Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Topics on the economics of international telecommunications Rieck, Olaf 2000

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2001-611647.pdf [ 6.39MB ]
Metadata
JSON: 831-1.0099675.json
JSON-LD: 831-1.0099675-ld.json
RDF/XML (Pretty): 831-1.0099675-rdf.xml
RDF/JSON: 831-1.0099675-rdf.json
Turtle: 831-1.0099675-turtle.txt
N-Triples: 831-1.0099675-rdf-ntriples.txt
Original Record: 831-1.0099675-source.json
Full Text
831-1.0099675-fulltext.txt
Citation
831-1.0099675.ris

Full Text

T O P I C S O N T H E E C O N O M I C S O F I N T E R N A T I O N A L T E L E C O M M U N I C A T I O N S By O L A F RIECK Dipl. Wi-Ing (Industrial Engineering), Universitat Karlsruhe (TH) M.A. (Economics), University of British Columbia A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R O F PHILOSOPHY in T H E F A C U L T Y OF G R A D U A T E STUDIES T H E F A C U L T Y OF C O M M E R C E AND BUSINESS ADMINISTRATION We accept this thesis as conforming to the required standard T H E UNIVERSITY^OF BRITISH C O L U M B I A November 2000 © O L A F RIECK, 2000 In presenting this thesis in partial fulfilment of the requirements for an ad-vanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Operations and Logistics Division, Strategy and Business Economics Division, Faculty of Commerce and Business Administration The University of British Columbia 2053 Main Mall Vancouver, Canada V6T 1Z2 Date: 2 7 . It. 2.0*0 Abstract This thesis comprises three essays of the economics of international telecom-munications. The first of the essays is concerned with the modeling of international telephone demand. Standard models of international telecommunications de-mand have exclusively focused on modeling the total demand for telephone minutes and ignored call durations as a separate dimension of calling behav-ior. Yet, average call durations and total volume of telephone traffic have developed in opposite directions over the past decade, suggesting that it is important to clarify the relationship between call duration and total traffic volume. This essay extends traditional telephone demand analysis to show how to handle call durations in the Point-to-Point telephone demand model, and presents estimation results using data of telephone traffic between the US and 42 countries. The remainder of this thesis analyzes the regulatory history of the Pro-portionate Return Requirement (PRR) and its effect on the US telephone markets and industry. The second essay is an institutional analysis of the P R R in the context of United States' regulation of international telephony. The main conclusion is that the actual effect of the P R R is not in line with its regulatory objectives, and that the P R R is not needed to prevent "whipsawing" of US carriers by foreign monopolists. The FCC's justifications of the P R R are found to be largely based on an erroneous interpretation of historic case law. The third essay, finally, attempts to provide an insight into the nature of economic distortions created by the PRR. It develops a competition model of an international telecommunications market with two types of carriers, facilities-based and resale carriers. The model is used to analyze and simu-late the effect of the proportionate return requirement (PRR) on the existence and properties of equilibria in the country where the P R R is implemented. In addition, this model analyzes the discriminatory effect of the P R R on re-sellers, and the extent to which facilities-based carriers can use the advantage of receiving incoming traffic to increase their market share at the expense of resellers in the end-user market. i i Contents Abstract ii Contents iii List of Figures vi List of Tables vii Acknowledgements ix Dedications x 1 Introduction 1 2 Call Lengths and International Telecommunications Demand 4 2.1 Introduction 4 2.2 Trends in International Telecommunications and Related Lit-erature 7 2.2.1 Focus and Limitations of International Telephone De-mand Analysis 7 2.2.2 Determinants of Call Duration 13 2.3 Theoretical Framework 19 2.3.1 The Classical International Telecommunications De-mand Model: Point-to-Point Demand without Connec-tion Charges 20 2.3.2 Point-to-Point Demand with Connection Charges . . . 22 2.3.3 The Case of Exogenous Call Duration 24 2.3.4 The Case of Endogenous Call Duration 26 2.3.5 Price Elasticities of Call Duration 28 in 2.4 Empirical Model, Estimation, and Results 32 2.4.1 Standard Empirical Specifications 32 2.4.2 Models with Exogenous Call Duration 42 2.4.3 Models with Endogenous Call Duration 46 2.4.4 Discussion 50 2.5 Summary and Conclusion 56 3 The Proportionate Return Requirement: Regulatory Overkill in International Telecommunications 58 3.1 Introduction 58 3.2 Institutional Characteristics and Regulation of International Interconnection 61 3.3 The Regulatory History of the Proportionate Return Require-ment 65 3.3.1 The Role of Proportionate Return in the Early Case Law 66 3.3.2 Proportionate Return and the Uniform Settlements Pol-icy 69 3.3.3 The Sprint and T L D cases 71 3.3.4 The Codification ofthe Proportionate Return Require-ment 73 3.3.5 Conclusion 75 3.4 The Proportionate Return Requirement and the Emergence of Competition : • 76 3.4.1 Competitive and Regulatory Challenges to the Account-ing Rate System 76 3.4.2 Regulatory Adjustments of the ISP 78 3.4.3 The Current and Future Regulatory Impact of the ISP 83 3.5 The Economic Effect of Proportionate Return in Switched Ser-vice Markets 85 3.6 Summary and Conclusion 88 4 The Economic Effect of the Proportionate Return Require-ment 91 4.1 Introduction 91 4.2 A Model of International Switched Service Competition under Proportionate Return 96 4.3 Home Market Equilibrium under Proportionate Return . . . . 104 4.3.1 An Equilibrium Characterization in the General Case . 104 iv 4.3.2 Stage Two Equilibrium Output 105 4.3.3 Stage One equilibrium wholesale rates 109 4.3.4 Market Equilibrium without P R R 113 4.3.5 An illustrative example: c f b = c r 115 4.3.6 Discussion 119 4.4 Simulating Equilibria under Differential Cost 121 4.4.1 Simulation inputs and assumptions 121 4.4.2 Simulated Equilibria 123 4.4.3 Simulating the effect of the P R R on welfare and market structure 127 4.4.4 Simulated comparative statics 130 4.4.5 Discussion 132 4.5 Policy Implications, Summary and Conclusions 133 Bib l iography 136 A p p e n d i x 146 A D a t a Descr ip t ion for Chapter 2 146 B Glossary of Ac ronyms for Chapter 3 and Chapter 4 150 List of Figures 2.1 Ratios of outgoing over incoming total traffic volume versus outgoing over incoming average call duration on routes be-tween the United States and selected O E C D countries, 1995 . 12 2.2 A T & T Collection charges and connection charges for tele-phone calls from the United States to The Netherlands, 1988-1996 16 2.3 The nonlinear budget constraint 24 4.1 International telecommunications relation between a compet-itive market and a monopoly market 97 4.2 Stage two optimal profit schedules contingent of the stage one wholesale price. Case I: Unique equilibrium 112 4.3 Stage two optimal profit schedules contingent of the stage one wholesale price. Case II: No equilibria 113 4.4 Stage two optimal profit schedules contingent of the stage one wholesale price. Case III: Multitple equilibria with resale. . . 114 4.5 Stage two optimal profit schedules contingent ofthe stage one wholesale price. Illustrative example of = cr 117 4.6 Simulation results: The effect of settlement rate changes on fb-carriers' profits and market shares, for A c = 0.12 US$ . . . 131 v i List o f Tables 2.1 Average annual traffic growth 1988-1996 of telephone traffic between the United States and selected countries, outgoing from and incoming into the U.S 8 2.2 Average annual change of average call durations 1988-1996 of telephone traffic between the United States and selected countries, outgoing from and incoming into the U.S 10 2.3 Average call durations of telephone traffic (in minutes) be-tween the United States and selected countries in 1995, out-going from and incoming into the U.S 11 2.4 Price and Income elasticities of international telephone de-mand as reported by selected studies 36 2.5 Estimation results for standard model of international tele-phone demand for traffic originating in the United States, us-ing 1988-96 data 38 2.6 Estimation results for standard model of international tele-phone demand for traffic with destination United States, using 1988-96 data 41 2.7 Estimation results for model of international telephone de-mand for calls originating in the United States, with exoge-nous call durations, using 1988-96 data 45 2.8 Estimation results for model of international telephone de-mand for calls originating in the United States, with endoge-nous call durations, using 1988-96 data 49 4.1 Simulation results: Existence and characterization of equi-libria in the case of four exemplary international routes, for A c = 0.12 US$ 123 vii 4.2 Simulation results: Existence and type of equilibria for differ-ent values of A c in the case of four exemplary international routes 126 4.3 Simulation results: Characterization of equilibria without P R R for three exemplary international routes, for A c = 0.12 US$ . 127 4.4 Simulation results: Welfare comparisons to equilibria under P R R in the case of four exemplary international routes, for A c = 0.12 US$ . . . 128 4.5 Simulation results: Fb-carriers profits, outputs, and market share for varying numbers of fb-carriers in the market based on data from the US-Chile route 130 viii Acknowledgements I would like to thank my two mentors, Professor Tae Hoon Oum, who is also my research advisor, and Professor Thomas Ross. Both of them gave me a great deal of advise, support, and motivation throughout the process of writing this thesis. I would also like to thank Professor Ken White for his excellent comments and Professor Bi l l Stanbury for all the spirited and insightful discussions. Thanks go to my fellow students and friends, Yuval Deutsch, Lorenzo Garlappi, Jon-Hun Park, Joshua Slive, Don Wagner, Darius Walzak, and Christoph Zott, for their advice and help, and in particular to the staff of the T^X-hotline. Acknowledgement goes to the financial support of the C N fellowship and the University Graduate Fellowship (UGF) of the University of British Columbia. My most special thanks go to Cindy, who supported me with her encour-agement, understanding, and care during a time when she was also working on her dissertation. ix To my parents, Wanda and Eckhard Rieck x Chapter 1 Introduction During the past 25 years, telecommunications has turned from a heavily regulated essential services industry with a public utility character to one of the fastest growing and most radically changing industries. It has become a key to industrial change and is forever altering the ways people interact and conduct business. The international segment of this industry, with its traditionally high profit margins and heavy use of cutting-edge technology, has been at the forefront of this development. As technology and industry structure are rapidly changing, so are the tasks and objectives of applied economics and business research on the telecom-munications industry and its regulation. Although the findings of much of the current research may be of little interest 20 years from now, they help to guide future ways to conduct business, facilitate a smooth transition towards a competitive business environment, and advance economic methodology for future telecommunications research. This dissertation is comprised of three essays that address issues of such 1 current and future relevance to businesses, regulators, and researchers of the international telecommunications industry. Chapter 2 is a study of telecom-munications demand, which extends the traditional analysis of demand for total telephone minutes to analyzing the structure of telephone demand, that is, the tradeoff between the number of telephone calls and their duration. In particular, it develops an empirical model to explore how this demand structure has adjusted due to technological change, and changes of telephone carriers' calling charge schedules. The main theme of both Chapters 3 and 4 relates to the Proportionate Return Requirement (PRR), a regulation that governs international tele-phone traffic on a significant number of international routes, typically be-tween countries with a competitive telecom market and countries with tradi-tional telecom monopolies. The analysis of this regulation is conducted from two different angles. Chapter 3 traces the historical and institutional background of the P R R . It shows that the regulatory process leading to the implementation of this important regulation is inconsistent, and sheds some serious doubts on its appropriateness for achieving the regulators' objectives. In addition, this chapter reveals the dangers and difficulties of the regulatory rulemaking pro-cess by which international telecommunications is currently regulated, and which may also be applied for future regulation of advanced telecommunica-tion services. Finally, Chapter 4 develops a competition model to analyze the impact of 2 the P R R on the telecommunications market structure in the country where it is implemented. It establishes, analytically and by simulations, how the P R R affects the profits and market shares of telecommunications carriers and resale carriers, and the welfare of the telephone end user. The chapter concludes with some policy recommendations regarding the future regulatory framework for international telephone services. 3 Chapter 2 Call Lengths and International Telecommunications Demand 2.1 Introduction In the literature on telecommunications demand analysis, demand for inter-national telephone services has received relatively little attention. One of the reasons is that international telephone service was often considered as a spe-cial type of long distance service, with essentially the same features. Only as late as at the mid-1980s have researchers started to develop theoretical and empirical frameworks that specifically captured the unique characteristics of international telephony. One aspect in which international telephone markets differ from domestic long distance markets is that international outgoing traffic and the respective return traffic may be subject to significantly different market conditions. For example, in 1988 the calling charge for a call from Turkey to the United States was more than twice the charge applicable in the opposite direction. 4 At the same time, Turkey's per-capita G D P amounted only to 8 percent of the American GDP. Asymmetric market conditions are more likely to be prevalent in international than in domestic long distance markets. They typically result in imbalanced traffic patterns and give rise to arbitrage, both of which significantly complicate demand analysis. Second, international telecommunications has been a starting point and prime target for competitive service providers and new types of telephone ser-vice concepts, including Internet Telephony, callback, International Simple Resale, Pure ("switchless") resale, corporate private networks, and refiling. As a result, collection charges1 have declined more rapidly in international than in domestic long distance markets, and the average growth of interna-tional telephone traffic has by far exceeded that of domestic traffic. A third distinctive feature of international telephone markets is the extent to which carriers levy call connection charges. In the United States, connec-tion charges were traditionally applied to all long distance calls, including domestic and international calls. They were phased out for domestic long distance calls in 1988, but remained in effect in international call markets throughout most of the 1990s, although in gradually decreasing magnitude. By 1998, most facilities-based carriers had phased out also the call connec-tion charges on international calls, however, significant connection charges remain in effect with US international resellers, as well as with.a number of 1 The term "collection charge" is used throughout this paper for the per-minute price of a telephone call. 5 non-US carriers. Finally, calling patterns in international telephone markets have conspic-uously shifted over the past decade, from longer average duration of calls to shorter average call duration. A similar shift is observable also in US domes-tic long distance demand; however, it appears by far not as pronounced as in the international call market. Moreover, average call durations significantly vary between different inter-country routes. While previous empirical research on international telecommunications demand has developed methods to investigate the implications of asymmetric market conditions on international call demand, it has so far neglected the joint analysis average call duration and total calling volume. Relatedly, there has been no research that focused on measuring the effect of call set-up charges on telephone demand and calling patterns. The objective of this research is therefore to explain the variations of calling patterns and to analyze the role of call durations and call set-up charges in telephone demand analysis. An investigation of international telephone demand along these new di-mensions would be useful for telecommunications policy analysis, as it the-oretically improves and refines previous empirical demand models and thus allows for a more accurate assessment of the effect of pricing policies on use, capacity, revenues and total welfare. This paper will contribute to such im-provement by extending the traditional telephone demand model to incorpo-rate the aspect of varying call durations and set-up charges. Section 2.2 fur-6 ther motivates this research by providing background information on trends in international telephone calling patterns and reviews the related academic literature. Section 2.3 develops two alternative demand models that both explicitly take call durations into account, and clarifies the assumptions un-der which these models are valid. Section 2.4 presents the empirical results of estimating both demand specifications with and contrasts them with pre-vious studies. Section 2.5 summarizes and concludes. A detailed description of data and data sources is included in Appendix A. 2 .2 Trends in International Telecommunica-tions and Related Literature 2.2.1 Focus and Limitations of International Telephone Demand Analysis Throughout the past decades, international telecommunications has arguably been one of the fastest growing market segments of the telecommunications industry. It has provided the essential infrastructure for international trade, tourism and migration, It has driven and been driven by globalization of businesses, technological and regulatory change. Yet, relatively little research has been conducted to understand the struc-ture of international telecommunications demand. Early research in this area, conducted by Lago (1970), and Rea and Lage (1978), simultaneously analyzed the demand for telephone service together with two of its closet 7 Table 2.1. Average annual traffic growth 1988-1996 of telephone traffic be-tween the United States and selected countries, outgoing from and incoming into the U.S. country from US into US country from US into US Australia 19.72% 12.46% India 37.64% 10.13% Chile 43.51% 34.95% Ireland 13.68% 14.98% Egypt 18.65% 11.03% Japan 15.04% 13.31% France 14.99% 8.66% Netherlands 18.74% 10.07% Germany 10.08% 8.36% New Zealand 20.80% 14.35% Hong Kong 31.88% 9.09% Spain 15.00% 14.29% Italy 12.79% 8.90% Turkey 14.88% 17.89% Iceland 6.59% 10.64% U.K. 11.77% 8.55% Source: Compiled from Statistics of Communications Common Carriers substitutes, telegram and telex services. Besides collection charges of these services, they found a number of statistically significant determinants of tele-phone demand, including household income, trade and number of main lines. Later studies, including Larson, Lehman and Weisman (1986), Appelbe et al. (1988), Acton and Vogelsang (1992), Garin-Muhoz and Perez-Amaral (1996, 1998, 1999), develop and use point-to-point models of telecommunications demand in order to include reciprocal calling effects as important determi-nants of demand. Acton and Vogelsang (1992) focus on call externalities and investigate whether incoming traffic is a substitute or a complement for outgoing traffic. Hackl and Westlund (1995), finally, allow for time-varying regression coefficients to estimate trajectories of price elasticities over time. At the center of attention of all previous studies is the aggregated total 8 demand for international telephone service2 in a given country. The total number of telephone minutes as a quantity measure of telephone demand has for obvious reasons been strictly preferred to another readily available quantity measure, that is, the total number of telephone calls. After all, the total number of call minutes appears to be a more instructive measure of information generated and transmitted during on any inter country relation than the actual number of calls made. Yet, simply ignoring the total number of calls as an output measure may result in the loss of valuable information. During the period 1988 to 1996, the total number of telephone calls has increased at a higher rate than the total number of call minutes on most international routes. In other words, while the total number of call minutes increased on all routes in this period of time (Table 2.1), the average call duration decreased significantly on most routes (Table 2.2). Comparing the average percentage change of outgoing traffic versus in-coming traffic, it appears that the total amount of US outbound traffic has grown faster than the inbound traffic on most routes. In the case of average call durations, there is no such clearly recognizable pattern. Average call durations vary widely between different countries. Among the routes displayed in Table 2.3, average call durations range from 2.76 minutes on the inbound route from India, to 8.34 minutes on the inbound route from Iceland. In the outbound direction, the shortest duration in the 2Expressed as the total number of call minutes. 9 Table 2.2. Average annual change of average call durations 1988-1996 of telephone traffic between the United States and selected countries, outgoing from and incoming into the U.S. country from US into US country from US into US Australia -2.80% 1.16% India -0.59% -6.67% Chile -2.84% -5.95% Ireland -3.75% -7.49% Egypt -5.13% -7.07% Japan -3.58% 0.58% France -4.62% -6.28% Netherlands -3.98% -4.25% Germany -5.56% -0.72% New Zealand -4.98% -1.78% Hong Kong -3.07% -1.71% Spain -4.65% -3.66% Italy -4.23% -3.36% Turkey -5.54% -1.49% Iceland -7.27% -4.49% U.K. -3.20% -1.71% Source: Compiled from Common Carriers sample is 4.47 minutes on the route to Hong Kong, the longest is 7.72 minutes on the route to Ireland. On most, but not all routes, the average telephone call in the outbound direction is longer than the average call in the inbound direction. The difference between outbound call durations and inbound call durations tends to be greater on routes to the less developed countries than on those to developed countries. Finally, focusing only on routes between the United States and industri-alized countries, Figure 2.1 shows a plot of relative (outbound over inbound) total telephone minutes (outbound over inbound) against relative call du-rations (outbound over inbound). The traffic volume ratios turn out to be greater than the duration ratios on any given route of the sample. However, there is otherwise again no pattern indicating a strong correlation between the ratios of total traffic volumes and the ratios of average call durations. 10 Table 2.3. Average call durations of telephone traffic (in minutes) between the United States and selected countries in 1995, outgoing from and incoming into the U.S. country from US into US country from US into US Australia 5.25 4.63 India 6.84 2.76 Chile 7.07 3.59 Ireland 7.72 4.52 Egypt 7.33 3.72 Japan 5.77 4.89 France 5.14 6.35 Netherlands 4.87 3.61 Germany 6.09 4.04 New Zealand 4.91 4.14 Hong Kong 4.47 3.72 Spain 5.78 7.31 Italy 5.59 3.56 Turkey 5.79 3.19 Iceland 6.58 8.34 U.K . 5.37 5.52 Source: Compiled from Statistics of Communications Common Carriers For example, an approximately balanced traffic volume in outbound and in-bound direction of a given route does not imply that the average duration of telephone calls in both directions will be approximately equal. Rather, calls may on average be longer, equal or shorter on the outbound versus the inbound direction. The traditional analysis of international telecommunications demand, with its focus on total traffic volumes as the main object of analysis, has as of today not made use of information that allow to gain deeper insights into the structure of international telephone traffic and its characteristics. The evidence suggests that determinants of international telecommunications de-mand do not affect average call durations in the same way as they effect total traffic volume. Rather, the relationship between these two output mea-sures of telephone service is far from obvious. A closer investigation into the 11 Figure 2.1. Ratios of outgoing over incoming total traffic volume versus outgoing over incoming av-erage call duration on routes between the United States and selected O E C D countries, 1995 ) Portugal • ) Turkey Ireland^ ^ Switzer-Nether- Italy land l a ^ d s ^ • -y \ Au „ New \ \ Germany Zealand \ itralm \ • ^ B e l g i u m \ ^ # 0 Greece 0 Japan ^ weden ) Icel • U K . 4 . Norway Austria • • ^Finland # tnd • France Spain ^ ^Denmark Luxemburg ) i 1 1 1 l 0.5 1.0 1.5 ^ 2.0 2.5 3.0 3.5 Note: 7=average duration of U.S. outgoing calls/average duration of U.S. incoming calls; //=total minutes of U.S. outgoing traffic/total min-utes of U.S. incoming traffic Source: Compiled from Statistics of Communica-tions Common Carriers structure of telecommunications demand could be valuable for a number of reasons. First, it may provide a deeper understanding of the mechanisms that drive telecommunications demand. Second, to the extent that the initiation of each telephone call utilizes network capacity, telephone carriers should be interested in learning more about how to forecast both total traffic volume and the number of calls. Third, it needs to be investigated whether including 12 additional information into the econometric demand model will improve the efficiency of estimators, and thus improve the quality of demand models. 2.2.2 Determinants of Call Duration Previous studies of aggregate international telephone demand have been rel-atively consistent in their selection of relevant determinants of demand, such as price, number of main lines, population, disposable income, trade, export and imports, tourism, or migration. Consequently, the effect of these deter-minants on international telephone traffic is empirically widely understood. In the context of analyzing both average call durations and total traffic volume, however, it is a priori unclear how these determinants can affect call durations and total traffic differently, such that an increase of one of the determinants over time would result in a shift of total traffic volumes and call durations in different directions, or such that differences of these determinants across countries affect volumes and durations in a different way. For example, a decrease in collection charges or increasing international trade flows should have a positive effect on both average call duration and total traffic volume. The underlying reasons for the observed traffic patterns may be found in a combination of technological and economic developments that have tradi-tionally not been accounted for in studies of international telecommunica-tions demand. Factors that may effect call durations and total traffic volume in opposite directions include certain technological developments in termi-13 nal equipment, telecommunications service innovation, decreasing connec-tion charges relative to marginal 'per minute' calling charges, calling behav-iors that try to capture arbitrage opportunities, telecommunications network quality, and changes of communications patterns due to cultural factors. The development of terminal equipment and telecommunications service innovation. Technological change in the terminal equipment market and the development of innovative telecommunications services may contribute to the reduction of call durations in a number of ways. Telefax, for example, allows transmitting complex text and graphical information via telephone lines. The transmission of an equivalent amount of information by means of conversation would arguably take significantly more time. Moreover, telefax transmissions are often preceded by an additional brief telephone conver-sation. Hence, telephone users can transmit the amount of information of a long telephone call by making two shorter calls: one telephone conversa-tion and one telefax transmission. Similarly, electronic mail is often a more efficient means of communicating factual complex information than verbal telephone conversation. To the extent that the communicating parties sup-plement their telephone conversations by exchanging electronic documents, e-mail is likely to be a contributing factor to an overall reduction of average call durations. Another possible explanation for decreasing call durations is the growing use of corporate private communications networks. Multinational corpora-14 tions typically establish private communications networks on routes with high data and telephone traffic volumes. To the extent that the duration of in-ternational business calls and data transmission exceed the average duration of international telephone calls, the transfer of corporate communications to private networks will effectively result in a reduction of average call durations on public telephone networks.3 On the call-receiving end of an inter-country relation, the use of voice mail, answering machines, and call waiting services may have an impact on call durations. Without voice mail, the attempt to establish a telephone connection will result in an uncompleted and unmeasured call, if the call recipient is away from the phone. With voice mail, it results in a telephone call that lasts only for the duration of voice box "greeting", plus the time it takes the caller to leave a voice mail. The call recipient may then return the call, or the caller will start another attempt at a later point in time. Similarly, a call waiting service at the call-receiving end of a connection eliminates the possibility of getting a busy signal. The call recipient is likely to answer an incoming call "on the other line", but only to agree with the caller upon who is calling back, and when. Hence, a higher penetration of voice mail and call waiting services will result in more and on average shorter telephone calls. Finally, service innovations such as call forwarding an cellular phones will increase the availability of call recipients, but may also result in shorter 3 A s imi lar effect is due to the use of resale carr iers by mu l t ina t iona l corporat ions. Note however, that the d a t a sample used in this s tudy is restr ic ted to the per iod 1988 to 1996. D u r i n g this per iod , resale services were proh ib i ted on most in te rnat iona l routes. 1 5 calls, since the call recipient is more likely to be in a situation where it is inconvenient to hold a long conversation. Moreover, even if mobile telephones are unattended, the mobile phone is likely to be attached to a voice mail or call forwarding system. Call connection charges. Some international telephone carriers, including carriers in the United States, have traditionally applied a pricing structure with a fixed connection charge per call, and a marginal charge per minute of call duration. Figure 2.2 shows the development of AT&T ' s average peak load connection charge and marginal per minute charge for calls from the United States to The Netherlands. It represents a typical trend for international Figure 2.2. A T & T Collection charges and connection charges for telephone calls from the United States to The Netherlands, 1988-1996 1.6 1.4 1.2 ,1.0 ' 0.8 0.6 0.4 0.2 _Qz -9- collection charge - O connection charge 1988 1989 1990 1991 1992 1993 1994 1995 1996 Source: Compiled from Statistics of Communications Common Carriers connection charges of US carriers throughout the late 1980s and 1990s. In 1988, connection charges were commonly set close to the level of the marginal 16 charges. In subsequent years, A T & T gradually reduced connection charges relative to marginal charges, and phased them out in 1997. Other major carriers, such as MCI and Sprint, adopted similar pricing policies. By reduc-ing connection charges relative to marginal charges, carriers provide the end users with economic incentives to change their calling behavior4. Arbitrage and callback. Differences in international telephone charges invite arbitrage and therefore influence the communication traffic patterns on international routes. Arbitrage opportunities, arising between high charge and low charge countries are commonly exploited in three ways. The party in the country with the higher calling charge can initiate the connection by making a short "call me back" call, and has the other party return the call. Second, the telephone call can be initiated at the end of the last call, through a mutual agreement about the time to call, and about what party is next to call. Third, the call initiating party can involve a callback carrier5. A l l three methods of arbitrage will have an influence on the structure of telephone traffic patterns, i.e., the duration and frequency of incoming and outgoing calls 6. 4For example, if price connection charges are lowered and marginal charges are increased such that the average charge remains constant, then one would expect the end users to reduce call duration and increase number of calls they make. 5There are a number of ways in which a call involving a callback carrier can be initiated. The most common form requires the calling person to dial the number of the operator and hang up after a few rings, without the call having been answered. The callback carrier will then initiate a call in the reverse direction, after which it establishes the connection to the actual call recipient. 6See Acton and Vogelsang (1992) for a more detailed discussion of the effect of arbitrage on telephone demand. 17 Technical Network Failures. A reduction of technical network failures will reduce involuntary interruptions of telephone calls and the subsequent need to redial. It will therefore result in an increase of average call duration, while total call volume should ceteris paribus remain approximately constant. Sig-nificant improvement of network quality are likely to be noticeable on routes from and to developing countries, rather than on routes between developed countries, where telecommunications networks are highly digitalized and net-works quality measures are usually stagnating at high levels. The problem with technical network failures is that they are empirically hard to measure. Indicators that are to some extent reflect the technical failure rate in net-works are the Answer Seizure Ratio 7 (ASR), and the Percentage of Network Digitalization. However, both measures are confounded with other effects, such as the presence of voice mail boxes on the receiving end of a call, the use of callback services, or limited network capacity. Cultural Differences in Communications Patterns. Finally, even with all of the above conditions being equal on all inter country relations and over time, differences in communications patterns due to cultural factors may lead to variations of calling patterns. As the communications behavior of both conversation partners determine number and duration of telephone calls between them, cultural factors on both sides of an inter country relation will shape calling patterns of both incoming and outgoing calls. 7 T h e percentage of calls that seize an international circuit and are unanswered at the terminating side. Quite obviously, the A S R is low on routes where many calls seize the international network, but are not answered due to technical failures in the network. 18 2.3 Theoretical Framework In his monograph on telecommunications demand analysis, Taylor (1980) considers both call number of calls and call duration in his discussion on how information is produced via telephone calls: "Suppose that a person is weighing making three-minute calls to a relative in the middle and at the end of a month or a six-minute call at the end of the month...the former would be more expensive, but could be of greater value because the information conveyed is more current. " In this view, therefore, the amount of information the agents derive from telephone calls not only depends on total number of minutes of outgoing and incoming calls, but also on the frequency and the duration of calls. While it has hence been recognized that call duration matters, there has been no systematic method developed for handling it, neither conceptually, nor empirically. The two alternative models developed in this section build up on work of Larson, Lehman and Weiss (1986) and Acton and Vogelsang (1992) , incorporating work of Cesario (1976), McConnell (1992), and Larson (1993) . They extend the existing conceptual framework for Point-to-Point demand models by specifying alternative empirical models of telecommuni-cations demand incorporating both call durations and total calling volume in a point-to-point market. 19 2.3.1 The Classical International Telecommunications Demand Model: Point-to-Point Demand with-out Connection Charges Consider a country pair telephone market with a home country, h, and a foreign country, / , and one economic agent in each country with access to the telephone network. Both agents are assumed to derive utility from in-formation, / , and from a composite good, z, that is, u = u(I,z). Following Larson et al (1986), the agents in h and / do not only derive information from calls they initiate, but also from calls they receive. Hence we can write jh = jh(th^tf) anc[ // = lf(tf,th), where th and tf are the total call min-utes initiated by the agents in country h and / , respectively. The utility maximization problem of the agent in h can then be written as maxtfczft uh(Ih(th,tf),zh) (2-1) s.t. yh = p\th + phzzh where p\ is the call charge , phz is the price ofthe composite good , and yh is the income of the agent in h. Under the standard assumptions on the curvature of the utility function in th and zh, the solution to this optimization problem then takes the form of the demand equation th* — f(p^,p^, yh, tf). Similarly, the foreign agent's demand equation is given by t^* = f{pt,pl,y*,th). The 20 Nash equilibrium therefore satisfies the equation system th* = f(p?,Phz,yh,tf*) (2.2) tf* = f(pft,Pfz,yf,th') The agent's choice variable is the total number of calls, i.e., the product of number of calls and call duration, th = xh • dh. An agent is commonly as-sumed to be indifferent between any combination of call number and average call length that results in a given total number of call minutes. Hence, an exogenous increase of the average call duration would not change the agent's choice of the total number of call minutes. ^=S^+**=° (2'3) that is, the agent would decrease the optimal choice of number of calls to compensate for an increase of average call duration dxh x h (2.4) ddh dh A common assumption in all of the previous literature on international telecommunications is that calling charges are exogenous. This assumption is also adopted for this study and can be justified on two accounts. First, in many of the countries included in the data set regulators still maintain some oversight over calling charges. The second reason is that even where regula-21 tors do not maintain such oversight, transmission capacities for international telephone services by far exceed actual demand. Supply can therefore be considered as perfectly elastic in the relevant range, making it unnecessary to deal with the estimation of the supply curve. 2.3.2 Point-to-Point Demand with Connection Charges The model described in the previous section serves as the basis for most of the recent empirical models of international telecommunications demand, such as Larsson et al. (1986) and Acton and Vogelsang (1992), and Garin-Muhoz and Perez-Amaral (1998, 1999). There has been little concern about the validity of this approach, in particular with respect to its implicit assumption that agents are indifferent between different combinations of call duration and number of calls, as long as the total number of call minutes remains the same (Equations (2.3) and (2.4)). The most natural way of modeling a setting as described by Taylor would be by using a dynamic framework. Such an approach would however exceed the limitation given by the data set that is available for this study. Instead, this research generalizes the static framework of Section 2.3.1 in order to encompass call durations. Let Ih = Ih(xh, dh, xf, df) and V = If(xf, df, xh, dh) be the information, from which agents in h and / derive utility. xh is the number of telephone cahs initiated by the agent in h, and dh is the average duration of each 22 of these telephone calls. Similarly, xf and df are the number and average duration of telephone calls initiated by the agent in / , respectively. The budget constraint of the agent in h is given by yh = xh(px+Pddh)+Pzz (2.5) where px is the connection charge, P(i is the (marginal) per-minute charge of a telephone call, and pz is the price of the composite good. This budget con-straint is non-linear, which makes this model different from standard models of consumer theory. Information can only be generated when neither call frequency, nor call duration are zero. Without any call duration, there is no information generated by making a call, " W ^ - . (2.6) and without making a connection, no information can be derived from longer call durations, dlh(0,dh,x',df) 3d* " ( j It is therefore reasonable to exclude the possibility of boundary solutions. In order to ensure an interior solution for the agents' utility maximization prob-lem, it is important to assume that the utility function uh(I(xh, dh, xf df), z) is such that the curvature ofthe indifference curves in (xh, dh) is greater than that ofthe budget constraint, as shown in Figure 2.3. Interior solutions occur 23 B x Figure 2.3. The nonlinear budget constraint with indifference curves that are shaped like U l . Without the restriction on the utility function, the concave shape of the budget constraint could lead to solutions at (B). This is the case when the utility function is shaped like U2. However, since the shape of the utility function is assumed to be such that boundary solutions cannot occur, only interior solutions (A) are relevant. 2.3.3 The Case of Exogenous Call Duration A simple way to introduce call durations in the standard framework is to consider a model where average call durations are exogenous. One could conjecture, for example, that average call durations are exclusively dictated by specific (exogenously determined) cultural traits, whereas the number of calls reflects economic optimization behavior. The utility maximization 24 problem of the agent in h is then max^fc uh(Ih(xh,dh,xf,df),zh) (2.8) s.t. yh = xh(phx + phddh) + phzzh Hence, when call duration are exogenous, the agent's choice variable is the number of calls, leading to the fundamental demand equation xh* = fh(phx+phdd\phz,d\y\xf ,d!) Similarly, the foreign agent's demand equation is given by = ff(Pfx + Pddf,p[AS, yf, x\ dh) Note that if call duration is exogenous but not constant, it must be included in the specification of the demand equation. The call duration enters the demand equation in two ways, once directly and once through the price of a call, px + phddh. Hence, when call duration exogenously increases, there are two additive effects. dxh* _ h dxh* dxh* ~ddF ~ P d d{phx + phddh) + 8dh { ' ] The first term is the price effect of longer call durations. As the call dura-tion increases, the price per call goes up, resulting in a lower call demand. 25 The second term is a pure utility effect. Longer call durations increase the marginal utility of a call, and in turn increase the demand for calls. The sign of the total effect is however ambiguous. The demand for total call minutes can be written as th*(phx + phdd\phz, dh, yh, xf, d^ = dhxh*{phx + p\d\ phz, dh, yh, xf, df) (2.10) The change of demand for total call minutes with an exogenous change of the call duration is non-zero a i _ Th* , ih h O X , d h a x (2 111 The Nash equilibrium outcome satisfies the following system of equations xh* = fh(ph+phdh,phz,dh,yh,xf*,df) (2.12) xf* = ff(pf+pfddf,pf,df,yf,xh*,dh) 2.3.4 The Case of Endogenous Call Duration Now consider the model when call duration is endogenous. Call duration then becomes one of the agent's choice variables, along with the number of calls and the quantity of the composite good. Hence, the agent's utility 26 optimization problem is m a x ^ ^ ^ k uh(Ih(xh, dh, xf, df), zh) (2.13) s.t. yh _ xh(ph + p / l ^ + ^ . j , The resulting indirect utility function can be written as V(pl,p{l,p{:, yh, xh, dh). The nonlinear budget constraint complicates the derivation of the demand system in this case. It can be easily verified, that Roy's identity holds in the case of the demand for number of calls, The demand function for call duration is given by1 While Equation 2.15 represents a standard Marshallian demand function in the sense that it expresses demand as a function of exogenous parameters, it does not relate the marginal utility of call durations to the marginal utility of income. However, this property is held by the demand function for the 8 A standard way of deriving equation 2.14 is by using the envelope theorem. Due to the envelope theorem, = —Xx and Vy = A, and hence Vj^/Vy — -x. The envelope theorem holds regardless of the specific form of the budget constraint. Therefore, Vpd = —Xdx. Dividing by = —Xx yields the following expression for the optimal call duration. x ,h* (2-14) 27 total number of call minutes, _yh yh yh th\=xh*dh* = = (2.16) yh yh yh v 1 y Px ry Hence, a complete demand system for the agent in the home country is given by xh* = fh(phx,phd,phz,yh,xf*,df) (2.17) . th* =gh(phx,phd,phz,yh,xf*,df) Again, with the respective demand equations for the foreign country, the equilibrium outcome of the complete point-to-point system is given by xh* = fh(phx,phd,phz,yh,xf*,df) dh* = gh(phx,phd,phz,yh,xf*,df) xft = ff(pfx,pfd,Pfz,yf,xh*,dh) df*=gf(pfx,pfd,pi,yf,xh*,dh) (2.18) 2.3.5 Price Elasticities of Call Duration The question arising from the previous analysis of Section 2.3.4 in the con-text of demand analysis is, whether the non-linear budget constraint alters standard economic results regarding the signs of demand elasticities. To de-rive demand elasticities in this case, consider again the utility maximization 28 problem of Equation 2.13. Dropping the country superscript, and assuming additive separability of the utility function of the form 9 U = I(x,d) + z (2.19) and Vz = 1 (2.20) the Lagrangian to Equation 2.13 can be written as L = I(x,d) + z + A(y — pxx — p^xd — z) (2-21) The first order conditions can be derived as h ~ HPx+Pdd) = 0 Id - Xpdx = 0 (2.22) pxx + Pddx + z = y A = 1 9 This restriction is needed for reasons of tractability. A n obvious feature of this specific form is that it implies a zero income elasticity. As explained below, the assumption of a zero income elasticity poses no empirical problem in the context of the data set used for this study. 29 Total differentiation and arranging the equations in matrix form yields 0 px + Pdd PdX 1 Px + Pdd Ixx ^xd Xpd 0 PdX Idx ~ ^Pd hd 0 1 0 0 0 dx dd \ d z J 1'dy — xdpx — dxdp^ Xdpx + Xddpd Xxdpd 0 V (2.23) Hence, using Cramer's Rule and substituting X = 1, the total differentials of x and d, respectively, are dx = ~TfMpx + dpdd) + Pdxdpd dd - Pd(dpx + dpdd) - jj^xdpd (2.24) (2.25) where I is the bordered Hessian matrix of Equation 2.23. The second order conditions of utility maximization require | / | < 0. Hence, the "own"-price elasticity of the number of calls can be signed as follows dx dpx dd < 0 (2.26) As it would be the case in the standard two goods model 1 0, the "own" price elasticity of number of calls is negative. The sign of the other price and cross 10I.e., where the budget constraint is y — p\Xi +P2X2 + z. 30 price elasticities, however, remain ambiguous 1112 dd ,Idx Pd Ixx to, O ' T N T~ = d—iTi xTn ^2-21> dpd \I\ \I\ doc ,Idd , Ixd Pd /r, oo\ dPd \I\ \I dd _ Ixd - pd dpx \I\ (2.29) unless more structure is imposed on the Information function. One way of imposing additional structure on the Information function is to assume that the agent ultimately cares only about the total number of call minutes and is indifferent between combinations of x and d that result in the same product t = xd13. Both price elasticities of the total number of call minutes can then be assumed as negative, ^d~^ < 0 and < 0. It turns out that d(xd) = d d ^ + x d d , = _dlM + hd^Pd = *X (2_3Q) dpx dpx dpx \I\ \I\ dp, d Therefore, < 0 implies ^ < 0, and hence, under the assumption that the agent maximizes over total call minutes, the cross-price elasticity of num-ber of calls would be negative. The own and cross-price elasticities of duration 1 1This is due to the fact that the expression Ixd — Pd cannot be signed. 1 2Recall that in the standard two goods case all elasticities can be signed: both own price elasticities would be negative, whereas the cross price elasticities would be positive. 1 3Note that this is one of underlying assumptions of the "standard" model of Sec-tion 2.3.1. 31 still remain ambiguous. 2.4 Empi r i c a l M o d e l , Es t imat ion , and Results In order to estimate an empirical specification based on models of Sec-tion 2.3.4 and Section 2.3.3, this paper employs a data set that has not been used in previously published demand studies. The question hence arising is, to what extent are the results of this study driven by the data, and to what extent are they influenced by the extended theoretical framework on which the empirical specification is based. This empirical section proceeds in two steps. The following subsection will report empirical results of estimating standard international telephone demand models with the current data. It will interpret the results and compare them to results of previous demand studies. The subsequent subsections, then, will estimate and analyze em-pirical specifications based on the new theoretical frameworks, based on the assumption of exogenous call durations (Section 2.3.3) and endogenous call duration (Section 2.3.4). 2.4.1 Standard Empirical Specifications Empirical models based on the framework of Section 2.3.1 and on previous work by, for example, Larson, Lehman and Weisman (1986), and Garin-Mufioz and Perez-Amaral (1999), are subject to the assumption that neither call durations nor connection charges are relevant in the context of telecom-32 munications demand analysis. Hence, a typical empirical specification of the "standard" demand model takes the following double-logarithmic form: In Tjl = fo + fo In + fo In T?t + fo In GDP£ + fo In TRADEit + fo\nTOUR^t + fo\nTOUR{t + folnIM* (2.31) + fo In MLht + fo In MLfit + foQ In POP/t + ehit In Tl = 50 + <S, In P/t + 52 In 7* + <53 In + & In TRADEit + 55\nTOUR{t + 56]nTOUR^ + 87lnIMtht (2.32) + £ 8 In ML[t + 59 In M L f -I- <J10 In POP/t + e{t where / i again stands for home country, in this particular case the United States, and / stands for foreign country. The variables are defined as follows: 33 GDP/t TRADEu TOURl TOUR(t 1Mb ML{t POPi Total number of call minutes from the US to country i in year t Charge for a (marginal) minute of a telephone call from the US to country i in year tu G D P of country i in year t , deflated by the CPI Total trade volume between the US to country i in year t Number of US tourists to country i in year t Number tourists from country i in year t to the US Six-year stock of immigrants from country i to the US in year t15 Number of main lines per inhabitant in the US, in year i 1 6 Number of main lines per inhabitant in country i and year t Number of Inhabitants of country i and year t17 Equation 2.31 represents the demand for total telephone minutes from the US to various countries, whereas Equation 2.32 describes the telephone demand in the reverse direction. The subscripts i, j, and t are country subscripts, 1 4 The definition of P§. is also used by Acton and Vogelsang (1992). Alternatively, Garin-Munoz and Perez-Amaral 1999 (among others) use the price of a three minute call. 1 5only migration into the US is included as an explanatory variable, whereas migration in the reverse direction is considered negligible. 1 6 The number of main lines per inhabitant ("telephone penetration") can be regarded as a proxy for the availability of advanced communications technology and services in a country's telephone system. Low penetration rates typically coincide with a low standard of technology and service availability. But also in countries with high penetration rates, further increases in penetration rates may indicate the use of telephone lines for advanced telecommunications applications, such as telefax, call waiting, credit card confirmation and debit card payment in retail. For an interpretation of the telephone penetration measure, see also OECD (1999). 1 7 The number of inhabitants is used as an indicator of market size. 34 i = 1..48, j = 1..2318, and t is the year subscript, t = 1988... 1996. The variables of Equation 2.32 are defined analogously. For a detailed de-scription of the data and data sources, see Appendix A. Some previous studies have included the G D P of the call originating country and the G D P of the destination countries as proxies for income. Others 1 9 have not included any measure for income at all. A particular empirical problem with includ-ing G D P in the empirical specification arises from nature of the panel data commonly used to estimate telephone Point-to-Point models. These panel data typically comprise observations of one call originating country and many destination countries. The overall G D P of the call originating country is not a useful proxy for income, because it provides no information about the rel-ative income levels of telephone users of different outgoing routes. The lack of cross-sectional variation in the data is particularly problematic where a data sets contains short time series for a large number of cross sections, as it is the case here. The same is also true with all other non-route specific explanatory variables, such as total population, main line penetration, and fax machines in the call originating country. Moreover, GDP, total population and main line penetration are highly correlated, such that including all three of them simultaneously will lead to multicollinearity. To deal with the problem of multicollinearity, the empiri-cal specifications used in this study will omit G D P and total population as 1 8 The country index in Equation 2.32 runs only to 23, since observations for Pjt were not available for all 48 countries. 19See Table 2.4. 35 explanatory variables2 0. Due to data restrictions, most previous studies of international telecom-munications have not succeeded estimating the equation for incoming and outgoing traffic as a simultaneous equation system. Rather, they focused on estimating only one of the equations, or looked at both equations separately. Table 2.4 summarizes a number of previous demand studies and their results. Estimated price elasticities ranged from -0.32 to -1.9. Previous findings on income elasticities are somewhat more inconsistent, they range from 1.39 to 2.662 1. Table 2.4. Price and Income elasticities of international tele-. phone demand as reported by selected studies Price Income Years Study Routes Elastic. Elastic. of Data Lago(1970) 23 routes -1.249 - 1962-64 to / f rom U S R e a and Lage 37 routes -1.7 to 1.92 to 1964-74 (1978) to / f rom U S -1.9 2.66 A c t o n and Vogel- 17 routes -0.36 1.39 1979-86 sang(1992) to / f rom U S G a r m - M u n o z and 27 routes -0.69 - 1981-91 Perez-Amaral(1997) to / f rom Spain Gar in -Muf ioz and 24 routes -0.32 - 1981-91 Perez-Amaral(1999) to / f rom Spain Source: Referenced Papers 2 0 T h e variables to be omitted were chosen by successively eliminating those variables that were not significantly different from zero. 2 1 Note that the elasticities at the high end of this range were estimated based on data from the 1960s and early 1970s, when international charges were considerably higher and international traffic volumes were lower than in the 1990s. 36 Table 2.5 shows the estimation results for US outbound demand, for two different estimation techniques. The first column reports the results using an empirical specification that assumes the presence of fixed route specific effects (fixed effects model) 2 2. Two important issues that also need to be addressed in the context of telecommunications panel data are the simultaneity between outgoing traf-fic23 and incoming traffic2 4, and the potential presence of autocorrelation. Since the available data are typically not rich enough to estimate both rela-tions simultaneously, previous empirical research has either ignored the prob-lem of simultaneity altogether, or employed instrumental variable regression techniques. Garin-Mufioz and Perez-Amaral (1999), for example, use the one period lagged incoming traffic variable as instrument for the incoming traffic. This approach is valid, as long as the error terms are not serially correlated. In the presence of autocorrelation, however, the lagged variables are still cor-2 2 The results of the fixed effects estimation were compared and tested against a ran-dom effects model. Since the sample of routes used for this study is not exhaustive, i.e., represents only a subset of all US outbound routes, the random effects model can be a more appropriate choice of model than the fixed effects model, if the model is meant to allow for inferences regarding the entire set of international routes. However, if the route specific errors of the random model are correlated with other regressors, then the random effect model will produce inconsistent estimates. A Hausman test can be used to check the appropriateness of the random effects model, (for details about the random effects model and about how to test for the appropriateness of the random effects model versus the fixed effects model, see, for example, Kennedy (1996)). In the present case, the Hausman test statistic is 77.04, which is significant compared to the critical value of Xio = 18.307. Hence, the GLS for the random effect model is inconsistent, and the OLS is the preferred estimator. 2 3 Represented by the variable T£. 24Represented by the variable T-t. 37 Table 2.5. Estimation results for standard model of international tele-phone demand for traffic originating in the United States, using 1988-96 data Dep var: lnT& OLS F G L S OLS F G L S regressor F . E . AR(1) regressor F . E . AR(1) InPtt -0.647* -0.441* lnIM% 0.036* 0.031* (-8.283) (-7.586) (2.382) (2.060) lnT/t 0.167* 0.168* InML'i 1.428* 2.069* (4.410) (4.348) lnML{t (4.477) (7.215) InGDP/t -0.872 0.019 0.437* 0.431* (-1.281) (0.220) lnPOP/t (7.636) (7.685) InTRADEu -0.042 -0.036 0.762* 0.606* (-1.428) (-1.685) (3.913) (3.908) lnTOUR% 0.004 0.004 constant 3.331 6.148* (0.094) (0.124) (0.781) (2.664) lnTOURfit 0.160* 0.076* (4.119) (2.728) obs 338 330 R2 0.990 rho - 0.522 related with the contemporaneous error term, their use as instrument is not appropriate. In the context of this study, a test for autocorrelation suggests that the error terms are autocorrelated of the order one, making the use of lagged variable instrument inappropriate. There are two possible alternatives to the lagged variable instrument. One is P/t, the calling charges in countries / . In fact, P/t is the only regressor in Equation 2.32 that was not also included in Equation 2.31. The problem with this instrument is that the relevant data are not available for the full set of observation, and including this instrument will significantly reduce the degrees of freedom in the estimation. The second 38 alternative is the number of incoming telephone calls, C(T . By definition, the number of incoming calls is correlated with the total number of incoming call minutes. Moreover, under the underlying assumption of this model type, that is, the number of calls do not need to be considered for estimating telephone demand, the number of incoming calls will be uncorrelated with the error terms, and can therefore serve as an instrument for the total traffic volume. Using either instrument, or both instruments in an instrumental vari-able regression, and performing a Hausman test, indicates that the incoming traffic does not need to be considered as endogenous in this model 2 5 . The second column of Table 2.5 presents the Feasible Generalized Least Squares (FGLS) results under the assumption of first order autocorrelation. Although fixed effects specification reported in the first column is unbiased in the presence of autocorrelation, greater efficiency of the model that accounts for autocorrelated disturbances suggests that the latter is preferred among all the empirical models discussed in this section. With the elasticity of -0.441, the first order effect of a price change on outgoing telephone traffic is highly significant and in the range of previous findings. Moreover, all estimates have shown to be relatively robust across different empirical specifications. The total volume of incoming traffic is positive and significant, which supports the hypothesis that the reciprocal 2 5 T h e x2 s tat is t ic in the case of inc lud ing C(t and pft as inst ruments is 2.99 and below the cr i t i ca l value of x\ = 5.991. Hence, the O L S remains a consistent est imator in this context. 39 calling effect is positive, and that it dominates the call substitution effect26. Other significant and positive variables are, the number of incoming tourists, immigration, the market sizes of the call receiving countries at the ends of a given relation, and the main line penetration at both ends of a given inter country relation. Not significant are G D P of the call receiving countries, trade and outgo-ing tourism. Some of these effects may however have been picked up by the correction for route specific effects in the regression. Table 2.6 reports the results of estimating the respective models for the in-coming traffic, whereby, due to the lack of data, fixed route effects could not be accounted for. In the OLS regression, the hypothesis of first order autocor-related residuals could not be rejected. Moreover, the results of a Hausman test performed on IV and OLS estimates suggested that the total volume of incoming traffic could not be considered as exogenous. OLS estimates are therefore both inconsistent and inefficient. The columns of Table 2.6 show the coefficient estimates of the model with a correction for first order auto-correlation 2 7 and the results of the instrumental variable regression, using Cl^t and P-J. as instruments for 7$. It is important to note that the reliability of a cross comparison between 2 6 For a detailed description of these effects, see Larson, Lehman and Weisman (1986), Acton and Vogelsang (1992), and Garfn-Mufioz and Perez-Amaral (1999). 2 7Due to the endogeneity problem, these estimates are still inconsistent. 40 Table 2.6. Estimation results for standard model of international tele-phone demand for traffic with destination United States, using 1988-96 data Dep var: InTft F G L S IV F G L S IV regressor AR(1) regressor AR(1) lnP/t -0.162* -0.262* -0.073* -0.127* (-3.464) (-5.472) lnML{t (-2.503) (-5.775) 0.704* 0.701* 0.299* 0.389* (12.908) (15.767) (2.929) (4.479) lnGDPft -0.196* -0.424* lnMLh -1.327* -1.951* (-2.836) (-6.776) InPOPl (-3.622) (-5.815) InTRADEtt 0.014 0.047* 0.138* 0.314* (0.614) (2.008) (2.301) (6.119) lnTOUR{t -0.353* 0.448* constant 3.399* 5.222* (-7.889) (11.058) (3.788) (6.426) lnTOUR% -0.013 -0.042* (-0.505) (-2.141) obs 197 186 R2 - 0.9863 rho 0.556 -the results for outbound demand estimation 2 8 and inbound demand estima-t ion 2 9 is only limited. For one, the empirical models are different: in the former case they include fixed effects, whereas in the latter they do not. More importantly, the characteristics of countries routes included in the es-timations are different. The inbound results include exclusively observations from O E C D countries, whereas the outbound demand estimations are based on observations from routes to both O E C D and non-OECD countries. The total effect, or long run equilibrium effect, of a price change on out-going telephone traffic can be computed by computing a closed form solution 28See Table 2.5. 29See Table 2.6. 41 for T$ from the system of equations 2.31 and 2.32. The total effect of a price change on the total traffic volume can then be computed from 0 1 n l 2 _ P\ ( 2 3 3 ) d\nP& 1-B2S2 d\nT/t 6, (2.34) d\nP/t I-S2P2 Hence, with the estimated coefficients of -0.441 and -0.262 for outgoing and incoming traffic, respectively30, total price elasticities are -0.500 and -0.296. While the elasticity estimate for incoming traffic is slightly below the most recent published findings (compare to table 2.4), the estimate of the price elasticity of outgoing call demand is well within the range of demand studies conducted throughout the previous decade. 2.4.2 Models with Exogenous Call Duration The empirical specification of Equations 2.31 and 2.32 change when the call duration is included as an exogenous variable. From Equation 2.12, the appropriate respective empirical specification becomes lnC? t = ^0 + A In P C ^ ++/3 2 In P M ^ + /33 In C/ t + ^ In A* + & In D{t + As In GDP/t + B7 In TRADElt + & In TOURp.3b) 3 0 G i v e n the incomparability of results, this should be merely regarded as a demonstra-tion exercise. 42 + B9 In TOUR{t + Bw In IM* + Bn In MLh + Pu In ML: + p13 In POi^{ + pu In FAX* + pl5 In FAX(t + lnC/ t = o^ + ^ l n P C / t + ^ l n F M ^ s l n C ^ + ^ l n D ^ + 55 In D{t + 66 In GDP/t + 5 7 In TRADElt + <58 In TOC//$2.36) + S9 In T O C / i $ + <Jio In IM* + <Jn In MLfu + 5 1 2 In MLj 1 + 5X 3 In P O P ^ + <Ji4 In FAX(t + S15 In FAX* + e{t where the newly introduced variables are defined as follows PC % Connection charge for a call from the US to country i in year t, deflated by the CPI PMit Marginal, per minute Charge for a call from the US to country i in year t C*t Number of calls from the US to country i in year t Average duration of calls from the US to country i in year t FAX* Estimated number of fax machines in the US in year t FAX[t Estimated number of fax machines in country i and year t31 and all other variables are defined analogously as in the context of Sec-tion 2.4.1. 3 1 Fax machines are included in this specification as they may contribute to explaining the variation of call durations. 43 The empirical strategy for estimating Equation 2.35 is also analogous to that of Section 2.4.1. Table 2.7 reports the empirical results for the fixed ef-fect model and the model with correction for first order autocorrelation3 2. In the context of this model, the Hausman specification test indicates that the random effects model is incorrectly specified (at Xis = 280.31). Again, the Hausman specification test is also used to test whether C{t can be considered as exogenous. The OLS is compared to the instrumental variable estimator with P/t and all exogenous variables of Equation 2.35 as instruments. As observed with the previous models, the hypothesis that C(t is endogenous is rejected33. As expected, the price elasticities with respect to both connection charge and marginal calling charge are negative and highly significant. The coeffi-cient on call duration is negative and highly significant, indicating that users substitute the number of calls with call duration, or, in the context of Equa-tion 2.9, the price effect of the duration change dominates the utility effect. The coefficient on call durations of incoming calls, however, is insignificant. The effect of an increase of incoming calls on outgoing calls is positive and significant, confirming the presence of positive reciprocal effect also with re-spect to number of calls. Analogous to the previous model, the coefficients 3 2 As specified in Equation 2.12, the dependent variable is C|£. T£ is deterministically dependent on C-lt in this specification (see Equation 2.10), such that the estimation of Tfl yields equivalent parameter estimates, except of the coefficient on D^, which then is estimated to be /34 = 0.572 (t = 4.329) and /?4 = 0.350 (t = 3.308) for the fixed effects model and the model with correction for autocorrelation, respectively. 3 3 A t x? 2 = 1-99. 44 Table 2.7. Estimation results for model of international telephone de-mand for calls originating in the United States, with exogenous call dura-tions, using 1988-96 data Dep var: lnC\t OLS F G L S OLS F G L S regressor F . E . AR(1) regressor F . E . AR(1) InPCl -0.037* -0.039* lnTOURht 0.057 0.018 (-2.175) (-3.577) lnTOUR{t (1.359) (0.597) InPMJl -0.587* -0.346* 0.024 0.019 (-6.391) (-5.103) (0.562) (0.683) lnC{t 0.132 0.123* lnMLh 0.593 1.277* (1.749) (2.112) lnMLftl (1.294) (3.423) lnD*t -0.428* -0.650* 0.591* 0.559* (-3.241) (-6.153) (7.460) (8.206) lnDfit 0.071 0.757 InFAX? 0.161* 0.142* (0.770) (1.071) lnFAX{t (3.810) (4.128) InGDPi 0.049 0.036 -0.014 -0.016 (0.664) (0.650) (-0.499) (-0.694) lnIM?t 0.044* 0.052* lnPOP/t 0.701* 0.595* (2.463) (3.058) (3.568) (3.892) InTRADEu -0.034 -0.011 constant 0.013 3.498 (-1.189) (-0.555) (0.004) (1.589) obs 266 266 R2 0.956 -rho - 0.533 F . E . = Fixed Effects on tourism (both directions), G D P ofthe call destination country, and trade are all insignificant, whereas significant positive effects were found for the mainline penetration of both countries, immigration, and destination coun-try market size. The number of fax machines in the US was found to have a significant positive effect, while that in the call destination countries had no significant effect. 45 2.4.3 Models with Endogenous Call Duration The final set of empirical models is based on the theoretical framework of Sec-tion 2.3.3. When call durations are assumed to be endogenous, the empirical specification for demand in country h becomes InC* = P* + P*\nPC* + P*\nPM* + p*\xiC[t + P*\nD(t + P* In GDP/t + p* In TRADEit + Pr I n T O U R * t (2.37) + p£ In TOUR(t + P% In IM* + p*0 In ML* + p*x In ML{ + P*2lnPOP/t+P*3\nFAX* + ph4\nFAXlt + e* \nD* = 5* + 5*\nPC* + 5*\xvPM* + 5*\nCfu + 5*\nD(t + 5* In GDP/t + <5g In TRADEit + 5* In TOUR*t (2.38) + 5* In TOUR{t + 5* In IM* + 5*0 In ML* + < In M L / + 5*2 In POP,{ + 5hu In FAX* + 5*u In F A A ^ + ^ An equivalent set of equations can be specified for incoming traffic, but is omitted at this point. Equations 2.37 and 2.38 are directly based on the system of equations 2.17. A structural equation specification would not be feasible in this case, because both equations contain the same set of exogenous variables. Including C*t on the right hand side of equation 2.38 and D*t on the right-hand-side of 46 equation 2.37, respectively, would hence leave both equations unidentified. Therefore, under the assumption that the number and duration of in-coming calls is exogenous to the above system, the equations would have no direct or indirect interaction. Where simultaneous Equations 2.37 and 2.38 do not interact, there may still be a relation between the error terms, and using the Seemingly Unrelated Regression (SUR) technique may improve the efficiency of estimation. Note however that because the set of right hand variables in both equa-tions is the same, no gains can be realized from the SUR technique, as it becomes identical to OLS. The crucial assumption here is, of course, that there is no simultaneous equation bias due to an interaction between incoming and outgoing traffic. For the previous models, results showed the effect of incoming traffic could in deed be considered exogenous when estimating the demand for US outgo-ing telephone traffic, and OLS produced consistent estimators. The further estimation strategy for this model will have to rely on the assumption that what was true in the standard models and in the models with exogenous call duration is still true here3 4. The two dependent variables, C*t and D*t determine the total number of call minutes, T$, in a non-stochastic fashion. In order to fully identify the 3 4 This assumption is needed because, first, the number of observations is not sufficient to estimate more complex specifications, second, instruments used in previous regressions will also be endogenous in this model, and third, error terms are autocorrelated such that the variables cannot be instrumented with their own lags. Hence, there are no appropriate instruments available to empirically test the assumption that OLS is consistent. 47 demand system of C*v D*t, and T*t, one needs to estimate only two of the three relevant demand equations. For the ease of comparison to previous re-sults, Table 2.8 reports the OLS estimates and the F G L S estimates under the assumption of first order autocorrelated disturbances for all three equations. Although the results of the two estimation techniques are qualitatively similar, the estimation with correction for autocorrelation is preferred, as it is more efficient than OLS. The main estimates of interest, that is, the coefficients on the price vari-ables, are highly significant in all regressions. Consistent with the frame-work of Section 2.3.5, the "own" price elasticity of calls 3 5 is negative (Equa-tion 2.26). As expected, the cross price elasticity of duration 3 6 was found to be positive, whereas the remaining elasticities are both negative3 7. The total effect of a price change on demand can be computed from . / dlnCl \ d In PCI dlnPM^ a In D\t it 3 In P C £ d\nPM^ a In 4 Sin C{t d In PC(t 9 In PM?t it dlnD{t V d\nPC(t din PM^ ) ( 1 0 -PI -Ph^ -1 'p\ 0 1 -SI -si 8\ 6* -Pi -Pi 0 1 Pi Pi -si -si 0 1 , u Si j (2.39) Assuming that outgoing traffic and incoming traffic are symmetrical 3 8, the 3 5That is, the coefficient on In PC^ in the regression of C% ' 3 6that is, the coefficient on In PCft in the regression of D^t 3 7Recall that this framework provided that the signs of these elasticities was ambiguous. 3 8That is, /?f = p{ and # = S{ for i = 3 ,4 . 48 Table 2.8. Estimation results for model of international telephone demand for calls originating in the United States, with endogenous call durations, using 1988-96 data OLS F G L S regressor F . E . AR(1) dep. var. l n £ £ I n P C S -0.052* 0.029* -0.023 -0.051* 0.022* -0.033* (-3.046) -0.474* (3.279) -0.159* (-1.359) -0.633* (-4.426) -0.327* (3.331) (-2.951) -0.373* In P Mjl -0.105* (-5.326) (-3.474) (-7.149) (-4.443) (-2.615) (-5.246) l n C / 4 0.076 -0.029 - 0.125* 0.000 -(1.576) (-1.161) - (1.971) 0.004 --0.010 0.029 -0.028 0.039 0.061 -0.025 (-0.138) (0.800) (-0.478) (0.508) (1.472) (-0.586) - - 0.047 - - 0.122* - - (0.981) - - (2.055) \nGDP/t 0.073 -0.059 0.014 0.067 -0.051 0.020 (0.986) (-1.545) (0.189) (1.106) (-1.532) (0.394) InTRADEu -0.014 -0.001 -0.015 -0.023 0.014 -0.006 (-0.477) (-0.066) (-0.513) (-1.033) (1.155) (-0.309) \nTOUR% 0.043 -0.010 0.033 0.026 -0.011 0.017 (1.022) (-0.447) (0.795) (0.767) (-0.628) (0.559) lnTOURfit 0.069 -0.000 0.069 0.021 -0.003 0.017 (1.691) 0.037* (-0.000) (1.696) (0.705) 0.055* (-0.020) (0.610) 0.050* InlM* -0.008 0.028 -0.007 (2.102) (-0.966) (1.610) (3.070) (-0.768) -0.895* (2.911) \nMLh 0.947* -0.838* 0.108 1.728* 0.958* (2.173) (-3.727) (0.251) (4.478) (-4.365) (2.620) In ML{t 0.501* 0.127* 0.629* 0.514* 0.089* 0.587* (7.521) (3.698) (9.468) (7.148) (2.364) (8.504) In POP/t 0.652* 0.058 0.710* 0.554* 0.072 0.624* (3.329) (0.579) (3.642) (3.356) (0.812) -0.090* (4.006) 0.111* In FAXf 0.267* -0.114* 0.152* 0.199* (8.331) (-6.928) (4.775) (5.584) (-4.761) (3.283) In FAX{t -0.043* 0.019 -0.023 -0.017 0.000 -0.016 (-2.109) (1.853) 4.444* (-1.158) (-0.662) (0.013) 3.119* (-0.687) 4.472* constant -1.540 2.903 1.423 (-0.428) (2.359) (0.811) (0.603) (2.461) (2.007) obs 298 298 298 266 266 266 R2 0.952 0.810 0.923 - - -rho - - - 0.463 0.337 0.531 F . E . = Fixed Effects 49 long run elasticities are then -0.057 -0.379 \ (2.40) V 0.023 -0.112 J A l l coefficients on main line penetration (in the US and in the call receiv-ing countries) and on the number of US fax machines are significant. The effect of the number of Fax machines is positive on the number of calls and on total call minutes, but negative on call durations. Also significant and positive is the effect of incoming number of calls on outgoing number of calls, and the effect of incoming total call minutes on outgoing total call minutes. Market size and Immigration have a positive significant effect on number of calls and total call minutes, but no significant effect on call duration. A l l remaining variables, including G D P of the call receiving countries, Trade, and Tourism, have no significant effect. However, some of these effects could have been picked up by the complete set of dummy variables included in the estimation 3 9. 2.4.4 Discussion In the light of the empirical results, there are three basic issues that call for a more detailed discussion. The first issue concerns the comparison of 3 9 9 of 41 dummy variables included in the estimation of total traffic volume are sig-nificant. In the estimation of call duration, only 2 out of 42 variables turned out to be significant. Omitting the insignificant variables would increase the significance most estimates, but qualitatively not alter the results. 50 the different model specifications. Is it necessary to consider call durations in telecommunications demand analysis, and if yes, should call durations be considered as exogenous or endogenous? Second, the results need to be tied to the discussion about the causes of decreasing call durations. Which are the factors that cause the variation of calling patterns? Finally, what are the main implications of this research for the telephone service providers with respect to pricing of international telephone services? Model Comparison. Given the data limitations in this study, the conclu-sions drawn with respect to the different model specifications must remain subject to future investigation. However, the empirical results clearly support the relevance of the extensions to the standard telecommunications demand model presented in this paper. The result that call duration is significant when used as an (exogenous) explanatory variable for the total minutes of telephone traffic (Equation 2.35) is not surprising. After all, total minutes of telephone traffic is the product of number of calls and call duration. More interestingly, however, are the results of the more general model -the model with endogenous call duration. Estimating the demand for call duration indicates that the key coefficients of Equation 2.37, that is, the co-efficients on the price variables, are significantly different from zero. One can therefore conclude, that call duration should be considered as endogenous40. The difference between the specification with endogenous call durations 4 0 O n the other hand, if the coefficients of Equation 2.37 were insignificant, then the hypothesis of endogenous call durations would have to be refuted in favor of the exogenous duration hypothesis. 51 and the standard empirical specification is that the endogenous specification involves including the connection price, PC^, and incoming call durations, D[u as relevant parameters, whereas the standard model does not. When estimated, the coefficient on the former parameter is significant but small, the coefficient on the latter parameter is not significantly different from zero. Hence, the standard demand model is misspecified to the extent that it omits the connection charge as a relevant variable. The good news for traditional "standard" demand modeling, on the other hand, is that the error of misspecification can be considered as small, because the coefficient on the connection charge variable, if included, is small. Determinants of Call Duration. The empirical results shown in Ta-ble 2.8 support several explanations for the observed traffic patterns. One of the central observations of Section 2.2 was that total telephone traffic increases throughout the period of observation, whereas call durations de-creases. Hence, the variables of interest are those where the coefficients in the demand equation for call duration significantly differ from those in the demand equation for total minutes of traffic. The variable with the greatest influence in this context is MLh, the main line penetration in the United States. The elasticity of main line penetration in the duration equation is -0.895, versus 0.958 in the total traffic equation. Main line penetration has a significant and strong negative effect on call duration, but a strong positive effect on total traffic volume. The respec-tive coefficients on the foreign countries' main line penetration, ML{t, do 52 not indicate a negative relationship with call duration. Yet, the coefficients are significant and small (0.089) in the case of call duration, and significant and large (0.587) in the case of total call volume. With main line penetra-tion interpreted as a proxy for the availability of advanced communications technology and services (see footnote 16), this result empirically supports the hypothesis that changes of traffic patterns are to a large extent due to technological change and service innovation. A more specific indicator for the development of terminal equipment tech-nology, the coefficient on FAX? is estimated to be -0.09 in the duration equation, but 0.11 in the total traffic equation. Another significant, albeit smaller effect is due to the pricing of interna-tional telephone services. The gradual decline of connection charges in the period of observation had a significant negative effect (-0.022) effect on call duration, yet a significant positive effect (-0.033) on total traffic volume. The effect of incoming traffic volume on total outgoing traffic volume is positive and significant, whereas the effect of incoming traffic on call duration is not significantly different from zero. Moreover, the duration of incoming calls has no significant effect on the duration of outgoing calls. Together, these results are consistent with the hypothesis of arbitrage, although it may also be partly due to a combination of other effects, such as call substitution and call stimulation. The two remaining reasons for changing calling patterns discussed in sec-tion 2.2.2 are Technical Network Failures, and Cultural Differences. None 53 of these explanations are supported by the empirical results. The impact of technical network failures could not be assessed, since no appropriate mea-sures for this effect were available. Finally, the majority of dummy variables included in the estimation was not significantly different from zero 4 1, indicat-ing that there were no significant route specific effects on most routes. The existence of cultural differences in calling patterns would have been reflected in such route specific effects, and can therefore be refuted. Implications for Pricing of Telephone Services. The changes of calling patterns described and analyzed in this paper are of interest in the context of pricing of telecommunications services. Call dial ups utilize costly network capacity. Shorter call durations result in shorter (charged) calls relative to (uncharged) dial-up time, and will therefore increase the per minute cost of a call. The shorter the call duration, the greater the relative importance of including their assessment for the purpose of demand forecasting and capacity planning. A n noteworthy aspect of this analysis is that it allows for a comparison between the consumers' price sensitivity to changes in connection charges, and their price sensitivity to changes in marginal calling charges. An average telephone call, that is, a telephone with mean duration (7.16 minutes), mean connection charge ($1.03) and mean marginal charge ($1.10), costs $8.91. A one percent increase of the total charge for such a call, solely by adjustment of the connection charge, would result in a decrease of the long-run demand 4 1 See footnote 39. 54 for telephone calls by 0.49 percent. The same increase of the total charge, solely by adjustment of the marginal charge, would result in a decrease of long-run demand by for calls 0.44 percent. While the difference between these elasticities is small, the results indicates that there is empirically no evidence that users ignore connection charges, or that they are less sensitive to changes of connection charges than to changes of marginal charges. In a market where service providers possess market power, connection charges are a useful tool to extract additional rents from users, if demand is less elastic with respect to changes of the connection charge than with respect to changes of the marginal charges. In the reverse, if telephone users are more sensitive to connection than to marginal charges, the usefulness of connection charges as a pricing tool is questionable. The estimates presented here empirically support the latter proposition. They help explaining, why service providers have gradually phased out connection charges on interna-tional routes. In the past two years, connection charges have had a renaissance with resale carriers. Their pricing schedule often includes a low, close to cost-level marginal charge, and a comparably high connection charge-42 The implica-tions of this demand study for the resale market may however be limited. First, the market conditions have considerably changed since 1996, the last year of the observation period. Second, users of resale services may exhibit 4 2 F o r example , i n 2000, C a n a d i a n resale carr iers offered cal ls to E u r o p e for as low as US$0,054, w i th a ca l l connect ion charge of $3.00. 55 different demand characteristics than users of facilities-based carriers. Third, charges of resale carriers are considerably different from those of those in-cluded in this study, that is, the typical connection charge of a resale carrier is far above the sample maximum of the dataset used for this study, while the typical marginal charge is far below the sample minimum. However, given the availability of appropriate disaggregated data, the methodology employed in this study can also provide the basis for an empirical investiga-tion of telephone resale service pricing. 2.5 Summary and Conclusion The paper extends traditional telephone demand analysis introducing the aspect of call durations into the Point-to-Point demand model framework. It develops the theoretical foundations and explores the implications for the cases of two alternative assumptions, the assumption that call average call durations are determined exogenously, and the assumption that telephone users simultaneously choose the number of calls they make and the average durations of these calls. The empirical results derived from these extended frameworks provide a number of insights into the structure of international telephone calling patterns. First, the results confirm the hypothesis that users regard call durations as a decision variable for their choice of international telephone service use. The findings do not support the hypothesis that telephone users 56 only value total minutes of traffic conversation, no matter what the average call duration is. Second, the results explain the observation that during the past decade call durations have significantly decreased on international routes, whereas both total traffic volume and the number of telephone calls have increased. The reasons for this shift in calling patterns can be found in both technolog-ical and economic factors. A large part of it can be attributed technological change and service innovation in the telecommunications industry and in the terminal equipment market. To a lesser extent, the gradual phasing out of connection charges for international telephone calls has also contributed to the decrease. The hypothesis that cultural factors or arbitrage behavior contributed to the shift of calling patterns was not be supported by the data. The analysis of call duration is important for telecommunications network planning, as the decrease of average call duration increases the significance of (unbilled) dial up time relative to the (billed) telephone conversation time. The framework presented in this paper can also be applied to assess the im-plications of currently popular pricing schedules with high connection charges and low marginal charges on telephone users' calling behavior. 57 Chapter 3 The Proportionate Return Requirement: Regulatory Overkill in International Telecommunications 3.1 Introduction The Proportionate Return Requirement (PRR) is a regulatory rule that gov-erns interconnection agreements between international telecommunications carriers. It stipulates that the incoming telephone traffic will be allocated among operators according to their respective share of the outgoing traf-fic. The P R R has been implemented in the United States as one of the three major features of the International Settlements Policy (ISP), a set of competitive safeguards with the objective of restricting the market power of foreign monopolists in the competitive US international telecommunications market. 58 The W T O Agreement on Basic Telecommunications Services (WTO 1997) has triggered a recent policy debate about the need for competitive safe-guards in international telecommunications. Some analysts argued that the W T O Agreement would lead to more competition, undercutting monopolists' market power and thus making competitive safeguards superfluous. Others argued that the economic cost of such safeguards would outweigh their eco-nomic benefits. The US Federal Communications Commission (FCC), on the other hand, holds that as long as a foreign carrier has market power in its home market, it will be able to exercise this market power to the detriment of US carriers and consumers, if competitive safeguards are not in effect. The F C C has therefore decided to continue applying the ISP on those international routes, where the foreign correspondent carrier possesses market power, while at the same time eliminating the ISP on routes between US carriers and competitive foreign carriers. As a result, the ISP is no longer in effect on a number of routes to countries with pro-competitive telecommunications policies, which together account for a significant portion of international traffic volumes and revenues. For the remainder ofthe routes, the F C C has in principle reasserted its settlement policy (FCC 1999b). Large US carriers have shared this view and pleaded in favor of competitive safeguards on routes to non-liberalized countries (AT&T 1998, Sprint 1998). The objective of this paper is neither to dispute the need for competitive safeguards in general, nor to dispute the need to prevent whipsawing of com-59 petitive US carrier by foreign monopolists. Rather, this paper is specifically concerned with the cumulative effect of regulations introduced under the ISP. It will show that the FCC's policy of requiring proportionate allocation of return traffic is flawed on two accounts. First, the FCC's arguments for the P R R are partly based on an erroneous interpretation of historic case law. Contrary to F C C claims, the P R R is not needed over and above other reg-ulations to prevent whipsawing. Second, the P R R distorts output decisions and price decisions in markets for outgoing and incoming international traf-fic to the detriment of US carriers. Foreign monopolists, on the other hand, may benefit from the PRR. On international routes where competitive or near competitive conditions on US outbound call markets prevail, the P R R has a negative effect on total US welfare. The remainder of this paper is organized as follows. Section 3.2 gives an overview of the institutional characteristics and regulation of international telecommunications and discusses previous literature on international inter-connection. Section 3.3 provides a historic perspective on the P R R in the United States, and critically assesses the justification of this policy. The F C C has adjusted its regulation of international telecommunications in re-cent years to account for global trends of liberalization and growing compe-tition. These adjustments are reviewed in section 3.4. Section 3.5 discusses the effect of the P R R on the market conduct of international carriers, and its welfare implications. Section 3.6 summarizes and concludes. A glossary of acronyms in included in Appendix B. 60 3.2 Institutional Characteristics and Regula-tion of International Interconnection The P R R is a policy that was specifically designed to regulate telecommu-nications markets in which interconnection arrangements are based on the Accounting Rate System1. Under the Accounting Rate System, the key economic parameter of any international interconnection agreement is the settlement rate, a call termination charge paid by the operator in the call-originating country to the operator in the country of the call destination2. Settlement rates are usually set to the same level in both directions. Settlement rates are determined through bilateral negotiations between international carriers. They therefore reflect the prevailing market condi-tions and policy objectives in each country as well as the relative bargaining position of the carriers involved. Within the general framework of the Ac-counting Rate System, operators are autonomous in negotiating the specifics of their operating agreements. Yet, a number of factors limit the operators' autonomy. First, the ITU recommendations themselves recognize that set-tlement rates in each direction of the same relation should be approximately 1The Accounting Rate System is a set of international regulations for inter-carrier op-erating agreements with the status of an international treaty. For a detailed discussion, see for example Ploman (1982), CCITT (1988), and OECD (1994). While the principles put forward by these regulations are general, the regulations note that operating agreements should take into consideration the relevant ITU recommendations, specifically the series D l to D.195. 2 Another term commonly used in this context is the term Accounting Rate. An Ac-counting Rate is the sum of the two settlement rates on a given relation, intended to reflect the total per minute cost of an international call from origin to destination. 6 1 symmetric, and that accounting rates should be cost based3. Second, many countries lack a clear distinction between operator, regulator and govern-ment. To the extent that governments can influence the bargaining process between operators, decisions are often subject to vested interests outside the operators' organizations, or even outside the telecommunications industry4. Third, even in countries with a clear institutional separation, national regu-lators may choose to intervene in the bargaining process. The ISP is an example of such an interventionist policy 5. The ISP is specifically concerned with the asymmetry of bargaining power between com-petitive US carriers on one side, and monopolistic carriers on the other side of a relation. With the objective to prevent the US carriers from being "whip-sawed"6 by foreign monopolists, it imposes three restrictions on inter-carrier bargaining outcomes: the Uniform Settlements Rule, the Even Splitting Rule and the Proportionate Return Requirement. The Uniform Settlements Rule 3 F o r detai ls on the I T U recommendat ions, refer to the I T U - D series of recommenda-t ions, in par t i cu la r D.140, D.150, D.151, and D155 ( I T U 1999). 4 M a n y countr ies, i n par t icu lar developing countr ies, use in ternat iona l telephone calls to generate " tax " revenues, wh ich are used to bu i l d up the te lecommunicat ions in f rast ructure. T h e carr iers of some developing countr ies derive up to 50% of their annua l prof i ts f rom in ternat ional sett lement payments. Others divert in ternat ional sett lement payments into the countr ies ' general budget. For this and more in format ion see www. i tu .org . 5 T h e U K government h is tor ica l ly used a set of compet i t ive safeguards to regulate inter-na t iona l te lecommunicat ions markets, but abol ished these safeguards in 1996, argu ing that the gains fo rm open compet i t ive markets outweigh the r isk that such rad ica l deregulat ion might give too much barga in ing power to monopol ies i n other countr ies. T h e B r i t i s h regu-lator however reserved the r ight to intervene subsequently, if i t finds that ant i -compet i t ive behavior between market par t ic ipants is occurr ing. (Ty ler and Bedna rczyk 1998). C a n a d a adopted a s imi lar po l icy in 1998. ( C R T C 1998). 6 T h e F C C defines "wh ipsawing" as "p lay ing U .S . carr iers against each other to the disadvantage of U .S . carriers and U .S . ratepayers" See, for example, F C C (1999b) at 8. 62 requires that the conditions of an interconnection agreement that a foreign carrier offers to one US carrier must also be available to all other US carriers. Under the Even Splitting Rule, the accounting rate is evenly split between the carriers, i.e., US carriers must receive the same settlement rate for ter-minating inbound traffic that they pay to the foreign carrier for terminating US outbound traffic. The Proportionate Return Requirement states that US carriers are al-lowed to accept incoming traffic only in the proportion of their market share of outgoing traffic. Faced with a soaring deficit in US settlement payment balance7 and with settlement rates that by far exceed the cost of international calls, the F C C has also become increasingly concerned with the absolute level of settlement rates. This concern resulted in the implementation of an additional restriction, which stipulates upper bounds ('benchmarks') for settlement rates. Under the Benchmark Restriction, US carriers are required to negotiate settlement rates at or below a level prescribed by the F C C . 8 With their ISP, the US arguably has implemented the most 'heavy handed' international telecommunications policy among all countries with competi-tive telecommunications environments. The F C C itself acknowledged some 7 T h e U S net sett lement payments amounted to $5.5 b i l l ion in 1997 ( F C C 1999a). 8 S e e F C C 1997c. Add i t i ona l l y , two supplements to the I S P prov ide that sett lement rates have to be made pub l ic ly avai lable ( F C C 1990), and that U S in ternat iona l carr iers are p roh ib i ted f rom accept ing "specia l concessions", defined as "any arrangement that affects traff ic or revenue flow to or f rom the Un i t ed States that is offered exclus ively by a foreign carr ier or admin is t ra t ion to a par t icu lar carr ier and not also to s imi la r ly s i tuated U S in ternat iona l carr iers author ized to service a par t i cu la r rou te " , see F C C (1997a). 63 of the anti-competitive effects of this approach (FCC 1996b at 7-19), but largely maintained its policy arguing that the benefits from limiting "whip-sawing" would clearly outweigh its economic cost. Analysts and researchers, on the other hand, have criticized the ISP on a number of grounds. One ma-jor point of criticism is that the benchmark requirement applies not only to US carriers but also to their correspondent foreign carriers. By unilaterally imposing benchmark rates, the F C C appears to have exceeded the bound-aries of its jurisdiction (Tyler and Bednarczyk 1998), and violated fair trade principles (Frieden 1998). More fundamental objections are raised by two theoretical studies. O'Brian (1988) finds that, contrary to common intuition, the Uniform Settlements Rule may in fact be the precise cause of the "whipsawing" problem, rather than its remedy. Hakim and Lu (1993) show that the Even Splitting Rule raises the collection charges when settlement rates are high. Both studies indicate that the Uniform Settlements and Even Splitting components of the ISP may not only cause some limited economic distortions, but also directly violate their very policy objectives. The effect of the P R R has so far received little attention by researchers. Galbi (1998) analyzes the effects of the P R R on facilities-based carriers' out-puts and total welfare in a general equilibrium framework. Scanlan (1998) assesses the effect of callback services in a country-pair market where propor-tionate return is stipulated. Using stylized accounts of international settle-ment payments, he finds that, contrary to common beliefs, a callback service 64 may benefit the monopolist in the country where callback is offered, and harm the carriers in the country from where the service is operated. 3.3 The Regulatory History of the Propor-tionate Return Requirement In order to analyze the economic effect of the P R R , it is important to un-derstand why this policy rule was developed and how it relates to the body of F C C regulations regarding international telecommunications. Although P R R was formally codified as a policy rule only in 1991 (FCC 1990, F C C 1991, F C C 1996b), proportionate return rules have been part of international interconnection agreements between foreign carriers and US record carriers at least since the 1950s9. The F C C made clear that it viewed the 1990 ruling not as a new policy, but rather as the affirmation of a "longstanding U.S. reg-ulatory policy," which had always been "imbedded" in the ISP (FCC 1996b at 24). This view is questionable. Although it is true that F C C has repeatedly stated concerns about foreign monopolists' market power over competitive US carriers over the past 60 years, it has by no means consistently applied 9In the FCC's terminology, a record carrier is a telecommunications carrier that spe-cializes in the transmission of documentary information, such as telegraphs, telegrams and telex. Record carriers traditionally used terrestrial (cable) or radio circuits for the transmission of information. Unlike in international public switched voice communications markets, which were exclusively served by AT&T until the mid 80s, the FCC has always allowed some degree of competition in markets for international record services. 65 the P R R as a policy tool to limit this market power. Rather, the P R R has developed as a generally accepted feature of international interconnection agreements, which the Commission considered to be compatible with, but not essential for, its regulatory objectives on the majority of international routes. 3.3.1 The Role of Proportionate Return in the Early Case Law The first and arguably most frequently cited F C C proceeding in the context of international telecommunications is the Mackay Radio case of 1936 (FCC 1936). Mackay Radio had filed an application to operate a direct radio circuit between New York and Oslo, a route already served by several cable carriers and one radio carrier, R C A Communications. The interconnection agreement proposed by Mackay was essentially equiv-alent to the agreement already in effect between R C A and the Norwegian Telecommunications Administration, encompassing an equal division of tolls between the two sides (FCC 1936 at 595). Nevertheless, the Commission denied Mackay's application, because it viewed the proposed contract as "contrary to the public interest" (FCC 1936 at 600). It argued that previous interconnection agreements between US cable operators and the Norwegian Telecommunications Administration had granted more favorable terms to the US side than to the Norwegian side. The proposed radio circuit, which effec-66 tively replaced an existing cable circuit, contemplated merely equal terms. The resulting loss of the US cable operators' contractual advantage over the Norwegian carrier was one of the FCC's major reasons for refusing to grant the operating permission. It is important to note that the application was denied despite a number of clauses in the proposed interconnection agreement that tried to address the potential concern over the possibility of "whipsawing" 1 0. The Commission's ruling in the Mackay Radio case was therefore not mainly aimed at limiting market power of foreign monopolists over US carriers, as is frequently implied (for example, F C C 1986), but largely an attempt to preserve the US carriers' contractual advantages over a foreign carrier 1 1. The F C C continued to deny any applications for radio circuits to countries already served by other radiotelegraph carriers until the outbreak of WWII (FCC 1951 at 696). This policy was abandoned as a wartime measure after 1939, when the Commission started to grant all applications for new circuits and later even adopted a policy of requiring parallel circuits to all overseas points. Therefore, by the beginning of the 1950s, many (but not all) foreign 1 0 B e s i d e s the prov is ion to share tol ls equal ly, the contract p rov ided a propor t ionate re turn clause ( F C C 1936 at 597) as wel l as a un i fo rm sett lement clause (Id. at 998). In other words, the proposed contract contained a l l elements that the F C C later deemed to be sufficient to prevent U S carriers f rom being whipsawed (see, for example, F C C 1996c). However, at the t ime of the Mackay proceeding the F C C concluded: " the contract provides no safeguard against the exact ion of terms f rom other A m e r i c a n carr iers less favorable to the A m e r i c a n carr iers and more favorable to the Norwegian A d m i n i s t r a t i o n " ( F C C 1936 at 598). 1 1 Ano the r ma jo r concern was the "wasteful and uneconomic dup l i ca t i on " of t ransmiss ion capaci ty, associated w i th the new c i rcu i t . 67 points were served by two radio carriers and several cable carriers. Proportionate return clauses were not necessarily part of interconnection agreements, nor did the F C C require them. Foreign carriers turned US-bound traffic to a US radio carrier 1 2 providing service to their respective country. If there was more than one radio carrier, then "such administrations divide(d) the traffic at their disposal among all United States radio carriers with which they communicate" (FCC 1955 at 1342). Consequently, the US carriers' market shares on the inbound routes were by no means proportional to their outbound market shares. Western Union, for example, transmitted 31% of all Portugal-bound traffic, but only 14% of the return traffic in 1953 (FCC (1955) at 1352). In 1951 and 1955 the F C C granted Mackay's application to operate cir-cuits to The Netherlands and Portugal, although the proposed interconnec-tion agreement was essentially identical to the one proposed for the New York - Oslo route 1 3. In its decision the Commission explicitly addressed the proposed proportionate return clause, and stated: "Since we have found competition by direct radio circuits to be beneficial, we feel that competing carriers ... should be permitted to share proportionately such inbound traffic 1 2 T h i s is because most foreign te lecommunicat ions admin is t ra t ions received less f inancial re turn f rom cable than f rom radio t ransmiss ion, (see, for example, Id. at 732). 1 3 T h e Commiss ion ' s gives a number of reasons for a r r i v ing at a different decis ion than in the Os lo case. One is that traffic volumes had increased, imp l y i ng the need for add i t iona l capaci ty ( F C C 1951 at 732). Second, the F C C argued that s tar t ing w i t h W W I I it had abandoned its str ic t "single c i rcu i t " po l icy on many routes (Id at 733). F ina l l y , the F C C pointed out the general ly posi t ive experience w i t h compet i t ion on para l le l rad io c i rcui ts ( F C C 1955 at 1340). 68 and the revenues therefrom...." (FCC 1955 at 1340) This statement is in-structive of the Commission's views at the time. It indicates that, first, the proportionate return clause was permitted, rather than required. Second, the proportionate return clause was seen as beneficial only in the context of com-petition between radio circuit carriers, but not in the context of preventing foreign carriers from "playing-off' US carriers against each other. By 1983, only a total of 15 out of 217 countries with direct interconnection to the US allocated traffic according to a proportionate return scheme (FCC 1983), and there is no indication that the F C C required this scheme to be part of the interconnection agreements14. 3.3.2 Proportionate Return and the Uniform Settle-ments Policy After the F C C opened international telephony to competition in 1985, US carriers requested a clarification of the FCC's position on competitive safe-guards in the case of telephony. In response to this request, the F C C released the Implementation and Scope of the Uniform Settlement Policy (USP) (FCC 1986), in which it essentially implemented two rules, the Uniform Settlements Rule, the Even Splitting Rule 1 5 . The F C C stated: "The policy ... limits the ability of a P T T to whipsaw US carriers" (FCC 1986 at 54). 1 4 One reason that propor t ionate re turn schemes were s t ipu la ted by on ly a smal l por t ion of foreign carr iers was that their networks were not suff iciently developed to app ly such schemes in pract ice. In any case, the point made here is that in 1983 propor t ionate re turn was not a very signif icant issue. 1 5 S e e Sect ion 3.2, supra. 69 The underlying economic rationale of the USP is based on the assertion that, without competitive safeguards, foreign carriers would play competing US carriers off against each other. They could threaten to divert traffic from one carrier to its rivals and force US carriers to pay a higher settlement rate for their outbound traffic than they receive for switching inbound traffic. A l -ternatively, even if the settlement rate were equal on inbound and outbound routes, the foreign carrier could negotiate different settlement rates with dif-ferent US carriers, and then send US bound traffic predominantly via the route with the lower settlement rate. This scheme would increase the foreign carriers revenues from settlement payments to the disadvantage of the US industry as a whole. The approach taken in the USP is not to remove the threat, but to pro-hibit US carriers from making concessions. The Commission believed that the USP would entirely eliminate the possibility of "whipsawing" 1 6. There was no need to prohibit traffic diversion, since the foreign monopolist would be left without incentive to prefer one carrier to another for routing US-bound traffic. It therefore seems logically consistent that the F C C introduced the USP as a sufficient set of competitive safeguards and refrained from also in-cluding the P R R in this new set of regulations. As we shall see, the F C C did not maintain this initially correct approach, but later added the P R R to the set of safeguards despite the fact that it was superfluous. 1 6 N o t e however O ' B r i a n ' s object ions to this v iew, O ' B r i a n (1988). 70 3.3.3 The Sprint and T L D cases Although the P R R had not been included in the 1986 USP regulation, the F C C subsequently decided to apply this principle in a number of individual cases. In 1988, Sprint filed an application for a direct interconnection agreement with Germany's monopolist D B P (FCC 1988a). The agreement principally complied with the USP, however, it raised concerns among competitors and the F C C , because D B P allowed Sprint to provide service to end-users as well as to resale carriers via the same gateway17. On the other hand, D B P re-fused to enter into direct interconnection agreements with small carriers, i.e., carriers with a market share below 13%. The F C C concluded that DBP's position effectively put Sprint into a competitive advantage over its resale carrier customers18, and reacted by establishing a number of regulatory mea-sures that were aimed at creating a "level playing field" for Sprint and the small carriers. The F C C ruled that "...unless US Sprint proposes to route a proportionate share of inbound traffic to the domestic network of its car-rier customers, we would expect the carrier rate to recognize the principle of proportionate return traffic, by providing a commensurate reduction in the price the carrier customer must pay to route its FRG-bound IMTS traffic through US Sprints switch" (FCC 1988a at 22). 1 7In this context, the term resale carriers refers to pure (swtichless) resale carriers, which do not operate leased lines. International Simple Resale, or resale via leased lines, was prohibited with Germany at the time of this case. 1 8Note that these resale carriers were Sprint's customers on the wholesale market, but at the same time also its competitors on the retail (end-user) market. 71 The Commission's ruling in the Sprint case suggests that, first, the Com-mission did not apply the P R R in order to limit DBP's market power over Sprint, but to limit Sprint's market power over its resale carrier customers. Second, this P R R exclusively concerns the market shares of Sprint's cus-tomer carriers on Sprint's circuits, not the market share of the incumbent, A T & T . Third, the P R R would be applied only if Sprint's advantage from re-ceiving inbound traffic were not reflected in its wholesale prices. Conversely, if wholesale prices had been competitive (and reflected the facilities-based carriers' advantage from receiving inbound traffic), then the P R R would not have been imposed. The P R R is therefore in this case neither a policy tool to regulate international interconnection agreements, nor a safeguard against "whipsawing" 1 9 . In 1992, the F C C had to reevaluate the service authorization of T L D (Telefonica Larga Distancia de Puerto Rico), when T L D was privatized and the majority of assets were sold to a wholly owned subsidiary of Telefonica (Spain) (FCC 1992). The Commission endorsed the Puerto Rican govern-ment's privatization efforts and recognized Telefonica as the only qualified buyer for T L D . Authorization was therefore granted despite the FCC's reci-procity policy 2 0 . However, the Commission imposed a number of competitive 1 9 T h e P R R was discussed and appl ied again in the case of F T C Commun ica t i ons , Inc., F C C (1989), where the P R R was appl ied essential ly for the same reasons as i n the Spr in t case. 2 0 U n d e r the rec iproc i ty pol icy, the F C C grants service au thor iza t ion to a foreign carr ier only if U S carr iers can obta in equivalent author izat ions in the respect ive foreign carr ier 's home country. A t the t ime of the ru l ing , Spain 's te lecommunicat ions market was not open to foreign investment. 72 safeguards, including the PRR, in order to ensure that Telefonica would not abuse its double-role as foreign monopolist and US carrier to discriminate against its competitors. 3.3.4 The Codification of the Proportionate Return Requirement In both the Sprint and the T L D case the Commission imposed the P R R un-der exceptional circumstances. The proceedings to both cases carefully and clearly lay out the particularities of each case and why it was applied as a competitive safeguard above and beyond the USP. It therefore comes as a surprise that the F C C adopted an explicit and general P R R in 1991 2 1. This policy applied to all international routes and was implemented in a compli-mentary fashion to the previously adopted set of international telecom regu-lations 2 2. The Commission justified the P R R Decision with its "longstanding US regulatory policy that US carriers should be afforded nondiscriminatory treatment in their traffic relation with a given country". This justification is flawed in a subtle yet important detail. It is true that the F C C had a longstanding policy to prevent "whipsawing", that is, 2 1 F C C (1991), hereafter referred to as P R R Dec is ion. In add i t ion , carr iers wou ld have to cert i fy that " they had not barga ined for, nor received any ind ica t ion that they wou ld re-ceive, more than their propor t ionate share of re turn traff ic" ( F C C 1990). A Propor t iona te R e t u r n Requ i rement had al ready been suggested in 1988, when the Commiss i on sol ic i ted comments concern ing the development of an " i dea l " regulatory regime for in ternat iona l te lecommunicat ions. T h i s proceeding was however closed w i thout ever adopt ing the P R R . ( F C C 1988b at 66, and F C C 1996a). 2 2 T h i s set of regulat ions had meanwhi le been renamed the In ternat ional Sett lements Po l icy . 73 discriminatory treatment of US carriers to the disadvantage of US carriers and ratepayers. However, as pointed out above, the P R R generally does not prevent "whipsawing" over and above the rules implemented under the USP. The allocation of return traffic per se (that is, which one of the car-riers receives the return traffic) does not affect the total US welfare23, since settlement rates must be uniform across carriers. Rather, it only the profits of individual carriers. The P R R therefore effectively requires adherence to some notion of fairness among carriers, a notion that the Commission had previously never defined or applied in this form. It is therefore incorrect to endorse the P R R as the "continuation of an existing policy" (FCC 1993b). Another frequently offered justification is that the proportionate return principle had been consistently applied and thus been "imbedded" in the ISP by reason of practice. Again, this justification is flawed. The cases in which the P R R had previously been applied were singular and not generalizable. They referred to particular market settings in which either no competitive re-sale market existed, or where particular foreign ownership interests prevailed (FCC 1988). Also, as pointed out above, the implementation of the P R R un-der the 1990 ruling differs significantly from the implementation under the Sprint model case. The Commission never indicated why it concluded that the reasoning in the Sprint and F T C cases applies to all other routes, or why the rules of implementation were changed2 4. 2 3 D e f i n e d as the sum of consumer and producer surplus. 2 4 T h e Commiss ion uses the above just i f icat ions also in the case of A t l an t i c Te le-Network Inc. ( A T N ) , when A T N seeks permiss ion to depart f rom the propor t ionate re turn pr inc i -74 To make the confusion complete, the Commission codified the P R R again in its 1996 Flexibility Order. The F C C states: "In this Report and Order we also codify our Proportionate Return Policy", although the P R R had already formally been codified in 1991 2 5. The repeated codification of the very same policy is not only an interesting detail of the regulatory process that shaped the ISP, but also indicates some confusion in the FCC's rulemaking process. 3.3.5 Conclusion The P R R has been called one of the cornerstones of the ISP (FCC 1996b). Yet, the Commission has provided surprisingly little and weak argumentative support for this important policy. The historic case law, on which the P R R decision is based, was largely misread and wrongly applied. The Commission failed to recognize that the P R R is superfluous as a competitive safeguard over and above the already existing safeguards. Instead, it appears that in implementing the PRR, the F C C confused the notion of economic welfare with the notion of fairness and equal treatment for US carriers. pie on the route to its monopo ly te lecommunicat ions subsid iary, G u y a n a Telephone and Telegraph. T h e case is prac t ica l ly analogous to the T L D case, and the F C C ' s refusal to grant the dev ia t ion f rom the P R R may be just i f ied. Paradox ica l l y , the F C C decides the case by defending the its general P R R pol icy w i th the usual flawed reasoning, rather than by po in t ing out the par t icu lar i t ies of the sett ing on the U S - G u y a n a route that in fact just i f ied the app l ica t ion of the propor t ionate re turn pr inc ip le in this par t i cu la r case. ( F C C 1996a). 2 5 F C C (1996b) at 4. In its jus t i f i ca t ion, the commiss ion refers to T L C and A T N cases and points out that the P R R had " long been a cornerstone of our I S P " , but fai ls to refer to the P R R Proposa l or P R R Order . 75 3.4 The Proportionate Return Requirement and the Emergence of Competition 3.4.1 Competitive and Regulatory Challenges to the Accounting Rate System As international markets become increasingly liberalized and competitive, the portion of international traffic that is routed under traditional intercon-nection arrangements is declining, and with it the portion of US traffic that is subject to ISP and PRR. Technological and regulatory change have been the two major factors driving the emergence of competition and challenged the role of the Accounting Rate System as the dominant framework for in-terconnection (Tyler and Bednarczyk 1998, Frieden 1998). Technological change has brought about six "alternative modes of op-eration" 2 6, in which interconnection arrangements substantially differ from the Accounting Rate System. Alternative modes of operation usually uti-lize economically inefficient ways to route traffic2 7. Their commercial success mainly stems from the fact that they bypass the Accounting Rate System, thus avoiding the payment of Settlement Rates that often by far exceed the real cost of international interconnection2 8. International Simple Resale 2 6 For a description of these alternative modes, see, for example, Frieden (1998), Taylor and Bednarczyk (1998), Levine (1998) and Scanlan (1996). 2 7 Alternative modes of operation evolved in a regulatory niche as services that could be offered exclusively to closed user groups, such as multinational companies. Starting in the mid-1990s, an increasing number of pro-competitive countries have allowed providers of these services to switch any type of traffic. 28Callback operators are often named in this context, however, they do not constitute an alternative mode of operation since they do not bypass the Accounting Rate System. Rather, Callback operators offer an arbitrage service within the Accounting Rate System. 76 (ISR) 2 9 , International Alliances, and Foreign Points of Presence are three alternative modes that are subject to regulatory approval and have been a driving force of competition in liberalized markets. Refile (Hubbing), Inter-net Telephony, and International Business Networks 3 0, on the other hand, are service concepts that allow consumers and carriers to bypass the Public Switched Network without prior regulatory approval. They have therefore been instrumental to imposing competitive pressures on regulators and op-erators in non-liberalized countries. The second factor that has had a profound impact on the liberalization of telecom markets is the W T O Basic Telecommunications Services Agreement. The 88 signatory countries, led by the United States, have committed in vary-ing degrees to open their markets to competition, to allow non-discriminatory access to foreign operators, and to remove the regulations that limited ser-vices provision under alternative modes of operation. As a result, flexible arrangements rather than the framework of the Accounting Rate System will shape future interconnection agreements between operators of these coun-tries. The Accounting Rate System has also increasingly become subjected to W h i l e ca l lback operators put signif icant pressure on establ ished telecoms to lower col lect ion charges and cost levels, they do not const i tute a challenge to the Accoun t i ng R a t e Sys tem per se. 2 9 In ternat ional S imple Resale involves the use of a pr ivate l ine (wi th a connect ion to the publ ic swi tched network at one or bo th ends) for the prov is ion of commerc ia l in ternat iona l telephone services. 3 0 A t the beginn ing of the 1990s, in ternat ional businesses s tar ted to route their inter-branch voice communicat ions v i a leased l ine da ta networks, and thereby considerably reducing its use of in ternat iona l P S T N services. 77 heavy criticism by academics31, and by regulators of countries with a large net payment deficit 3 2. Responding to these various pressures and criticisms, a number of multinational organizations including the ITU and O E C D have been working on a reform of the Accounting Rate System. 3.4.2 Regulatory Adjustments of the ISP The liberalization and growing competition in various national and inter-national markets during the past decade has raised concerns about adverse economic effects of the ISP on emerging competitive environments. One of the most prominent concerns is that the P R R may impede competitive behavior in outbound and inbound markets. The F C C recognizes three ways in which this may happen. First, the P R R may generally deter U.S. terminating carriers from offering innovative pricing and supply arrangements. Second, it may constitute a barrier to entry. When 3 1 It has been well understood that the structure of the Accounting Rate System, in conjunction with above cost settlement rates, creates economically inefficient incentives, by limiting the operators' incentive to lower collection charges, and by limiting net recipients' incentives to promote output growth. (See, for example, Tyler and Bednarczyk (1998), p. 802). 32See footnote 9, supra. Other major deficit countries are, for example, Switzerland, Australia and Japan (ITU 1998b). A number of economic studies have tried to establish a theoretical and empirical link between the inefficiencies due to the Accounting Rate System and international traffic imbalances. Ergas and Paterson (1991) identified the Accounting Rate System as the major cause of these imbalances. Chong and Mullins (1991), on the other hand, empirically showed that traffic imbalances may be related to long-term economic factors, rather than price-related factors. Domestic regulatory policies, not the Accounting Rate System per se, are the underlying determinants of traffic imbalances. Yun, Choi and Ahn (1996) developed a Cournot competition model to show how differences in demand, technology and market structure affect Accounting Rates as well as traffic and payment imbalances. They demonstrated that in some cases, high Accounting Rates might even have the role of reducing traffic imbalances. 78 entrants start serving a route, they have no past record of outbound traffic and thus no credits for the lucrative inbound traffic. For an initial period after entry, therefore, the new carrier will face a competitive disadvantage (FCC 1998a). Also, most foreign carriers will not start sending a US carrier the lucrative return traffic until that US carrier's outbound traffic volume reaches a certain threshold. Third, the Commission points out that inbound and outbound markets have different attributes, and that carriers differ in their effectiveness in serving these attributes 3 3. These and other concerns led to a series of regulatory adjustments of the ISP 3 4 . The main premise of the adjustments was that the US regulator did not intend to generally relax its regulatory oversight in international mar-kets 3 5. Rather, it decided to selectively allow for flexible interconnection arrangements where it perceived no potential for abuse of market power by 3 3 F C C (1996b). Ano ther basic concern, raised by A T & T , is the enforceabi l i ty of this pol icy. If foreign monopol is ts wanted to favor one U S carr ier over another , they could do this w i thout formal ly v io la t ing the P R R , for example, by sending the favored carr ier higher propor t ions of non-peak hour traff ic, more valuable operator -handled cal ls, or by sending calls that can be terminated in the U S over short distances. ( F C C 1998a, A T & T 1998 at footnote 12 and Spr in t 1998 p.4). 3 4 I n add i t ion to the specif ic concerns associated to the P R R , the Commiss i on also rec-ognizes a number of economic distor t ions caused by the U n i f o r m Sett lements R u l e and the even sp l i t t ing rule. ( F C C 1996b at 18). For example, the I S P cou ld reduce a carr ier 's incent ive to barga in for the lowest possible sett lement rate. Due to the U n i f o r m Sett le-ment R u l e , the carr ier w i l l be unable to achieve a cost advantage over i ts compet i to rs , no mat ter how aggressively it negotiates. Moreover , under the I S P , sett lement rates have to be made pub l ic , thus reveal ing a signif icant part of the carr iers ' cost s t ructure to their compet i tors and inh ib i t ing compet i t ion . T h e object ive of the regulatory adjustments was to e l iminate economic distort ions associated w i th the ISP , whi le at the same t ime foster ing compet i t ion and thereby impos ing downward pressure on sett lement rates. ( F C C 1996b at 17) 3 5 T h e F C C approach sharp ly differs f rom the one adopted i n other pro-compet i t ive countr ies, such as B r i t a i n and Aus t ra l i a . See also footnote 3, supra . 79 foreign carriers, and to continue enforcing the ISP on all other routes. In pursuit of this policy, the F C C developed a set of rules that define market conditions under which the ISP will be waived. These rules have been con-tinuously fine-tuned and adapted to the various technological and regulatory developments that have challenged the Accounting Rate System and the reg-ulatory framework surrounding it. Three regulatory decisions in this series of adjustments are directly concerned with the ISP. The Flexibility Order (FCC 1996b) permitted interconnection arrange-ments that deviated from the ISP between US carriers and any foreign cor-respondent in of countries that satisfied the ' E C O ' 3 6 test. For countries that failed to pass the test, a deviation from the ISP was permitted only if "the US carrier can demonstrate that deviation from the ISP will promote market-oriented pricing and competition, while precluding abuse of market power by the foreign correspondent."37 (FCC 1995 at 2) In 1997 the Commission revised the Flexibility Order in order to bring the ISP in accordance with the open market commitments made under the W T O Agreement. The E C O test was abandoned and replaced by a "presumption 3 6 The "Effective Competitive Opportunities" (ECO) test attempts to determine the degree of competitiveness of a county's telecommunications industry, taking into consid-eration a number of regulatory and market conditions in a given country, including the degree of liberalization, the existence of non-discriminatory entry conditions, the exis-tence of an effective regulatory framework and appropriate competitive safeguards, see FCC (1996) at footnote 65, or FCC (1995). 3 7 The FCC however continued to (1) impose a filing and publication requirement; (2) maintain the right to intervene in arrangements affecting more than 25% of inbound or outbound traffic, if these contained 'unreasonable discriminatory terms and conditions'. (FCC 1996b). 80 in favor of alternative settlement arrangements on routes servicing W T O Members", which could however "be rebutted with a showing that there are no multiple facilities-based competitors operating in the foreign market for international services" 3 8 . Finally, a further set of regulatory rulings was released on May 6, 1999 as part of the FCC's 1998 biennial review of regulations3 9. In the Biennial Review Order the F C C abolished the Flexibility Order and provided that the ISP will be revoked in two circumstances: (1) for settlement arrangements between US carriers and foreign carriers that lack market power4 0; and (2) on routes where US carriers are able to terminate at least 50% of the their US billed traffic in the foreign market at settlement rates that are at least 25% below the applicable benchmark settlement rate 4 1 , 4 2 . 3 8 F C C ( 1 9 9 7 a ) at 20. 3 9 F C C ( 1 9 9 9 b ) , i n the fo l lowing referred to as B ienn ia l Rev iew Order . 4 0 T h e F C C presumes that " a carr ier that possesses less than 50% market share in a foreign market lacks the ab i l i ty to exercise market power in that marke t " ( F C C 1999b at 39). 4 1 T h i s includes a l l rates for te rmina t ing traff ic, inc lud ing rates of In ternat ional S imple Resale ( ISR, resale of services v i a pr ivate leased lines) and sett lement rates. T h e F C C ' s I S R restr ict ions obv ious ly became superf luous under c i rcumstances where the I S P no longer appl ies. ( F C C 1999b) Under these restr ict ions, carr iers may engage in I S R on routes to W T O signatory countr ies only where "sett lement rates for at least 50% of the sett led, U .S . b i l led traff ic on the route are at or below the appropr ia te benchmark or where the foreign market offers equivalent resale oppor tun i t ies" . For service to n o n - W T O signatory countr ies, I S R is author ized on ly "where 50% of the traffic is sett led at benchmark rates and where the foreign market offers equivalent resale oppor tun i t ies" ( F C C 1999b at 15). Hence there are routes where the F C C allows I S R , whi le at the same enforcing the ISP. O n such routes, faci l i t ies-based carr iers have the choice of negot ia t ing an I S R agreement w i th the foreign correspondent carr ier and rout ing its traff ic v i a I S R , or of car ry ing traffic under a t rad i t iona l sett lement arrangement that is subject to the ISP . It is a pecul iar feature of the I S P that it main ta ins a d is t inc t ion between I S R and " t rad i t iona l sett lement arrangements" even where this d is t inc t ion has become blurry, if not meaningless. 4 2 I n the Not ice of P roposed Ru lemak ing ( F C C 1998a) leading to the B ienn ia l Rev iew 81 With their recent adjustments of the ISP rules, the F C C followed mainly three objectives. First, the F C C has eliminated the ISP rules where they have become insignificant due to the realities of a competitive telecom en-vironment 4 3. Second, the Commission has attempted to reduce the adverse effects on emerging competition and new entrants in selected markets, by permitting alternative settlement arrangements with non-dominant foreign carriers. Third, the F C C believes that the Biennial Review Order provides foreign carriers with a significant incentive to lower their settlement rates be-low benchmark level. "As competitive pressures develop in foreign markets, foreign carriers will have an incentive to lower their rates to take advantage of increased opportunities to enter into innovative arrangements as a result of lifting the ISP" (FCC 1999b). Order , the F C C had or ig ina l ly proposed to waive the I S P (1) on routes between the U S and W T O signatory countr ies where foreign carriers are author ized to prov ide I S R , or where the correspondent carr iers lack market power. (2) for in terconnect ion agreements between U S carr iers and carr ier f rom W T O s ignatory countr ies, i f these affect less than 2 5 % of the traffic on the given route. T h e re laxat ion of the I S P made in the B ienn ia l Rev iew Order tu rned out to be much less far reaching than the re laxat ion or ig ina l ly proposed. T h e fact that the F C C largely main ta ined the I S P ma in ly goes back to the comments of the large faci l i t ies-based carr iers A T & T and M C I W o r l d C o m , wh ich c la imed that the proposed modi f icat ions wou ld leave the U S consumers w i th too l i t t le protect ion f rom foreign market power. ( F C C 1999b). 4 3 F o r example, in ternat ional routes that qual i f ied for a complete removal of the I S P as of M a y 1999 were the routes to C a n a d a , the U K , Sweden, Germany , France, H o n g K o n g , the Nether lands, D e n m a r k and Norway. ( F C C 1999b at 55). 82 3.4.3 The Current and Future Regulatory Impact of the ISP Despite this relaxation of regulatory oversight and despite the growing impor-tance of alternative interconnection arrangements, the impact of Accounting Rate System, ISP and P R R remains significant. It is true that, as of May 1999, a total of 43 governments had formally committed to opening their markets to International Simple Resale (ISR) 4 4 by the year 2010, reducing the significance of traditional interconnection arrangements on routes to their respective countries. However, it is unclear whether these governments will meet their commitments in time. The liberalization process in the W T O signatory countries has been slower than anticipated. As of Apri l 1999, the F C C listed 22 countries as approved for ISR (FCC 1999a). 9 out of the 31 countries that committed to open their markets to ISR immediately failed to do so by F C C standards. Between July 1999 and May 2000, the F C C classified merely another 3 countries as ISR-approved, namely Singapore, the Netherlands Antilles, and Poland. While settlement rates have already dropped significantly on most inter-national routes, only few settlement rates have reduced to a level that is even close to the FCC's benchmarks. As of August 1999, the average settlement rate was 67 cents per minute on the 51 routes between the US and W T O sig-natory countries that are subject to the ISP. The FCC's benchmarks ranged 4 4 T h e s e 43 Governments represent countr ies that account for 69% of the to ta l U S out-bound traff ic. 83 from 15 to 23 cents per minute, depending on the correspondent country's level of development On 42 of those 51 routes, the settlement rates exceed the benchmark rates by more than one hundred percent. As of May 2000, the ISP restrictions have been completely removed on routes to only 11 countries (FCC 2000), up from 9 countries in 1999 (FCC 1999b at 55). The 1998 US outbound traffic volume on routes between the United States and these countries was 8,828 million minutes, or 36% of the total outbound (settled) traffic. On 25 routes, either ISP or ISR restrictions have been removed. 11,602 million outbound minutes, accounting for 47% of the total US outbound traffic, are transmitted on these routes. The US carriers' revenues from all routes with both ISP and ISR restrictions still in place amount to US$ 5.2 billion, or 55% of the total P S T N revenues earned by US carriers 4 5. Despite all pro-competitive rhetoric, even the FCC's approach towards market liberalization can be described as cautious. The underlying premise of the ISP is clearly to allow for full competition and alternative interconnection arrangements only on routes to countries that have liberalized their market as well. On all other routes, the Commission has considered the potential harm of market power abuse by foreign carriers to be greater than the potential benefits of competition. Hence, the US regulatory approach has to a large extent not attacked, but conserved the structure of the Accounting Rate System including its anti-competitive effects. 4 5 FCC (1998b). 84 3.5 The Economic Effect of Proportionate Re-turn in Switched Service Markets The recent modifications of the FCC's regulatory framework constitute a significant step toward removing unnecessary regulations and impediments to emerging competition created by the ISP. The Commission has however stopped short of identifying and eliminating all the economic distortions cre-ated by the PRR. One immediate consequence of linking the inbound traffic allocation to outbound market shares is that facilities-based carriers take the profits asso-ciated with the inbound traffic in consideration for their output and price de-cisions on the outbound markets. In order to capture these profits, facilities-based carriers try to expand their outgoing minutes beyond the level they would choose without consideration of inbound calls in their output deci-sion. This distortion aggravates the US traffic imbalance and affects both the wholesale and the end-user market. In the end-user market, the P R R creates a (downward) pressure on calling charges. In the wholesale market, it reduces the wholesale prices that facilities-based carriers charge to pure resellers46 for the transmission of international traffic. Galbi (1998) points out that the output distortion created by the P R R will act in the opposite direction of the output distortion from imperfect com-4 6 P u r e ("Switchless") resellers do not own or lease facilities. Instead, they route traf-fic into the facilities-based carriers' network in exchange for an agreed-upon per minute transmission charge. Pure resellers neither directly pay settlements to foreign carriers, nor do they receive any return traffic. 85 petition. In imperfectly competitive markets, the P R R can therefore serve as a regulatory device to bring prices in line with marginal cost. In a near competitive market, on the other hand, the P R R may push wholesale prices and retail prices below marginal cost of outbound service. Facilities-based carriers then cross-substitute outbound service with profits from inbound ser-vice, thus forgoing profits from inbound service which they could have made without proportionate return traffic allocation 4 7. Moreover, the P R R has a negative effect on the retail (end-user) market shares of facilities-based car-riers. Market shares of facilities-based carriers will be lower and the market shares of the resale industry resale industry will be higher under the PRR, than without it. While consumers still profit from the P R R under this sce-nario, the overall welfare effect will be negative 4 8 . The economic evaluation of the P R R therefore largely depends on the degree of competitiveness in the international switched service market. The recent developments in the United States indicate that the telecommunications industry is rapidly mov-ing towards a high degree of competition, particularly in the resale segment of the market. In 1996, the combined (retail) market share of all US pure resale carriers accounted for 37% of the total outbound call market, up from 6.3% in 1993 (FCC 1997b; Blake k Lande 1998). 4 7 A n alternative return traffic allocation scheme is, for example, a scheme under which each US carrier receives a fixed share of return traffic. Clearly, a US carrier that would be supplied with little return traffic or no return traffic at all under such a fixed allocation scheme would be worse off than under the P R R . Nevertheless, the facilities-based service industry as a whole is better off under a fixed allocation scheme. 4 8 A detailed formal analysis of the economic effect of the P R R is provided in Chapter 4. 86 The number of resellers is large, and only few resellers have succeeded in obtaining a significant market size. In 1996, of the estimated 1000 or more pure resale carriers, the largest 2 served 12.3%, and 7.3% of the total pure resale market, respectively. Only 19 pure resellers had a market share of more than 1%; the typical market share of a pure resale carrier was be-low 0.01%. Wholesale competition between the three major facilities-based carriers serving these resale carriers is fierce, such that, in some instances, wholesale charges have been reported to be below the termination price at the corresponding destination (Galbi 1998). This price distortion can be di-rectly attributed to the P R R and indicates that the P R R may already have adverse welfare effects, at least on some international routes4 9. The effect of the P R R on the correspondent foreign monopolist is less ob-vious. If the foreign monopolist's demand were independent of US wholesale rates and end-user charges, then the foreign monopolist would clearly ben-efit from the proportionate return allocation, since this rule would result in higher outbound traffic volumes and thus higher settlement revenues for the foreign monopolist. However, even with demand interdependencies, and par-ticularly of callback services, there is some evidence that foreign monopolists may still profit from proportionate return traffic allocation. 4 9 I t is impor tan t to emphasizes that the effect of resale, pa r t i cu la r l y of pure resale on the market outcome under the P R R is an indirect . A s resellers do not have direct in terconnect ion agreements, they are not subject to the P R R . However, resellers compete w i th faci l i t ies-based carr iers in end-user markets. T h e h igh degree of compet i t ion in these markets , resul t ing f rom the presence of resellers, then in tu rn aggravates the effect of the P R R . 87 Scanlan (1998) demonstrated that in the presence of the PRR, callback might in fact positively contribute to the profitability of monopoly operators in countries where a callback service is offered50. If this is the case, than the P R R effectively strengthens, rather than weakens, foreign monopolists' position vis-a-vis the US facilities-based carriers, and foreign operators may have an incentive to use proportionate return schemes even without an ex-plicit requirement to do so. The appropriate regulatory reaction would be to prevent proportionate return allocation, rather than enforce it. While more comprehensive research needs to be done in this field, it is reasonable to state that the P R R creates a number of economic distortions of outputs and prices, which are not taken into consideration in the relevant regulatory rulemaking (FCC 1999b), and which may counteract the very objectives of the ISP. 3.6 Summary and Conclusion The Proportionate Return Requirement is a regulation that governs the al-location of US inbound telephone traffic in a significant, albeit decreasing portion of the US international P S T N market. It was implemented in order to prevent foreign monopolistic carriers from "whipsawing" competitive US carriers. Although the P R R has been removed on a number of routes, the 5 0 S c a n l a n arr ives at th is result by employ ing a mode l of s ty l ized accounts of in ternat iona l revenue streams, tak ing into considerat ion the various elast ic i t ies and feedback effects impac t ing in ternat iona l telephone demand. 88 F C C has made clear that it does not intend to remove the P R R on routes to non-liberalized countries unless settlement rates on those routes fall signifi-cantly below the FCC's benchmark rates. The objective of this paper is to provide an analysis of the regulatory history of the PRR, its economic rationale and its effect on the US telephone industry. The main conclusion obtained from this analysis is that the actual economic effects of the P R R are not in line with its regulatory objectives. On one hand, the P R R is not needed to prevent "whipsawing" of US carriers by foreign monopolists. The FCC's justifications of the P R R are largely based on an erroneous interpretation and application of historic case law. On the other hand, the P R R creates excessive competition for incoming traffic. This competition aggravates the US traffic imbalance, distorts price and output decisions of facilities-based carriers and resellers, and may ultimately benefit the foreign monopolists. As US markets for outbound international calls rapidly move towards more intensive competition, the P R R may have a negative effect on total US welfare in the near future. Large carriers, such as A T & T , M C I and Sprint have traditionally supported the F C C on their International Settlements Pol-icy (including the implementation of the PRR) , and even opted for an adap-tation of similar rules in Canada, although, ironically, the P R R reduces the US facilities-based carriers' industry-wide profits and market shares. There are four plausible reasons that could explain why the P R R has nevertheless been created and has existed for the past decade. One is that 89 neither the F C C , nor the large facilities-based carriers has fully understood the implications of this regulation. In this view, the P R R is a temporary regulatory inconsistency that is caused by miscommunication and short "or-ganizational memory" within the F C C and the large carriers, in conjunction with fast technological change and complexity of regulatory tasks. Second, the persistence of the P R R could indicate an agency problem within the facilities-based carriers' organizations. By tying together incoming and out-going telephone markets, the P R R significantly reduces the number of man-agerial decision parameters to be considered relative to a situation where both markets would have to be managed separately. Furthermore, the P R R protects management from having to take responsibility for incoming market shares, which are largely due to the uncontrollable actions of foreign carriers. Hence, in view of the management of large carriers, abolishing the P R R may lead to an increase in managerial complexity and risk at little if any payoff for the managers. The managers may personally prefer an established regulatory system under which all or a part of the companies' potential windfall profits from incoming traffic are competed away. A third explanation, already noted and explained in section 3.4.2, is that established carriers may try to use the P R R as a means to prevent entry. Finally, the F C C could have established and maintained the P R R as a regulatory device to bring wholesale and retail prices of outbound calls in line with marginal cost. If this were the case, the F C C would have failed to publicly convey the true regulatory objectives of the P R R . 90 Chapter 4 The Economic Effect of the Proportionate Return Requirement 4.1 Introduction Throughout the last decade, the Federal Communications Commission (FCC) has been paying considerable attention to the reduction of settlement rates in international telecommunications markets. Settlement rates1 have tradition-ally been negotiated bilaterally between international telecommunications carriers. Where carriers from a country with a competitive telecommunica-tions market interconnect with a telecommunications monopolist, the foreign monopolist has an incentive to play the competitive carriers against each other on the terms and rates of interconnection. The F C C has identified this asymmetry of market power in international telecommunications mar-1 A settlement rate is a per-minute charge, paid by an international carrier to a foreign carrier for terminating an international telephone call in its respective country. 91 kets as the leading cause of inflated settlement rates2. In conjunction with the prevailing traffic imbalances3, high settlement rates have led to signifi-cant wealth transfers from US carriers and consumers to foreign monopoly carriers. The FCC's regulatory responses to the market power asymmetry had two major thrusts. Starting in 1986, the F C C designed a set of restrictions on settlement rates that effectively formed a cartel between the competitive US carriers against the foreign monopolists. These restrictions, also called "Uni-form Settlements Policy" 4 , provided equal settlement rates among carriers ("uniform settlements rule"), same level of settlement rates for inbound and outbound traffic ("equal sharing rule"), and the sharing of incoming traffic among US carriers in proportion to their respective market share of outgo-ing traffic ("proportionate return requirement (PRR)") . By 1997 it became clear that in spite of the Uniform Settlement Policy, settlement rates were getting increasingly out of line with actual cost levels. As a result, the F C C further introduced "benchmarks" for international settlement rates5. These 2 I n their 1997 Benchmark Order ( F C C 1997), the F C C stated: " U . S . consumers pay on average 88 cents per minute for in ternat ional calls and they pay on average 13.5 cents per minute for domest ic long distance cal ls. Ye t , the difference in cost of the under ly ing faci l i t ies between the two services is m in ima l . Indeed, as we stated in the Not i ce , the costs of p rov id ing telephony have been decreasing and are becoming v i r tua l l y distance insensit ive due to recent technological advances." 3 U S outgoing traff ic volumes are wel l above incoming traffic volumes on most in terna-t iona l routes. 4 T h e key F C C orders are F C C ( 1 9 8 6 ) and F C C ( 1 9 9 1 ) . Fo r more deta i led in forma-t ion about a l l relevant F C C decisions w i th regard to the Un i f o rm Sett lements Po l i cy , see Chap te r 3. 5 S e e F C C (1997). 92 benchmarks capped settlement rates to levels that the F C C perceived as reflecting foreign carriers' actual cost levels. Ever since the implementation of international telecom regulations, eco-nomists have debated these measures and tried to assess whether they re-ally help to achieve the intended regulatory objectives. O'Brian (1988) uses a Rubinstein-type bargaining model to analyze the effect of the Uniform Settlements Policy on the inter-carrier bargaining process. He finds that, contrary to common intuition, the Uniform Settlements Rule may in fact be the precise cause of the "whipsawing" problem, rather than its remedy. Hakim and Lu (1993) show that the equal sharing rule raises the collection charges when settlement rates are high. Cave and Donnelly (1996) argue that under the settlement rate system, rates often end up at the higher of the rates proposed by the two parties. Yun, Choi and Ahn (1996) develop a Cournot competition model to show how differences in demand, technology and market structure affect settlement rates as well as traffic and payment imbalances. They demonstrated that in some cases, high settlement rates might even have the role of reducing traffic imbalances. Among the few studies that explicitly consider the P R R , Scanlan(1999) analyzes the incentives to provide callback services under the P R R , and Galbi (1998) discusses some welfare effects of proportionate traffic allocation. F i -nally, Chapter 3 of this thesis argues that the P R R is in fact not needed over and above the uniform settlement rate and the equal sharing rule as a com-petitive safeguard against foreign monopolists. If settlement rates are forced 93 to be equal for all carriers and in both directions, there are no economic in-centives left for the monopolist to prefer one carrier over the other, nor would a traffic diversion have any effect on the magnitude of settlement rates. Also, Chapter 3 argues that the regulatory process leading to the codification of the P R R exhibits a number of flaws and inconsistencies, such that there ex-ist reasonable doubts about the proper justification of this regulatory rule. The questions arising are therefore: what impact does the P R R have on the different economic entities involved in the international telecommunications service markets6, and to what extent may these entities have an interest in a continued P R R regulation. The objective of this paper is to investigate these questions by supple-menting the institutional analysis in Chapter 3 with a formal economic anal-ysis of the P R R . Extending international telecommunications competition models by Sharkey (1994) and Yun, Choi and Ahn (1996), it develops a two-stage model of inter-national telecommunications markets. This model is used to simulate market outcomes with 1998 data of the US international telephone industry. At the center of this analysis is not the interaction between the carriers of 6The economic entities affected by the PRR are the consumer, international telephone carriers owning transmission facilities (facilities-based carriers, fb-carriers), international switchless resale carriers (resellers), and the government. Fb-carriers assume a dual role in international telephone markets, as providers of end-user (retail) service, and wholesale service for resellers. Under the PRR, they receive the incoming traffic in proportion to the total volume of traffic they transmit to a foreign country, including the traffic originnating from resale carriers. Resellers generate a significant share of international outbound telephone traffic. Yet, they do not receive any share of incoming traffic. 94 two countries, as in the studies by Yun, Choi and Ahn. Rather, it focuses on the effect of proportionate return traffic allocation on the existence of equi-libria and on their properties in the country where the P R R is implemented. In addition, this paper analyzes the discriminatory effect of the P R R on re-sellers, and the extent to which fb-carriers can use the advantage of receiving incoming traffic to increase their market share at the expense of resellers in the end-user market. Section 4.2 presents the basic structure ofthe model. Section 4.3 provides a general analysis of international telecommunications markets under the P R R . It shows that the P R R may preclude the existence of equilibria, if fb-carriers and resellers incur different marginal costs of providing retail service. Welfare implications and some comparative statics results are discussed for the special case in which both types of carriers face the same marginal service cost. Section 4.4 uses simulation techniques to illustrate equilibrium outcomes, and the effect of the P R R for the general case where fb-carriers and resellers incur different costs. It demonstrates that realistic parameter ranges exist where there is no equilibrium. If an equilibrium exists, then equilibrium prices and wholesale charges are lower under the P R R than they would be without it. Also, the resellers' market share is higher under the P R R , in spite the fact that they are apparently discriminated against regarding the alloca-tion of incoming traffic. The simulation results also provide support for the claim that, although fb-carriers have continuously been in favor of the PRR, 95 their profits may in fact be higher under a regulatory regime without pro-portionate return traffic allocation. Section 4.5 discusses policy implications, summarizes and concludes. 4.2 A Model of International Switched Ser-vice Competition under Proportionate Re-turn Consider a country pair market for international switched services between two countries, / (foreign) and h (home). The market in / is served by a monopolistic telecommunications provider. The settlement rate is considered to be exogenous7. The telecommunications industry of h is competitive, but subject to com-petitive safeguards on the route to and from / . These safeguards are of the kind imposed by the F C C . In particular, they require settlement rates to be uniform among carriers in h, and equal for carriers in h and / . Furthermore, regulation requires return traffic to be allocated among the home country's carriers in proportion to their share of outgoing traffic (PRR). There is no International Simple Resale (ISR) 8. 7However, the carriers' incentive structure for setting settlement rates in the context of this model will be discussed below. 8 International Simple Resale involves the use of private lines (with a connection to the public switched network at one or both ends) for the provision of commercial interna-tional telephone services. ISR interconnection agreements are neither subject to the FCC's uniform settlements policy, nor to the benchmark order. Note, however, that the FCC prohibits ISR on routes to countries with non-liberalized telecom markets. In particular, 96 Figure 4.1. International telecommunications relation between a com-petitive market and a monopoly market The end-user market in h is served by two types of carriers: resellers (de-noted by superscript r ) , and fb-carriers (denoted by superscript fb). Resellers do not operate international transmission facilities. Instead, they exclusively offer end-user (retail) service, and route traffic into the fb-carriers' network in exchange for an agreed-upon per minute transmission charge. Fb-carriers, on the other hand, offer both end-user and wholesale service. They own or lease international transmission facilities and have direct interconnection agreements with foreign carriers. Figure 4.2 depicts the market structure in the case of two fb-carriers (FBI, FB2), two resellers (RI, R2), and one monopoly carrier ( M ) in the foreign market. Fb-carriers derive profits from 3 distinct sources, domestic retail, I S R is proh ib i ted on most (but not all) of the routes that are subject to the P R R . 97 domestic wholesale, and from switching incoming traffic. TR = TR 4- T V i i,retail ' i,- i, incoming (4.1) whereas resellers derive profits only from their retail operations IT = HT retail (4.2) Without the PRR, the profits from incoming traffic could be simply regarded as a windfall profit from regulation, a fixed amount of money that could be retained by the government9, allocated to carriers according to some fixed sharing scheme, or the allocation scheme for the return traffic could be en-tirely left up to the foreign monopolist. No matter how it is distributed, the total surplus from incoming traffic is maximization problem as a constant and hence play no further role. Under the PRR, the profits from incoming traffic are allocated among 9For example, the FCC could negotiate all settlement rates with foreign monopolists, and then auction the incoming traffic off to the fb-carriers that is willing to switch incoming traffic at the lowest per-minute charge. cl)Qf (4.3) were s is the settlement rate, cl is the transmission cost, and is the (fixed) total incoming traffic volume. T[fnc would then enter the carriers' 98 fb-carriers according to their relative market shares: n * ; = q-^F~ • (s - <?)Qf (4-4) where qf is fb-carrier i's output, Qr is the total output of all resellers, fa is the fb-carrier i's market share of the resale market, and Qh is the total home country output, i.e., the sum of all resellers' and fb-carriers' outputs. To complete the model specification, the game theoretic framework used to model the interaction between the carriers in h is represented by the following two-stages. In stage one, fb-carriers set the per-minute "wholesale" transmission charge for the telephone calls they transmit on behalf of their reseller clients, assuming Bertrand competition between the fb-carriers in the wholesale mar-ket. In stage two, both fb-carriers and resellers compete in quantities on the retail (end-user) market 1 0. 1 0 The assumption Cournot competition in stage two of the game allows for equilibria where the fb-carriers' and resellers' market shares differ, even when they charge the same collection charge. The solution to stage two profit maximization problem will however be algebraically complex. For instance, the first order condition to the profits from incoming traffic alone (equation 4.3) will be a polynomial of degree 3 in qf'. The assumption of Bertrand competition in stage one of the game then has two merits. One is that it retains the tractability of the model. Under the Bertrand assumption, optimal wholesale charges of stage one can be computed without having to maximize a prohibitively complex profit function. Second, the assumption of Bertrand competition in the wholesale market reflects the realities of these markets reasonably well. In this wholesale market, a few fb-carriers with large excess capacity are offering wholesale service to a large number of resellers, and resellers can instantaneously and costlessly switch from one wholesale service provider to 99 The most general form of the fb-carriers and resellers' profit functions can then be written as -(pfb(Qr,Qfb)-cfb-ct-s)q! + (w- cl- s) faQr+ = (pr(Qr,Qfb)~cr- w)q] (4.5) (4.6) for fb-carriers i = 1,..., m and resellers j = 1,..., n. Analogously to equa-tion 4.1, the three terms of the fb-carriers' total profit function represent the profits from retail service, wholesale service, and incoming traffic. The fb-carriers' and resellers' calling charges are represented by p^ and pr, re-spectively. Resellers face two types of cost, the retail service cost, and the wholesale transmission charge, levied by the fb-carriers. The retail service cost, cr, is the long run marginal cost of operating the local switch to handle calls, and transmitting these calls to a fb-carrier. The wholesale transmission charge, w, is a per-minute price a reseller pays to fb-carriers for having its calls routed to the recipient in the call destination country. Under the assumption of Bertrand competition in the wholesale market, the resellers will route all their traffic via the cheapest available network connection, hence w = min^-fu^}, where Wi is the wholesale transmission charge levied by fb-carrier i. another, if a better wholesale rate is offered. 100 Consequently, fb-carrier i's share of the wholesale market, fa, can only take the following discrete values: 1 if Wi < wk for all fb-carriers k ^ i where r is the number of facilities-(4.7) r if Wi = W based carriers that charge w 0 otherwise The basic structure of the model is similar to a model with an upstream (wholesale) and a downstream (retail) market, where the upstream firms also directly sell to the retail customers. The unusual feature of this setting is that the incoming call market is both price regulated and tied to the out-going call market. By expanding service in the end-user market, fb-carriers can therefore not only increase their revenue from the end-users, but also their share of the incoming traffic. Alternatively, they can attract more traf-fic from resellers (i.e., more wholesale business) in order to increase their share of incoming traffic1 1. It is precisely this sharing rule, which confounds the analysis of international telephone markets, and leads to a number of surprising and counterintuitive market outcomes. 1 1 It is important to note that revenues from switching incoming calls are exorbitantly high on routes to countries with non-liberalized markets, more so than switching outgoing calls. The "best-practice settlement rates", which can be assumed to reflect an upper bound to the true cost of switching inbound traffic, are currently at $0.10 per minute (FCC 1998), compared to a (worldwide, non-weighted) average settlement rate of $0.48 per minute received for switching incoming calls. 101 The following restrictions are necessary in order to make the model tractable: (1) End-user telecommunications service is assumed to be a homoge-nous product - hence, resellers and facilities-based retail prices are identical, i.e. p(Qh) = pr(Qr,QJb) = pSb{Qr,QIb). The two types of carriers differ only with respect to their cost levels 1 2. (2) The resellers marginal cost are at least as high as those of the fb-carriers, cr > c^13. (3) The demand for telephone service is linear, i.e., p = a — bQh (4) The number of fb-carriers serving a given inter-country route is typically small compared to the number of resellers14. This model will focus on the limiting case where the number of fb-carriers is finite and the number of resellers approaches infinity. 1 2 In reality, resale service typically differs from fb-carriers' service in at least two aspects. One aspect is that the use of resale service imposes various "nuisance" cost on the end-user, such as dialing into the resellers switch, dialing pin-numbers, or buying pre-paid phonecards. In compensation, resellers levy lower per-minute calling charges then fb-carriers. This model collapses all of the above differences between fb-carriers and resellers into the differences between their the retail service cost, and c r. One could hence interpret this setting as one where the price reflects both the end-user charge and the effort of using the service, and where the reseller grants the end-users a fixed per-minute rebate that exactly compensates for the nuisance of using the resale service. This rebate would then be reflected in the resellers' higher cost. 13Although this model does not incorporate fixed cost, it assumes that the resellers'long run marginal cost of servicing the end-user (i.e., the retail service cost) are higher than those of the fb-carriers, due to smaller-sized and less efficient technology. 1 4 The largest three fb-carriers, AT&T, MCI Worldcom and Sprint together account for 95.6% of the 1998 international facilities-based service market (based on net revenues) (FCC 1999). On the other hand, the number of resellers reporting to the FCC in 1998 was 349. Most of these carriers have a market share of below 1% of the US resale market. 102 (5) The wholesale charges Wi are non-negative. If w were allowed to be negative, then resellers would be able to earn money by generating telephone calls. Hence, the profit functions of equation 4.5 and 4.6 can be rewritten as T T / 6 =(a-bQh-c»-ct-s)qt + (« , , - c<- a) faGf + (q!b+QtQr) {s - 0 Qf (4-8) irrj = (a-bQh-cr-w)qrj (4.9) It is obvious from the setup of this problem that corner solutions will play an important role when deriving the market equilibrium. This is due to two reasons (1) The linear demand function may lead to negative prices when output is high (2) Second and more importantly, the presence of incoming traffic may lead to situations in which the wholesale price or the carriers' output is driven to zero. 103 4.3 Home Market Equilibrium under Propor-tionate Return 4.3.1 An Equilibrium Characterization in the General Case The home market carriers' two-stage optimization program is solved back-wards, starting in the second stage, in which fb-carriers and the resellers simultaneously maximize their profit functions for any combination of whole-sale prices with respect to the output quantities, given the other carriers' out-puts. Formally, carrier i's and reseller j's stage two optimization problems can be written as m a x { 9 / b } ir{b(qjb \ vr,qf,ql-*,k = 1... m, j = 1. . . n , k^i) (4.10) s.t. gf > 0 and max{qfl,} <(?[ | w, q£\ q]*, k = 1... m, j = 1... n, j^i) (4.11) s.t. q? > 0 where w = (IUJ, . . . , wm) is a vector of transmission charges. The first stage equilibrium, if it exists, can then be characterized by an optimal wholesale transmission charge, w*, given the stage two solutions. In equilibrium, 7 r * « , w ^ ,q?(w* , w^) ,ql(w*, w^ ) ) > 104 TTi(wi, w^, q?(Wi, w ^ ) , ql(wi, w ^ ) ) (4.12) for all and for alH, j = 1 , . . . , m and for all Wi, where w^j = (to*,..., w*_x, w*+l, . . ., w^) for i = 1 . . . m and w^, w* > 0 for i = 1 , . . . , m. 4.3.2 Stage Two Equilibrium Output From Equation 4.9, reseller A;'s stage two optimality condition is given by a r r m S = a - c t - w - 9; + E i t + Qk) = 0 (4-13) 0(ik j=l i=l d2irr The SOC Q2^- = —2b < 0 guarantees that the optimality condition charac-terizes a global maximum. Symmetry among the resellers implies that each reseller's optimal output can be written in its nonnegative range as, q> b{r + l) { } where Qfb — YALI qf- Then, in the limiting case of the number of resellers approaching infinity Of* = lim rqr* = I a—C —w if Qfl> < ^ » " " "= b (4.15) 0 otherwise The expression a ^ w represents the output of a competitive industry with the resellers' marginal cost. Equation 4.15 states that, as long as the fb-105 carriers produce less than that competitive resale output, the resale industry will provide - in the limit - the difference between the competitive resale output and the total combined output of all fb-carriers. It follows that Q f c » = Qr* + Qfb= a ~ c T - w (4.16) for all output combinations, and in particular for all stage two equilibria (q\* (w) , . . . , «£* (w), qt (w),..., q£* (w)), as long as Q^* < . Hence, as the number of resellers becomes infinitely large, the total indus-try output approaches the level of a competitive industry with the resellers' marginal cost. This marginal cost is the sum ofthe marginal cost of providing end-user service, and the per-minute wholesale charge. Equivalently, the retail price of a call minute from h to f approaches the marginal cost of the resale industry: p*(w) = w + cr (4.17) At Qfi* = a~°Tb~w, the fb-industry produces exactly the output of a competi-tive industry with the resellers' marginal cost. This situation occurs when the resellers' marginal cost, w + cr are high compared to the fb-carriers marginal retail service cost, c^, such that the resellers are driven out of the retail market. In general, there will always be a wholesale price beyond which the resellers are no more willing to provide retail services. Let w be this critical 106 wholesale charge, i.e., Q^iw) = a-crb~™ (4.18) For w > w, the resellers will offer no retail service, and there is hence no meaningful wholesale market. The two-stage game will then collapse into a one stage Cournot game between the fb-carriers. In order to exclude de-generate solutions, only solutions in the range 0 < w < w will hereafter be considered15. The relatively simple form of the total output function in equation 4.16, Qh*, makes it possible to derive the facilities based carrier i's stage two equilibrium outputs, contingent on the three relevant classes of stage one outcomes, namely fa = 0, fa = ^ , and fa — \. The case of fa = 0 occurs if fb-carrier % offers wholesale service at a price above the minimum wholesale price in the market. In this case, no reseller will route traffic via carrier i's network, and i's wholesale market share is zero. If every carrier offers the same wholesale charge Wi = w for i = 1... m, then each fb-carrier will have the same market share, and fa = ^ . Finally, if carrier i undercuts all other carriers' wholesale prices, then it will obtain all resellers traffic and fa = 1. Depending on the fb-carriers stage one wholesale prices, w, and carrier i's resulting wholesale market share, fa, the optimal fb-carriers' output schedules 1 5This restriction can be justified by noting that the F C C has explicitly required fb-carriers to offer wholesale service at reasonable wholesale charges, see, for example, (FCC 1988a at 22). 107 can be computed as 16 < = o H 1 rn {Qh*)2(w + Ac - s - c*) + Q ^ Q ^ s - c1) b(Qh*)2 + Qf(s-ct) w + Ac-s-c* m-lQf^s-c*) + m bQh* (4.19) (4.20) w + A c - s - cl (m-i)Qf{s - V ) ^ where A c = c r — c-^ . Note, that the optimal quantities of equations 4.19, 4.20 and 4.21 do not necessarily satisfy the positive quantity constraint ofthe maximization problem 4.10. Hence, a complete specification ofthe solution to 4.10 for w < w is given by C U H i f<; : = 0 H>o 0 otherwise (4.22) -.fb* 0 otherwise ?5t=iW i f g / ; : = o H > o (4.23) (4.24) 0 otherwise The corresponding optimal stage 2 profit schedules, contingent on the whole-1 6 T h e SOC for profit maximization are satisfied, since 108 sale rate, and for the domain of w for which Qr* > 0 can be written as: (4.25) (4.26) (4.27) 4.3.3 Stage One equilibrium wholesale rates The number and type of stage one equilibria in the wholesale market obvi-ously depends on the shape and position of the profit schedules 4.25, 4.26 and 4.27 relative to each other. In general, the stage one equilibrium - if it exists - is characterized by a wholesale charge w* with the following proper-ties: (1) Trf* (™*) > 7 r f ; i = 1 H for all w < w\ w* > 0 (2) ^ = i K ) > ^ = o K ) The first condition postulates that no carrier can improve profits by under-cutting w*. The second condition guarantees that in equilibrium none of the carriers would be better off by charging a wholesale rate, which is higher than w*. While the exact shape and relative position of the profit schedules depend on the specific set of parameters, it can in general be shown that 1 7 1 7The proof is by simply finding and evaluating the intersections between the profit schedules 4.19 to 4.21 and 4.22 to 4.24, respectively. While this exercise is purely alge-109 (1) Each of the profit schedules intersects every other schedule at least one more time at a w < w. However, these intersections may or may not fall on the same point, they may occur in the negative profit range, or in the negative part of the domain of w. (2) If all three intersections fall into the positive profit range, and the output constraints are non-binding, then they will all intersect in the same point. In this case, the intersection of the three profit schedules constitutes a unique equilibrium in the wholesale market. In order to illustrate this, the set of possible stage one equilibria can be classified into three stylized cases. Figures 4.2 to 4.4 depict these cases by schematically representing the (stage two) optimal profits as functions of the wholesale rate. The thick lines represent a fb-carrier's profits for the case that all carriers choose the same wholesale rate. The solid thin lines represent the profits of a carrier that has undercut its competitors' wholesale charge, and therefore is the only provider of wholesale services. The dotted lines depict the profits of a carrier that charges a wholesale rate above the minimum rate in the market, and therefore has zero wholesale market share. braic, it involves finding and comparing the solutions of polynomials of degree 5. The computations can be conveniently carried out using mathematical software such as Maple V or Mathematica. Due to the algebraic complexity, an presentation of the relevant ex-pressions is omitted at this point. 110 Figure 4.2 illustrates the first of the stylized cases, Case I. Three zones can be identified in this figure. If the wholesale rate is high, i.e., for wholesale rates w > w, the resellers total cost (relative to the fb-carriers' cost) are high to an extent that they would not provide any retail service. To exclude degenerate solutions, the feasible set of wholesale rates was previously limited to the range 0 < w < w. Hence, point B is not a feasible outcome by assumption. At any point in the area between the points A and B, capturing the entire wholesale market will provide a carrier with benefits of supplying all resellers, and with associated benefits in the incoming traffic market. Note that a carrier will be willing to make losses in its direct dealing with resellers, if these losses are compensated or overcompensated by gains from incoming traffic, and from their own retail operations. Point A is reached at the wholesale rate, at which undercutting the whole-sale rate by an infinitesimal small amount would bring no further profits, as the additional gains in the incoming traffic and resale markets would be exactly compensated by additional losses in the wholesale operations. For w < w*, a carrier would be better of by not offering wholesale service at all, than by offering it. The only equilibrium point in this market is A, since at this point no carrier has an incentive to defect from the equilibrium. Undercutting the wholesale rate would result in lower profits and quitting the wholesale market altogether would not change the profits. In Case II, shown in Figure 4.3, 111 w* w w Figure 4.2. Stage two optimal profit schedules contingent of the stage one wholesale price. Case I: Unique equilibrium. no stage one equilibrium (and therefore no stage two equilibrium) exists. None of the points on the 4>i = ^ profit schedule between C and B is an equilibrium, as at each of these points a carrier would have the incentive to undercut the wholesale price in order to receive a larger portion of the incoming traffic. However, C itself is not an equilibrium either, since for all w < wE the carriers can make higher profits by only serving the end-users, than they could make by competing for wholesale business and for the associated gains from incoming traffic. There is no point, at which the two equilibrium conditions are met 1 8. Finally, Case III (depicted in Figure 4.4) describes a situation where an infinite number of sage one and stage two equilibria exists. A l l points on the thick line between the E and C are 1 8 Note that there is also no asymmetric equilibrium at D , since at any point between D and E,—(wE ,-K^*=1(WE )), a monopolist on the wholesale market would be better of by increasing the wholesale price w by in infinitesimal small amount. 112 Figure 4 . 3 . Stage two optimal profit schedules contingent ofthe stage one wholesale price. Case II: No equilibria. equilibria, but there exists no equilibrium without resale. On the other hand, if 7T^.=1(VJ) is non-decreasing left of point B, there will also be an equilibrium without resale. Situations like in Figure 4.3 and 4.4 can occur only if the output restriction 1 9 is binding at any of the points C , D, or E . 4.3.4 Market Equilibrium without P R R Consider now a telecommunications market that operates under the same conditions as the market described in section 4.2, except that the government gets all the profits from incoming traffic2 0. Then the profit functions of fb-carrier i, reseller j and surplus from incoming traffic (which can either be 19I.e., the restriction that outputs be positive, see equations 4.22 to 4.24. 2 0 The alternative of allocating an equal fixed share of incoming traffic to each fb-carrier would be equivalent in terms of equilibrium outcome and welfare implications. In the context of this model, both alternatives differ only with respect to who gets the fixed sum of profits from incoming traffic. In the following, either setting will be referred to as a market without PRR. 113 wB wDwc w w Figure 4.4. Stage two optimal profit schedules contingent of the stage one wholesale price. Case III: Multitple equilibria with resale. allocated among the fb-carriers according to a fixed scheme, or retained by government) can be written as Tff=(a - 6 Q - c^-c*- s)qf + (w- c*- s ) & Q r (4.28) 7 f ; = (a - b { Q - c r - ( 4 . 2 9 ) Vfinc =(s - c 4 ) ^ (4.30) As in Equation 4.8, the first two terms of Equation 4.28 represents carrier i's profit in the home market, and its profit from transmitting resold traf-fic, respectively. It is easy to see that the equilibrium outcome of the two-stage game between fb-carriers and resellers (characterized by the equilibrium wholesale charge w*, and outputs Q*, q^*) corresponds to the outcome of the game under proportionate return but without incoming traffic, i.e., = 0. 114 At Qf = 0, the fb-carriers' optimal stage two output schedules contingent on the stage one outcome (equation 4.19 to 4.21) are always identical and the non-negativity constraint for the fb-carriers' output is never binding. Hence, there exists a unique equilibrium without proportionate return, and the equilibrium outcome simplifies to w* = c< + s, Q* = a - C r - S ~ C \ -q^ = f (4.31) Without the P R R , fb-carriers will set the wholesale charge to the marginal cost of transmitting outbound traffic. Due to their cost advantage over the resellers, fb-carriers will retain a profit from their retail service, 7T, fi, _ (Ac)' (4.32) 4.3.5 An illustrative example: c{b = cr The market equilibrium, a number of comparative statics results, and the welfare implications of the P R R can be analytically illustrated for the special case that fb-carriers and resellers face the same retail service cost. In the case of c = cfb = cr, the fb-carriers' stage one profit functions take the following simple form: 7 T & = { w - c t - s + ( H ( s - c')) (ft, oV*(w) + &Qr(w)) (4.33) 115 where = is country h's traffic imbalance. If ((w*) < J^J, then there exists an equilibrium wholesale charge, w*, at which the fb-carriers' equilibrium profits are zero, regardless of their wholesale market share, fa, and regardless of their second stage output decision. w * = s + c* - C K ) ( s - c*) if C K ) < (4.34) v r f = 0ifCK)<^4 (435) s — c On the other hand, if ((w*) > ~t, then the non-negativity constraint of the wholesale charge is binding, and the fb-carriers equilibrium profits remain positive. w* = 0 if CK) > ^ 7 (4.36) s — c* Trf * = 1(Q/( S - c<) - Qh*(s + c*)) if C K ) > ^ (4.37) Carrier i's optimal second stage output can be computed as fb* if/"* m s + c L 1 1 S - m - 1 s -c ' (4.38) m * - c « - 8 + m ^ l Q ^ - c ' ) o t h e r w i s e 6 m bQn* Hence, in the case of = cr there always exists unique equilibrium, char-acterized by the equilibrium wholesale charge, w*. Figure 4.5 illustrates the stage two profit schedules. The relative positions of the profit schedules cor-respond to Case I of section 4.3.3. In this particular setting, however, the 116 The case of low incoming traffic volume C(w*) < The case of high incoming traffic volume C(W) > *g Figure 4 . 5 . Stage two optimal profit schedules contingent of the stage one wholesale price. Illustrative example of = cr equilibrium is always a corner solution, either at the point at which the fb-carriers' profits are zero, or where the equilibrium wholesale charge is zero. Pont A in the left hand graph of figure 4.5 is obviously an equilibrium, as all three profit schedules intersect in A . In the right hand side graph, A is an equilibrium, because no carrier can undercut the wholesale charge of w* = 0. If the equilibrium traffic imbalance in is small (i.e., CK) < f^i21), then the wholesale charge takes on a positive value, but always remains below the marginal cost of transmitting outgoing calls, s+c*. A l l profits from incoming 2 1 I n order to i l lust rate this boundary cond i t ion , assume that the sett lement rate is at $0.20 and the t ransmiss ion cost are at $0.02. T h e n the wholesale pr ice w i l l be zero if the incoming traffic vo lume exceeds the outgoing traffic vo lume by 22%. O n routes where the home country t ransmi ts more traffic to the foreign count ry than it receives in re turn , the traffic rat io w i l l always be below the thresho ld , ((w*) < . T h i s is true for a l l but two in ternat ional routes or ig inat ing and te rmina t ing in the Un i t ed States. One of these routes is the l ink between the U S and Sweden, on wh ich the I S P is waived. T h e other is l ink between the U S and the French Overseas Depar tments , wh ich carr iers a relat ively insigni f icant traffic volume. 117 traffic are passed on to the resellers and from the resellers to the end users. The fb-carriers, in spite of receiving all the lucrative incoming traffic, are not able to retain any profits. Furthermore, the fb-carriers offer no end-user (retail) service, but rather exclusively focus on the wholesale business.22 On the other hand, if the traffic imbalance is large (i.e., the ratio of incoming to outgoing traffic exceeds the threshold jz^t), then the wholesale price is driven to zero. Only in this case are fb-carriers able to retain some profits. If the traffic imbalance exceeds the threshold of ^ r [ j z $ ) , the fb-carriers will also retain a positive retail market share. An immediate consequence from equation 4.38 is that each fb-carrier's output increases with the number of fb-carriers in the market, > 0, if g{b* > 0. The intuition for this result is that an increase in the number of car-riers has an effect in both the outbound call market as well as in the inbound call market. On the outbound market, more carriers compete to serve the end-users and resellers. The carriers' optimal reaction is to contract output. On the other hand, the revenues from incoming calls are split among more carriers, and the carriers will compensate lost revenues by expanding out-bound output. The latter effect dominates the former one since the number of carriers splitting incoming traffic is smaller than the number of carriers-2 2It is intuitively clear that the fb-carriers will not offer retail services: In Cournot-type competition models, the equilibrium total industry output increases with the number of carriers in the market. Fb-carriers will therefore have an interest in using resellers to service the retail market - so as to outsource and "unmerge" their retail business in order to capture a larger market. Yet, they have to compete with other fb-carrier for the privilege to service customers through resellers. In the course of this competition, all potential gains are competed away. 118 competing on the outbound market. Finally, the fb-carriers' equilibrium profits do not change with a marginal change of settlement rates as long as ((w*) < Otherwise, if the incoming traffic volume is large compared to the outgoing traffic, i.e., ((w*) > then the fb-carriers' profits will increase in the settlement rates. The effect of a change of settlement rates on the resellers' market share is ambiguous. 4.3.6 Discussion The previous section highlighted some of the effects of the P R R on the equi-librium outcome for the simple case that both types of carriers face the same retail service cost. Linking incoming traffic to the carriers' outgoing traffic volume causes fb-carriers to set their wholesale charge below the marginal cost of transmitting outbound calls, which in turn will induce the competi-tive resellers to increase the retail output beyond the economically efficient levels, and set the retail price below the cost of producing the outbound ser-vice. Consumers benefit from the low retail price, but the fb-carriers compete away at least some of the potential gains from incoming traffic. On routes where incoming traffic volumes are lower than outgoing volumes, all gains will be competed away. In a setting where the resale market is competitive, fb-carriers and re-sellers face the same cost, and incoming traffic volume is lower than outgoing traffic volume, fb-carriers may therefore have no advantage of receiving in-coming traffic under the PRR, neither in terms of profits, nor in terms of 119 retail market shares. It is easy to see that the P R R increases consumer surplus, but total welfare is higher under a regime where the profits from incoming traffic are retained by the government, or distributed among the carriers according to a fixed allocation scheme. While the conclusions drawn from the case of c r = are intuitive, and they can help demonstrate some equilibrium properties under the PRR, the insights gained from this special case are somewhat limited. First, the as-sumption that both carriers face the same marginal retail service cost is unrealistic, given their differences in size, technology, experience, and image with consumers. Second, the market equilibrium under the assumption of cr = will be a corner solution, where the non-negativity restriction for outputs is binding, the non-negativity restriction for the wholesale price is binding, or the fb-carriers' equilibrium profits are zero 2 3. For the general case where cr ^ c^, section 4.3.3 demonstrated that the effect of the P R R may reach beyond distorting the outputs and calling charges in international telephone markets. Rather, the P R R can preclude the existence of equilibria in international telecom market. Under the PRR, there may exist a unique interior equilibrium 2 4, infinitely many equilibria, or no equilibria at all, depending on the specific set of cost and demand parameters. The arithmetic complexity of the general model, however, makes it impos-2 3 Comparative statics exercises are mostly trivial in the case of corner solutions, since the relevant equilibrium outcomes do not change with a marginal change of model parameters. 2 4I.e., equilibria where fb-carriers' profits and outputs are strictly positive. 120 sible to derive a tractable closed form solution of the equilibrium outcome, or even to specify the range of parameters, for which one, several, or no equilib-ria exist 2 5. The following section develops further perspective on the nature and comparative static properties of telecom market equilibria under propor-tionate return and differential retail service cost, by simulating equilibrium outcomes and their properties. 4.4 Simulating Equilibria under Differential Cost 4 . 4 . 1 Simulation inputs and assumptions In order to make simulations useful and credible, it is important that inputs to the simulations be observable or estimable, and that the simulation results are consistent with the observable features of the market. The exogenous parameters of the model characterized by equations 4.8 and 4.9 are the demand curve intercept and slope a and b, the marginal cost parameters c r , c^, and c*, the settlement rate s, incoming traffic volume , and the number of fb-carriers m. The endogenous model parameters are the outgoing traffic volume Qh*, the wholesale charge w*, the fb-carriers market share in the retail market Q^/Qh*, and the elasticity of demand e 2 6 . 2 5 Note that the arithmetic complexity is solely due to the proportionate return rule, equation 4.3, and not to the number of parameters in the model. In fact, assuming that any of cost parameters are zero will not significantly reduce complexity. 2 6 T h e elasticity of demand in equilibrium can be directly inferred from the demand 121 Some, but not all of the model primitives are directly observable on a route-by-route basis. The observable data that will be used as simulation inputs are based on four exemplary routes between the United States and Chile, China, Laos, and the United Arab Emirates ( U A E ) 2 7 . Other exogenous model parameters are not directly observable, however, their value can be conjectured or inferred from observable information. The following values for non-route-specific parameters are assumed for all simu-lations: (1) the number of facilities-based carriers is m — 3, reflecting the fact that the largest three facilities-based carriers in the 1998 US telecom market accounted more than 95% of the outgoing international facilities-based service. (2) The transmission cost and the fb-carriers retail service cost are c* = 0.02$ and = 0.08$, respectively2 8. (3) the price elasticity of demand is e = —0.6 on all routes2 9. function parameters and the equilibrium output. 2 7 The criteria for selecting the exemplary routes were that (1) all routes be subject the PRR; (2) bypassing the international settlement rate mechanisms be illegal on all routes, that is, all traffic between the United States and these countries is subject to the PRR; (3) the four country routes reflect a broad spectrum of routes in terms of applicable settlement rate, and traffic imbalance (that is, the route specific ratio of incoming over outgoing traffic). 2 8 Where international telecom market are highly competitive on both sides of an inter-country relation, the settlement rate may be regarded as a proxy for the real cost of providing telephone service. In 2000, the lowest US settlement rate of 0.10$ applied on routes to and from Australia. It is therefore reasonable to assume that the facilities-based carriers' marginal cost components, c* and add up to 0.10$. 2 9 This value is in the range of elasticity estimates for US outbound telephone demand published in recent years (see Chapter 3) . 122 4.4.2 Simulated Equilibria If the parameters of the demand function were directly observable, then the computations of model equilibria would be straightforward. This is how-ever not the case. Rather, the underlying demand parameters can only be inferred, for example, from an observed output quantity and an estimated price elasticity. The output quantity itself is, on the other hand, endogenous to the model. The approach taken in the context of these simulations is to calibrate the demand parameters a and b, such that the resulting simulated optimal total industry output coincides with the observed 1998 traffic volume on the respective route. For example, the top row of Table 4.1 shows the simula-Table 4.1. Simulation results: Existence and characterization of equilibria in the case of four exemplary international routes, for A c = 0.12 US$ country o s bserval specif Qf )le route c data Qh C a sir b nulat w* i o n r esults equilibrium us$ min *103 min *103 io- 9 us$ Chile 0.38 81,068 128,217 0.63 1.11 5.429 0.212 0.36 unique China 0.53 79,290 431,863 0.18 1.78 2.57 0.467 0.27 unique Laos 0.67 9 1,536 0.01 2.37 963.5 0.687 0.24 unique U A E 0.74 32,331 60,389 0.54 - - - - no equilibrium tion results based on data from the US-Chile route. The first four columns contain the observable route-specific data, the settlement rate, and the total minutes of incoming traffic, the total minutes of outgoing traffic, and the 123 traffic imbalance C = QTT- The next two columns report the parameters of the underlying demand function, a and b. These parameters were calibrated such that Qh, which is the endogenously determined stage two equilibrium outcome, coincides with the observed number of telephone minutes on the route from the United States to Chile, and such that the equilibrium price elasticity of demand equals —0.6. In the case of the US-Chile route, there exists a unique pair of demand parameters (a, b) that support an equilibrium at the observed quantity, and at the estimated elasticity. The final three columns, labeled w*, and "equilibrium" show the key simulation results, that is, the characteristics of the simulated equilibrium. In the case of the US-Chile route, for example, the wholesale charge was simulated to be w* = 0.212, and the fb-carriers market share in the retail market is at 36%. Note that the equilibrium wholesale charge is at a level well below the settlement rate. On the routes US-China and US-Laos, the traffic ratio £ is lower than on the US-Chile route. The distortion created by the proportion-ate allocation of incoming traffic is therefore less severe, and the equilibrium wholesale charge is closer to the marginal cost of transmitting outgoing calls. An extreme example is the route US-Laos, with only 9,000 minutes of incom-ing traffic from Laos. The simulated wholesale charge applicable outgoing traffic on this route is only little below the marginal transmission cost, s + c*. This simulated values of the market share, provide some indication that the simulation results are approximately consistent with the observed 124 market structure. The simulated values were 36% for the US-Chile route, 27% for the US-China route, and 24% for the US-Laos route. In compari-son, the obvserved resale industry's average market share (averaged over all international routes) was 43% in 19983 0. The first three rows of Table 4.1, based on data from routes to Chile, China and Laos correspond to Case I and figure 4.2 of section 4.3.3. For each of these routes, there exists a set of demand function parameters that supports a unique equilibrium. The example of the U A E , on the other hand, corresponds to Case II and figure 4.3 of section 4.3.3. There is no set of demand function parameters that supports an equilibrium with the observed traffic volumes, under the assumed values of the remaining model primitives. One of the crucial parameters determining the existence of equilibria is Ac, the cost difference between fb-carriers and resellers. The results shown in Table 4.1 were based on the somewhat ad-hoc assumption that Ac=0.12$. The following Table 4.2 exhibits different values of A c for which equilibria exist on the same four exemplary routes. A black dot indicates a unique equilibrium at the given value of Ac. No black dot signifies that there is no equilibrium, at any given set of demand parameters. Case III of multiple equilibria did not occur in any of the simulations, although the possibility of its occurrence could analytically not be ruled out. At high values of Ac 3 0 T h e F C C reports only the to ta l market share of resale carr iers for a given year. Marke t share da ta on a route-by-route basis are not avai lable. 125 Table 4.2. Simulation results: Existence and type of equilibria for different values of A c in the case of four exemplary international routes Countrj • c * A c s .02 .07 .12 .17 .22 .27 .32 .37 .42 .47 .52 .57 .62 .67 .72 .77 .82 .87 .92 .97 Chile 0.63 0.38 China 0.18 0.53 Laos 0.01 0.67 U A E 0.54 0.74 • A unique equilibrium exists the fb-carriers cost advantage is large, such that resellers are driven out of the market. There exists no feasible equilibrium 3 1. The more interesting scenario occurs at low values of Ac, where the fb-carriers' market share in the retail market is low. The wholesale market, in conjunction with the asso-ciated incoming traffic, becomes so lucrative, that a fb-carriers would chose not to provide any retail output, if it were the exclusive wholesale provider 3 2. Hence, the output constraint of equation 4.21 is binding, and there will be no equilibrium. 3 1 T h a t is, there exists no equ i l ib r ium in which resellers ho ld a posi t ive market share, as required by assumpt ion . 3 2 T h a t is, if it had undercut a l l compet i tors in the wholesale market and op t im ized the reta i l outputs accord ing to the profit schedule 7 T ^ . _ 0 ( w ; ) (equat ion 4.27). 126 4 .4 .3 Simulating the effect of the P R R on welfare and market structure Using the demand parameters from table 4.1, it is easy to compute the equi-librium outcome of the model without the P R R (equations 4.28 to 4.30), and perform welfare comparisons between the two regimes33. The first four Table 4.3. Simulation results: Characterization of equilibria without P R R for three exemplary interna-tional routes, for A c = 0.12 US$ country obser route-s da s vable pecifk ta Qf dem par an (from T i a and leters i b l e 4 .1) b sin r with c lulati esult out I w* on s ' R R nh us$ min * 1 0 3 io- 9 us$ Chile 0 . 3 8 8 1 , 0 6 8 1 .11 5 . 4 2 9 0 . 8 6 0 . 4 0 0 . 7 1 China 0 . 5 3 7 9 , 2 9 0 1 .78 2 . 5 7 0 0 . 2 0 0 . 5 5 0 . 3 5 Laos 0 . 6 7 9 2 . 3 7 9.63.5 0 . 0 1 0 . 6 9 0 . 2 4 column of table 4.3 show the exogenous model parameters, as they appear in table 4.1. The equilibrium outcome without P R R is reported in the three right hand columns. Comparing the simulated values of C, w*, ^ with the respective figures in table 4.1, it is easy to see that the wholesale charge is higher under the regime without PRR. The traffic imbalance ratio is higher without the PRR, since the increased wholesale charge results in a smaller volume of outbound traffic, while the inbound traffic volume remains con-3 3 Obv ious ly , welfare compar isons can on ly be per formed for those routes where an equi-l i b r i um under P R R exists. T h e U S - U A E route is hence not considered in the subsequent discussion. 127 stant. A less intuitive simulation result is that the simulated fb-carriers' retail market share, is higher without the P R R on all three routes3 4. Con-trary to intuition the P R R therefore does not necessarily help reinforcing the fb-carriers dominance over the resellers in the telephone retail market. Rather, it can provide the fb-carriers with an incentive to decrease their retail market share and place relatively more emphasis on the provision of wholesale service. The competition in the wholesale market under the P R R Table 4.4. Simulation results: Welfare comparisons to equi-libria under P R R in the case of four exemplary international routes, for A c = 0.12 US$ country welfa wi PS re am th PR CS dysis ,R W w \ elfare vithou PS analys t PRI CS is tW AW US$*10 6 US$*10 6 US$*10 6 US$*10 6 US$*10 6 US$ *10 6 US$*10 6 US$*10 6 Chile 10.54 44.63 55.17 29.18 7.96 23.93 61.07 -6.10 China 18.22 239.66 257.88 40.43 16.81 206.63 263.87 -6.01 Laos 0.05 1.14 1.19 0.06 0.05 1.13 1.24 -0.05 forces fb-carriers to pass a significant share of their windfall profits from in-coming traffic on to the resellers, which in turn will pass it on to consumers. The figures shown in table 4.4 illustrate that the P R R results in a higher producer surplus and a higher consumer surplus than the alternative regime without P R R , assumed they receive no share of the profits from incoming 3 4 T h e same results hold not only for the route and specific parameter values exhibited in table 4.1, but for all reasonable parameter combinations examined during the course of this research. 128 traffic under the alternative regime. However, the total welfare effect of the PRR, is negative. For example, for the three exemplary routes simulated for this study, the combined welfare loss from the P R R amounts to 12.16 million US$. Note that in each of the exemplary cases, the consumers have the largest benefit from the P R R . A change from the P R R to a regime without the P R R would reduce the producer surplus by a relatively small amount, compared to the reduction of consumer surplus. Consider, for example, a situation where the F C C would exempt the US-Chile route from the P R R , and would leave the allocation of US-bound traffic to the discretion of the Chilean carrier. Then the fb-carrier industry, as a whole, would now have a Producer Surplus of 37.12 million US$, improved from 10.54 million US$ under the P R R . Depending on the specific allocation of US bound traffic among the fb-carriers, each single one of the three fb-carriers, which had a third of the incoming traffic prior to the exemption, could face a loss of at most US$ 860,000. However, a carrier that receives only 2.9% or more of the US bound traffic after the exemption from the P R R would be better off relative to the situation before. If the incoming traffic were distributed in equal parts among the fb-carriers, each fb-carrier would always earn more profits without the PRR. Hence, under the assumptions of this model and for realistic sets of pa-rameter values, a fb-carrier benefits from the exemption of a route from the P R R , unless it is almost entirely excluded by the foreign carrier from receiv-129 ing incoming traffic. 4.4.4 Simulated comparative statics The final set of simulations illustrates the effect of increasing the number of fb-carriers in the market and of increasing the settlement rate on the equilib-rium outcome. Section 4.3.5 showed analytically that, in the case of — cr, the fb-carriers' output increases with their number in the market. The same result can be demonstrated in the general model with different marginal re-tail service cost. Figure 4.5 shows each carrier' profits, output, and market share in the case where 3, 4, and 5 fb-carriers serve the US-Chile route 3 5. The Table 4.5. Simulation results: Fb-carriers profits, outputs, and market share for varying numbers of fb-carriers in the market based on data from the US-Chile route Number of fb-carriers < minutes us$ m=3 m=4 m=5 0.12 0.15 0.16 15,354,580 18,839,000 20,929,650 3,515,071 3,515,071 3,515,071 simulation results indicate that for the given set of parameters, fb-carriers outputs and market shares increase with their number in the market, whereas each fb-carrier's profits were not affected by the number of competitors. 3 5 T h e parameters used were the same as in table 4.1. 130 Finally, figure 4.6 demonstrates the percentage change of fb-carriers prof-its and market shares, based on data from the US-China route, if the set-tlement rate is changed from the base rate of 0.53 US$. Both profits and market shares increase with an increase of the settlement rate. Based on these results, fb-carriers can be expected to oppose a decrease of settlement rate, even in a setting with highly competitive wholesale and retail markets. F igure 4.6. Simulation results: The effect of settlement rate changes on fb-carriers' profits and market shares, for A c = 0.12 US$ 5.00% -4.00% J — —••• settlement rate 131 4.4.5 Discussion Simulations can be useful tools to gain insights into economic models that are algebraically too complex to solve. On the downside, there always remains some doubt about the general validity of results obtained under specific pa-rameter assumptions. The approach presented in this section attempts take such concerns into account in two ways. First, the parameter values are chosen incorporating all available data and information on the international telecommunications market. Second, an array of different values were tested for those parameters, for which no obvious proxies were available. The results reported in this paper represent only a snapshot the feasi-ble set of parameter values. Yet, they can be considered a representative sample of results obtained from a reasonable range of parameters. The ef-fects obtained from using alternative values may differ in magnitude, but qualitatively they remain the similar. The simulations presented in this section served three objectives. First, it is analytically not possible to specify parameter ranges that support a unique equilibrium, multiple equilibria, or that do not support any equilibrium un-der the P R R . The simulation approach illustrated that, within reasonable parameter ranges, proportionate return traffic allocation may bring about situations in which no equilibrium exists. Second, the simulations demonstrated the welfare implications of the pro-portionate return requirements. It provided some approximation of the rela-tive magnitude of the impact of proportionate traffic allocation on fb-carriers' 132 profits, consumer surplus, and total welfare. Finally, it illustrated some counterintuitive effects of the P R R on the car-riers' market shares, and some counterintuitive comparative statics properties of equilibria under the PRR. 4.5 Policy Implications, Summary and Con-clusions The Proportionate Return Requirement was implemented by the F C C as one of three measures to prevent US carriers from being whipsawed by for-eign monopolists. Large carriers, such as A T & T and Sprint have supported this regulation, or even suggested to implement it in other countries, such as Canada (CRTC 1998). Yet, it is not clear why and how the P R R should prevent whipsawing in a setting where settlement rates are regulated to be uniform, and equal in both directions. Chapter 3 argues that the P R R does in deed not help to prevent whipsawing over and above the FCC's uniform settlement policy regulation, and that the PRR, as a codified regulation, is a product of a series of regulatory inconsistencies and misconceptions. The current paper supplements the institutional analysis of Chapter 3 by analyz-ing the economic effect of the P R R on the international telecommunications market in the country where it is implemented. This analysis is intended to clarify the impact of the P R R on international telecommunications markets, and on the different economic players involved. 133 The formal framework developed in this paper is a two-stage model with a wholesale and a retail market. In the wholesale market, fb-carriers compete to provide wholesale service to resale carriers. At the same time, fb-carriers compete with resellers in the end-user market. The P R R creates an incentive for fb-carriers to increase output beyond the economic efficient amount in order to secure a larger portion of the lucrative incoming traffic. The effect of this additional incentive on the equilibrium outcome may differ, depending on the specific model parameters. For certain parameter ranges and under the assumptions of the model, the P R R may preclude the existence an equilibrium altogether. If an equilibrium exists, then the equilibrium wholesale price will be below the marginal cost of providing outbound service. Outbound traffic volume will be inflated beyond economic efficient levels, and prices pushed below efficient level. The consumers are the main beneficiaries of the PRR. Resellers earn no profits with our without the P R R under the assumptions of the model, however, they will obtain a higher market share under the P R R then under a regime where inbound traffic is not allocated among fb-carriers in proportion to their relative share of outbound traffic. Paradoxically, the P R R appears to work against the interests of its two main proponents: The fb-carriers earn lower profits and capture a lower market share under the P R R . The regulator would increase total welfare by eliminating the PRR. The effect of the P R R on the foreign monopolist depends on the spe-cific demand structure, and calling charges in the foreign country. However, 134 from the view of the foreign monopolist, the proportionate allocation of its outbound traffic among the receiving carriers will result in higher volumes of return traffic, and with the higher levels of settlement payment transfers, paid in hard currency. The foreign monopolist may therefore have an interest in allocating traffic proportionally, even if it were not required to do so. A regime without P R R can be implemented by the regulator in either of two ways. One way would be to have the regulator (rather than the carri-ers) negotiate settlement rates with foreign monopolists, and auction off the right to switch incoming traffic to the fb-carrier that is willing to do so at the lowest per-minute charge. The profits from incoming traffic would then be retained by the government. An administratively less burdensome approach is to simply eliminate the P R R on all routes, or even prohibit the use of proportionate return clauses in interconnection agreements. The regulator could then still continue to monitor the traffic allocation, and step in only in cases where traffic allocation is in a severe imbalance, without however affecting the carriers' output and pricing decisions on the margin. The prof-its from incoming traffic would then remain with the fb-carriers, but total welfare would be improved as compared to a regime with the P R R . 135 Bibliography Chapter 2 [1] Acton J. P. and I. Vogelsang (1992). Telephone Demand over the At-lantic: Evidence from Country Pair-Data. Journal of Industrial Eco-nomics, 40(3), 305-22. [2] Appelbe, T. W. , Sinihur, N . A. , Dineen, C , Franes, D. and R. Giordano (1988). Point to Point Modelling: An Application to Canada-Canada and Canada-US Long Distance Calling. Information Economics and Pol-icy, 3(4), 311-31. [3] Cesario, F. J. (1976), "Value of Time in Recreation Benefit Studies," Land Economics, 52(1), 32-41. [4] Federal Communications Commission (1989), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [5] Federal Communications Commission (1990), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [6] Federal Communications Commission (1991), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [7] Federal Communications Commission (1992), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [8] Federal Communications Commission (1993), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [9] Federal Communications Commission (1994), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. 136 [10] Federal Communications Commission (1995), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [11] Federal Communications Commission (1996), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [12] Federal Communications Commission (1997), Statistics of Communica-tions Common Carriers. F C C , Washington. D.C. [13] Garin-Muhoz, T. and T. Perez-Amaral (1996), "Demand for Interna-tional Telephone Traffic in Spain: an Econometric Study Using Provin-cial Panel Data," Information Economics and Policy, 8(4), 289-315. [14] Garin-Muhoz, T. and T. Perez-Amaral (1998), "Economic Modeling of Spanish Very Long Distance International Calling," Information Eco-nomics and Policy, 10(2), 237-52. [15] Garin-Muiioz, T. and T. Perez-Amaral (1999), "A Model of Spain-Europe Telecommunications," Applied Economics, 31, 989-97. [16] Hackl, P. and A . H. Westlund (1995), "On Price Elasticities of interna-tional Telecommunications Demand," Information Economics and Pol-icy, 8(1), 27-36. [17] ITU (1998), World Telecommunications Indicators Database '97. Inter-national Telecommunications Union, Geneva. [18] Lago, A . M . (1970), "Demand Forecasting Model of International Telecommunications and their Policy Implications," Journal of Indus-trial Economics, 19(1), 6-21. [19] Kennedy, P. (1996), A Guide to Econometrics. 3rd edition. MIT Press, Cambridge, Mass. [20] Larson, D. M . (1993), "Joint Recreation Choices and Implied Values of Time," Land Economics, 69(3), 270-86. [21] Larson, A . C. , Lehman, D. E. and D. L. Weisman (1986), "A General Theory of Point-to-Point Long Distance Demand," In A . De Fontenay et al. (eds.), Telecommunications Demand Modeling. North Holland, Ams-terdam, pp. 299-318. 137 [22] McConnell, K . E. (1992), "On-Site Time in the Demand for Recreation," American Journal of Agricultural Economics 74(11), 918-25. [23] O E C D (1999), Communications Outlook 1999. O E C D , Paris. [24] O E C D (1998), Main Economic Indicators. O E C D , Paris. [25] O E C D (1993), Mam Economic Indicators. O E C D , Paris. [26] Rea, J. D. and T. M . Lage (1978), "Estimates of Demand Elasticities for International Telecommunications Services," Journal of Industrial Economics, 26(4), 363-81. [27] U.S. Immigration and Naturalization Service (1996), Statistical Yearbook of the Immigration and Naturalization Service, 1994- U.S. Government Printing Office, Washington, D.C. [28] Taylor, L. D. (1980), Telecommunications Demand: A Survey and Cri-tique. Ballinger Publishing Co., Cambridge, M A . [29] Taylor, L. D. (1994), Telecommunications Demand in Theory and Prac-tice. Kluwer Academic Publishers. [30] Varian, H. R. (1984), Microeconomic Analysis. Norton & Company, New York. [31] World Tourism Organization (2000), World Tourism Database. www.world-tourism.org. Chapter 3 [1] A T & T (1998), In the Matter of 1998 Biennial Regulatory Review - Re-form of the International Settlements Policy and Associated Filing Re-quirements, Docket No C C 90-337, Comments of A T & T Corp. [2] Blake, L. , J. Lande (1998), Trends in the US International Telecommu-' nications Industry, Federal Communications Commission, Washington, D.C. [3] Brecher, R. (1998), "Ask the Expert: What is a P ICC Charge & Who Pays?," TeleCompete, Vol. 4, No.3, 30-31. 138 [4] Ergas, H. and P. Paterson, (1991), "International Telecommunications Settlement Arrangements: An Unsustainable Inheritance?" Telecommu-nications Policy, 15 (1), 29-48. [5] C C I T T (1988), General Tariff Principle Charging and Accounting in International Telecommunications Services, Recommendations of the D Series, C C I T T Blue Book, Volume II - Fascicle I L L [6] Cheong, K . and M . Mullins (1991), "International Telephone Service Imbalances: Accounting Rates and Regulatory Policy," Telecommunica-tions Policy, 15 (2), 107-118. [7] C R T C (1998), Regulatory Regime for the Provision of International Telecommunication Services, Telecom Decision C R T C 98-17. Canadian Radio and Telecommunications Commission, Ottawa. [8] F C C (1936), In the Matter of Mackay Radio and Telegraph Company Inc, C C Docket No 8777, 2 F C C 592. [9] F C C (1951), In the Matter of Mackay Radio and Telegraph Company Inc, C C Docket No 8777, 15 F C C 690. [10] F C C (1955), In the Matter of Mackay Radio and Telegraph Company Inc, C C Docket No 8777, 19 F C C 1321. [11] F C C (1983), In the Matter of Interconnection Arrangements Between and among the Domestic and International Record Carriers, C C Docket No 82-122, Memorandum Opinion, Order and Request for Further Com-ments, 93 FCC.2d 845. [12] F C C (1985), In the Matter of Implementation and Scope of the Uniform Settlements Policy for Parallel International Communications Routes, C C Docket No 85-204, Notice of Proposed Rulemaking, F C C 85-332. [13] F C C (1986), Common Carrier Services; Implementation and Scope of the Uniform Settlements Policy, Al C F R Part 43. [14] F C C (1988a), In the Matter of US Sprint Communications Company, Memorandum Opinion, Authorization and Certificate, 3 F C C Red 1484. 139 [15] F C C (1988b), In the Matter of Regulatory Policies and International Telecommunications, C C Docket No. 86-494, Report and Order and Sup-plemental Notice of Inquiry, 4 F C C Red 7378. [16] F C C (1989), In the Matter of FTC Communications Inc, Memorandum Opinion and Order, 4 F C C Red 5633. [17] F C C (1990), In the Matter of Regulation of International Accounting Rates, C C Docket No 90-337, Notice of Proposed Rulemaking, 5 F C C Red 4948. [18] F C C (1991), In the Matter of Regulation of International Accounting Rates C C Docket No 90-337 Phase I, Report and Order, 6 FCC.Red 3552. [19] F C C (1992), In the Matter of Telefonica Larga Distancia de Puerto Rico and LD Acquisition Corporation, Memorandum Opinion, Authorization and Certificate, 8 F C C Red 106. [20] F C C (1993a), In the Matter of Regulation of International Accounting Rates, C C Docket No 90-337 Phase II, Second Report & Order and Second Further Notice of Proposed Rulemaking, 7 F C C Red 8040. [21] F C C (1993b), In the Matter of Atlantic Tele-Network, Inc., Application for Authority to Acquire and Operate Facilities for Direct Service between the U.S. and Guyana, Order on Review, 8 F C C Red 4776. [22] F C C (1994), In Re Application of ACC Global Corporation, Application for Authority Pursuant to Section 214 of the Communications Act of 1934, As Amended, to Operate as an International Private Line Carrier, F C C 94-236. [23] F C C (1995), Market Entry and Regulation of Foreign-Affiliated Entities, IB Docket 95-22, Report and Order, 11 F C C Red 3873. (Foreign Carrier Entry Order) [24] F C C (1996a), In the Matter of Regulation of International Accounting Rates Phase II, C C Docket No 90-337, Opinion: Order on Reconsidera-tion, 11 F C C Red 6332. 140 [25] F C C (1996b), In the Matter of Regulation of International Accounting Rates, Docket No C C 90-337 Phase II, Forth Report and Order, 11 F C C Red 20063. (Flexibility Order) [26] F C C (1996c), In the Matter of Policy Statement on International Ac-counting Rate Reform, 11 F C C Red 3146. [27] F C C (1997a), In the Matter of Rules and Policies on Foreign Partic-ipation in the U.S. Telecommunications Market; Market Entry and of Foreign-Affiliated Entities, IB Docket No 97-142, Report and Order un-der Reconsideration, 12 F C C Red 23891. (Foreign Participation Order) [28] F C C (1997c), In the Matter of International Settlement Rates, IB Docket No 96-261, Report and Order, FCC97-280. (Benchmark Order) [29] F C C (1998a), In the Matter of 1998 Biennial Regulatory Review - Re-form of the International Settlements Policy and Associated Filing Re-quirements, Docket No C C 90-337, Notice of Proposed Rulemaking, 13 F C C Red 15,320. [30] F C C (1998b), Statistics of Communications Common Carriers, Federal Communications Commission, Washington, D.C. [31] F C C (1999a), International Simple Resale (ISR), http://www.fcc.gov/ib/isr.html. [32] F C C (1999b), In the Matter of 1998 Biennial Regulatory Review - Re-form of the International Settlements Policy and Associated Filing Re-quirements, Docket No C C 90-337, Report and Order and Order on Reconsideration, F C C 99-73. [33] F C C (2000), Routes exempt from the ISP, http://www.fcc.gov/ib/ [34] Frieden, R. (1998). "Falling Through the Cracks," Telecommunications Policy, 22 (11), 963-975. [35] Galbi, D. (1996), "Cross-Border Rent Shifting in International Telecom-munications", Information Economics and Policy, 10 (4), 515-536. [36] Hakim and Lu (1993), "Monopolistic Settlement Agreements in Interna-tional Telecommunications," Information Economics and Policy, 5 (2), 145-57. 141 [37] ITU (1998a), Work of ITU-T Study Group 3 on International Ac-counting Rates, Documents of the plenipotentiary conference, 12.10. -6.11.1998, Minneapolis, Plenary Assembly, Melbourne 14-25. Nov. 1988. http: / / www.itu.int/itudoc/gs/plenipot/pp98 / conf/docs3/index.html [38] ITU (1998b), World Telecommunications Indicators Database '97, In-ternational Telecommunications Union, Geneva. [39] ITU (1999), ITU-D Series Recommendations in Force, Various Imple-mentation Years, www.itu.ch [40] Levine, J. (1998), From Tribal Space to Cyberspace, Ph.D. Dissertation, University of Toronto, Toronto. [41] O'Brian, D. P. (1989), The Uniform Settlements Policy in International Telecommunications: A Non-cooperative Bargaining Model of Interme-diate Product Third Degree Price Discrimination, Ph.D. Dissertation, Northwestern University, Evanston, IL. [42] O E C D (1994), "International Telecommunications Tariffs: charging Practices and Procedures," Information Computer Communications Policy, 34, O E C D , Paris. [43] O E C D (1997), "International Telecommunications Pricing Practices and Principles," Information Computer Communications Policy, 36, O E C D , Paris. [44] Ploman, E. W. (1982), International Law Governing Communications and Information: A Collection of Basic Documents, Greenwood Press, Westport Connecticut. [45] Scanlan, M . (1996), "Why is the International Accounting Rate System in Terminal Decline, and what might be the Consequences?" Telecom-munications Policy, 20 (10), 913-930. [46] Scanlan, M . (1998), "Using Call-back to Demonstrate the Discrimina-tory Nature of the Proportionate Return Rule," Telecommunications Policy, 22 (11), 913-930. [47] Sprint (1998), In the Matter of 1998 Biennial Regulatory Review - Re-form of the International Settlements Policy and Associated Filing Re-quirements, Docket No C C 90-337, Comments of Sprint Corp. 1 4 2 [48] Tyler, M . and S. Bednarczyk (1998), "International Economic Rela-tionships in Telecommunications," Telecommunications Policy, 22 (10), 797-816. [49] W T O (1997), Highlights of Commitments and M.f.n Exemptions Result-ing from the Negotiations, http://www.wto.org/services/tell3.htm. [50] Yun, L . -L . , H.-W. Choi, and B. -H. Ahn (1997), "The Accounting Rev-enue Division in International Telecommunications: Conflicts and Inef-ficiencies," Information Economics and Policy, 9 (1), 71-92. Chapter 4 [1] Acton, J. P. and I. Vogelsang (1992), "Telephone demand over the At-lantic: Evidence from Country-Pair Data," Journal of Industrial Eco-nomics, Vol. 40, 305-323. [2] A T & T (1998), "In the Matter of 1998 Biennial Regulatory Review -Reform of the International Settlements Policy and Associated Filing Requirements," Docket No C C 90-337, Comments of A T & T Corp. [3] Blake, L. , J. Lande (1998), "Trends in the US International Telecom-munications Industry, " Federal Communications Commission, Wash-ington, D.C. [4] Brecher, R. (1998), "Ask the Expert: What is a PICC Charge and Who Pays?" TeleCompete, Vol. 4, No.3, 30-31. [5] Ergas, H . and P. Paterson, (1991), "International Telecommunications Settlement Arrangements: An Unsustainable Inheritance?" Telecommu-nications Policy, February, 29-48. [6] C C I T T (1988), CCITT Blue Book, Volume II - Fascicle III, General Tariff Principle Charging and Accounting in International Telecommu-nications Services, Recommendations of the D Series, IX. [7] Cheong, K . and M . Mullins (1991), "International Telephone Service Imbalances: Accounting Rates and Regulatory Policy," Telecommunica-tions Policy, Vol. 15, 107-118. 143 [8] C R T C (1998), Regulatory Regime for the Provision of International Telecommunication Services, Telecom Decision C R T C 98-17. [9] F C C (1986), Common Carrier Services; Implementation and Scope of the Uniform Settlements Policy, 47 C F R Part 43. [10] F C C (1988a), In the Matter of US Sprint Communications Company, Memorandum Opinion, Authorization and Certificate, 3 F C C Red 1484. [11] F C C (1991), In the Matter of Regulation of International Accounting Rates. CC Docket No 90-337 Phase I, Report and Order, 6 F C C Red 3552. [12] Frieden, R, (1998), "Falling Through the Cracks," Telecommunications Policy, Vol. 22, No. 11, 963-975. [13] Galbi, D. (1996), "An Economic Model of International Interconnec-tion", Draft. [14] Hakim and Lu (1993), "Monopolistic Settlement Agreements in Inter-national Telecommunications," Information Economics and Policy, Vol. 5, No. 2, 145-57. [15] Levine, J. (1998), From Tribal Space to Cyberspace, Ph.D. Dissertation, University of Toronto, Toronto. [16] O'Brian, D. P. (1989), The Uniform Settlements Policy in International Telecommunications: A Noncooperative Bargaining Model of Interme-diate Product Third Degree Price Discrimination, Ph.D. Dissertation, Northwestern University, Evanston, IL. [17] O E C D (1994), "International Telecommunications Tariffs: charging Practices and Procedures," Information Computer Communications Policy, 34, O E C D , Paris. [18] O E C D (1997), "International Telecommunications Pricing Practices and Principles," Information Computer Communications Policy, 36, O E C D , Paris. [19] Ploman, E. W. (1982), International Law Governing Communications and Information: A Collection of Basic Documents, Greenwood Press, Westport Connecticut. 144 [20] Rubinstein, A . (1982), "Perfect Equilibrium in a Bargaining Model," Econometrica 50, 97-109. [21] Rieck, 0 . (2000), "The Proportionate Return Requirement: Regulatory Overkill in international Telecommunications," C S G B Working Paper #99-04, The University of British Columbia, Vancouver. [22] Scanlari, M . (1998), "Using Call-back to Demonstrate the Discrimina-tory Nature of the Proportionate Return Rule," Telecommunications Policy, December, 913-930. [23] Sharkey, W. (1994), "Oligopoly in International Tjlecqmmunications'', in Noam, E. and G. Pogorel (Eds.), Asymmetric Deregulation. The Dy-namics of Telecommunications Policy in Europe and the United States, Ablex Publishing Company, Norwood, N . J . [24] Sprint (1998), In,the Matter of 1998 Biennial Regulatory Review - Re-form of the International Settlements Policy and Associated Filing'Re-quirements, Docket No C C 90-337, Comments of Sprint Corp. [25] Tyler, M . and S. Bednarczyk (1998), "International Economic Relation-ships in Telecommunications," Telecommunications Policy, December, 797-816. [26] Yun, L . -L . , H.-W. Choi, and B. -H. Ahn (1997), "The Accounting Rev-enue Division in International Telecommunications:- Conflicts and Inef-ficiencies," Information Economics and Policy, Vol. 9, 71-92. 145 Appendix A Data Description for Chapter 2 This paper analyzes the annual telephone traffic between the United States and 43 countries over the period 1988 to 1995. Traffic data, that is, the total minutes of US outbound and inbound telephone calls, and the total num-ber of outbound and inbound telephone calls, are obtained from the FCC's Statistics of Communications Common Carriers (FCC, 1989-1997). Data on AT&T's collection charges, including connection charges and marginal 'per minute' charges, are obtained form the SOCC. MCI's and Sprint's collection charges, as well as foreign carriers' collection charges for calls to the US, were provided by the telecommunications consulting group E U R O D A T A . There are several problems arising from the use of publicly collected and reported US pricing data. One is that reported collection charges are list prices, and do not reflect discounts. With increasing competition in inter-national telephone markets throughout the 1990s, discounts have become an increasingly important and widely used pricing tool among US carriers. The dramatic reduction of collection charges in international telephony is in fact 146 entirely reflected in discounts, while listed calling charges have been increas-ing. Discount schemes usually grant a percentage reduction on total charges for calls to one or several routes. The discounted routes may be determined by the caller, as part of an individual calling plan, or they may be determined by calling volume1. Second, US carriers commonly levy higher collection charges during peak hours than during off-peak hours. While the peak and off-peak rates as such are reported and readily available, carriers do not report what portion of traffic is transmitted during peak and off-peak periods. A third and similar problem exists with the carrier specific traffic data. Price data are available on a per carrier basis, whereas traffic data are not. Finally, a further com-plication is due to the recent emergence of resale carriers. Resale carriers are usually small enough such that they do not fall under the FCC's reporting re-quirements. It is well-known that their calling charges are significantly lower than those of the facilities-based carriers, however, there exists neither any publicly collected information on their calling charges, nor on their per-route traffic volumes. Although prices may vary widely depending on call discount schedule, time of the day, or choice of carrier, the levels of the connection charges rel-ative to the marginal charges do not necessarily vary. Let px be the listed connection charge and pd the listed marginal charge per minute of call du-1 For example, the caller may get a certain discount on calls to the country she calls most often. 147 ration. Discount schedules, which give percentage discounts on the total monthly charges on a given route, equally affect both charges. They lower the absolute level of the charges, but do not alter the level of connection charges relative to marginal charges. Also, in the data sample used for this study, ratio of connection charges over marginal charges is typically similar for peak and off peak schedules. Price differences between reporting carri-ers, finally, are entirely reflected in their discount schemes. The reported collection charges of facilities-based carriers are virtually identical. The gross (before discount) revenues can be computed from the carriers list prices and from the number of calls and average call durations reported in the SOCC, R — pxx + pdd. The net (after discount) revenues, R, are directly reported in the SOCC. Hence, the average per route discount is R/R, and the average discounted calling charges can be computed as px — (R/R)px and pd = (R/R)pd. px and pd as computed above are used as price indices throughout the study. Collection charges of resale carriers are not considered in these price indices. Rather, they include the charges that facilities-based carriers ob-tain from resale carriers.' For the purpose of the study, therefore, resellers are regarded as business customers without impact on telecommunications demand. This is equivalent to assuming that the resale market are perfectly competitive, and the cost of resellers are the same as the cost of facilities-based carriers2. 2Moreover, the effect of the Proportionate Return Requirement is ignored. 148 Collection charges for telephone calls in the reverse direction, i.e., from various countries to the US, are not as readily available as US charges. The collected data set is balanced for only 23 out of 48 inter country routes3. The remainder of the Data is taken from a variety of sources. Data on total population, number of main lines, number of fax machines, G D P (as a proxy for income) and the Consumer Price Index are obtained from the ITU World Telecommunications Indicators Database (ITU 1997). Data on Tourism are provided by the World Tourism Organization (WTO 1999). The level of trade between the United States and the 43 countries is computed as the sum of imports and exports, reported by the O E C D (OECD 1998, O E C D 1993). Data on immigration into the US are obtained from the American Statistics Index (U.S. Immigration and Naturalization Service 1996). This study uses the 4-year stock of immigrants. For example, the 1988 figure for immigration from the U K is the total number of immigrants from the U K from 1985 to 1988. Immigration in the reverse direction, i.e., out of the United States, is assumed to be negligible. A l l financial data are deflated using the Consumer Price Index. 3 T h a t means, a complete set of p r ic ing , output and exp lanatory da ta was avai lable w i th regard to traff ic in bo th direct ions on these routes. For the rema in ing 25 routes, col lect ion charges appl icable to U S bound traffic were unavai lable. 149 Appendix B Glossary of Acronyms for Chapter 3 and Chapter 4 D B P Deutsche Bundespost, former German monopolist of postal and telecommunications services E C O Effective Competitive Opportunities, Test developed at the US Federal Communications Commission to assess the degree of telecommunications market liberalization in foreign countries F C C Federal Communications Commission, US telecommunications regulator F R G Federal Republic of Germany ISP International Settlements Policy, a set of regulations developed by the F C C for the international telecommunications market ISR International Simple Resale, service concept that involves the use of a private line (with a connection to the public switched network at one or both ends) for the provision of commercial international telephone services ITU International Telecommunications Union, a civil international organization established to promote standardized telecommuni-cations on a worldwide basis O E C D Organization of Economic Co-operation and Development P R R Proportionate Return Requirement 150 P S T N Public Switched Telephone Network, A telecommunications net-work publicly accessible by telephones, key telephone systems, private branch exchange trunks, and data arrangements P T T In countries having nationalized telephone and telegraph ser-vices, the organization, usually a governmental department, which acts as its nation's common carrier. T L D Telefonica Larga Distancia de Puerto Rico, Puerto Rican Telecommunications Carrier USP Uniform Settlements Policy, F C C Regulation for International Telecommunication Services, set out in F C C (1986), preceded the International Settlements Policy W T O World Trade Organization 151 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0099675/manifest

Comment

Related Items