i.e., the quality of the interconnection between / and j is min{&,j, For example, if one firm chose a capacity of 45 megabits per second and the other chose a capacity of 30 megabits per second, only 30 megabits per second of information could be passed between the networks. The proof to Lemma Al in Appendix A shows that this results in an infinite number of Nash equilibria. In all instances in this chapter, the interconnecting firms make identical choices when optimizing external quality, so I use this as a focal point for the equilibrium choice. Customers make their purchasing decisions after firms have made their quality and quantity choices. I assume no price discrimination and further assume that prices ^ Technically, the term ^,^in (1) should be qi,fj-\ because customers do not obtain value from accessing themselves. I suppress the -1 and assume that q is sufficiently large that it does not affect the results. If the -1 were included, its effect would be to lower each firm's internal quality relative to external quality.

PAGE 22

13 adjust to firms' and customers' equilibrium choices. Lemma 1 describes customers' network preferences in equilibrium. Lemma 1. Given the assumptions of the model, each customer of type ris indifferent between networks at equilibrium; i.e., r + 5, ^ p, ^ = t + Sj ^ ^ for every i,j = 1, 2, 3 and / ^ j. It follows from Lemma 1 that if firm / attracts customers, it has a quality-adjusted price I define the marginal customer to be the customer that, in equilibrium, is indifferent between buying and not buying network service. Such a customer exists because I assume that x e [0, x], costs are strictly positive, and v, the constant marginal value of connectivity, is sufficiently large relative to firms' costs to ensure that q^^ > 0 for every firm. At equilibrium, the marginal customer will receive zero net surplus and so will have a value of r = p^. Because the distribution of customers is uniform, the quantity of customers that firms choose to serve in equilibrium is simply the total number of customers that lie between the upper bound of customer preferences and p^. Normalizing the density of customers to 1 in each market, the quantity of customers served in market // is

PAGE 23

14 Combining (2-1), (2-2), and (2-3) gives the customers' inverse demand curve for firm / in market // J -3 J (2-4) j*> J Existing (incumbent) firms have identical innate cost functions. New entrants have a cost penalty because they must build networks. Actual costs for some firms may exceed innate costs because incumbents in formerly monopoly markets can raise rivals' costs. For example, incumbents may provide rivals with inferior access to essential facilities or withhold critical network information (Economides, 1998). Entrants cannot raise incumbents' costs in this way because they do not have essential facilities nor do they have network information that the incumbent does not have. Also, incumbents might deny access to rights of way by either refusing access or by using all the capacity of the rights of way. Incumbents might also require points of interconnection to be in locations that are far from its rivals' customers, causing the rivals to incur extra costs to reach their customers. The US Federal Communications Commission recently fined GTE $2.7 million for allegations that GTE denied rivals access to GTE facilities for locating equipment. Incumbents might also delay rivals' market entry by raising legal objections. Because the quality choices are restricted to quality that can be affected by interconnection, the costs of adding customers and of providing quality are separable. Incumbents incur a constant marginal cost c > 0 of serving a customer. Entrants' marginal costs are c + co > 0, where co > 0. For simplicity, I assume a symmetric, continuous cost function K{0ij) > 0 of providing quality e &,2, for i = 1, 2, 3. Quality costs are separable from quantity costs and among connections. The

PAGE 24

15 assumption that quality costs are separable is reasonable because it is common for telecommunications firms to have dedicated network equipment for each network interconnection and to have separate equipment for connections for its own customers. I further assume that K{di,j) > 0 for all Oij > 0, K{0) = O,K0> 0, K0e> 0, and Km< 0. Extending Economides' (1998) model of raising rivals' costs and Mandy and Sappington's (2000) model of sabotage, I assume that an incumbent and former monopolist, which I call firm 1, in market A incurs a cost ^{rq) ^ 0 to raise firm 2's and firm 3's costs by Vgqj.A and rqqxA, respectively.' I assume 2(^,) is continuous, ^r^) > 0 if rq > 0, 2(0) = 0,Zr> 0> Zrr > 0, and > 0. Furthermore, firm 1 incurs a cost ^r^) > 0 to raise rivals' costs of external quality for interconnecting with firm 1 by K{O,r0) K{0)>O, where ^r^) is continuous, (^rS) > 0 if r^> 0, ^0) = 0, 4 > 0, > 0, K(0,r0) >K{0)if rg > 0, Ki&,j,0) = K{0ij), Kr > 0, Krr > 0, and r^> 0 for all i,j = 1, 2, 3. As a result, firm 1 's cost of serving qi,A + q\.B customers with quality choices 9i = 0\^2, is c[q,,A > 9,,. , e, ) c{q,^, + J+ X )+ k(0,,)+ zir, ) + ^r,) , (2-5) and rival fs cost of serving qjj + qj^B customers,^ 1, is c(qj,A . ^j.B , e J = (c + Co + )qj, + cq^, + k[0j, , rj+ ^ ) • I assume a two-stage game in which the incumbent chooses its cost-raising activities and then firms compete in quantity and quality. Presuming Nash behavior, each firm takes its rivals' quantity and quality choices as given when it chooses its quantity and ^ Economides (1998) assumed that raising rivals' costs was costless for the incumbent. This can lead to the incumbent preventing all entry, so I impose a cost on the incumbent.

PAGE 25

16 quality levels. Therefore, (2-4) and (2-5) imply that firm I's profit maximization problem can be written as: max = l\,A r c X (l )q,,, (l v^,,, )q,, + vj] ^ _ c ^(l v6',_^. )qj ,, (l + vX'^i,,^;,, 7=2 7=1 subject to j e [0,l] for ; = 1, . . . ,3 q,^>0 for// =^,5. Assuming for simplicity that co = 0 in market B, firm fs profit maximization problem,y ^ 1, can be written as: (2-6) max;r,. = r c X (l v0., )q,^, (l v0^_j )q^^, + k*j k=\ k^k,A ^7. -Z^(^7J-^fe'^.) subject to 4 G [0,l] for k = 1,2,3 qj^>0 for /i^A,B. To ensure internal solutions that satisfy second order conditions, I assume (2-7) , and T -c-Cq > 0 . Also, throughout the dissertation, I assume that identical firms choose identical levels of quantity.

PAGE 26

17 2.3. Symmetric. Single Market Case In this section I consider the case where three identical firms compete in a single market, which I call market B. I designate the firms as 2, 3, and 4 and, because no firm is an incumbent market B, the firms have symmetric cost functions and no firm can raise its rivals' costs. I consider the equilibrium in which these identical firms choose identical levels of quantity. Proposition 1 provides this section's primary result. Proposition 1. In the symmetric, single market setting, each firm sets all of its external quality levels equal to its selected internal quality level; i.e., 0]^ = 0] j for j ^i, for all i,j = 2, 3, 4. In choosing external quality, each firm considers its quantity choice, the quantity choice of the network with which it is interconnecting, and v, the value a customer places on communicating with another customer; i.e., 0'j = K^^(yq'^q* g). Firms consider their own quantity choice because this determines the number of customers that are willing to pay prices that reflect the value of the external quality. Firms consider the other firm's quantity choice because more customers on other networks increase the value of the intercormection. Because firms have symmetric quantity choices, they have symmetric external quality choices. Furthermore, firms determine internal quality based on v and their quantity choice squared; i.e., = K^^ {yq'i,Bl],B ) ^{^',b J\ Their quantity choice is squared because more customers on the firm's own network increase the value of the network, and each customer represents someone who will pay a price that reflects that value. Because in a symmetric equilibrium, all quantity choices are equal, internal quality equals external quality.

PAGE 27

18 From Proposition 1, the symmetric equiUbrium quantity for a representative firm / is Q^R = —I n ' where Firms that serve more customers also choose higher quality. 2.4. Monopolist Entry into a Competitive Market In this section I consider the case where firm 1 has a monopoly in market A and firms 1, 2, and 3 compete in market B. This might represent a situation where the monopoly has merged with firm 4 to enter market B. Costs are symmetric in market B and, because no firm is an incumbent in market B, no firm can raise its rivals' costs. As in the symmetric, single market case, I consider the equilibrium in which identical firms' choices are symmetric. Proposition 2 provides this section's primary result. Proposition 2. In the setting in which a monopolist from one market enters a second market which is competitive, the monopolist and its rivals choose the same levels of external quality for a given interconnection; i.e., 0'j = <^ for j ^ i, for all i,J = 1,2,3. Each firm in the monopolist entry setting considers its quantity choice and the quantity choice of the network provider with which it is interconnecting when choosing external quality. Because value increases with the number of customers reached through an interconnection, the firms make symmetric external quality choices even though their quantity choices may be asymmetric. The monopoly firm does not strategically degrade

PAGE 28

19 the quality of its interconnection with smaller rivals. Corollary 1 further describes firms' quality choices. Lemma 2 is useful for Corollary 1 . Lemma 2. In the monopolist entry setting, the monopolist in market A chooses a greater quantity in market B than does its rivals; i.e., q'^ > q] g for ally 1. Firm 1 chooses a higher quantity in market B than does either of its rivals because it internalizes some network externalities. In other words, its quantity choice in market B has synergistic effects with its quantity choice in market ^-higher output in market B increases the value of the monopolist's network in market A. Furthermore, higher output in market A increases the value of firm I's network in market B. Corollary 1 describes the firms' internal and external quality choices. Corollary 1. In the monopolist entry setting: a. The monopolist's internal quality choice exceeds its external quality choices; b. The rivals implement an internal quality below the external quality they implement for interconnecting with the monopolist; c. The rivals implement an internal quality that is equal to the external quality they implement for interconnecting with each other; and d. The rivals implement an external quality for interconnecting with each other that is lower than the external quality they implement for interconnecting with the monopolist. That is to say, > $1. = 0] , > O] . = O) ^ for j ^ k, for all j,k = 2,3. The monopolist provides the highest quality because its optimal quantity choices are higher than its rivals' optimal quantity choices. These higher quantity choices make

PAGE 29

20 the monopolists competitive network more valuable than its rivals' networks. Furthermore, for connection to a network of a given size, the monopolists' higher quantity choice makes quality more profitable for it than for its rivals. The rivals choose internal qualities that are lower than the quality of their intercormections with the monopolist because connection with the monopolists' network provides more value to their customers than do their own networks. For the same reason, the rivals choose higher interconnection qualities with the monopolist than with each other. 2.5. New Competitive Entry into a Formerly Monopoly Market In this section, I examine a new entry setting, a situation where firms 2 and 3 enter market ^. Because firm 1 is an incumbent, it has a cost advantage of co in A and has the ability to raise its rivals' costs. Lemma 3 shows that interconnecting firms choose the same external quality if the incumbent does not raise its rivals' interconnection costs. Lemma 3. lfrff=0 in the new entry setting, then all firms implement the same external quality for the same interconnection; i.e.,^*^. = . for j ^ i, for all i,j 1,2,3. Proposition 3 provides this sections main findings. Proposition 3. In the new entry setting, the incumbent will raise the entrants' quantity costs. However, it will not raise the entrants' interconnection costs. That is to say, rg* > 0 and = 0. The incumbent chooses to raise its rivals' quantity costs because raising these costs causes the rivals to reduce their quantities, which allows the incumbent to increase

PAGE 30

its output and its profits. However, the incumbent will never raise its rivals' interconnection costs because raising these costs would cause the rivals to implement lower external quality than the level preferred by the incumbent. In some countries, incumbents have restricted interconnection capacity for rivals. This restriction causes calls or messages between the rival networks to be delayed, dropped, or not completed. This would appear to be in conflict with the conclusion that an incumbent would not raise its rivals' interconnection costs nor strategically degrade interconnection quality. But field interviews with customers revealed that customers generally did not understand network interconnection and believed that the service problem was caused by the rivals' failing to provide adequate internal quality. This perception lowered the demand for the rivals' services, which caused rivals to incur extra costs to obtain and keep customers. Therefore, even though the incumbents' choices related to intercormection, the effect was to raise its rivals' quantity costs. Corollary 2 further describes the firms' quantity choices and quality choices. Corollary 2. In the new entry setting: a. The incumbent chooses higher quantities than does each of its rivals; b. The incumbent implements an internal quality that exceeds its external quality; c. The rivals implement the same external quality when connecting with the incumbent as the incumbent chooses; d. The rivals implement internal quality that is equal to the external quality that they implement for interconnecting with each other; and

PAGE 31

22 e. The rivals' implement internal quality that is lower than the external quality that they implement for interconnecting with the incumbent. The incumbent chooses higher quantities because it has a cost advantage, and its higher quantity choices drive its higher quality choices. As in previous cases, firms that are interconnecting make equal external quality choices when interconnecting, so no firm prefers a higher or lower interconnection quality than the other firm prefers. Corollary 3. In the new entry setting, the entrants' quantity choice in market / is a strategic substitute for the incumbent's quantity choice in market j, for all / ^ j and i,j = A, B. However, the incumbent's quantity choice market / is a strategic complement to the entrants' quantity choice in market j. The incumbent and entrants have opposite responses to each other's other-market production because the incumbent's higher internal quality causes it to have a greater response to its internalized network externalities. A firm internalizes more network externalities when it serves two markets than when it serves only one market. Furthermore, higher internal quality internalizes more network externalities than does a lower internal quality. As a result, when a larger firm connects with a smaller firm and both firms serve two markets, the larger firm quantity choice in the first market affects its quantity choice in the second market more than does the smaller rival's quantity choice in the first market. Conversely, the larger firm's quantity choice in the first market affects the smaller firm's quantity choice in the second market more than does the smaller firm's and 0l > 01 = e], > 0'j = e), for j ^ k, for ally, = 2, 3.

PAGE 32

23 own quantity choice in the first market. These opposite reactions resuh from the firms' quahty choices. The larger firm implements a higher internal quality than it does external quality for interconnecting with the smaller firm. Therefore, the larger firm's quantity choice has a greater feedback effect than does the smaller firm's quantity choice. Conversely, the smaller firm implements an internal quality that is lower than its external quality with the larger firm. Therefore, the larger firm's quantity choice has a greater affect on the smaller firm's quantity choice. For example, assume the entrants' increase their quantity in market .4. This higher quality causes the incumbent to choose a lower quantity in market A because the entrants' quantity choice is a strategic substitute for the incumbent's market-y4 quantity. With respect to the incumbent's quantity choice in market B, the entrants' higher quantity in market A has a direct positive effect on the incumbent's quantity choice in market B. However, this direct effect is weaker than the indirect effect, which is caused by the incumbent lowering its market-5 quantity because of its lower market-^ quantity. Just the opposite happens for the entrants. Assume that the incumbent unilaterally chooses a higher quantity in market ^. In response, the entrants' lower their quantity choice in market y4. Their lower quantity choice in market A lowers the positive network externalities that result from their serving both markets. However, their internal quality is lower than their external quality with the incumbent. Therefore, the positive network externality caused by the incumbent's higher quantity choice in market ^4 is greater than the effect of the entrants' lower quantity choice in market ^4. As a result, they have a higher quantity choice in market B.

PAGE 33

2.6. Conclusion 24 In this dissertation, I examine incentives for discrimination in network interconnection. I find that a large firm implements the same external quality for a given network interconnection as does its smaller rival. Furthermore, I find that smaller firms offer their own customers an internal network quality that is lower than the interconnection quality that the smaller firms implement with the larger firm. My results conflict with the findings of Cremer et al. (2000). Their model assumes that differences in firm size result from differences in numbers of attached customers, and find that these differences in attached customers cause firms with a higher number of attached customers to prefer a lower quality interconnection than a firm with a lower number of attached customers. They obtain this result because firms ignore the preferences of their own attached customers when choosing interconnection quality, but take into account unattached customers' preferences. The assumption that attached customers have entered into contracts that do not reflect the firm's quality choices, but that unattached customers' contracts do reflect the firm's quality choices, seems difficult to support. If it were optimal for firms to ignore preferences of captive customers, such as when the customers were in a monopoly market, then the monopoly in my model would have done so. That the monopoly in my model did not ignore its captive customers' preferences indicates that it is not optimal for firms to enter into contracts in which customers believe the firms will ignore their preferences when making quality choices. Therefore, these customers are not attached in the sense of Cremer et al. (2000). Because it is difficult to construct a scenario in which customers and firms choose contracts that result in attached customers in the sense of Cremer et al. (2000), I assume

PAGE 34

25 that asymmetry in firm size results from access to monopoly markets and cost asymmetries. I further assume that firms simultaneously choose qualities and outputs. This dissertation indicates that the EU may not have been justified in its conclusion that a larger Internet firm resulting from the MCIWorldCom merger or the WorldCom-Sprint merger would discriminate against European firms for access to the Internet backbone. It appears true that the larger firm would provide its own customers with a higher quality service than it would provide its competitors, but the larger firm's interconnection quality choice would be no different than the smaller firms' interconnection quality choice for connecting with the larger firm. Furthermore, the interconnection quality the larger firm would choose for connecting with the smaller firms would be higher than the quality the smaller firms would choose for connecting with each other.

PAGE 35

CHAPTER 3 NETWORK INTERCONNECTION AND TELECOMMUNICATIONS COMPETITION: THE CASE IN THE U.S. 3.1. Introduction Opening telecommunications markets to competition has sparked controversy over what role, if any, the government should play in facilitating entry, regulating prices, and reforming traditional methods of subsidizing services. Difficult issues in writing the US Telecommunications Act of 1996 (Act) included how Bell Operating Company (BOC) long distance restrictions would be lifted,' how and when local telephone markets^ would be opened to competition, and how state and federal regulators would share authority. A primary economic issue has been standards for regulating prices that incumbent local exchange carriers (incumbents) charge new entrants for exchanging calls ' The AT&T divestiture agreement of 1982 restricted the BOCs from providing long distance service, except in limited areas (called Local Access Transport Areas, or LATAs). ^ The prohibitions on local telephone competition restricted customers and nonlocal exchange carriers from using telecommunications facilities that were owned by someone other than the franchised local exchange carrier, to complete voice and data communications within a local exchange. Customers and nonlocal exchange carriers were allowed to place telephone facilities in local exchanges before the Act, but in most cases, they could use them only for nonlocal communications or for private communications, such as communications within a company. This use restriction made constructing local facilities less economical than it would have been without the restriction. Throughout this dissertation, I use the terms local competition and local telephone service competition to refer to the right to use nonincumbent local exchange company facilities to complete these formerly prohibited local communications. 26

PAGE 36

27 and for using portions of the incumbents' networks to provide services.^ Call traffic exchange between competing carriers is necessary if customers of one company are to be able to call customers of another company. Payment for exchanging calls is called reciprocal compensation in the US. "Reciprocal" means that both companies involved in an interconnection are obligated to make payments. "Symmetric" reciprocal compensation means the companies charge the same prices to each other. Reciprocal compensation prices are generally symmetric in the US, so I assume symmetry in my models. The Act provides three methods of market entry, illustrated in Figure 3-1. Some entrants use more than one method. Entrants can build their own facility-based network, lease portions of an incumbent's network, or buy an incumbent's services and resell them. Leasing portions of an incumbent's network is called purchasing unbundled network elements or UNEs. Figure 3-1 shows how an entrant would lease a local telephone line from the incumbent. The line would connect to the incumbent's building. It could then connect to the incumbent's switch or the entrant's switch, depending on how the entrant wishes to use the incumbent's facilities. Entrants that have their own switches must interconnect their switches with those of an incumbent and pay reciprocal compensation for terminating telephone calls on the incumbent's network. Figure 3-1 shows how lines, called trunks, would connect the incumbent and entrant central offices. Likewise, an Traditional voice telecommunications networks consist of lines and switches. Lines either connect customers to the network or connect switches in the network. Switches route calls between customers. Switches are of two types: local switches (also called central offices) that customers connect to and that switch local calls, and long distance switches (also called tandem or toll offices) that route long distance calls from one local switch to another.

PAGE 37

28 incumbent must pay reciprocal compensation for terminating calls on entrants' network. Regarding resale, Figure 3-1 shows an entrant's customer using a resold service. The facility arrangement is just as if it was an incumbent's customer. In a sense, reselling is little more than rebranding the incumbent's service. Figure 3-1. Intercormection Arrangements Economists disagree on how prices for UNEs and reciprocal compensation should be set. Supporting the idea that incumbents should be allowed to charge "high" prices to competitors, Baumol and Sidak (1994a, 1994b, 1995), Kahn and Taylor (1994), Hausman and Tardiff (1995), Larson and Parsons (1994), and Larson (1997) argue that incumbents' prices should be based on the Efficient Component Pricing Rule (ECPR). The ECPR sets the price of an upstream input equal to its incremental cost plus the retail profits that the

PAGE 38

29 incumbent loses to downstream competitors. The ECPR has two purposes: First, to ensure that regulated prices do not provide entrants with an artificial price umbrella that encourages inefficient entry. The second purpose is to avoid giving incumbents price signals that encourage them to hinder efficient competitors. Arguing for "low" prices, Mitchell et al. (1995), Albon (1994), Economides and White (1995), Tye and Lapuerta (1996), and Tye (1994) point out numerous flaws with the ECPR theory and argue for prices based only on incremental cost. Laffont and Tirole (1994, 1996) and Laffont, Key, and Tirole (1998a, 1998b) show that the ECPR can be efficient in some circumstances, but that incumbent prices below incremental cost could be efficient in other circumstances. Despite the voluminous economic debate on these issues, empirical research has been lacking. This is understandable given how recently the Act was passed and given that telecommunications liberalization in other countries is still recent. In this dissertation, I take an initial step in filling this void in the literature by examining how the Act and regulators' interpretation of the Act have affected the early development of local telephone service competition. I find that regulators' adoption of prices based on incremental cost have given incumbents an incentive to hinder competitive entry, as the ECPR theory would predict. However, these prices have not led incumbents to hinder entrants' in their efforts to gain market share. The data do not reveal whether this is the result of a lack of incentive for incumbents, a lack of ability, or a lack of opportunity. I further find that the opportunity for entrants to place facilities in incumbents' buildings (called collocation) is important to encouraging facilities-based entry and causes entrants to substitute facilities and unbundled network elements for resale-based entry. Higher

PAGE 39

symmetric reciprocal compensation prices discourage entry and cause entrants to select customers that receive more calls than they send. High wholesale discounts for resold services encourage entry, but do not encourage entrants to substitute resale-based entry for facilities-based supply. This study provides guidance for US policy makers as they review the progress made under the Act and consider further steps in its implementation. Congress has held several hearings on the implementation and effects of the Act. The Act requires the Federal Communications Commission's (FCC) to review its rules every two years to determine whether the rules are necessary. The Act also requires the FCC to forbear from applying any regulation or any provision of the Act if the FCC finds that the regulation or provision is unnecessary or inconsistent with the public interest. The Act also prohibits state public utility commissions from applying regulations or provisions of the Act that the FCC has decided to forbear from applying. In addition to providing information for US policy makers, this study suggests strategies that incumbents and entrants might choose in advocating policy positions before policy makers, and provides insights that other countries might use in establishing their own telecommunications policies. The World Trade Organization Agreement on Basic Telecommunications is prompting numerous countries to open telecommunications markets to competition. In this dissertation, I contribute to the literature on competition in telecommunications by providing an empirical analysis of how US regulatory policies for interconnection affect the development of local competition. Section 3-2 is a review of the literature. In Section 3-3, 1 provide historical background on telecommunications in the US, including an explanation of portions of the Act. In Section 3-4, 1 describe my

PAGE 40

31 models and their underlying theory. In Section 3-5, 1 discuss the estimation results. Lastly, I conclude with policy implications and recommendations for further research. 3.2. Literature Review Before presenting my results, I examine and summarize the empirical research on the development of competition in telecommunications. One of the earliest contributions in this area is Crandall (1991). He finds that the difference between total telecommunications equipment sales and telephone company equipment purchases increased from 1984 through 1988, indicating an increase in private network investment. Tomlinson (1995) shows that the development of competition from competitive access providers, who provide fiber optic networks for large customers primarily in metropolitan areas, prompts incumbents to build advanced fiber optic rings. Ros (1999) performs a cross-country analysis of how privatization and liberalization affect teledensity (telephone lines per 1000 population) and telecommunications investment. Without distinguishing between different forms of liberalization,'* he finds that competition increases teledensity and decreases investment in the sample countries with liberalization. Dekimpe et al. (1998) find that increased numbers of cellular competitors increases the penetration rate and the rate of diffusion. Blank et al. (1998) show that entry by long distance companies, such as AT&T, into BOC intraLATA long distance markets lowers BOC intraLATA long distance prices. The markets open to competition varied across countries. Countries in the sample liberalized various combinations of local telephone service, domestic long distance, and international long distance.

PAGE 41

32 In an examination of the effects of incentive regulation on incumbent investment, Greenstein et al. (1995) find that entry by competitive access providers has no measurable affect on incumbents' investment levels, but that removal of regulatory restrictions on entry encourages incumbents' investment in fiber optics. They also show that removal of restrictions on long distance company provision of intraLATA long distance and on resale of local services decreases incumbent fiber optic investment. Woroch (2000) finds that competitive access provider entry into markets for digital fiber optics in metropolitan areas prompts incumbents to make competing investments. Ros and McDermott (2000) find that repricing local telephone service to remove traditional subsidies from business to residential customers encourages entry by new competitors. In a study on the effects of incentive regulation, Ai and Sappington (1998) find that higher levels of competitive access provider investment in fiber optics is associated with higher incumbent investment in ISDN switches^ and lines and with longer delays by incumbents in resolving customer trouble reports. They also find that intraLATA long distance competition encourages incumbents to invest more in fiber optics and is also associated with long delays by incumbents in resolving customer trouble reports. In a cross-country comparison, Spiller and Cardilli (1997) use a case study approach to conclude that the absence of clear rules on interconnection and parity for long distance competitors in terms of the how customers can use their services, causes delays in entry ' ISDN stands for Integrated Services Digital Networic and was a technology associated with advanced telecommunications.

PAGE 42

33 and disadvantages new competitors. They also find that Hmiting rights for entrants to use incumbents' networks encourages investment by entrants. Using event analysis, MacAvoy (1997) finds that the California Public Utility Commission's decisions in 1995 to allow local competition caused a 15.73% decline in Pacific Telesis' stock price. Using behavioral tests, MacAvoy (1998) concludes that long distance companies' price-cost margins increased during the late 1980s and through most of the 1990s, even though market concentration declined. 3.3. Historical Background Today's new era of telecommunications competition has its roots in the introduction of competition for telephone equipment in the 1960s and in long distance in the 1970s and 1980s. AT&T's primary tool for protecting its monopoly against each new competitive threat was the company's reftisal to allow would-be competitors to connect to the BOCs' networks. Because AT&T owned the BOCs, AT&T had end-to-end control of the telephone network and was able to use this control to deny rivals access to the telecommunications market. This tactic led the US Department of Justice in the early 1970s to renew its earlier antitrust suit against AT&T. In 1982, AT&T and the Department of Justice settled the suit, with AT&T agreeing to divest its ownership of the BOCs. The divestiture was completed on January 1, 1984. The underlying theory of the divestiture was that local telephone service markets are natural monopolies while other market segments are at least potentially competitive. Therefore, the thinking went, dividing the industry into monopoly and competitive companies would facilitate the orderly development of competition. To ensure that the

PAGE 43

34 BOCs did not use their local monopolies to hinder competition in other markets, the divestiture agreement restricted the BOCs from manufacturing telephone equipment and from providing information services and interLATA long distance. In a separate antitrust settlement, the government imposed interLATA long distance restrictions on GTE, too. But the divestiture's underlying theory was fatally flawed. The boundaries between local telecommunications and all other telecommunications are artificial and arbitrary. The division between long distance and local telecommunications made long distance companies heavily dependent on incumbent local exchange companies, who were natural competitors to long distance companies. Long distance companies' payments to incumbent local exchange companies for originating and terminating long distance calls constituted nearly 50% of a long distance company's costs. Furthermore, almost 99% of long distance companies' long distance calls passed through the local exchange companies' networks (Jamison, 1995). In part to relieve the resulting strategic tension, long distance companies sometimes bypass portions of the BOCs' local telephone networks by placing long distance network facilities in high traffic areas and connecting directly to large customers. BOCs would find it in their interest to enter long distance markets. The interLATA restrictions prevent the BOCs from competing for large customers by providing seamless end-to-end services that cross LATA boundaries. Also, BOC provision of long distance would eliminate the double marginalization that occurs when long distance companies take BOC access prices as given when they (the long distance companies) make their profit maximizing decisions (Weisman 1995). Furthermore, BOCs need to enter the interLATA markets if they are to be able to offer the

PAGE 44

35 combinations of local and long distance products that are comparable to the combinations the long distance companies' provide when they enter the local exchange markets. Lastly, the BOCs' traditional monopoly position in local telephone service is being eroded by the convergence of telephone, computers, and media. The arbitrariness of the distinction between local telephone and all other communications finally spelled the end of the local telephone monopoly. In the early 1990s, some states began allowing competition for local telephone service. By 1996, the pressure for change was overwhelming and Congress passed the Act. Among other things, the Act removes legal restrictions on local telephone service competition for all geographic areas except those served by small rural telephone companies. In order to facilitate local competition, the Act requires all incumbent local exchange companies to provide entrants nondiscriminatory interconnection and UNEs at any technically feasible point within incumbents' networks at cost-based prices. Creating UNEs allows entrants to enter markets more quickly than they could if they had to build all of their own facilities. The Act further requires incumbents to offer for resale at wholesale rates any telecommunications service that they provide to retail customers. Wholesale prices must be based upon retail prices minus the portion attributable to marketing, billing, collection, and other costs avoided by the incumbent when it does not provide the retail service.

PAGE 45

36 This is essentially the ECPR discussed earlier^ (Larson, 1997). The essential trade-off in the Act is that the BOCs and GTE are permitted to offer interLATA long distance service in exchange for giving up their local monopolies. GTE was permitted to offer long distance immediately upon passage of the Act, as were the BOCs with respect to long distance outside their regions. With respect to long distance within their regions, the key market for BOCs because of their traditional customer base, the BOCs are to be permitted into the market once they have satisfied certain preconditions. Tomlinson (2000) provides details on these preconditions. The FCC began implementation of the local competition provisions of the Act by adopting rules in August 1996 that incumbents and entrants are to follow to be in compliance with the Act.^ As of 1999, the FCC had issued 231 orders or similar actions related to its implementation of the Act. The FCC determined that incumbents' prices should be based upon a measure of incremental cost, which the FCC created and calls Total Element Long Run Incremental Cost (TELRIC). TELRIC incorporates specific assumptions about the incumbents' technology and input prices, which cause TELRIC to understate incremental cost ' Consider a firm that produces and sells and input, called the upstream product, that is used to produce a final product, that is called the downstream product. This firm also produces the downstream product in competition with its customers for the upstream input. Under the ECPR, the difference between the firm's price for the upstream input and its price for the downstream product is simply the incremental cost of producing the downstream portion of the product. ' FCC 96-325, The First Report & Order In the Matter of Implementation of the Local Competition Provisions in the Telecommunications Act of 1996, CC Docket No. 96-98 and CC Docket No. 95-185, August 8, 1996. * It became clear during these proceedings that some items are more controversial than are others. Consensus quickly emerged on items such as white pages, dialing parity, and access to signaling and databases. On other issues, such as prices that incumbents would charge to entrants and collocation there is disagreement among stakeholders. (Harris and Kraft, 1997; Kennard, 1998)

PAGE 46

... 37 (Jamison, 1999; Weisman, 2000; Mandy, 2000). This decision, and the FCC's extensive unbundUng requirements, caused incumbents and state public utility commissions to object to the FCC's local competition rules. Incumbents objected to the FCC's decisions because the decisions gave entrants the right to make extensive use of incumbents' networks at prices below incremental cost. The state commissions objected to the FCC's rules because states believed that the Act leaves these decisions to the states. The incumbent local exchange companies and the state commissions appealed the FCC's rules to the United States Court of Appeals for the Eighth Circuit. On July 18, 1997, the court released its decision generally agreeing with the states and the incumbents and vacating, among other things, most of the FCC's pricing and unbundling rules. The FCC, MCI, and AT&T petitioned the United States Supreme Court to review the Eighth Circuit's decision. On January 25, 1999, the Supreme Court largely reversed the lower court, holding that the FCC has general jurisdiction to implement the Act's local competition provisions and upholding almost all of the FCC's unbundling requirements. On July 18, 2000, the Eighth Circuit Court of Appeals vacated the FCC's TELRIC methodology and remanded it back to the FCC for revision. Prior to the Supreme Court's ruling, states adopted an array of regulatory policies for implementing the Act. On relatively noncontroversial issues, such as white page listings and access to signaling and databases, states were reasonably uniform. On controversial issues, such as incumbents' prices, states differed from each other and from the FCC. Based on a review of state commission interconnection decisions located on the National Regulatory Research Institute's web site (NRRI, 1998), 16% of the states chose an incremental cost measure called total service long run incremental cost as their

PAGE 47

basis for pricing reciprocal compensation. Total service long run incremental cost generally gives higher estimates of incremental cost than does TELRIC (Jamison, 1999b). Thirteen percent of the state commissions chose TELRIC and the rest chose bill and keep.^ With bill and keep, incumbents and entrants do not pay each other for exchanging minutes. Some states that adopt bill and keep apply the policy only as long as the traffic exchange is relatively balanced. Similarly, 77% of the states chose total service long run incremental cost as their cost standard for UNEs and 19% chose TELRIC. One state chose an accounting cost allocation approach called flilly distributed cost as its method. The Act also mandates collocation, the process by which entrants locate their equipment in incumbents' buildings. There are two types of collocation, physical collocation and virtual collocation. With physical collocation, entrants physically place their equipment in incumbents' buildings where incumbents also have telephone equipment. With virtual collocation, incumbents place equipment in their buildings and dedicate the equipment to use by entrants. This equipment is owned and maintained by the incumbents. Collocation is valuable to entrants because it decreases their costs of interconnection and of accessing UNEs, relative to having to place their equipment some distance away from the incumbents' facilities. Entrants prefer physical collocation because it gives them greater operational control than virtual collocation over their equipment and services. ' The FCC's policies allow for bill and keep, but do not mandate it.

PAGE 48

39 3.4. The Models I analyze how regulatory policies affect entrants' entry decisions and the quantities they sell. 1 summarize the entry decisions by extending Bresnahan and Reiss's (1991) and Berry's (1992) ordered probit models for entry, which apply a zero-profit equilibrium for entrants. They show that, as additional firms enter a market, quantity produced increases, which causes per-unit profits to decrease. At the equilibrium number of firms, the addition of one more firm makes per-unit profits negative, which makes additional entry unprofitable. I use a standard regression model to summarize entrant system expansion and the resulting effects on entrant market share. Because of data restrictions, I consider each incumbent's traditional local exchange areas in a state to be a market. This causes distortions because for a given incumbent, an entrant may choose to enter some of the incumbent's local exchanges and not enter others, and may choose to supply only some areas of a local exchange. Because regulators generally require incumbents to average retail prices across exchanges and to charge lower retail prices in rural areas than in urban areas, and because per customer costs are generally lower in high density, urban areas than in rural areas, I expect entrants to serve urban areas first. Therefore, regions with higher than average proportions of urban areas should have more entry than areas that are below average. To control for this effect, I include in my models a measure of customer density. Higher customer density indicates markets with higher than average proportions of urban areas. I consider three types of models. The first describes entry. The second describes how entrants expand their market share by constructing their own facilities and by using

PAGE 49

40 UNEs. The third describes how entrants expand their market share by reselling incumbents' services. I first describe the market entry models. 3.4. 1. Entry Models More entry is expected in larger markets, in markets where regulatory policies are conducive to entry and in markets where incumbents have created fewer barriers to entry. Bresnahan and Reiss (1991) explain how market size and the effects of minimum efficient scale affect the number of competitors that can profitably enter a market. They show that in 'small' markets, only a monopolist can profitably serve the market. Then as market size increases, the monopolist's profits increase. At some level of market size, it becomes profitable for a new entrant to enter the market. This new entrant enters when its expected ex post profits are strictly posifive; i.e., Ei/rdn^^l)) > 0, where E is the expected value and ^c{n=l) is an entrant's profits if n = 1, where n is the number of entrants. Only one entrant enters the market if E(/rc{n=l)) > 0 and E{;rc(n=2)) < 0. At even larger market sizes, more entrants will find it profitable to enter the market. In general, entry occurs up to the point where £'(;rc 0 and E{;rc(n+l)) < 0. For a given market, price-cost margins and profits decrease as the number of firms increases, the quantity supplied increases, or both increase. Regulatory policies and incumbent responses to entry affect entry by affecting the profitability of entrant activity in a given market. Policies favorable to entrants and that provide incumbents with little or no incenfives to hinder entry increase the probability of entry, while unfavorable policies and aggressive incumbent responses have the opposite effect. In general, low UNE prices favor entrants because entrants can substitute UNEs for their own facilities when facilities are more costly. Readily available collocation

PAGE 50

41 opportunities decrease entrants' costs for using their own facilities and for using UNEs. Likewise, high resale discounts result in low wholesale prices, which make it more economical for entrants to resell incumbents' services, particularly when entrants' facility costs and UNE prices are relatively high. Reciprocal compensation prices are more complex. Entrants receive more reciprocal compensation payments than they pay (i.e., are net receivers) if they obtain customers who are net receivers of minutes of calling (which I simply refer to as minutes). An Internet Service Provider (ISP) is an example of a customer that receives more minutes than it sends. A residential customer who uses telephone lines to reach an ISP, and does so for several hours per month, is an example of a customer that makes more minutes than it receives. High reciprocal compensation price-cost margins encourage entry if entrants are better able than incumbents to attract customers who are net receivers of minutes. This effect of high reciprocal compensation price-cost margins decreases as additional entrants enter the market because the number of available netreceiver customers decreases. Low symmetric reciprocal compensation price-cost margins encourage entry when entrants' are more successful than incumbents in attracting customers who are net senders of minutes. My dependent variable measures the total number of entrants in a market, which includes those that use their own facilities, those that use UNEs, those that are pure resellers, and those that use some combination of the three entry methods. Because entry is new, it is unlikely that the data represent a long-run equilibrium. I adjust for this disequilibrium by including as an explanatory variable the amount of time that has elapsed since entrants were allowed to enter each market.

PAGE 51

42 I use an ordered probit to estimate how regulatory policies affect entry decisions. Ordered probit models are used when dependent variables are discrete and represent ordered outcomes. Underlying the model is a 'virtual' model in which there is an unobserved, continuous dependent variable y* whose conditional mean is assumed to be a linear function of the independent variables (Hausman et al., 1992). In my model, y* measures entrants' abilities and propensities to enter, and incumbents' abilities and propensities to limit entry. Although;;* is unobserved, it is related to my observed discrete random variable for entry. For example, no entry is observed if>'* < 0. Likewise, one entrant is observed ifO 0 and (« + 1) < O) = 0(^c («)) " ^i^c (« + 0) • The residual probability of observing n or more firms is

PAGE 52

•i . . Pr(;re(«)>0) = O(^c(«)). My entry models explain the number of entrants be examining parameters that affect entrants' expected profits. 3.4.2. Quantity Models Now consider the second and third types of models, those that describe how entrants expand their market share. These models follow closely the models for entry. Larger entrant market share is expected in markets where regulatory policies are conducive to entrants and in markets where incumbents have created fewer barriers to entrants serving the market demand. Let ^'represent a customer's type; i.e., the degree to which the customer is a net receiver or a net sender of minutes. '° Assuming cost minimizing behavior, an entrant / provides positive output for customers of type /$'when its expected ex post profits from increasing supply beyond zero are strictly positive; i.e.. dTT, V > 0 , where , represents the quantity that entrant / provides to ??,,=« ) customers of type 0. Such an entrant would choose its methods of supply based on its expected ex post profits. For example, an entrant in markets with lower prices for UNEs and higher prices for wholesale services, relative to other markets, should use higher proportions of UNEs to provide services. Greater availability of collocation increases Customers can be viewed as being distributed on a two-dimensional plane that represents their calling volumes. One axis represents the total number of minutes for calls a customer receives and the other axis represents the total number of minutes for calls a customer sends. 0 represents a customer location on this two-dimensional plane.

PAGE 53

44 entrants' market penetration and causes entrants to use less resale when collocation lowers entrants' costs of UNEs or facilities. Also, if the regulated prices for UNEs or wholesale services provide the incumbent with a lower price-cost margin than the incumbent retail services that are displaced by entrants, incumbents could be expected to take steps to hinder entrants' expansion (Ordover, Sykes, and Willig, 1985). As in the entry models, the effects of reciprocal compensation prices on entrant market share are complex. High price-cost margins for reciprocal compensation encourage entrants to attract customers who are net receivers. Entrants can do this if, for example, their marketing costs for obtaining these customers are low relative to the incumbents' costs of retaining these customers. On the other hand, low price-cost margins for reciprocal compensation prices encourage entrants to increase their use of resale and to increase their supply for customers who are net senders. Entrants can do this if their marketing costs for obtaining these types of customers are low relative to the incumbents' costs of retaining these customers. 1 use ordinary least squares regression to examine entrants' market share. I consider two measures of entrant output. The first measure is the number of entrant interconnections to the incumbent. These interconnections, called trunks, are necessary for exchanging minutes between entrant and incumbent customers. Figure 3-1 illustrates how these trunks interconnect central offices. Higher numbers of trunks indicate higher amounts of entrant output for customers served by entrant facilities and, to a certain extent, UNEs. Entrants need trunks for customers served by UNEs only when the entrant is using its own switch; for example, when the entrant has a switch and uses an incumbent's local line UNEs to connect customers to the entrant switch. Trunks

PAGE 54

45 understate entrant supply if the entrants are large because, as an entrant grows, the number of minutes that stay on the entrant's network generally increase. This happens because the probability of a call originating on an entrant's network also terminating on that entrant's network generally increases as the entrant attracts more customers. Also, trunks overstate entrant supply if there are many small entrants, or entrants whose customers are dispersed geographically. This happens because these entrants' low traffic volumes keep them from making efficient use of their local interconnection trunks. Figure 3-1 also illustrates how entrants use a local loop UNE. The local loop UNE connects the entrant customer to the incumbent's building that houses the incumbent's central office. At the central office, the entrant can either connect the customer to its own central office or use switching UNEs to route customer traffic. My second measure of entrant supply is the number of business resold lines; i.e., the number of units of local telephone service that entrants resell to businesses. Figure 31 shows how entrant customers of resold services use incumbent facilities in the same way that the incumbents' own customers do. 3.4.3. Estimation Models and Data Table 3-1 describes the data I use for my dependent variables, which are from the United States Telephone Association's (USTA) report to Congressman Thomas Bliley on December 9, 1998 (USTA, 1998)." This sample omits Sprint's local exchange " Observations include Ameritech (all states). Bell Atlantic (all states), BellSouth (all states), GTE (California, Florida, Hawaii, Illinois, Indiana, Kentucky, Michigan, North Carolina, Ohio, Oregon, Texas, Virginia, Washington, and Wisconsin), SBC (Arkansas, California, Oklahoma, and Texas), and US West (all states) for 1998.

PAGE 55

operations because Sprint was not part of USTA at the time the data were reported. COMPT is the number of entrants that are in each market in 1998. TRUNKS is the number of interconnection trunks between the entrants and the incumbent. RESOLDB is the number of the incumbent's business telephone lines that were being resold in 1998. For the entry models, 1 create an ordinal variable that represents levels of entry. The Appendix explains the development of this variable. Table 3-1. Descriptions of Dependent Variables Variable Name in Model Mean Minimum Maximum Standard Deviation Observations Number of Entrants COMPT 26.15 0 164 30.10 59 Number of Interconnection Trunks TRUNKS 30,018.81 0 289,299 50,611.44 59 Number of Business Resold Lines RESOLDB 27,009.02 0 183,594 42,696.88 59 For models of entrant output, I estimate entrant market share by dividing the dependent variables and certain explanatory variables by the number of incumbent telephone lines in the market. This indexes these variables according to market size. I explain below which explanatory variables are divided by the number of lines. TRNKPLN represents TRUNKS per line. RSLDBPLN represents RESOLDB per line. RSLDRPLN represents RESOLDR per line. Table 3-2 describes the data for the explanatory variables. I use the price per month for a 2-wire local line in urban areas, UNEPRC, as the representative UNE price. '^A two-wire local line has two wires twisted together that are used to connect the customer to the telephone company central office.

PAGE 56

47 Incumbents have many UNEs. I choose local line prices to represent UNE prices because the sunk nature of line investment and the need for right-of-way and conduit space make lines the most difficult facilities for entrants to construct themselves. Two-wire lines are Table 3-2. Descriptions of Explanatory Variables Variable Name in Model Mean Minimum Maximum Standard Deviation Observations UNE Prices UNEPRC $16.82 $3.72 $32.00 $6.16 59 Prices for Reciprocal Compensation RCP $0.0057 $0 $0.0283 $0.0067 59 Business Resale Discount BUSRSL 0.1794 0.0800 0.2601 0.0374 59 Residential Resale Discount RESRSL 0.1788 0.0700 0.2500 0.0405 59 Incumbent Local Service Revenues (000) REVL $749,773 $71,863 $4,731,829 $902,587 59 Incumbent Line Cost Level USFCOST $265.72 $65.68 $408.81 $57.71 59 Collocation Availability for Voice Lines COLCV98 23.66% 0% 72.62% 17.54% 59 Quarters entrants in the market before 1996 NUMQRT 10.75 0 18 15.44 59 1997 Incumbent Service Quality Complaints SCMPAM97 274.29 6 2,637 466.18 59 1998 Incumbent Service Quality Complaints SCMPAM98 339.08 8 2,473 506.33 59 Incumbent Total Revenue 1998 (000) REVT $1,511,887 $170,463 $8,460,236 $1,721,406 59 Incumbent Total Assets (000) TPIS $4,347,098 $493,351 $27,585,598 $4,929,606 59 Incumbent Total Billable Lines LINES 2,286,606 235,862 16,071,707 2,785,059 59 Number of Incumbent Central Offices COS98 230.1 29 752 174.2 59 Local Calling Minutes for Incumbent (000) MOU98 7,056,890 820,771 36,441,427 7,740,706 59 Incumbent Central Office Investment (000) COETPIS $760,976 S59,52l $4,581,901 $864,958 59 Presence of Universal Service Fund Reform USFFUND 0.2881 0 0.4568 59 GTE GTE 0.2373 0 0.4291 59 Bell Atlantic BA 0.2203 0 0.4180 59 BellSouth BS 0.1525 0 0.3626 59 SBC SBC 0. 0678 0 0.2536 59 US West USW 0.2373 0 0.4291 59 Ameritech AM 0.0847 0 0.2809 59

PAGE 57

48 the most common technology used for local telephone service. I choose urban prices because I expect most entrants to serve urban areas. Data on these prices are from state commissions (NRRI, 1998; Alabama, 1998), X-Change (1998-1999), and interviews with incumbents. In cases where there are discrepancies, I employ the data provided in state commissions' decisions (NRRI, 1998) when available, and other public data when the commission data is not available. RCP is the price per minute for symmetric reciprocal compensation. Data on these prices are from the same sources as UNEPRC. Some states have multipart prices for reciprocal compensation. There may be separate prices for terminating minutes at a central office or a tandem office and for different times of the day. To express these prices as a single price, I follow the convention of assuming 6.25% of the minutes terminate at local central offices and the remainder terminate in a tandem. Entrants often connect to the incumbent's network at a tandem office because it gives them access to numerous central offices. BUSRSL and RESRSL represent the discounts that regulators give entrants for buying wholesale services for business services and residential services respectively. In states where discounts vary by service, I use the smallest percentage discount. In states where residential service discounts are different from business service discounts, I use the residential discount for the models explaining entry and output using UNEs and entrantowned facilities. BUSRSL and RESRSL are highly correlated (R^ = 0.88). In tests of various models, business and residential discounts have similar results. To avoid multicollinearity, I use only the residential discount for these models. Sources for

PAGE 58

49 percent discounts are the same as the sources for reciprocal compensation prices and UNE prices. I use the ratio of total revenues for basic local telephone services (REVL) and total billable lines (LINES) to indicate incumbents' price levels for local exchange services. I call this ratio LCLPLN. Firms charge many prices, so using a single price is inappropriate. I use incumbents' Universal Service Fund costs per line, USFCOST, to represent incumbents' costs for telephone lines. Universal Service Fund costs are the local line costs that incumbents report to the FCC for estimating subsidies that are given to small incumbents and rural incumbents for 1998. USF costs are based on incumbents' regulatory accounting records and provide an average cost for all of an incumbent's operations in a state. I use the percent of incumbent voice telephone lines accessible by entrants through collocation in 1998, COLCV98, to represent the ease of collocation. Data are from Tables 3.6 and 3.7 of the FCC's 1998 Local Competition report. (FCC, 1998) There is a risk of endogeneity because higher entrant interest in a market should increase entrant demand for collocation. Higher demand for collocation should increase the incidence of collocation, which could cause a higher percentage incumbent lines to be in central offices with collocation. However, differences between markets should also reflect the ease of obtaining collocation. Regulatory costs are an imperfect measure of economic costs. The accounting processes (Gabel, 1967) and distortive efficiency incentives caused by methods of regulation (Sappington and Weisman, 1 996) cause these regulatory costs to deviate from economic costs.

PAGE 59

50 I use the number of quarters in a state from the time the first entrant was given telephone numbers until the Act took affect, NUMQRT, to represent the amount of time that entrants have been operating in a state. Data are from Table 4.8 of the FCC's December 1998 Local Competition report (FCC, 1998). I use total numbers of customer complaints to state and federal regulators for 1997 and 1998 in metropolitan statistical areas, SCMPAM97 and SCMPAM98, as my measures of incumbent service quality. Low incumbent service quality should encourage customers to buy from facilities-based entrants. Data are from the FCC's ARMIS reports.'" I use SCMPAM97 to examine entry because it represents ex ante entry information. I use SCMPAM98 per line (which I call SCMP98PL) to examine entrant output. I choose 1998 service complaints for my supply models because this represents the quality that customers and entrants experienced at the time supply and purchasing decisions were put into effect. I use REVT, the incumbent's total operating revenues for 1998 for the market, to indicate market size. For output models, REVTPL represents incumbent total operating revenues per line. REVT data are from the FCC ARMIS reports. Also for supply models, I use several measures of incumbent costs and quantities supplied to analyze incumbent responses to price-cost margins on inputs sold to competitors. TPIS represents the incumbent's total plant in service for 1998, a measure of the incumbent's total investment in facilities. I express TPIS per line as TPISPL. My measure of the number of incumbent lines is LINES, the number of billable telecommunications lines in ''' All ARMIS data are from http://fcc.gov and were downloaded between March 1999 and August 2000.

PAGE 60

51 1998. COS98 is the number of incumbent central offices in 1998. MOU98 is the number of incumbent local telephone minutes in 1998. COETPIS is the amount of incumbent investment in central office switches in 1998. TPIS, LINES, COS98, MOU98, and COETPIS are from FCC ARMIS reports. COS98PL, MOUPL, and COTPISPL are the per-line expressions of COS98, MOU98, and COETPIS respectively. For the entry models, I include two price-cost ratios, PRCSTUNE and PRCSTRCP, as explanatory variables. PRCSTUNE is the ratio of the incumbent's UNE price-cost ratio and the incumbent's average retail price-cost ratio. The UNE price-cost ratio is the ratio of UNEPRC to USFCOST. The incumbent's average retail price-cost ratio is the ratio of the incumbent's 1998 total operating revenues and total plant in service. PRCSTRCP is the ratio of the incumbent's RCP price-cost ratio and the incumbent's average retail price cost ratio. The RCP price-cost ratio is the ratio of RCP and the incumbent's total investment in central office switches in 1998 divided by the total number of local exchange minutes in 1998. These ratios reflect the relationships between the price-cost ratios for inputs incumbents sell to entrants and the price-cost ratios for the incumbent's retail services that the entrants displace. Low values of PRCSTUNE and PRCSTRCP indicate that UNE and RCP price-cost ratios are low relative to incumbents' retail price-cost ratios. Ordover et al. (1985) explain that incumbents have an incentive to hinder competitors' success if competition lowers incumbents' profits.'^ Examples of incumbents' efforts to hinder entrants might be providing poor service quality to entrants, delaying collocation, and delaying BOCs have a countervailing incentive to cooperate with entrants so that the interLATA restrictions are lifted sooner.

PAGE 61

52 interconnection negotiations. Reciprocal compensation provides a special case for this incentive. If PRCSTRCP is low, incumbents would like to be net payers of reciprocal compensation. In the next section, I examine the coefficients of these ratios to test whether incumbents or entrants have the stronger propensities and abilities to respond to the profitability of using UNEs and paying or receiving reciprocal compensation. Including the variables PRCSTUNE and PRCSTRCP creates multicoUinearity because the variables are constructed from other variables. Therefore, I exclude these variables from the entrant output models and test hypotheses of nonlinear combinations of UNE prices, RCP prices, and incumbent revenues, costs, and quantities supplied. That is to say, I test incumbent responses to UNE price-cost margins by testing the significance of the combination UNEPRC/REVTPL, and I test incumbent responses to RCP price-cost margins by testing the significance of the combination RCP/(REVTPL/MOUPLN). Following Ros and McDermott (2000), I examine the effects of telephone subsidy reform. The Act requires regulators to develop subsidies that are competitively neutral. Traditionally, telephone subsidies, called universal service subsidies, were embedded in telephone company prices and only incumbents' prices were affected by subsidies (Jamison, 1995). Following the passage of the Act, regulators began taking steps to I use UNEPRC/REVTPL to test UNE price-cost margins because the actual ratio of interest, (UNEPRC/USF)/(REVTPL/USF) solves to UNEPRC/REVTPL. Similarly, the reciprocal compensation ratio of interest, (RCP/(COTPISPL/MOUPL))/(REVTPL/COTPISPL) solves to RCP/(REVTPL/MOUPLN).

PAGE 62

53 develop mechanisms for collecting funds for subsidies from all telecommunications service providers and making the subsidies available to all qualified service providers. According to a National Regulatory Research Institute survey (Rosenberg and Wilhelm, 1998), fourteen states had revised or were revising their subsidy policies in 1998. USFUND is a dummy variable that indicates whether the market is in one of these fourteen states. Lastly, I include dummy variables to identify incumbents. I have a dummy variable for each incumbent, but omit the Ameritech dummy from models to avoid multicoUinearity. Incumbent dummy variables may reveal differences in how incumbents processed entrant requests interconnection from 1996 through 1998. Soon after the passage of the Act, incumbents differed in how they processed these entrant requests. One incumbent required entrants to fax their requests for telephone numbers or local lines. Another required entrants to call with their requests, but assigned only one employee to the task of taking entrant orders. Other incumbents worked on electronic methods of taking entrant orders. These differences, as well as differences in regulatory scrutiny and enforcement, could cause incumbent dummy variables to be significant. To examine potential multicoUinearity, I regress all explanatory variables on each other. I also examine the linear correlation of each pair of explanatory variables. MulticoUinearity problems occur between the reciprocal compensation variables RCP and PRCSTRCP (R^ = 0.97), between the service complaint variables SCMPAM97 and SCMPAM98 (R^ = 0.92), between the resale discounts for business and residential services (BUSRSL and RESRSL, R^ = 0.88), among variables that indicate market size (for example, REVT and TPIS), and between PRCSTUNE and the variables that are

PAGE 63

54 included in it. To avoid multicollinearity in the entry models, I include no more than one variable from each of the collinear groups, with the exception of coUinear groups involving PRCSTUNE and PRCSTRCP. I need these variables in some entry analyses to perform likelihood ratio tests of entrant and incumbent incentives. Regarding multicollinearity in the entrant output models, dividing market size indicators such as REVT by LINES resolves much of the multicollinearity. Otherwise, I include in each model no more than one variable from each collinear group. 3.5. Model Results In this section I examine the results of my models. I examine the entry models first. I then investigate the models for entrant facilities and UNEs. Lastly, I examine models for resale. Tests of log linear models did not improve the overall fit, so I report only the linear results. Table 3-3 provides the coefficients and t-statistics for the entry models. One asterisk (*) indicates significance at the 0.10 level. Two asterisks (**) indicate significance at the 0.05 level. Three asterisks (***) indicate significance at the 0.01 level. Table 3-4 shows the marginal effects of explanatory variables. The first three rows show how markets are grouped into nine categories. Rows 2 and 3 show the range of entrants and the number of categories. The remaining rows show how a marginal change in each explanatory variable affects the probability of a market being in one of the nine categories. I estimate three models. Model 1 examines how UNE prices, reciprocal compensation prices, resale discounts, customer density, local service price-cost margins,

PAGE 64

service quality, market size, and UNE price-cost margins affect entry. The coefficients for UNE and RCP prices are negative, the coefficient for resale discounts is positive, and all are statistically significant. These results indicate that higher UNE, reciprocal Table 3-3. Regression Results for Entry COMPT) Explanatory Variable Model 1 Model 2 Model 3 **-0.1523 **-0.1900 **-0.1964 UNEPRC (-2.000) (-2.322) (-2.307) *-56.038 -70.7291 -51.4000 RCP (-1.879) (-0.733) (-0.501) **10.0708 **11.1879 **10.7780 RESRSL (2.230) (2.404) (2.288) ***0.0001 *0.0001 0.0001 LNSPCO (2.794) (1.934) (1.409) ***-1.9855 ***-2.0848 **-1.9342 LCLPUSF (-2.939) (-2.940) (-2.509) 0.0088 0.0126 COLCV98 (0.653) (0.901) -0.0622 -0.0643 NUMQRT (-1.234) (-1.108) *-0.0010 -0.0010 -0.0010 SCMPAM97 (-1.697) (-1.615) (-1.509) ***1.77e-06 ***1.92e-06 ***2.00e-06 REVT (5.038) (5.108) (5.039) ***14.4117 *** 17.435 ** 16.7705 PRCSTUNE (2.287) (2.589) (2.360) PRCSTRCP 0.3575 (0.099) -0.0792 (-0.021) USFUND 0.2729 (0.668) 0.2489 (0.605) GTE 0.6403 (0.730) **1.1397 0.9940 1.447 BA (2.065) (1.665) (1.867) ***2.2862 ***2.255 ***2.845 BS (3.761) (3.477) (3.121) ***5.0165 ***4.8481 ***5.5051 SBC (4.769) (4.554) (4.183) usw 0.8116 (0.939) 113.70 116.07 116.97 d.f. = 1 1 d.f = 15 d.f = 17 Log Likelihood -66.2809 -65.0953 -64.6450 compensation, and wholesale prices discourage entry. The marginal effects in Table 3-4, which are based on Model 1, show that a marginal increase in UNE prices or reciprocal

PAGE 65

56 compensation prices, or a marginal decrease in wholesale discounts, causes the probability of a market having more than 25 entrants to decrease, the amount of which varies by level of entry. No particular method of entry appears to dominate the others in importance. Table 3-4. Marginal Effects for Entry (COMPT) Markets Category Number 0 1 2 3 4 5 6 7 8 Entrants in Market 0-2 3-4 5-7 814 15-25 26-38 39-44 45-60 >60 Number Markets 8 8 5 4 12 11 3 4 4 Marginal Effects UNEPRC 0.00013 0.00051 0.00687 0.01454 0.01889 -0.00169 -0.03841 -0.00075 -8.20e-05 RCP 0.04838 0.18599 2.52867 5.34936 6.94956 -0.62294 -14.13170 -0.27727 -0.03001 RESRSL -0.00870 -0.03342 -0,45444 -0.96135 -1.24893 0.11195 2.53967 0.04983 0.00539 LNSPCO -8.63e-08 -3.30e-07 -4.50e-06 -9.50e-06 -1.20e-05 l.lle-06 2.52e-05 4.95e-07 5.36e-08 LCLPUSF 0.00171 0.00659 0.08959 0.18954 0.24623 -0.02207 -0.50071 -0.00982 -0.00106 SCMPAM97 8.63e-07 3.32e-06 4.51e-05 9.55e-05 0.00012 -l.lOe-05 -0.00025 -4.90e-06 -5.40e-07 REVT -1.53e-09 -5.90e-09 -8.00e-08 -1.70e-07 -2.20e-07 1.97e-08 4.46e-07 8.76e-09 9.48e-10 PRCSTUNE -0.01244 -0.04783 -0.65032 -1.37573 -1.78727 0.16021 3.63436 0.07131 0.00772 BA -0.00048 -0.00208 -0.05143 -0.10880 -0.14134 0.01267 0.28741 0.00564 0.00061 BS -7.13e-06 -5.90e-05 -0.10316 -0.21824 -0.28352 0.02541 0.57654 0.01131 0.00122 SBC -6.67e-13 -2.30e-ll -0.22637 -0.47887 -0.62212 0.05577 1.26507 0.02482 0.00269 The coefficient for the variable for customer density, LNSPCO, is positive and highly significant, indicating that more entry occurs in densely populated markets. The coefficient for LCLPUSF is negative and highly significant, indicating that higher pricecost margins for local telephone services decrease the number of entrants. This is contrary to the conventional wisdom that economic pricing for local telecommunications services is important for local network competition and indicates that such pricing may actually hinder entry by encouraging incumbents to hinder entry to protect profits. The

PAGE 66

57 coefficient for the variable indicating incumbent service quality in the previous year (SCMPAM97) is negative, but only significant at the 0. 10 level. This indicates that poor incumbent service quality hinders entry. This may be because of the importance of resale to entry. If an incumbent's retail services have poor quality, then it should be true that the wholesale versions of these same services will have just as poor, or poorer, service quality. The UNE-retail price-cost margin ratio, PRCSTUNE, has a positive and highly significant coefficient. This indicates that more entry occurs when incumbents' UNE price-cost margins are high relative to their retail price-cost margins. This is consistent with the theory of Ordover et al. (1985); i.e., incumbents are more likely to hinder entry if the inputs sold to entrants have low price-cost margins relative to the retail services that the entrants replace. The remaining variables in Model 1, those being for market size (REVT) and selected incumbents (Bell Atlantic, BellSouth, and SBC), are all highly significant. The importance of market size confirms that there are some economies of scale for entrants. The coefficients for the incumbents are positive, possibly indicating that these incumbents have responded less aggressively to entry than have other incumbents. Consistent with this result. Bell Atlantic and SBC are the only BOCs at the time of this writing to receive permission to enter interLATA long distance markets. The marginal effects in Table 3-4 indicate that SBC has a more positive effect on entry in larger markets than do Bell Atlantic and BellSouth, reflecting the large number of entrants in SBC's Texas and California markets. .

PAGE 67

Model 2 in Table 3-3 provides the results of adding the other regulatory policy variables to Model 1 , namely collocation (COLCV98), time (NUMQRT), price-cost margins for reciprocal compensation (PRSTRCP), and universal service reform (USFUND). Using a likelihood-ratio test, I fail to reject at the 0.10 level the joint hypothesis that the coefficients for these variables are all zero (X (4) = 2.37). Greater access to customers through collocation does not appear to prompt entry. Time does not appear to have been important, perhaps indicating that the passage of the Act was the watershed event triggering entry. Lower price-cost margins for reciprocal compensation have not resulted in incumbents hindering entry. This is probably because, as the market share models indicate, incumbents are sometimes net payers of reciprocal compensation. Lastly, I do not find that subsidy reform has encouraged entry. Model 3 in Table 3-3 shows entry model results after adding other variables for factors that shift demand or supply. Using a likelihood-ratio test on the hypothesis that the coefficients for all of these added variables and the added policy variables from Model 2 are zero, I am unable to reject the hypothesis at the 0. 10 level (x^(6) = 3.27). I conclude that Model 1 is the most appropriate model for examining how regulatory policies affect entry decisions. Tables 3-5 and 3-6 provide the results for the entrant market share models. Model 4 in Table 3-5 shows the results of including all of the explanatory variables in a model for entrants using UNEs and entrant-owned facilities. The primary purpose of this model is to form the basis for testing the signs and significance of nonlinear combinations of explanatory variables. Using an F-test, I fail to reject the hypothesis at the 0.10 level that coefficients for price-cost margins for UNEs, reciprocal compensation, and local

PAGE 68

59 telephone services, resale discounts, subsidy reform, and for the incumbent dummy variables for GTE, BellSouth, and US West are equal to zero (F(8, 40) = 1.19). This indicates that incumbents either do not hinder this form of entrant output, or are unsuccessful in doing so. This also indicates that neither higher margins on local telecommunications services nor subsidy reform prompt incumbents to increase their supply using UNEs or facilities. Indeed, the positive and significant coefficient for reciprocal compensation prices indicates that entrants are increasing UNE and facilitybased supply primarily to customers such as ISPs, which are net receivers of minutes. Model 5 in Table 3-5 provides results for the TRNKPLN after dropping variables that prove to be insignificant; namely, the resale discount, universal service reform, and the dummy variables for GTE, BellSouth, and US West. These results are comparable to the results of Model 4. Higher UNE prices limit entrant output, as well as entry. Higher local telecommunications services prices encourage customers to buy more from entrants, even though the incumbent's price-cost margin for these services is unimportant. The coefficient for ease of collocation (COLCV98) is positive and highly significant, indicating that collocation is important for entrants that want to build their own facilities that want to use UNEs. The coefficient for NUMQRT is positive and highly significant, showing that that building facilities and assembling UNEs takes time, which gives entrants a cost disadvantage relative to incumbents. Incumbent service quality has a positive and significant coefficient in Model 5. This implies that entrants choose to use their own facilities and UNEs, and that customers choose entrant services, when incumbent service quality is poor. The positive and highly significant coefficient for number of minutes per line is consistent with the conclusion that entrants use their own

PAGE 69

facilities and UNEs to target customers based on calling patterns. The negative and highly significant coefficient for Bell Atlantic may indicate that this incumbent has the most aggressive response to entrants targeting ISPs as customers. Table 3-5. Regression Results for Output Using UNEs allU -TaClllliva ^ 1 Ivii l\x Iji^^ Explanatory Variable Model 4 Model 5 **-0.0004 **-0.0003 UNEPRC (-2.254) (-2.026) *0.2213 0.1900 RCP (1.801) (1.724) 0.0119 RESRSL (0.680) 0.00004 0.00003 LCLPRLN (1.927) (2.244) ***0.0002 0.0002 COLCV98 (3.536) (3.546) **0.0005 0.0005 NUMQRT (2.535) (3.020) 6.9692 9.3746 SCMP98PL (1.534) (2.360) 0.0243 0.0183 REVTPL (1.248) (1.199) -0.0077 -0.0054 TPISPL (-1.605) (-1.375) 0.0022 0.0034 MOUPLN (1.213) (2.119) 0.0192 0.0250 COTPISPL (1.085) (1.872) 0.0014 USFUND (0.797) 0.0047 GTE (0.961) -0.0015 -0.0053 BA (-0.450) (-2.634) 0.0043 BS (1.146) **0.0077 0.0053 SBC (2.190) (1.986) 0.0063 usw (1.540) ***-0.0316 -0.0301 Constant (-3.559) (-3.665) 4.58 6.39 F d.f. = 17, 40 d.f. = 12, 45 R^ 0.6608 0.6302

PAGE 70

61 Table 3-6 shows the resuhs for entrant market share using resold business services. Model 6 shows the results using all explanatory variables. This model forms the basis for a joint hypothesis test of the significance of price-cost margins of UNEs and local telecommunications services, and of incumbent investment, minutes, central office investment, reciprocal compensation prices, and dummy variables for GTE, BellSouth, Bell Atlantic, and SBC. Using an F-statistic, I fail to reject at the 0.10 level the null hypothesis that the coefficients for all of these variables and combinations of variables are zero (F(10, 40) = 1.72). Consequently, I show the results for Model 7, which omits the insignificant variables. The variable for reciprocal compensation prices is in Model 7 because I fail to reject the joint hypothesis when the price-cost margin for reciprocal compensation is included. Comparing the results of Models 6 and 7, it is apparent that higher UNE prices decrease resold business lines. Comparing this result with the coefficients for collocation and subsidy reform (which are negative and significant), it appears that entrants primarily use resale of business lines as part of an entry strategy whose main intent is to use UNEs and facilities, and that availability of collocation and universal service subsidies causes entrants to move away from reselling business lines more quickly than they otherwise would. The business resale discount itself is insignificant, indicating that entrant market share using a pure resell strategy is either minor or effectively nonexistent. Consistent with Models 4 and 5, more time for entrants to expand their systems increases the amount of business line resale. Consistent with the discussion of Model 1, poor incumbent service quality decreases entrant use of resale, for reasons stated above. The positive and highly significant coefficient for US West's dummy variable indicates that this

PAGE 71

incumbent may be more cooperative with resellers than with facilities and UNE-based entrants, relative to other incumbents. Table 3-6. Regression Results for Resold Business Lines (RSLDBPLN) Explanatory Variable Model 6 Model 7 **-0.0010 ***-0.0011 UNEPRC (-2.204) (-2.818) 0.4136 0.4809 RCP (1.173) (1.498) 0.0319 0.0275 BUSRSL (0.591) (0.543) 0.00002 *0.00008 LCLPRLN (0.382) (1.707) ***-0.0003 **-0.0003 COLCV98 (-2.353) (-2.115) **0.0013 *0.0009 NUMQRT (2.206) (1.736) -21.0764 *-2 1.9800 SCMP98PL (-1.612) (-1.723) -0.0774 *-0.0640 REVTPL (-1.410) (-1.751) 0.0067 TPISPL (0.493) 0.0011 -0.0022 MOUPLN (0.213) (-0.545) 0.0598 0.0079 COTPISPL (1.181) (0.217) **-0.0110 *-0.0084 USFUND (-2.175) (-1.988) -0.0220 GTE (-1.578) -0.0140 BA (-1.462) -0.0078 BS (-0.736) -0.0015 SBC (-0.151) 0.0169 ***0.0312 USW (1.442) (4.752) 0.0311 0.0387 Constant (1.231) (1.623) 2.88 3.71 F d.f. = 17,40 d.f. = 12,45 R^ 0.5500 0.4974

PAGE 72

63 3.6. Conclusion This dissertation indicates that entrants use a mixture of entry and supply strategies, but that most entrant output is based on using UNEs and entrant-owned facilities. Reselling incumbent services appears to be important for entry, but not for growth in entrant market share. Prices for UNEs and reciprocal compensation are important determinants of entry, marketing, and modes of supply. Higher UNE and reciprocal compensation prices discourage entry, but higher reciprocal compensation prices also cause entrants to market to customers who are net receivers of minutes. Low price-cost margins for UNEs encourage incumbents to hinder entry, but incumbents have little effect on entrant market share. Also, the ease of using UNEs and placing facilities appears to be more important than relative prices for UNEs and wholesale in influencing entrants' choices of methods of supply. My findings indicate that, if regulators want to increase market penetration for facilities-based entrants, they should focus on unbundling networks, prices for UNEs, and collocation. Resale discounts for services, while important for entry, have little effect on entrant market share. Additional work is also needed on this topic. I was unable to test whether the interLATA restriction, which is to serve as a carrot for BOCs to open their markets to competition, is having its intended effect. Also, I was unable to test whether the policies that encourage entrant supply also improve social welfare. A more complete model of how policies affect industry costs and prices is needed to determine welfare effects. Lastly, international comparisons are lacking, as are more in depth analyses of entrant strategies for data and voice services.

PAGE 73

CHAPTER 4 NETWORK EXTERNALITIES, MERGERS, AND INDUSTRY CONCENTRATION 4.1. Introduction The welfare effects of mergers and market concentration have been analyzed extensively. (See, for example, Salant et al. (1983), Farrell and Shapiro (1990a, 1990b), and Gaudet and Salant (1991).) However, this literature does not examine how network externalities affect incentives to merge and the welfare effects of mergers and market concentration. These issues have taken center stage in recent merger and antitrust cases in network industries. Concerned with market dominance in the Internet backbone, the US and European Union (EU) regulators required MCI and WorldCom to divest Internet assets as a precondition for their merger and later blocked the merger of Sprint and WorldCom. In Silicon Graphics, Inc. 's purchase of software makers Alias and Wavefront, the US was concerned that the horizontal merger of the two software makers would result in higher prices despite the depressing effect that higher prices would have on sales of Silicon Graphics, Inc.'s computer workstations. The US placed restrictions on Silicon Graphics, Inc. to ensure that its products were compatible with competitors' products. (Shapiro, 2000) Similarly, the EU required Vodafone to divest assets in the UK before approving Vodafone 's purchase of Mannesman. A US District Court ordered the breakup of Microsoft based in part on the theory that Microsoft attempted to leverage 64

PAGE 74

65 its dominance in the market for PC operating systems to monopolize the Internet browser market. I extend the Katz and Shaprio (1985) model for a network industry to examine the welfare implications of mergers in network industries. I examine an oligopoly model in which firms choose output levels, quality within a network (internal quality), and quality between networks (external quality). There are multiple markets and positive network externalities among the markets. Examples of such situations include the following: (1) Telecommunications markets in separate countries or cities (Mergers between SBC and Ameritech, or between Bell Atlantic and GTE, fall into this category.); (2) Internet providers in separate geographic areas (Mergers between US and European Internet providers fall into this category.); and (3) Computer operating software (Windows Me and Windows 2000 fall into this category).' These network externalities imply that an increase in output in one market stimulates demand in other markets. When a single firm serves multiple markets with network externalities, the firm internalizes network externalities and chooses higher levels of production than separate firms would provide. This fact implies that a merger may increase welfare even if it raises marginal production costs, increases market concentration, or extends market dominance. To analyze this result, I consider mergers that combine firms that operate in different markets, which I call pure cross-border mergers, and mergers that combine firms that operate in some of the same markets, but not all of the same markets. I call these latter mergers mixed mergers. ' Windows Me serves the consumer market for PCs while Windows 2000 is for Internet servers. These two operating systems contain features that allow customers using one to communicate more easily with customers using the other than with customers using a non-Microsoft operating system.

PAGE 75

66 Even if a pure cross-border merger raises marginal production costs for the merging firms, they will choose higher levels of production after the merger as long as the revenue from internalizing the network externalities exceeds the higher marginal production costs. When a single firm serves multiple markets with network externalities, its additional revenue from an increase in output in market A reflects not only the effects on price and quantity sold in market A, but also the effects that higher market demands have on prices in other markets in which the firm operates. The higher prices in these other markets provide additional revenue for every level of output in market A. I call the effect of a change in the output in A on this additional revenue the marginal extra-market revenue. Positive marginal extra-market revenue causes the merged firms to produce more even if the merger increases marginal production costs. Also, a pure cross-border merger between large firms may increase welfare more than does a pure cross-border merger between small firms. Recall that a cross-border merger internalizes the positive effects that higher output in one market has on prices in other markets. Large firms profit more from these higher prices than do small firms, and so increase their output more in response to a cross-border merger than small firms do. Although competitors of the merged entity may reduce their outputs (Farrell and Shapiro, 1990b), the output reduction does not offset the merging firms' production increase.^ Pure cross-border mergers cause higher Herfindahl-Hirschman indices even while improving welfare. The indices increase with cross-border mergers because the inside Furthermore, in the network setting, rivals in one market may actually increase their production if the merged firms' higher production in other markets stimulates demand.

PAGE 76

firms produce more output after the merger than they did before the merger. This gives the inside firms higher market shares when the outside firms reduce their production. My findings combine the existing literature on mergers and on network effects. In their seminal article on mergers, Salant et al. (1983) describe how exogenous mergers may decrease joint profits of the merging firms. Farrell and Shapiro (1990a) identify industry conditions and asset transactions among firms that increase industry concentration and worsen industry performance. Farrell and Shapiro (1990b) develop sufficient conditions for profitable mergers that raise welfare and show that the Herfindahl-Hirschman indices can be misleading. Gaudet and Salant (1991) provide rules for determining when a merger increases or decreases welfare. In their seminal article on network economics, Katz and Shapiro (1985) show how consumption externalities cause demand-side economies of scale and that firms may choose less product compatibility than is socially optimal. Katz and Shapiro (1994) analyze competition between systems, a collection of two or more components that work together. They describe how customer expectations about the future popularity of systems can result in multiple equilibria or the absence of equilibria. They also show how systems markets make coordination among firms difficult. Liebowitz and Margolis (1994) describe the limits of network externalities. Cremer et al. (2000) show how large networks may discriminate against small networks in terms of the quality of network interconnection. I develop and explain my results as follows. I develop a model in which firms serve three markets, which have network externalities between the markets. In the first stage of the game, consumers form expectations about the size of each firm's network and

PAGE 77

68 each firm's quality choices. In the second stage, firms choose quahties simultaneously. Firms play a quantity game in the third stage and these quantities generate marketclearing prices. Lastly, each customer chooses its network provider by comparing for each provider the value the customer places on the provider's service with its price. I examine pure cross-border mergers in two settings a monopoly setting and an oligopoly setting. In the monopoly setting, all three markets are initially served by separate monopolies. I then consider a merger between two of the three monopolies. If quality is unchanged, the merger improves welfare as long as marginal production costs do not increase, or, if they do increase, the increase is less than the marginal extra-market revenue. I show that quantities and welfare would increase even more if all three monopolies merged. In the oligopoly setting, output and net consumer surplus increase as long as marginal production costs do not increase. Quality and welfare unambiguously increase if the additional output stimulates demand sufficiently to cause the merged firms' rivals to increase their production. Mergers between larger firms have greater welfare effects than mergers between smaller firms because larger firms profit more from extramarket revenue than do smaller firms. I then examine a mixed merger. I consider a model in which there are initially two firms. Each firm is a monopoly in one market and the firms compete in a third market. Using a simulation, I examine the welfare effects of a merger that results in all three markets being served by a single monopolist. I find that welfare increases as long as the effects of internalizing additional network externalities plus the effect of quality changes dominate the effect of losing a competitor in the duopoly market, a situation that

PAGE 78

may occur if the duopoly market is small relative to the other markets or if customers place high value on network externalities. The analysis proceeds as follows. Section 4.2 describes the model. Section 4.3 describes how quality is chosen. Section 4.4 presents the results for pure cross-border mergers. Section 4.5 provides the results for mixed mergers. Section 4.6 is the conclusion. All proofs are in Appendix C. A detailed description of the market demand is also in Appendix C. 4.2. The Model 4.2.1. Demand and Supply Extending Katz and Shapiro (1985), I consider an oligopoly model in which there are three markets for the network service, which I designate as markets A, B, and C. For purposes of this model, a market is a customer group, such as customers located in a specific geographic region. To keep the markets distinct, I assume that customers cannot migrate between markets for purposes of buying the service; for example, a bank in New York cannot purchase telecommunications service from a provider that does not operate in New York. As many as four firms may operate in the model. The firms are labeled I, 2, 3, and 4. A firm may serve more than one market. Let q' "' > 0 denote the number of customers that firm / serves in market m, q' will denote the vector of all q' '" for a single firm /, and q will denote the vector of all outputs of all firms in all markets. For simplicity, I assume that each customer buys at most a single unit of output.

PAGE 79

70 I assume that firms "interconnect" their networks.^ In the setting of physical communications networks, this interconnection would be the lines and technical arrangements that allow customers to communicate across the networks. In the setting of virtual networks, such as computer software, this interconnection could be interpreted as a feature that allows customers of one network to benefit from customers of other networks. For example, software providers may create a feature that allows spreadsheet users to exchange data with database users. Extending Cremer et al. (2000), let 9' ' g [0,^] represent firm /'s quality choice for communications between its customers, for all i= 1, 2, 3, and let 6' ' e [0,6^] represent firm /'s quality choice for external interconnection between its network and /s network, for j ^ For simplicity, I assume that each firm chooses a single internal quality and a single external quality choice for interconnecting with each other firm. Let 9' represent the vector of /'s quality choices. Quality includes such things as capacity for customers of physical networks to exchange messages, and features, such as instant messaging and technical protocols. A choice of zero represents a refiisal to interconnect. I assume that network quality is perfectly observable to firms and customers alike. Furthermore, only one quality choice can prevail for each network interconnection. This is because only a single physical capacity can exist at a single point of intercoimection, ^ In some industries, regulators require firms to interconnect their networks. Regulated telecommunications is an example of a network industry where regulators require interconnection. The Internet is generally unregulated. Refusal to interconnect is rare in the Internet, but controversy over type of interconnection is common. See Kende (2000) for an excellent overview. " Note that the market designations are irrelevant if each firm has a single interconnection with each other firm.

PAGE 80

71 and if a technical feature cannot be used by customers of network i to communicate with customers of network j, then neither can customers of network j use the feature to communicate with customers of network /. Section 4.3 describes how differences in quality choices are resolved. The inverse demand curve for firm / in market m can be expressed as I consider a market-clearing price, given output and qualities, and I assume that price decreases continuously with quantity sold in the market. This will be the setting, for example, when there is a continuum of customer types, and higher type customers place more value on communications than lower type customers so that p'^'^„ < 0 for i,j =1,2, 3, where the subscript represents the first partial derivative. Price increases with quality because customers value higher quality network service more than lower quality network service, all other things being equal, i.e., p^" > 0 for i,j = 1, 2, 3. I assume that the marginal value of quality decreases as quality increases. Marginal willingness to pay in one market also increases with quantity sold in other markets because of positive network externalities, i.e., p^™^ > 0 for i,j= I, ...,4,andm^m . Furthermore, the marginal value of these positive network externalities decreases with the quantity sold in other markets. Quality and network externalities interact in that the marginal value of the size of the system increases with quality, i.e., p'^'", > 0 for all i,j = 1, 4, and m, m = A, B, C. To ensure that an internal solution exists for output choices, I assume that each firm's marginal revenue in a market declines as its rivals increase their output in the market, that customers who value the network service the least value it at zero, and that (4-1)

PAGE 81

each firm's residual demand curve intersects its marginal cost curve fi-om above (Dixit, 1986).^ I assume a customer makes purchasing decisions after firms have made their quality and output choices, which are based are customer expectations. Each customer chooses to purchase fi-om the firm that provides the greatest net consumer surplus. The marginal customer is the customer that is indifferent between buying and not buying the network service in equilibrium when quantity sold is strictly positive (Katz and Shapiro, 1985).^ At equilibrium, the marginal customer will receive zero net surplus and, for each firm that sells a positive amount, price is equal to the value of the network service to the marginal customer. Prices vary among firms if quality and network size vary among firms. All other things being equal, marginal customers value larger networks and networks with higher internal and external qualities. I assume that costs for production are separable from costs for quality. Firm / incurs fixed costs A^' "" > 0 for each market m and a constant marginal cost c' "' > 0 of production. All fixed costs are assumed to be sunk costs. For simplicity, I assume that there are no economies of scope across markets. Let G'(l ,0) represent firm /'s cost function for interconnection, where I is a vector such that the element of I is the number of interconnections that firm / has with firm j or has the value 1 . In the case of physical networks, firm /' establishes a costly physical cormection with each firm or with each firm in each market. In such cases, the {ij) element of I is the number of ' Underlying this inverse demand curve is a customer value function that decreases as customer types increase, increases as quality increases, and increases as the system grows. * The assumptions that price is zero for the lowest customer type and that marginal costs are strictly positive ensure that there are unserved customers in equilibrium.

PAGE 82

73 interconnections that firm / has with firm j. In the case of software, there may be no additional costs for a technical feature, such as exporting pure text files, to be available for interfacing with additional software packages. In such cases, the element of I is 1 . I assume that the cost of interconnection is weakly increasing in quality and is quasiconvex and weakly increases with the number of interconnections, i.e., G'^j > 0 , G'^j > 0 , and • GV, > 0 for i,j=\,..., 4, where /-^ is the (/,;) element of I. For simplicity, I assume that interconnection costs are independent of the number of customers. The assumption simplifies notation and does not affect results because the effects of customers on quality costs could be included in c' ™. Each firm takes its rivals' quantity choices as given when it chooses its own quantity levels. Firm /'s production in market m is zero if firm / does not operate in the market. Firm Vs profit maximization problem can be written as: When firm /' chooses its profit maximizing outputs, its output choice for each market reflects the effects of internalizing network externalities. For example, if firm / operates in both markets A and B, its marginal benefit from increasing q''^ includes the marginal extra-market revenue, p'^t^l''^ , which is strictly positive and represents the portion of network externalities between markets A and B that firm / internalizes when it operates in both markets. If firm / did not operate in both markets, its output in market A would create network externalities and higher profits for market B, but these profits subject to ^•^€[0,^J for 7 = 1,. ..,3 q'-"'>0 for m = A, B,C. (4-2)

PAGE 83

74 would not directly benefit firm /. Therefore, firm / does not consider these profits when choosing its output for market yi. 4.2.2 Welfare The surplus a customer receives from purchasing depends on the irmate value the customer places on the network service, on the internal and external quality choices of the customer's network supplier, and the total number of customers who purchase the network services. In each market, a customer only purchases if he values the service at least as much as the marginal customer does. Integrating over all customers who purchase and summing over all firms and all markets gives the net consumer surplus: -n, (4-3) where q is the vector of all q'^"" , and the weighted social welfare: Z.^r-+(l-^)X(^'+7:), (4-4) 1=1 where <3r= [1/2, 1] is the weight given to net consumer surplus, assuming that the social planner may give preference to consumers, and T'^ is a transfer payment from consumers to firm i. If a social planner chooses q and 0 to maximize weighted social welfare subject to a non-negative profit constraint for firms, she would ensure that the marginal consumer surplus, plus the positive network externality, are equal to marginal production cost and that the marginal consumer surplus from quality is equal to the marginal cost of quality. Formally, using pointwise optimization:

PAGE 84

75 4 C 4 (4-5) for all z = 1, 3 and/n =A,B, C, and: (4-6) forall/,y= 1, ...,3 mdm=A,B, C. 4.3. Quality Choices Choosing quality allows a firm to increase or decrease the consumer value of network externalities for its service and for its rivals' services. Consider oligopoly setting 1 pre-merger described in Table 1 . Firm 1 operates in markets A and B and firm 2 operates in markets B and C. There are no other firms. Firm 1 can increase network externalities for itself by increasing internal quality or increasing external quality. First consider the situation where firm 1 increases network externalities by increasing internal quality. Increasing internal quality raises demand for firm I's output. Firm 1 in turn increases output for two reasons. First, the higher demand increases marginal revenue for each level of output, which also causes firm 1 to increase output. Second, firm 1 adds customers in one market to increase the value of its network in its other market because the higher internal quality increases the marginal value from network externalities of adding a customer. The cost of adding these network externalities is the cost of the higher internal quality plus the marginal production costs of increasing output. Compare this result to the result if firm 1 increases network externalities by increasing external quality with firm 2. As in the situation of raising internal quality, increasing external quality increases the demand for I's service, which causes 1 to

PAGE 85

increase output. However, in contrast to the situation where 1 increases internal quaHty, 1 does not increase its output to add network extemahties. Firm 2's output does this for firm 1 . This leads to Lemma 1 . Table 1. Arrangements of Firms Before and After Merger Between Firms 1 and 2 Settings Firms in Marliets before and after Mergers Marlcet A Market B Market C Oligopoly Setting 1 Pre-merger Firm 1 Firms 1 and 2 Firm 2 Post-merger Firm 1 Firm 1 Firm 1 Oligopoly Setting 2 Pre-merger Firm 1 Firms 2 and 3 NA Post-merger Firm 1 Firms 1 and 3 NA All-monopoly Setting Pre-merger Firm 1 Firm 2 NA Post-merger Firm 1 Firm 1 NA Lemma 1. When firms 1 and 2 increase the external quality between their networks, firm 2's higher output increases the value of firm I's network, which saves firm 1 marginal production costs relative to the situation where 1 raises internal quality. Because of the saving in marginal production costs, firm 1 prefers higher external quality with non-rivals than it does internal quality. In addition to adding network externalities for firm 1, firm 2's higher output competes with firm I's output in markets where 1 and 2 are rivals. This decreases the profitability to firm 1 of increasing external quality. Therefore, a firm prefers higher external quality with a rival than it does internal quality if the proportion by which the firms' markets overlap is small relative to the value customers place on network externalities. Proposition 1 summarizes this result for Oligopoly Setting 1 Pre-merger. Assumption 1 describes how a change in external quality might affect firm I's revenue.

PAGE 86

77 Assumption 1: The effect of an increase (respectively, decrease) in firm I's external quality on firm I's revenue is greater than the effect of an increase (respectively, decrease) in I's internal quality. That is to say, n=A.B m=A.B where 9' "' represent firm I's optimal output in market m and p is the resulting price, given the firms' quality choices. and ^''^ p\\ — ,, In (4-5), q "' p and q p 't represent the increase in revenue 1,5' 1,5 • 2,B for firm 1 that results from firm 2's greater output and q ' p^^.a "^^iT ^presents the decrease in firm I's revenue. Proposition 1. The relationship between firm I's preferred internal quality and its preferred external quality with firm 2 depends upon the aggregate sizes of their individual markets relative to the aggregate size of the market that they both serve. More specifically: a. Firm I's preferred internal quality is lower than its preferred external quality with firm 2 if and only if Assumption 1 holds; and b. Firm I's preferred internal quality is greater than its preferred external quality with firm 2 if and only if Assumption 1 does not apply and the inequality sign in (4-7) is reversed. The relationship between firm I's preferred internal and external qualities depend on the relative sizes of the markets and the value customers place on exchanging

PAGE 87

78 information. For a given value of exchanging information, Assumption 1 will hold if market y4 or C, or both, is sufficiently large relative to market B. Furthermore, a sufficient condition for Assumption 1 to apply is for firm 2's higher output in B to stimulate more revenue in absolute value for firm 1 in market^ than it decreases firm I's revenue in market B, i.e., 9'"^ p^'i^s > 9''* This leads to the following corollaries. Corollary lA. Firm 1 prefers a higher external quality than internal quality if the ratio of its outputs in A and B is greater than the negative ratio of the slope of its demand in B P to the marginal value of firm 2's output in market A, i.e., — — r > Corollary lA applies regardless of the size of firm 2, so even if firm 1 is larger than firm 2, firm 1 would prefer a higher external quality than internal quality. Corollary IB. In the special setting where markets and firms are symmetric, a sufficient condition for firms to prefer higher external quality than internal quality is for the marginal value of network externalities to be greater in absolute value than the slope of the demand curve, i.e., > 1 I assume that two firms engage in efficient bargaining to determine external quality between their networks if the firms prefer different interconnection qualities. The bargaining maximizes the profits of the two interconnecting firms given the quality choices of all other firms. Also, it may be difficult in some industries for firms to adopt an external qualities that are higher than internal qualities. An example would be telecommunications networks if quality represented capacity for communications. In

PAGE 88

these industries, if Assumption 1 applied to one or more of the firms such that the bargaining would result in a firm having an external quality that was higher than its internal quality, the firm would choose a single quality for internal and external communications. ^ 4.4. Analysis of Pure Cross-Border Mergers In this section I consider pure cross-border mergers, mergers between firms that are not rivals. After the merger, customers form new expectations about the sizes of the firms' networks and all firms select their outputs and qualities to maximize their profits. I first consider the all-monopoly setting described in Table 1 , the situation in which each market is served by a monopoly. The merger that 1 analyze is a merger of firms 1 and 2. I then consider oligopoly setting 2, the situation in which at least one of the markets has more than one firm. Consider the all-monopoly setting in which markets A and B are initially served by separate monopolies. Now suppose the monopolies merge. Before the merger firm 1 operates in market A and firm 2 operates in market B. To maximize profits in this permerger situation, each firm equates its marginal revenue from expanding output with ^ From first order conditions, the firms choose a negotiated external quality defined by (ff'^'p (^-'' ))7^'^ }=-(!^)^;, [^''^'p (^''^ * )^' )+ t;-^ }, which maximizes joint profits. Also fi-om first order conditions, the transfer payment from j to / is equal to the weighted average of the effect of negotiated quality on their individual profits; i.e., n-' = +^ fef )i^j!^ h (1 ^P ^ii Y ^' fe I

PAGE 89

i 80 marginal cost, given quality levels. Now assume that firms 1 and 2 merge and let 1 represent the merged firm. Assumption 2 describes a possible effect of the merger on marginal production costs. Assumption 2: The change in marginal production costs in market^ (respectively, market B) is less than the effect that positive network externalities have on the combined firm's marginal extra-market revenue in A (respectively, B) from firm 2's (respectively, firm 1 's) original market plus the effect that quality changes have on marginal revenue, i.e., {p'^:*p','.,,.: -V'^l?" >?'-^ where A^'' is the merger's effect on firm I's internal quality. The pre-merger marginal revenue for firm 1 is + P^m "^'"^ ' [py' + P^M ^1..* • <7' '^'}^<9''' is the effect of a change in internal quality on marginal revenue, [p^^' + p^'^t ^u* • 9'"^* 1^ ^''^ ) and is the effect of changing the external quality between firms 1 and 2 to internal quality for the merged firm. Proposition 2 provides this section's initial result. Proposition 2. Suppose Assumption 2 holds in the all-monopoly setting. Then a pure cross-border merger results in higher output and welfare. If Assumpfion 2 does not hold such that inequality in (4-8) is reversed, then a pure cross-border merger reduces output and welfare.

PAGE 90

81 If Assumption 2 holds, a pure cross-border merger increases output because the merged firm intemaUzes positive network externalities. Before the merger, if firm 1 were to increase its output in market A, firm 2's profit in market B would increase because demand in market B would increase. However, because firm 2's profits do not directly benefit firm 1, firm 1 does not take this increase in firm 2's profits into consideration when choosing quantity in market ^. After the merger, firm T internalizes the effect that the quantity sold in market A has on demand in market B, and visa versa. Consequently, firm T increases output in both markets A and B above the levels delivered by firms 1 and 2 before the merger, all other things being equal. Welfare increases (respectively, decreases) when a pure cross-border merger increases (respectively, decreases) quantity because both customers and producers are better off (respectively, worse off) when quantity increases (respectively, decreases). To illustrate, consider a pure cross-border merger that increases output of all firms. All new customers enjoy increased net surplus, because these customers received zero net surplus before the merger and now receive a positive net surplus.^ Also, because the marginal value of network externalities is decreasing in quantity, the difference between consumer value and price increases for existing customers when quantity increases, causing an increase in net surplus for existing customers. Profits are also higher after the merger because both inside and outside firms can always choose the same outputs after the merger as they did before the merger, but both types of firms voluntarily increase output. * All customers who purchase receive a positive net surplus, except for the marginal customer, who receives zero net surplus.

PAGE 91

82 Because the pure cross-border merger increases both net consumer surplus and profits, the merger increases welfare. Because of these external effects, pure cross-border mergers can increase the merged firm's output even if the merger raises marginal production costs. If the crossborder merger for firms 1 and 2 increases marginal production costs in market A by an amount less than the marginal increase in demand in market B that firm 1 internalizes, then firm T increases output in market >4 above the level delivered by firm 1. If the increase in marginal production costs exceeds the marginal increase in demand that the merged firm internalizes, then output decreases. A merger changes how the merging firms choose quality because the connection between markets A and B changes from an external connection to an internal connection. This change may cause quality to decrease or increase depending on how the cost of quality varies with the number of interconnections, the numbers of customers in each market, and the value that customers place on the network service. The following corollaries provide rules of thumb for the effects of pure cross-border mergers on quality. Recall that is equal to either one or the number of interconnections between firms / and jCorollary 2A. A pure cross-border merger in an all-monopoly situation increases (respectively, decreases) internal quality for a merging firm if the marginal effect on revenue of increasing internal quality for the merged firm is greater than (respectively, less than) the ratio of the post-merger to the pre-merger times the

PAGE 92

83 pre-merger marginal effect on revenue of increasing internal quality, i.e., internal quality increases for firm / if P-tj^'"^ + Pf^r^^^'^ > ~JIT P'^'^ "" ' If the cost of quality is independent of the number of interconnections, it is sufficient for internal quality to increase for the output in market m to be unchanged by the merger. If the cost of quality is linear with respect to the number of interconnections, then internal quality increases only if the marginal effect on revenue of increasing quality after the merger is three times what it was before the merger. In the special case where the value of network externalities is linear with respect to quality and output, internal quality for customers in market m increases only if the post-merger output in markets A and B is three times the pre-merger output in market m. In some situations, quality may be fixed by industry standards or by a regulatory authority, or the cost of quality may be sufficiently low so that firms choose the maximum quality. The following corollaries provide two additional rules of thumb for pure cross-border mergers in the all-monopoly situation when quality is not equal to zero and is unaffected by the merger: Corollary 2B. A pure cross-border merger in an all-monopoly situation increases output and improves welfare as long as it does not increase marginal production costs if quality is unchanged. Corollary 2C. In an all-monopoly situation where quality is unchanged by the merger, increasing the number of markets involved in a pure cross-border merger increases output and welfare, all other things being equal.

PAGE 93

Corollary 2C is simply a special setting of Proposition 2. Corollary 2D follows directly from Proposition 2. Increasing the number of markets involved in the merger increases the amount of positive network externality that the monopoly is able to internalize. As a result, the monopoly chooses higher output, which improves welfare. Now consider the oligopoly setting 2 in which, initially, firms 2 and 3 serve market B and firm 1 is a monopoly in market ^. Now let firms 1 and 2 merge and let 1 represent the merged firm. Proposition 3 and the following corollaries provide the initial results for a pure cross-border merger in this oligopoly setting. Proposition 3. Suppose Assumption 2 holds in oligopoly setting 2. Then a pure crossborder merger results in higher output and welfare. If Assumption 2 does not hold such that inequality in (4-8) is reversed, then a pure cross-border merger reduces output and welfare. When Assumption 2 holds, the merging firms increase their output because they are internalizing network externalities. Depending on the value customers place on network externalities and the relative sizes of markets, rivals to the merged firm either increase or decrease their output in markets. Corollaries 3A and 3B describe sufficient conditions for the rivals to increase their output. Corollary 3A. If Assumption 2 holds in oligopoly setting 2, a pure cross-border merger causes the merging firm's rival in market B to increase its output if the effect of positive network externalities from greater output in market A, plus the effect of the change in external quality, is greater in absolute value than the effect of greater output

PAGE 94

85 in market B, i.e., an increase in output in market A, \p^~, , + P^l.a gi.y 9 is the effect of the change in external quahty between firm 3's customers and the negative effect of an increase in the merged firm's output in market B. Corollary 3A describes the situation where network externalities are sufficiently great to cause profits to increase for the merged firms' rival. Welfare increases because net consumer surplus and all firms' profits increase as a result of the merger. Even if the merged firms' rival decreases its output in response to the merger, output increases in market B because the rival would decrease its output by an amount less than the merging firms' increase. (Farrell and Shapiro, 1990b) If follows from Corollary 3 A that the merging firms' rival increases its output more (respectively, decreases it less) if the number of markets that provide the network service increases, all other things being equal. Likewise, increasing the number of markets where the rival and the merging firms compete causes the rival to increase its customers of the merged firm, and represents the

PAGE 95

86 output less (respectively, decrease it more) in response to an increase in the merging firms' output. These results from Proposition 3 provide additional rules of thumb, which are summarized in the following corollaries. Corollary 3B. A pure cross-border merger in oligopoly setting 2 improves net consumer surplus as long as it does not increase marginal production costs and quality remains unchanged. Corollary 3C. Increasing the number of markets involved in a pure cross-border merger improves the welfare affects of the merger if marginal production costs and quality remain unchanged. In oligopoly setting 2, cross-border mergers between larger firms are more beneficial for merging firms and customers than mergers between smaller firms. Recall that before the merger, firm 1, which operates in market A, can choose whether to merge with firm 2 or firm 3, which compete in market B. Further assume that firm 2 is larger than firm 3 before the merger because firm 2 has lower marginal production costs than firm 3 (Farrell and Shapiro, 1990b) and that the merger has the same effect on marginal production costs regardless of which firm merges with firm 1 . If firm 1 merges with firm 2, the larger firm in market B, it internalizes more network externalities than if it merges with the smaller firm because the network externalities that it internalizes are determined by the output that the merged firm produces in market B. This quantity is greater if firm 1 merges with the larger firm in market B. The following corollary summarizes these results.

PAGE 96

87 Corollary 3D. A large-firm pure cross-border merger in oligopoly setting 2 increases net consumer surplus more than a small-firm merger. A large-firm merger reduces industry costs more than does a small-firm merger. The firms in the cross-border merger increase their production relative to the outside firms. Because the larger firm in market B has lower marginal production costs than the smaller firm, increasing the larger firm's market share provides lower industry costs than increasing the smaller firm's market share. This decrease in industry production costs occurs even though the sum of the marginal costs and the variance of marginal costs are both constant. This modifies the results of Bergstrom and Varian (1985) and Salant and Shaffer (1999), who respectively found that a market's output is determined by the sum of the marginal costs of the firms in the market and that there is a negative correlation between aggregate producfion costs and the variance of marginal costs. Others have found that higher Herfindahl-Hirschman indices can be associated with higher welfare in Coumot models (Farrell and Shapiro, 1990b; and Salant and Shaffer, 1999). This finding also holds for pure cross-border mergers in network industries. Specifically, when any increase in marginal production cost for the merging firms is more than outweighed by positive network externality effects of the merger, the market share for the merging firms increases. This causes the Herfindahl-Hirschman index and net consumer surplus to increase. Per Corollary 3A, if the merging firm's rivals operate in markets where the merging firms do not and if the rival's output in these markets is large relative to its output in the market where it competes with the merging firm, then the merger unambiguously increases welfare even when the HerfindahlHirschman index has increased.

PAGE 97

88 4.5. Mixed Mergers In this section I consider an oligopoly setting where a merger decreases the number of firms in one market and the number of markets served by the merged firm is greater than the number of markets served by individual merging firms before the merger. Mixed mergers, such as the one in this model, raise issues of market concentration and extending monopoly or dominant market position for regulators. Consider oligopoly setting 1 where firm 1 is a monopolist in market A and a duopolist in market B. Also suppose firm 2 is a monopolist in market C and serves as the second duopolist in market B. Now assume that firms 1 and 2 merge and let 1 represent the merged firm. After the merger, customers form new expectations and the merged firm selects its quality and output to maximize its profits. I call market B the common market because both merging firms operated in it before the merger, and I call the markets A and C the non-common markets because only one merging firm operates in each of these markets before the merger. The mixed merger has two effects on output. First, the merger removes a competitor in the common market. All other things being equal, this causes output to decrease in this market. Second, the merger increases the network externalities that merging firms internalize in non-common markets. To illustrate, consider how the merging firms choose output in market y4. Before the merger, when firm 1 chooses its output, it internalizes network extemalifies from its output in market B. After the merger, when the merged firm chooses its output in market A, it internalizes network externalities

PAGE 98

89 from its output in both markets B and C. Not only does the merged firm internalize network extemahties from an additional market, but it also internalizes network externalities from a larger output in the common market, market B, because the merged firm serves the entire market. As a result, unless the post-merger output in markets C and B are less than firm I's pre-merger output in market 5, the merged firm internalizes more network externalities than the pre-merger firm when choosing output in market A. The corresponding analysis for market C would provide analogous conclusions. The mixed merger does not necessarily increase network externalities that the merged firm considers when it chooses its output in the common market. Before the merger, the merging firms jointly internalize network externalities from markets A and C when choosing their output for market B. The merger does not change the total number of markets served by merging firms. As a result, unless output in non-common markets increases as a result of the merger, the network externalities internalized for output in market B are no greater after the merger than before the merger. However, the merger does increase the number of markets whose network externalities the merged firm internalizes when choosing output in non-common markets. Now consider the effect of the mixed merger on quality. Proposition 4 describes how the mixed merger affects quality. Assumptions 3 and 4 provide conditions for Proposition 4. Assumption 3: The difference between — — , the ratio of the post-merger and pre-merger correlation between quality costs and internal connections, times the pre-merger marginal revenue from internal quality and the post-merger

PAGE 99

90 marginal revenue from internal quality in firm f's pre-merger markets is less than the post-merger marginal revenue from internal quality in the market that i did not serve before the merger, i.e., where is the set of markets served by / before the merger and miM' . Assumption 4: The difference between the ratio of the post-merger and premerger correlation between quality costs and internal connections, times the pre-merger marginal revenue from external quality in firm i's pre-merger markets, is less than the post-merger marginal revenue from internal quality in all markets, i.e., p^^ -r' -r^ < • (4-10) Proposition 4. A mixed merger in oligopoly case 1 increases internal quality if Assumption 3 holds and results in an internal quality that is higher than the premerger external quality if Assumption 4 holds. Alternatively, the merger decreases internal quality if Assumption 3 does not hold and the inequality (4-9) is reversed and results in an internal quality that is lower than the pre-merger external quality if Assumption 4 does not hold and the inequality in (4-10) is reversed. The merger increases the number of markets that a single firm considers when choosing internal quality. As a result, if the cost of quality is independent of the number of interconnections, the merger increases internal quality for the markets served by firm / unless the merger causes a decrease in output sufficient to lower marginal revenue from

PAGE 100

91 internal quality in all markets to less than what it was for i's pre-merger markets. The merger also changes from external to internal the communications between markets A and C, between market C and firm I's customers in B, between market A and firm 3's customers in B, and between firm I's customers in B and firm 3's customers in B, which changes how quality is chosen for these interconnections. It follows from Proposition 1 that changing an interconnection between markets from external to internal may lower quality. I use a simulation to illustrate the mixed merger's effect on output and welfare. For the simulation, I adopt the supply and demand functions explained in the Appendix and assume that the merger has no effects on marginal production costs.^ Figure 1 shows the effect the merger has on net consumer surplus for various sizes of the common market relative to the non-common markets and for various values of network externalities. The curved plane shows the region where the effect of the merger is to increase net consumer surplus. The region where no plane is shown represents a decrease in net consumer surplus. The light colored portion of the plane represents the area where quality is less than one, but more than zero. The merger always decreases net consumer surplus when quality is zero because consumers and firms receive no benefits from network externalities. The merger increases industry profits for all regions, so the curved plane also shows the region for which the merger unambiguously increases welfare. The change in net consumer surplus increases with the value customers place on network externalities because higher values increase the firm's marginal revenue from ' The Appendix provides further details of the simulation.

PAGE 101

92 Figure 1. Effect of Mixed Merger on Net Consumer Surplus internalizing network externalities, which causes the firm to increase output. The higher output increases net consumer surplus because customers who did not purchase before purchase now and receive a positive net surplus. Customers who did purchase before receive a net increase in their surplus because their value of the network service increases relative to price. The change in net consumer surplus becomes large at high values of network externalities because higher values placed on network externalities cause the marginal revenue from internalizing these externalities to always exceed marginal production cost, which gives the firm an incentive to serve the enfire market.

PAGE 102

93 The change in net consumer surplus decreases as the size of the common market increases relative to the combined sizes of the non-common markets. This occurs because, as the relative size of the common market increases, the effect of losing a competitor in the common market rises relative to the benefit of additional internalized network externalities in the non-common markets. Losing the competitor allows the merged firm to lower output and raise price in the common market. When the common market is small relative to the non-common markets, the increased output in the noncommon markets (from additional internalized network externalities) causes an increase in demand in the common market. When the common market is relatively small, this demand increase dominates the effect of losing the competitor. 6. Conclusion In this paper, I extend Katz and Shapiro (1985) to examine how mergers affect industry performance in network industries. I find that internalizing network externalities causes a cross-border merger to improve efficiency. This has several results that contradict conclusions the current literature. These mergers may improve efficiency even when they increase marginal production costs. Also, mergers between larger firms have more beneficial affects on welfare than do mergers between smaller firms. This study implies that conventional views of mergers in network industries may be inappropriate and that careful study of proposed cross-border mergers and mixed mergers is needed before mergers are rejected on grounds of extending monopoly or dominant market positions or increasing industry concentration.

PAGE 103

94 Additional research is needed. I do not analyze whether all cross-border mergers improve welfare. I also do not examine the case where products provided by merging firms are imperfect substitutes rather than completely separate products. Lastly, empirical research is needed to determine the size of network externalities, how they vary across markets and networks, and whether network firms that serve multiple markets respond to the incentives that I model.

PAGE 104

CHAPTER 5 CONCLUSION This dissertation consists of three studies of the restructuring of the telecommunications industry. In the first study, I examine incentives for discrimination in network interconnection. I find that a large firm implements the same external quality for a given network interconnection as does its smaller rival. Furthermore, I find that smaller firms offer their own customers an internal network quality that is lower than the interconnection quality that the smaller firms implement with the larger firm. My results conflict with the findings of Cremer et al. (2000). They examine a sequential game and their model assumes that differences in firm size result from differences in numbers of attached customers, and find that these differences in attached customers cause firms with a higher number of attached customers to prefer a lower quality interconnection than a firm with a lower number of attached customers. They obtain this result because firms ignore the preferences of their own attached customers when choosing interconnection quality, but take into account unattached customers' preferences. The assumption that attached customers have entered into contracts that do not reflect the firm's quality choices, but that unattached customers' contracts do reflect the firm's quality choices, seems difficult to explain. If it were optimal for firms to ignore preferences of captive customers, such as when the customers were in a monopoly market, then the monopoly in my model would have done so. Because the monopoly in 95

PAGE 105

96 my model did not ignore its captive customers' preferences, it is not optimal for firms to enter into contracts in which customers believe the firms will ignore their preferences when making quality choices. Therefore, these customers are not attached in the sense of Cremer at al. (2000). Because it is difficult to construct a scenario in which customers and firms choose contracts that result in attached customers in the sense of Cremer et al. (2000), I assume that asymmetry in firm size results from access to monopoly markets and cost asymmetries. These conclusions indicate that the EU may not have been justified in its conclusion that a larger Internet firm resulting from the MCIWorldCom merger or the WorldCom-Sprint merger would discriminate against European firms for access to the Internet backbone. It appears true that the larger firm would provide its own customers with a higher quality service that it would provide its competitors, but the larger firm's interconnection quality choice would be no different than the smaller firms' interconnection quality choice for connecting with the larger firm. Furthermore, the interconnection quality the larger firm would choose for connecting with the smaller firms would be higher than the quality the smaller firms would choose for cormecting with each other. My study of the second issue indicates that entrants in telecommunicafions markets in the US use a mixture of entry and supply strategies, choosing to build their own network facilities, lease facilities or UNEs from incumbents, or resell incumbent services, but that most entrant output is based on using UNEs and entrant-owned facilities. Reselling incumbent services appears to be important for entry, but not for growth in entrant market share. Prices for UNEs and reciprocal compensation are

PAGE 106

97 important determinants of entry, marketing, and modes of supply. Higher UNE and reciprocal compensation prices discourage entry, but higher reciprocal compensation prices also cause entrants to market to customers who are net receivers of minutes. Low price-cost margins for UNEs cause incumbents to hinder entry, but incumbents have little effect on entrant market share. Also, the ease of using UNEs and placing facilities appears to be more important than relative prices for UNEs and wholesale in influencing entrants' choices of methods of supply. These results indicate that, if regulators want to increase market penetration for facilities-based entrants, they should focus on unbundling networks, prices for UNEs, and collocation. Resale discounts for services, while important for entry, have little effect on entrant market share. Additional work is also needed on this topic. I was unable to test whether the interLATA restriction, which is to serve as a carrot for BOCs to open their markets to competition, is having its intended effect. Also, I was unable to test whether the policies that encourage entrant supply also improve social welfare. A more complete model of how policies affect industry costs and prices is needed to determine welfare effects. Lastly, international comparisons are lacking, as are more in depth analyses of entrant strategies for data and voice services. In my fourth chapter, I extend models by Katz and Shapiro (1985) and Farrell and Shapiro ( 1 990b) to examine how mergers affect industry performance in network industries. I find that internalizing network externalities causes cross-border mergers to improve efficiency. Furthermore, I find that cross-border mergers increase the efficiency of horizontal mergers. This study implies that competition regulators should favor cross-

PAGE 107

border mergers. Also, pure cross-border mergers between large firms should be viewed more favorably than pure cross-border mergers between smaller firms. Additional research is needed. I do not analyze whether a cross-border merger needs to be profitable to improve welfare. The effects on the efficiency of horizontal mergers implies that this externality may make it possible for even unprofitable crossborder mergers to be welfare improving. I also do not examine the case where products provided by merging firms are imperfect substitutes rather than completely separate products. Lastly, empirical research is needed to determine the size of network externalities and how they vary across markets and networks.

PAGE 108

APPENDIX A PROOFS AND LEMMAS FOR CHAPTER 2 Lemma Al. Profit maximizing behavior for choices of external quality results in a set of Nash equilibria &,j = e [0, min {^ij, ^p}], where ^ij represents /'s profit maximizing choice for external quality with fs network if i were allowed to dictate • the quality, and ^jj represents fs profit maximizing choice for external quality with i's network if j were allowed to dictate external quality. Proof: Recall that the firm with the lowest quality choice determines the effective quality of the interconnection. Now consider strategy choices that have unequal quality for communications established by I's customers. The firm with the higher quality can decrease its costs and not affect revenues by choosing a lower quality, so all unequal capacity choices are not Nash equilibria. Now consider the choices 0 < &,j= < min {^ij, Neither firm would increase its quality because it would incur costs and gain no revenues. Neither firm would decrease its quality because, at a lower quality, its marginal revenue from quality would be greater than its marginal cost of quality. Therefore, all of these choices are Nash equilibria. Now consider the choices = (^,, > min { ^ ij, For at least one of the firms, marginal revenue from the quality choice is less than the quality choice's marginal cost of quality, so this firm can improve its profits by decreasing its quality. Therefore, equal quality choices in this range are never a Nash equilibrium. 99

PAGE 109

100 Proof of Lemma 1. Assume that Lemma 1 is false. Then at equihbrium, for some i,j 1 , 2, 3 and / r + 5,^ p,^ > r + 5^,^ p^,^ « P,,^ > ^j,, ' Pj,, ' which means that customers would strictly prefer network / over network;. This conflicts with the assumption that customers buy from both networks in equilibrium, so Lemma 1 must be true. Proof of Proposition L Because there is no incumbent firm, = 0 and r0= 0. From (27), the first order conditions for an internal solution include: T .c-2(i-v^,,K,z(i-v^,,;fcr^,«=o, dTC. 5^ = vql,-KXo,,)=0,m
PAGE 110

101 S^l.B > = 2 ^v^f., +2vq,,q,,+vql -^,(^,,,)= 0, and (A4) = v9,,,fe,,,+9,,J-/^.k>0. (A5) First order conditions for an internal solution for firmy 1 include: — = T-cdn X(l-v^,,k« -2(l-v^„>„. +2v^^,,^,,, =0, (A6) -vql,-K,(0j=O, (A7) =^qj,Bq,B-Ko[0j,)-^ for/^7,and (A8) 30, From(A5), v^;,!^', +^;, for j = 2,3. Now consider the external quality choices for the interconnection between firms 2 and 3. From (A8), 01, = 0^^ = ^^'l^^lfi^Is)This confirms Proposition 2.0 Proof of Corollary 1 and Lemma 2: Because I assume a symmetric equilibrium for firms 2 and 3, (A7) and (A8) imply K^' (v^ 2,5^3% )= ^0 {^(^Ib )^ )= ^0 (^(^3, which implies that ^'j 3 = 0^ 2 0\^i = ^3,3 •

PAGE 111

102 Now consider quality choices for interconnection between asymmetric networks. Solving (A3) for qi^ in terms of q\j, and solving (A6) for qj^ in terms of qij gives: . (r c)[3(l v^;, )2(1 v0l )J+ ql [Gv^I (l v<, )2v<, (l v^l )] 6(l-v^,^v^J-2(l-v^J and 6(l-v^,)(.-v<,)-2(l-v^J Consider how changes in the cost of quality affect the optimal quantity choices in (A 10) and (All). For simplicity, assume for purposes of this proof that cost functions for quality are such that comer solutions are obtained for quality. If the cost of quality is sufficiently low that firms choose the comer solution ^ ij = 1, then the monopolist chooses a greater quantity than its rival in market B; i.e., As the cost of quality becomes higher for every level of quality, firms lower their quality choices. The rivals lower their intemal quality and their extemal quality with each other before they lower their extemal quality choices with the monopolist and before the monopolist lowers its quality choices. To demonstrate this result, note that at the comer solution where quality is equal to 1, v(q:J (A13) for j ^ k, and ij, k = 2,3. Combining (A 13) with (A4), (A5), (A7) through (A9), and (A 12), gives,

PAGE 112

103 kM^i +^:JhK-;(vg],{gi +^:,.))>^.'fe,;,j=^,-(v(,;,,y). (am) (A 14) contains weak inequalities because the maximum value for quality is 1. Because A:^'(v9*.b9I.5) and K'g'i^{q'gJ ) are the lowest values in (A3), ,and(?* , are the first to move below the comer solution ^ij ^ 1 as the cost of quality increases for each quality choice. 0' j andff'j^ decrease next as the cost of quality choices increases because K^^ (vq* g (q^^ + q[g )) is the next lowest value in (A3). 5^*1 is the last quality choice to decrease. Because > 6\ ^ > 5'*^ = q'j B = ql g for all levels of cost of quality. These quantity choices cause the pattern of quality choices to persist until costs are so high that the comer ^ ^ ^ ^ ^ solution ^ij = 0 is reached, at which point ql g = ql g q\ g = . This confirms Corollary 1 and Lemma 2. Proof of Lemma 3: The incumbent's first order conditions for an intemal solution include: = r c X ((l )qj,, + v^,,,9,,, )2(1 v^,,, + 2v0,,q,g = 0 , (A 1 5) d^l,A J=2 = f c }}j.s + y^ujqj2(1 v^,,, \g + 2v^,,9,,, 0 , (A16) vfou -'^^(^i,i)=0,and First order conditions for an intemal solution for entrant j ^ 1 include:

PAGE 113

104 T -c-c^-r. + 2v^,,,^,., =0 dn = r c v^,, k,. + v^,,^,,, )2(1 v^^,>,« + 2v^^,^.^^,, = 0 , (A19) dn: ^ = vfe,,, +^,,)-^.(^,,„r,)=0. (A21) Consider (A 17), (A20), and (A2 1). Recalling that Ki^j,0) = K(6i.j), the proof of Proposition 2 confirms Lemma 3. Proof of Proposition 3: First consider the incumbent's optimal choice of r^. If r^> 0, then 0* >0' , V / 1 . Because the firm with the lowest external quality preference determines the effective interconnection quality, the interconnection quality between the incumbent and the entrants is lower than the incumbent's optimal external quality choice. Therefore, the incumbent's optimal choice for r^is re = 0. Now consider firm 1 's optimal choice of r. Combining (A15) and (A16) gives: 9Mfc..>9.J=^^r^2(l-2v<,) 2(l-2v<,)^--'"'^'^''^ . i \ T-c 2[\-v0l-v0l) 2[v0l-vel) 9.,.fe,.^9M)=^(^r^2(i-2v^;,) 2[i-ive:y"^^"^^^ Similar calculations for (A 18) and (A 19) give:

PAGE 114

105 . ( N r c (l v^;, h + ^J(\ 0 . This confirms Proposition 3.

PAGE 115

106 Proof of Corollary 2: The rivals' higher costs cause each to choose lower quantity than the incumbent. Recalling that firms 2 and 3 make symmetric choices in equilibrium, the proof to Corollary 1 confirms > > O] ^ = e]j^ for; k for ally, k=2,7>. This confirms Corollary 2. Proof of Corollary 3: Consider (A23). An increase in the entrants' quantity choice in market y4 decreases the incumbent's quantity choice in market B because 6'', > <9,V for all j = 2, 3. Analysis of (A22) shows that the same holds for the entrants' quantity choice in market B and the incumbent's quantity choice in market /I. Now consider the entrants' response to the incumbent. From (A24), the entrants' increase their quantity choice in market A if the incumbent increases its quantity choice in market B because O^j > 0] j for all j = 2, 3. Analysis of (A25) shows that the same holds for the incumbent's quantity choice in market B and the entrant's quantity choice in market^. This confirms Corollary 3.

PAGE 116

APPENDIX B APPENDIX FOR CHAPTER 3 For the ordered probit models, I group observations in categories. This is necessary because an ordered probit requires multiple observations for each value of the dependent variable. Figure B-1 graphs the entry levels in my data set, from the markets with the fewest entrants to those with the most entrants. The number of entrants ranges from 0 to 164. Over 130 entry levels in this range are not observed and many have only one observation. Using these data directly in an ordered probit model causes the model to estimate entry probabilities. To illustrate the problem, consider an economy with four markets. The markets are designated 1 through 4 and they have 0, 1, 5, and 10 entrants respectively. If these markets provided the data for an ordered probit model of entry, the model would estimate probability of entry levels as follows. The model would first estimate the probability of no entry and positive entry. Then it would estimate the probability of one entrant and more than one entrant, given that entry was positive. Next it would estimate the probability of two entrants or more than two entrants, given that entry was positive and more than one. The model would continue this process until it reached the observation with 10 entrants, where the model would treat as the probability of having more than 9 entrants, given that entry was more than 8 and more than 7, etc. As a result, the model would estimate ten probabilities of entry from four observations. The results would be questionable at best. 107

PAGE 117

108 Figure B-1. Entry Observations To avoid the problem of estimating probabilities for numerous entry levels for which I have no data or too little data, I group the observations into nine categories. The dashed lines in Figure B-1 show the deliniation of the categories. I chose the categories boundaries and the number of categories based upon visual observation of this graph. I chose boundaries based upon what appeared to be natural groupings of markets, based on the number of entrants, and on the number of observations that would be in a category. I attempted to avoid creating categories that were very similar in their number of entrants and categories with few observations. I tested other arrangements of categories based on

PAGE 118

the same criteria. All gave similar results and none gave stronger results than the results present. Table B-1 shows the statistics for the entry categories. It also shows the statistics for the categories for TRUNKS. I present the results for ordered probit models of TRUNKS below. Table R1. Descriotion of Grouped Dependent Variables Group Ordinal Group Mean Group Min Group Max Standard Deviation Number Observations Value Grnunpd Dpnendent Variable COMPT 0 1.250 0 2 0.7071 8 1 3.625 3 4 0.5175 8 2 6.600 6 7 0.5477 5 3 11.000 9 14 2.1605 4 4 22.167 18 25 2.6227 12 5 33.091 28 38 3.4192 11 6 42.667 41 44 1.5275 3 7 52.750 50 60 4.5000 4 8 113.500 74 164 42.1228 4 Groaned Dependent Variable TRUNKS 0 84.3 0 264 107.1 9 1 1,891.4 996 2,905 779.8 7 2 5,547.1 4,090 7,135 1,059.2 9 3 9,553.4 9,260 10,203 396.7 5 4 12,696.7 11,924 13,363 725.4 3 5 18,540.7 16,011 20,488 1,676.1 6 6 25,011.0 21,770 28,595 3,396.4 4 7 39,717.8 36,521 42,249 2,428.6 4 8 57,469.4 55,685 61,938 2,655.0 5 9 137,689.1 72,641 289,299 82,240.4 7

PAGE 119

APPENDIX C PROOFS AND LEMMAS FOR CHAPTER 4 Explanation of demand. In each market, customers are identified by their type r e [r" , r'" J, where r is the highest customer type in market m and r" is an arbitrarily low number which ensures that the lowest type customer would receive a negative surplus even if all other customers purchased the network service, quality was equal to 1, and prices were zero. A type r customer has a willingness to pay v(r, /, 9', q) for firm z's service and obtains a net surplus from buying from firm / at price p'"' in market //equal to v(r,/,e' ,q)p'-"" . For convenience, I express v(r, /, 0', q) as v'. I assume that higher customer types place greater value on the service, customers strictly prefer higher quality, customers strictly prefer for more customers on the system of networks (reflecting positive network externalities), and that the marginal value of additional customers is decreasing; i.e., > 0 , v^,j > 0 , v^, ,„ > 0 , and v' , „ < 0 , for all / = 1 , . . . , 4. The assumption that v > 0 causes marginal revenue in a market to be increasing in quality. I fiirther assume that the marginal value of network externalities is weakly higher for higher customer types; i.e., v[ ^, „ > 0 and the demand curve is quasiconvex. If all customer types are identical in their marginal values of network externalities, then the demand curve is linear. I assume that, all other things being equal, a customer is indifferent about which other 110

PAGE 120

Ill customers are connected to networks and the networks to which these other customers are connected. To avoid tipping effects for a single market, the situation where network externalities create a comer solution in which firms serve the entire market, I assume that v\ > v^,,„ for all; = 1, . . ., 4. I assume a customer makes purchasing decisions after firms have made their quality and quantity choices. Assuming no price discrimination, each customer of type rchooses the firm for which v' p' " is the greatest. I define the marginal customer to be the customer that, in equilibrium, is indifferent between buying and not buying network service.' At equilibrium, the marginal customer will receive zero net surplus and so will have a value of v' = p' "' . I assume that customers in market m are distributed according to the cumulative distribution ftinction f). The quantity of customers that firms choose to serve in market m in equilibrium is simply the total number of customers that lie between the upper bound of customer types and the marginal customer; i.e., 2^9' '"' = (^^ )' where n is the number of firms, q' "" = 1=1 * 0 if / is not in market m and r" , which is the type of the marginal customer m market m in equilibrium, is defined by Proof of Proposition 1 and Corollaries lA and IB. From (4-2), firm 2's first order conditions for optimally choosing output include: p'-' + P'fi^'-' -c''' + p'fiq''' = 0, and ' The assumptions that v' = 0 for the lowest customer type and that marginal costs are strictly positive ensure that there are unserved customers in equilibrium.

PAGE 121

112 Totally differentiating these first order conditions gives: > 0 , and Firm I's first order conditions for optimally choosing 6''' and 5^'^ include: * , * m=A,B l,m .2.8 and (A4.1) 2,C A (A4.2) From (A4.1) and (A4.2), if firm 1 prefers a higher external quality than internal quality, then G^u l^' '' )> (^'•'' ), m=/l,5 l,m \,m , „l,m or o 2,B a„2,cA > (A4.3) n=/f,B This confirms Proposition 1. The term p'f/ is the only negative element in (A4.3). Therefore p'f, * — — must be larger in absolute terms than all of the other elements on the left-hand side of (A4.3) combined for firm 1 to prefer a lower external quality than internal quality. This confirms Corollaries lA and IB. Proof of Proposition 2 and Corollaries 2A-C. Consider firm I's first order condition for choosing output before the merger: (A4.4)

PAGE 122

113 and firm T'5 first order condition for choosing output if it serves two markets: f-'+pl6f-'-c'''+pl^^f-'=0, (A4.5) where the ~'s denote poster-merger prices and quantities. Combining (A4.4) and (A4.5) gives: A merger can decrease marginal revenue by decreasing quality, increasing output, or both. The effect a change in quality on marginal revenue is which is positive if quality increases and negative if quality decreases for customers in market y4. If the marginal revenue after the merger is less than the original marginal revenue plus (A4.7), then the merger increases output. This confirms Proposition 2. Before the merger, firm i's first order condition for choosing internal quality includes: p'^":,q'-"' = /' 'G;., . (A4.8) After the merger, firm T 's first order condition for choosing internal quality includes: p^q^-'^phf,q'-'=l''Gy,,. (A4.9) If the internal quality is higher after the merger, then the marginal cost of quality is higher after the merger. Using this and combining (A4.8) and (A4.9) gives: . _ . /I'.i \.A 1,/f , \,B ],B ^ ^i.m „i,m which confirms Corollary 2A.

PAGE 123

114 If quality and marginal production costs are unaffected by the merger, then (A4.7) is zero and the right-hand side of (A4.6) is zero. Because p'^tq''" > 0 , the merger decreases marginal revenue, which means that output increases because quality is constant. This confirms Corollary 2B. Furthermore, increasing the number of markets increases the marginal extra market revenue. This causes marginal revenue to decrease more as a result of a merger, which, because quality is constant, means that output is higher when the pure cross-border merger involves more markets. This confirms Corollary 2C. Proof of Proposition 3: The proof for Proposition 2 applies. Proof of Corollary 3A: This proof first examines how changes in output in markets A and B affect firm 3's output. It then examines the effects on external quality and combines these effects to identify necessary conditions for firm 3's output to increase. From firm 3's first order conditions for optimally choosing output, the effects of changes in output in markets A and B are: I s :? ~3. (A4.11) and the effects of a change in external quality on a firm / include dq'-" 30'2p'Z+p'Z ..q >0, (A4.12)

PAGE 124

115 Before the merger, firm 3's first order conditions for optimally choosing external quality include: i.B „XB . „3.fl „3,g ^1 _ 1-1.3/73 J ^^-^ Su,ana 2,5 Because firm 2's higher output in market B decreases firm 3's price, firm 3's preferred external quality with firm 2 is lower than its preferred quality with firm 1. Firms 1 and 2's preferred external qualities with firm 3 before the merger are defined by: a 3,B 1,B 1,B , 2,B 2.B _ T2.3ri2 From pre-merger bargaining, the pre-merger external qualities are: P>' ' + P',W^-^*P'U-' * % = /'-'fe.. + g;,..). and (A4.13) p'^a'+ fj'.,'' + P:-.'.9'" + P)l.
PAGE 125

116 quantities are constant, lower after the merger than before the merger if ^3,5^3.5 + p^'*,^'-^ > pl^'q''"' . Combining (A4.1 1) through (A4.15) confirms Corollary 3A. Proof of Proposition 4: Consider the pre-merger and post-merger first order conditions for choosing internal quality: pl^r-'' + pU'^-'' + Plf'f^' = ^''Gl,, , and (A4.16) p'^:^''''^p',!':^'''=i'K-'If internal quality is higher after the merger, then the marginal cost of quality is also higher after the merger. Combining (A4.16) and (A4.17) into an inequality that reflects this difference in marginal costs gives (4-9). Now compare first order conditions for pre-merger internal quality with postmerger internal quality. Before the merger, firms 1 and 2 engage in efficient bargaining to determine internal quality. Firm 1 and 2's first order conditions for • choosing external quality are: , and Bargaining gives:

PAGE 126

117 ^ ^^;v^v-^^:^v--V^(^:^^"--^-v--) ,,,,, 2/'-' which defines the agreed upon quaUty in terms of the average marginal cost that the firms incur. Because the firms choose a single quality, the left-hand side of (A4.18) is simply Gj,,,, . After the merger, the merged firm's internal quality choice is defined by (A4.I6). If this quality is higher than the pre-merger external quality, then gItt > G'.3 Combining (A4.16) and (A4.18) in such an inequahty gives (A10). This confirms Proposition 4. Description of Simulation: From Chapter 2, 1 adopt the following linear demand fiinction: ;=1,3 m=A,B.C\j=].3.j*i J I omit fixed costs, assume constant marginal costs, and assume that quality costs follow an exponential fiinction. Before the merger, firm 1 and 3's respective profit fiinctions are: ^' =(p'' -cV'^ -cV'' -gee'" -g4/\and

PAGE 127

118 where c is the constant marginal cost that is common for all firms and is unaffected by the merger, and g is a constant that I take times the antilog of quality to represent the cost of quality. The merged firm's cost function is: r -cY' Af-' -^Y' -^Y' -^12/'. To perform the simulation, 1 express optimized outputs in terms of quality. Solving for optimized qualities is difficult because quality enters the output-optimized profit functions in non-algebraic ways. Therefore, to obtain optimized quality, I graph output-optimized profits against quality choices and find that, for most ranges of output for the parameters given below, optimal quality is equal to 1 for both internal and external quality. For larger sizes of market B and lower levels of v, optimal quality changes from one to zero over a narrow region of market B size and v, so I identify the point where optimal quality is zero and extrapolate between the comer solutions of one and zero. To conduct the simulation, I assume the following values for parameters: f = r*^ = 100 , c = 0.1, and g = 80. I estimate net consumer surplus and profits before and after the merger, assuming ranges from 0 to 200 and v ranges from .01 to 0.3.

PAGE 128

REFERENCES Ai, C, and Sappington, D. E. M. "The Impact of State Incentive Regulation on the US ' Telecommunications Industry." Mimeo, Department of Economics, University of Florida, 1998. Alabama Public Service Commission. Order In the Matter of Generic Proceedings: Consideration of TELRIC Studies, Docket No. 26029, August 25, 1998. Albon, R. "Interconnection Pricing: An Analysis of the Efficient Component Pricing Rule." Telecommunications Policy, Vol. 18 (1994), pp. 414-420. Baumol, W. J., and Sidak, G. "The Pricing of Inputs Sold to Competitors." Yale Journal on Regulation, Vol. 11 (1994a), pp. 171-202. Baumol, W. J., and Sidak, G. Toward Competition in Local Telephony, Cambridge, Massachusetts: MIT Press, 1994b. Baumol, W. J., and Sidak, G. "The Pricing of Inputs Sold to Competitors: Rejoinder and Epilogue." Yale Journal on Regulation, Vol. 12 (1995), pp. 177-186. Bergstrom, T., and Varian, H. R. "When are Nash Equilibria Independent of the Distribution of Agents' Characteristics." Review of Economic Studies, Vol. 52 (1985), pp. 715-718. Berry, S. T. "Estimation of a Model of Entry in the Airline Industry." Econometrica, Vol. 60(1992), pp. 889-917. Blank, L., Kaserman, D. L., and Mayo, J. W. "Dominant Firm Pricing with Competitive Entry and Regulation: The Case of IntraLATA Toll." Journal of Regulatory Economics, Vol. 14 (1998), pp. 35-53. Bresnahan, T. F., and Reiss, P. C. "Entry and Competition in Concentrated Markets." Journal of Political Economy, Vol. 99 (1991), pp. 977-1009. Broadview. 1998. Technology M& A Report: First Half 1998 http://www.broadview.eom/info/na_ma98fh.html#here . Downloaded January 1999. 119

PAGE 129

120 Capron, L., and Mitchell, W. "Outcomes of International Telecommunications Acquisitions: Analysis of Four Cases with Implications for Acquisitions Theory," European Management Journal, Vol. 15 (1997), pp. 237-251. Cremer J., Rey, P., and Tirole, J. "Connectivity in the Commercial Internet." Journal of Industrial Economics, Vol. 48 (2000), pp. 433-472. Crandall, R. After the Breakup: US Telecommunications in a More Competitive Era, Washington, D.C.: Brookings Institution, 1991. Dekimpe, M. G., Parker, P. M., and Sarvary, M. "Staged Estimation of International Diffusion Models: An Application to Global Cellular Telephone Adoption." Technological Forecasting and Social Change, Vol. 57 (1998), pp. 105-132. Dixit, A. "Comparative Statics for Oligopoly." International Economic Review, Vol. 27 (1986), pp. 107-122. Economides, N. "The Incentive for Non-Price Discrimination by an Input Monopolist," International Journal of Industrial Organization, Vol. 16 (1998), pp. 271-284. Economides, N., and White, L. J. "Access and Interconnection Pricing: How Efficient is the 'Efficient Component Pricing Rule?"' Antitrust Bulletin, Vol. 40 (1995), pp. 557-579. Farrell, J., and Shaprio, C. "Asset Ownership and Market Structure in Oligopoly." RAND Journal of Economics, Vol. 21 (1990a), pp. 275-292. Farrell, J., and Shapiro, C. "Horizontal Mergers: An Equilibrium Analysis." American Economic Review, Vol. 80 (1990b), pp. 107-126. Federal Communications Commission. "Local Competition." Report by the Industry Analysis Division, Common Carrier Bureau, 1998. Gabel, R. Development of Separations Principles in the Telephone Industry, East Lansing, Michigan: Institute of Public Utilities, Michigan State University, 1967. Galbi, D., and Keating, C. "Global Communication Alliances: Forms and Characterizations of Emerging Organizations." Mimeo, International Bureau of the Federal Communications Commission, 1996. Gaudet, G., and Salant S. W. "Increasing the Profits of a Subset of Firms in Oligopoly Models with Strategic Substitutes." ^menca/z Economic Review, Vol. 81 (1991), pp. 658-665. Global One. "Global One, International Joint Venture of Deutsche Telekom, France Telecom, and Sprint, Begins Operations." Press Release, January 31, 1996.

PAGE 130

121 Greene, W. H. Econometric Analysis, Upper Saddle River, New Jersey: Prentice Hail, 1997. Greenstein, S., McMaster, S., and Spiller, P. "The Effect of Incentive Regulation on Infrastructure Modernization: Local Exchange Companies' Deployment of Digital Technology." Journal of Economics and Management Strategy, Vol. 4 (1995), pp. 187-236. Harris, R. G., and Kraft, C. J. "Meddling Through: Regulating Local Telephone Competition in the United States." The Journal of Economic Perspectives, Vol. 1 1 (1997), pp. 93-112. Hausman, J. A., Lo, A. W., and MacKinlay, A. C. "An Ordered Probit Analysis of Transaction Stock Prices " Journal of Financial Economics, Vol. 31 (1992), pp. 319-379. Hausman, J. A., and Tardiff, T. J. "Efficient Local Exchange Competition." Antitrust Bulletin, Vol. 40 (1995), pp. 529-556. Jamison, M. A. "A Competitive Framework for Pricing Interconnection in Global Telecommunications Markets." Denver Journal of International Law and Policy, Vol. 23 (1995), pp. 513-533. Jamison, M. A. "Emerging Patterns in Global Telecommunications Mergers and Alliances." Industrial and Corporate Change, Vol. 7 (1998), pp. 695-713. Jamison, M. A. "Business Imperatives." In J. R. Schement, ed.. The New Global Telecommunications Industry & Consumers. University Park, Pennsylvania: Penn State Institute for Information Policy, 1999a. Jamison, M. A. "Does Practice Follow Principle? Applying Real Options Principles To Proxy Costs in US Telecommunications." In J. AUeman and E. Noam, eds.. Real Options: The New Investment Theory and its Implications for Telecommunications Economics, Boston: Kluwer Academic Publishers, 1999b. Jamison, Mark A. " Impacts of Mergers and Global Rivalry on Local Network Competition." Mimeo, Department of Economics, University of Florida, 1999c. Jasinski, P. "Competition Rules and Regulations in Telecommunications: The Case of Poland's Intent to Join the EU." In D. J. Ryan, ed., Privatization and Competition in Telecommunications, Westport, Connecticut: Praeger, 1997. Kahn, A. E., and Taylor, W. E. "The Pricing of Inputs Sold to Competitors: A Comment." Yale Journal on Regulation, Vol. 1 1 (1994), pp. 225-240. Katz, M. L., and Shapiro, C. "Network Externalities, Competition, and Compatibility. American Economic Review, Vol. 75 (1985), pp. 424-440.

PAGE 131

122 Katz, M. L., and Shapiro, C. "Systems Competition and Network Effects." Journal of Economic Perspectives, Vol. 8 (1994), pp. 93-1 15. Kende, M. "The Digital Handshake: Connecting Internet Backbones." Mimeo, Office of Plans and Policy, Federal Communications Commission, 2000. Kennard, W. E. Letter to Senators John McCain and Sam Brownback. http://www.fcc.gov/Bureaus/Cotnmon_Carrier/in -region applications/. March 20, 1998. Downloaded March 10, 1999. Kennard, W. E. Letter to Mr. Richard C. Notebaert, Chairman and Chief Executive Officer of Ameritech Corporation, and to Mr. Edward E. Whitacre, Jr., Chairman and Chief Executive Officer of SBC Communications, Inc., April 1, 1999. Laffont, J., and Tirole, J. "Access Pricing and Competition." European Economic Review, Vol. 38 (1994), pp. 1673-1710. Laffont, J., and Tirole, J. "Creating Competition Through Interconnection: Theory and Practice." Journal of Regulatory Economics, Vol. 10 (1996), pp. 227256. Laffont, J., Rey, P., and Tirole, J. "Network Competition: I. Overview and Nondiscriminatory Pricing." The Rand Journal of Economics, Vol. 29 (1998a), pp. 1-37. Laffont, J., Rey, P., and Tirole, J. "Network Competition: II. Price Discrimination." The Rand Journal of Economics, Vol. 29 (1998b), pp. 38-56. Larson, A. C. "Wholesale Pricing and the Telecommunications Act of 1996: Guidelines for Compliance with the Avoided Cost Rule." University of Florida Journal of Law and Public Policy, Vol. 8 (1997), pp. 243-260. Larson, A. C, and Parsons, S. G. "An Economic Analysis of Transfer Pricing and Imputation Policies for Public Utilities," In M. A. Crew, ed. Incentive Regulation for Public Utilities, Boston: Kluwer Academic Publishers, 1994. Liebowitz, S. J., and Margolis, S. E. "Network Externality: An Uncommon Tragedy." Journal of Economic Perspectives, Vol. 8 (1994), pp. 133-150. MacAvoy, P. W. "Regulation's Role in Creating Competition in Telephone Service markets: The Experience of the California Public Utilities Commission." University of Florida Journal of Law and Public Policy, Vol. 8 ( 1 997), pp. 151200. MacAvoy, P. W. "Testing for Competitiveness of Markets for Long Distance Telephone Services: Competition Finally?" Review of Industrial Organization, Vol. 13 (1998), pp. 295-319.

PAGE 132

123 Mandy, D. M. "TELRIC Pricing with Vintage Capital." Mimeo, Department of Economics, University of Missouri, 2000. Mandy, D. M., and Sappington, D. E. M. "Incentives for Sabotage in Vertically-Related Industries." Mimeo, Department of Economics, University of Missouri, 2000. McDavid, J. L. "Antitrust Issues in Communications Industries," In J. Sims and P. L. Verveer, eds., Competitive Policy in Communications Industries: New Antitrust Approaches. Little Falls, New Jersey: Glasser LegalWorks, 1997. Mitchell, B., Neu, W., Neumann, K., and Vogelsang, I. "The Regulation of Pricing of Interconnection Services." In G. W. Brock, ed.. Toward a Competitive Telecommunications Industry: Selected Papers from the 1994 Telecommunications Policy Research Conference. Hillsdale, New Jersey: Lawrence Erlbaum Associates, 1995. National Regulatory Research Institute. Interconnection Arbitration Decisions, , downloaded September December 1998. Olbeter, E. R., and Robison, M. Breaking the Backbone: The Impact of Regulation on Internet Infrastructure Deployment, Mimeo, iAdvance, 1999. Ordover, J. A., Sykes, A. O., and Willig, R. D. "Nonprice Anticompetitive Behavior by Dominant Firms toward the Producers of Complementary Products." In F. M. Fisher, ed.. Antitrust and Regulation: Essays in Memory of John J. McGowan, Cambridge, Massachusetts: MIT Press, 1985. Ros, A. J. "Does Ownership or Competition Matter More? The Effects of Telecommunications Reform on Network Expansion and Efficiency." Journal of Regulatory Economics, Vol. 15 (1999), pp. 65-92. Ros, A. J., and McDermott , K. "Are Residential Local Exchange Prices Too Low? Drivers of Competition in the Local Exchange Market and the Impact of Inefficiently-Set Prices." In M. Crew, ed.. Expanding Competition in Regulated Industries, Boston: Kluwer Academic Press, 2000. Rosentberg, E. A., and Wilhelm, J. D. State Universal Service Funding and Policy: A Overview and Survey. Columbus, Ohio: National Regulatory Research Institute, 1998. Salant, S. W., and Shaffer, G. "Unequal Treatment of Identical Agents in Coumot Equilihrium." American Economic Review, Vol. 89 (1999), pp. 585-603. Salant, S. W., Switzer, S., and Reynolds, R. J. "Losses from Horizontal Merger: The Effects of an Exogenous Change in Industry Structure on Coumot-Nash Equilibrium." The Quarterly Journal of Economics, Vol. 98 (1983), pp. 185-199.

PAGE 133

124 Sappington, D. E. M., and Weisman, D. L. Designing Incentive Regulation for the Telecommunications Industry, Cambridge, Massachusetts: MIT Press, 1996. Shapiro, C. "Competition Policy in the Information Economy." In E. Hope, ed., Competition Policy Analysis, London: Routledge Studies in the Modem World Economy, 2000. Spiller, P. T., and Cardilli, C. G. "The Frontier of Telecommunications Deregulation: Small Countries Leading the Pack." Journal of Economic Perspectives, Vol. 1 1 (1997), pp. 127-138. Tomlinson, R. G. "The Impact of Local Competition on Network Quality." In W. Lehr, ed.. Quality and Reliability in Telecommunications Infrastructure, Mahwah, New Jersey: Lawrence Erlbaum, 1995. Tomlinson, R. G. Tele-Revolution: Telephone Competition at the Speed of Light, Camden, Maine: Penobscot Press, 2000. Tye, W. B. "The Pricing of Inputs Sold to Competitors: A Response." Yale Journal on Regulation, Vol. 1 1 (1994), pp. 203-224. Tye, W. B., and Lapuerta, C. "The Economics of Pricing Network Interconnection: Theory and Application to the Market for Telecommunications in New Zealand." Yale Journal on Regulation, Vol. 13 (1996), pp. 419-500. United States Telephone Association. Report to House Commerce Committee Chairman Thomas Bliley, 1998. US Department of Commerce and US Department of Agriculture. Advanced Telecommunications in Rural America: The Challenge of Bringing Broadband Services to All Americans, Washington, D.C., 2000. Weisman, D. L. "Regulation and the Vertically Integrated Firm: The Case of RBOC Entry into InterLATA Long Distance." Journal of Regulatory Economics, Vol. 8 (1995), pp. 249-266. Weisman, D. L. "Footprints in Cyberspace: Toward a Theory of Mergers in Network Industries." info. Vol. 1 (1999), pp. 367-369. Weisman, D. L. "The (In)Efficiency of the 'Efficient Firm' Cost Standard." The Antitrust Bulletin, Vol. 45 (2000), pp. 195-21 1. Woroch, G. A. "Competition's Effect on Investment in Digital Infrastructure." Mimeo, Department of Economics, University of California at Berkley, 2000. X-Change. "Status of State Local Competition Proceedings." X-Change, 1998-1999.

PAGE 134

BIOGRAPHICAL SKETCH Mark Jamison is the Director of Telecommunications Studies for the Public Utilities Research Center at the University of Florida. His responsibilities include developing training programs and conducting research on telecommunications issues. The PURC training program has trained over 800 infrastructure executives and government officials from 100 countries. He is also Associate Director for Business and Economic Studies for the Center for International Business Education and Research, a Research Associate with the Center for Public Policy Research, the Special Academic Advisor to the Chair of the Florida Governor's Internet task force, and an Affiliated Scholar with the Communications Media Center at New York Law School. From February 1993 through June 1996, he was a manager of regulatory policy for Sprint where he developed policies on pricing, costing, and market structure issues. Prior to joining Sprint, he worked nine years for state regulatory commission staffs in the US. From February 1987 through February 1993, he was with the Iowa Utilities Board as a telecommunications analyst and finally as Head of Research. During this fime, he served as chairperson of the National Association of Regulatory Utility Commissioners (NARUC) Staff Subcommittee on Communications, as chairperson of the state staff for the Federal/State Joint Conference on Open Network Architecture, and as a member of the state staff for the Federal/State Joint Board on Separations. From June 1984 through February 1987, he was the Communicafions Economist for the Kansas Corporation 125

PAGE 135

126 Commission. Mr. Jamison has served on the faculty of NARUC Annual Regulatory Studies Program. His current research covers the globalization of telecommunications, convergence of information industries, and the international development of telecommunications competition.

PAGE 136

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. DAvdd^ M./Saj^itfgton, Chairperson LaBzilotti-McKethan Eminent Scholar of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Steven M. Slutksy Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Chunrong Ai Associate Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fiilly adequate, in scope and quality, as a dissertation for the degree of Doctor of Ph^'osophy. Heathar Elms Assistant Professor of Management This dissertation was submitted to the Graduate Faculty of the Department of Economics in the College of Business Administration and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. May 2001 Dean, Graduate School